Merge tag 'drm-xe-next-2025-09-19' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-next

UAPI Changes:
- Drop L3 bank mask reporting from the media GT on Xe3 and later. Only
do that for the primary GT. No userspace needs or uses it for media
and some platforms may report bogus values.
- Add SLPC power_profile sysfs interface with support for base and
power_saving modes (Vinay Belgaumkar, Rodrigo Vivi)
- Add configfs attributes to add post/mid context-switch commands
(Lucas De Marchi)

Cross-subsystem Changes:
- Fix hmm_pfn_to_map_order() usage in gpusvm and refactor APIs to
align with pieces previous handled by xe_hmm (Matthew Auld)

Core Changes:
- Add MEI driver for Late Binding Firmware Update/Upload
(Alexander Usyskin)

Driver Changes:
- Fix GuC CT teardown wrt TLB invalidation (Satyanarayana)
- Fix CCS save/restore on VF (Satyanarayana)
- Increase default GuC crash buffer size (Zhanjun)
- Allow to clear GT stats in debugfs to aid debugging (Matthew Brost)
- Add more SVM GT stats to debugfs (Matthew Brost)
- Fix error handling in VMA attr query (Himal)
- Move sa_info in debugfs to be per tile (Michal Wajdeczko)
- Limit number of retries upon receiving NO_RESPONSE_RETRY from GuC to
avoid endless loop (Michal Wajdeczko)
- Fix configfs handling for survivability_mode undoing user choice when
unbinding the module (Michal Wajdeczko)
- Refactor configfs attribute visibility to future-proof it and stop
exposing survivability_mode if not applicable (Michal Wajdeczko)
- Constify some functions (Harish Chegondi, Michal Wajdeczko)
- Add/extend more HW workarounds for Xe2 and Xe3
(Harish Chegondi, Tangudu Tilak Tirumalesh)
- Replace xe_hmm with gpusvm (Matthew Auld)
- Improve fake pci and WA kunit handling for testing new platforms
(Michal Wajdeczko)
- Reduce unnecessary PTE writes when migrating (Sanjay Yadav)
- Cleanup GuC interface definitions and log message (John Harrison)
- Small improvements around VF CCS (Michal Wajdeczko)
- Enable bus mastering for the I2C controller (Raag Jadav)
- Prefer devm_mutex of hand rolling it (Christophe JAILLET)
- Drop sysfs and debugfs attributes not available for VF (Michal Wajdeczko)
- GuC CT devm actions improvements (Michal Wajdeczko)
- Recommend new GuC versions for PTL and BMG (Julia Filipchuk)
- Improveme driver handling for exhaustive eviction using new
xe_validation wrapper around drm_exec (Thomas Hellström)
- Add and use printk wrappers for tile and device (Michal Wajdeczko)
- Better document workaround handling in Xe (Lucas De Marchi)
- Improvements on ARRAY_SIZE and ERR_CAST usage (Lucas De Marchi,
Fushuai Wang)
- Align CSS firmware headers with the GuC APIs (John Harrison)
- Test GuC to GuC (G2G) communication to aid debug in pre-production
firmware (John Harrison)
- Bail out driver probing if GuC fails to load (John Harrison)
- Allow error injection in xe_pxp_exec_queue_add()
(Daniele Ceraolo Spurio)
- Minor refactors in xe_svm (Shuicheng Lin)
- Fix madvise ioctl error handling (Shuicheng Lin)
- Use attribute groups to simplify sysfs registration
(Michal Wajdeczko)
- Add Late Binding Firmware implementation in Xe to work together with
the MEI component (Badal Nilawar, Daniele Ceraolo Spurio, Rodrigo
Vivi)
- Fix build with CONFIG_MODULES=n (Lucas De Marchi)

Signed-off-by: Dave Airlie <[e-mail]>

From: Lucas De Marchi <[e-mail]>
Link: https://lore.kernel.org/r/c2et6dnkst2apsgt46dklej4nprqdukjosb55grpaknf3pvcxy@t7gtn3hqtp6n

drivers/gpu/drm/drm_gpusvm.c

···
  *
  * This function initializes the GPU SVM.
  *
+ * Note: If only using the simple drm_gpusvm_pages API (get/unmap/free),
+ * then only @gpusvm, @name, and @drm are expected. However, the same base
+ * @gpusvm can also be used with both modes together in which case the full
+ * setup is needed, where the core drm_gpusvm_pages API will simply never use
+ * the other fields.
+ *
  * Return: 0 on success, a negative error code on failure.
  */
 int drm_gpusvm_init(struct drm_gpusvm *gpusvm,
···
 		    const struct drm_gpusvm_ops *ops,
 		    const unsigned long *chunk_sizes, int num_chunks)
 {
-	if (!ops->invalidate || !num_chunks)
-		return -EINVAL;
+	if (mm) {
+		if (!ops->invalidate || !num_chunks)
+			return -EINVAL;
+		mmgrab(mm);
+	} else {
+		/* No full SVM mode, only core drm_gpusvm_pages API. */
+		if (ops || num_chunks || mm_range || notifier_size ||
+		    device_private_page_owner)
+			return -EINVAL;
+	}
 
 	gpusvm->name = name;
 	gpusvm->drm = drm;
···
 	gpusvm->chunk_sizes = chunk_sizes;
 	gpusvm->num_chunks = num_chunks;
 
-	mmgrab(mm);
 	gpusvm->root = RB_ROOT_CACHED;
 	INIT_LIST_HEAD(&gpusvm->notifier_list);
 
···
 			drm_gpusvm_range_remove(gpusvm, range);
 	}
 
-	mmdrop(gpusvm->mm);
+	if (gpusvm->mm)
+		mmdrop(gpusvm->mm);
 	WARN_ON(!RB_EMPTY_ROOT(&gpusvm->root.rb_root));
 }
 EXPORT_SYMBOL_GPL(drm_gpusvm_fini);
···
 	range->itree.start = ALIGN_DOWN(fault_addr, chunk_size);
 	range->itree.last = ALIGN(fault_addr + 1, chunk_size) - 1;
 	INIT_LIST_HEAD(&range->entry);
-	range->notifier_seq = LONG_MAX;
-	range->flags.migrate_devmem = migrate_devmem ? 1 : 0;
+	range->pages.notifier_seq = LONG_MAX;
+	range->pages.flags.migrate_devmem = migrate_devmem ? 1 : 0;
 
 	return range;
+}
+
+/**
+ * drm_gpusvm_hmm_pfn_to_order() - Get the largest CPU mapping order.
+ * @hmm_pfn: The current hmm_pfn.
+ * @hmm_pfn_index: Index of the @hmm_pfn within the pfn array.
+ * @npages: Number of pages within the pfn array i.e the hmm range size.
+ *
+ * To allow skipping PFNs with the same flags (like when they belong to
+ * the same huge PTE) when looping over the pfn array, take a given a hmm_pfn,
+ * and return the largest order that will fit inside the CPU PTE, but also
+ * crucially accounting for the original hmm range boundaries.
+ *
+ * Return: The largest order that will safely fit within the size of the hmm_pfn
+ * CPU PTE.
+ */
+static unsigned int drm_gpusvm_hmm_pfn_to_order(unsigned long hmm_pfn,
+						unsigned long hmm_pfn_index,
+						unsigned long npages)
+{
+	unsigned long size;
+
+	size = 1UL << hmm_pfn_to_map_order(hmm_pfn);
+	size -= (hmm_pfn & ~HMM_PFN_FLAGS) & (size - 1);
+	hmm_pfn_index += size;
+	if (hmm_pfn_index > npages)
+		size -= (hmm_pfn_index - npages);
+
+	return ilog2(size);
 }
 
 /**
···
 			err = -EFAULT;
 			goto err_free;
 		}
-		i += 0x1 << hmm_pfn_to_map_order(pfns[i]);
+		i += 0x1 << drm_gpusvm_hmm_pfn_to_order(pfns[i], i, npages);
 	}
 
 err_free:
···
 EXPORT_SYMBOL_GPL(drm_gpusvm_range_find_or_insert);
 
 /**
- * __drm_gpusvm_range_unmap_pages() - Unmap pages associated with a GPU SVM range (internal)
+ * __drm_gpusvm_unmap_pages() - Unmap pages associated with GPU SVM pages (internal)
  * @gpusvm: Pointer to the GPU SVM structure
- * @range: Pointer to the GPU SVM range structure
+ * @svm_pages: Pointer to the GPU SVM pages structure
  * @npages: Number of pages to unmap
  *
- * This function unmap pages associated with a GPU SVM range. Assumes and
+ * This function unmap pages associated with a GPU SVM pages struct. Assumes and
  * asserts correct locking is in place when called.
  */
-static void __drm_gpusvm_range_unmap_pages(struct drm_gpusvm *gpusvm,
-					   struct drm_gpusvm_range *range,
-					   unsigned long npages)
+static void __drm_gpusvm_unmap_pages(struct drm_gpusvm *gpusvm,
+				     struct drm_gpusvm_pages *svm_pages,
+				     unsigned long npages)
 {
-	unsigned long i, j;
-	struct drm_pagemap *dpagemap = range->dpagemap;
+	struct drm_pagemap *dpagemap = svm_pages->dpagemap;
 	struct device *dev = gpusvm->drm->dev;
+	unsigned long i, j;
 
 	lockdep_assert_held(&gpusvm->notifier_lock);
 
-	if (range->flags.has_dma_mapping) {
-		struct drm_gpusvm_range_flags flags = {
-			.__flags = range->flags.__flags,
+	if (svm_pages->flags.has_dma_mapping) {
+		struct drm_gpusvm_pages_flags flags = {
+			.__flags = svm_pages->flags.__flags,
 		};
 
 		for (i = 0, j = 0; i < npages; j++) {
-			struct drm_pagemap_addr *addr = &range->dma_addr[j];
+			struct drm_pagemap_addr *addr = &svm_pages->dma_addr[j];
 
 			if (addr->proto == DRM_INTERCONNECT_SYSTEM)
 				dma_unmap_page(dev,
···
 		/* WRITE_ONCE pairs with READ_ONCE for opportunistic checks */
 		flags.has_devmem_pages = false;
 		flags.has_dma_mapping = false;
-		WRITE_ONCE(range->flags.__flags, flags.__flags);
+		WRITE_ONCE(svm_pages->flags.__flags, flags.__flags);
 
-		range->dpagemap = NULL;
+		svm_pages->dpagemap = NULL;
 	}
 }
 
 /**
- * drm_gpusvm_range_free_pages() - Free pages associated with a GPU SVM range
+ * __drm_gpusvm_free_pages() - Free dma array associated with GPU SVM pages
  * @gpusvm: Pointer to the GPU SVM structure
- * @range: Pointer to the GPU SVM range structure
+ * @svm_pages: Pointer to the GPU SVM pages structure
  *
  * This function frees the dma address array associated with a GPU SVM range.
  */
-static void drm_gpusvm_range_free_pages(struct drm_gpusvm *gpusvm,
-					struct drm_gpusvm_range *range)
+static void __drm_gpusvm_free_pages(struct drm_gpusvm *gpusvm,
+				    struct drm_gpusvm_pages *svm_pages)
 {
 	lockdep_assert_held(&gpusvm->notifier_lock);
 
-	if (range->dma_addr) {
-		kvfree(range->dma_addr);
-		range->dma_addr = NULL;
+	if (svm_pages->dma_addr) {
+		kvfree(svm_pages->dma_addr);
+		svm_pages->dma_addr = NULL;
 	}
 }
+
+/**
+ * drm_gpusvm_free_pages() - Free dma-mapping associated with GPU SVM pages
+ * struct
+ * @gpusvm: Pointer to the GPU SVM structure
+ * @svm_pages: Pointer to the GPU SVM pages structure
+ * @npages: Number of mapped pages
+ *
+ * This function unmaps and frees the dma address array associated with a GPU
+ * SVM pages struct.
+ */
+void drm_gpusvm_free_pages(struct drm_gpusvm *gpusvm,
+			   struct drm_gpusvm_pages *svm_pages,
+			   unsigned long npages)
+{
+	drm_gpusvm_notifier_lock(gpusvm);
+	__drm_gpusvm_unmap_pages(gpusvm, svm_pages, npages);
+	__drm_gpusvm_free_pages(gpusvm, svm_pages);
+	drm_gpusvm_notifier_unlock(gpusvm);
+}
+EXPORT_SYMBOL_GPL(drm_gpusvm_free_pages);
 
 /**
  * drm_gpusvm_range_remove() - Remove GPU SVM range
···
 		return;
 
 	drm_gpusvm_notifier_lock(gpusvm);
-	__drm_gpusvm_range_unmap_pages(gpusvm, range, npages);
-	drm_gpusvm_range_free_pages(gpusvm, range);
+	__drm_gpusvm_unmap_pages(gpusvm, &range->pages, npages);
+	__drm_gpusvm_free_pages(gpusvm, &range->pages);
 	__drm_gpusvm_range_remove(notifier, range);
 	drm_gpusvm_notifier_unlock(gpusvm);
 
···
 EXPORT_SYMBOL_GPL(drm_gpusvm_range_put);
 
 /**
+ * drm_gpusvm_pages_valid() - GPU SVM range pages valid
+ * @gpusvm: Pointer to the GPU SVM structure
+ * @svm_pages: Pointer to the GPU SVM pages structure
+ *
+ * This function determines if a GPU SVM range pages are valid. Expected be
+ * called holding gpusvm->notifier_lock and as the last step before committing a
+ * GPU binding. This is akin to a notifier seqno check in the HMM documentation
+ * but due to wider notifiers (i.e., notifiers which span multiple ranges) this
+ * function is required for finer grained checking (i.e., per range) if pages
+ * are valid.
+ *
+ * Return: True if GPU SVM range has valid pages, False otherwise
+ */
+static bool drm_gpusvm_pages_valid(struct drm_gpusvm *gpusvm,
+				   struct drm_gpusvm_pages *svm_pages)
+{
+	lockdep_assert_held(&gpusvm->notifier_lock);
+
+	return svm_pages->flags.has_devmem_pages || svm_pages->flags.has_dma_mapping;
+}
+
+/**
  * drm_gpusvm_range_pages_valid() - GPU SVM range pages valid
  * @gpusvm: Pointer to the GPU SVM structure
  * @range: Pointer to the GPU SVM range structure
···
 bool drm_gpusvm_range_pages_valid(struct drm_gpusvm *gpusvm,
 				  struct drm_gpusvm_range *range)
 {
-	lockdep_assert_held(&gpusvm->notifier_lock);
-
-	return range->flags.has_devmem_pages || range->flags.has_dma_mapping;
+	return drm_gpusvm_pages_valid(gpusvm, &range->pages);
 }
 EXPORT_SYMBOL_GPL(drm_gpusvm_range_pages_valid);
 
···
  *
  * Return: True if GPU SVM range has valid pages, False otherwise
  */
-static bool
-drm_gpusvm_range_pages_valid_unlocked(struct drm_gpusvm *gpusvm,
-				      struct drm_gpusvm_range *range)
+static bool drm_gpusvm_pages_valid_unlocked(struct drm_gpusvm *gpusvm,
+					    struct drm_gpusvm_pages *svm_pages)
 {
 	bool pages_valid;
 
-	if (!range->dma_addr)
+	if (!svm_pages->dma_addr)
 		return false;
 
 	drm_gpusvm_notifier_lock(gpusvm);
-	pages_valid = drm_gpusvm_range_pages_valid(gpusvm, range);
+	pages_valid = drm_gpusvm_pages_valid(gpusvm, svm_pages);
 	if (!pages_valid)
-		drm_gpusvm_range_free_pages(gpusvm, range);
+		__drm_gpusvm_free_pages(gpusvm, svm_pages);
 	drm_gpusvm_notifier_unlock(gpusvm);
 
 	return pages_valid;
 }
 
 /**
- * drm_gpusvm_range_get_pages() - Get pages for a GPU SVM range
+ * drm_gpusvm_get_pages() - Get pages and populate GPU SVM pages struct
  * @gpusvm: Pointer to the GPU SVM structure
- * @range: Pointer to the GPU SVM range structure
+ * @svm_pages: The SVM pages to populate. This will contain the dma-addresses
+ * @mm: The mm corresponding to the CPU range
+ * @notifier: The corresponding notifier for the given CPU range
+ * @pages_start: Start CPU address for the pages
+ * @pages_end: End CPU address for the pages (exclusive)
  * @ctx: GPU SVM context
  *
- * This function gets pages for a GPU SVM range and ensures they are mapped for
- * DMA access.
+ * This function gets and maps pages for CPU range and ensures they are
+ * mapped for DMA access.
  *
  * Return: 0 on success, negative error code on failure.
  */
-int drm_gpusvm_range_get_pages(struct drm_gpusvm *gpusvm,
-			       struct drm_gpusvm_range *range,
-			       const struct drm_gpusvm_ctx *ctx)
+int drm_gpusvm_get_pages(struct drm_gpusvm *gpusvm,
+			 struct drm_gpusvm_pages *svm_pages,
+			 struct mm_struct *mm,
+			 struct mmu_interval_notifier *notifier,
+			 unsigned long pages_start, unsigned long pages_end,
+			 const struct drm_gpusvm_ctx *ctx)
 {
-	struct mmu_interval_notifier *notifier = &range->notifier->notifier;
 	struct hmm_range hmm_range = {
 		.default_flags = HMM_PFN_REQ_FAULT | (ctx->read_only ? 0 :
 			HMM_PFN_REQ_WRITE),
 		.notifier = notifier,
-		.start = drm_gpusvm_range_start(range),
-		.end = drm_gpusvm_range_end(range),
+		.start = pages_start,
+		.end = pages_end,
 		.dev_private_owner = gpusvm->device_private_page_owner,
 	};
-	struct mm_struct *mm = gpusvm->mm;
 	void *zdd;
 	unsigned long timeout =
 		jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
 	unsigned long i, j;
-	unsigned long npages = npages_in_range(drm_gpusvm_range_start(range),
-					       drm_gpusvm_range_end(range));
+	unsigned long npages = npages_in_range(pages_start, pages_end);
 	unsigned long num_dma_mapped;
 	unsigned int order = 0;
 	unsigned long *pfns;
 	int err = 0;
 	struct dev_pagemap *pagemap;
 	struct drm_pagemap *dpagemap;
-	struct drm_gpusvm_range_flags flags;
+	struct drm_gpusvm_pages_flags flags;
+	enum dma_data_direction dma_dir = ctx->read_only ? DMA_TO_DEVICE :
+							   DMA_BIDIRECTIONAL;
 
 retry:
 	hmm_range.notifier_seq = mmu_interval_read_begin(notifier);
-	if (drm_gpusvm_range_pages_valid_unlocked(gpusvm, range))
+	if (drm_gpusvm_pages_valid_unlocked(gpusvm, svm_pages))
 		goto set_seqno;
 
 	pfns = kvmalloc_array(npages, sizeof(*pfns), GFP_KERNEL);
···
 	 */
 	drm_gpusvm_notifier_lock(gpusvm);
 
-	flags.__flags = range->flags.__flags;
+	flags.__flags = svm_pages->flags.__flags;
 	if (flags.unmapped) {
 		drm_gpusvm_notifier_unlock(gpusvm);
 		err = -EFAULT;
···
 		goto retry;
 	}
 
-	if (!range->dma_addr) {
+	if (!svm_pages->dma_addr) {
 		/* Unlock and restart mapping to allocate memory. */
 		drm_gpusvm_notifier_unlock(gpusvm);
-		range->dma_addr = kvmalloc_array(npages,
-						 sizeof(*range->dma_addr),
-						 GFP_KERNEL);
-		if (!range->dma_addr) {
+		svm_pages->dma_addr =
+			kvmalloc_array(npages, sizeof(*svm_pages->dma_addr), GFP_KERNEL);
+		if (!svm_pages->dma_addr) {
 			err = -ENOMEM;
 			goto err_free;
 		}
···
 	for (i = 0, j = 0; i < npages; ++j) {
 		struct page *page = hmm_pfn_to_page(pfns[i]);
 
-		order = hmm_pfn_to_map_order(pfns[i]);
+		order = drm_gpusvm_hmm_pfn_to_order(pfns[i], i, npages);
 		if (is_device_private_page(page) ||
 		    is_device_coherent_page(page)) {
 			if (zdd != page->zone_device_data && i > 0) {
···
 					goto err_unmap;
 				}
 			}
-			range->dma_addr[j] =
+			svm_pages->dma_addr[j] =
 				dpagemap->ops->device_map(dpagemap,
 							  gpusvm->drm->dev,
 							  page, order,
-							  DMA_BIDIRECTIONAL);
+							  dma_dir);
 			if (dma_mapping_error(gpusvm->drm->dev,
-					      range->dma_addr[j].addr)) {
+					      svm_pages->dma_addr[j].addr)) {
 				err = -EFAULT;
 				goto err_unmap;
 			}
···
 			addr = dma_map_page(gpusvm->drm->dev,
 					    page, 0,
 					    PAGE_SIZE << order,
-					    DMA_BIDIRECTIONAL);
+					    dma_dir);
 			if (dma_mapping_error(gpusvm->drm->dev, addr)) {
 				err = -EFAULT;
 				goto err_unmap;
 			}
 
-			range->dma_addr[j] = drm_pagemap_addr_encode
+			svm_pages->dma_addr[j] = drm_pagemap_addr_encode
 				(addr, DRM_INTERCONNECT_SYSTEM, order,
-				 DMA_BIDIRECTIONAL);
+				 dma_dir);
 		}
 		i += 1 << order;
 		num_dma_mapped = i;
···
 
 	if (pagemap) {
 		flags.has_devmem_pages = true;
-		range->dpagemap = dpagemap;
+		svm_pages->dpagemap = dpagemap;
 	}
 
 	/* WRITE_ONCE pairs with READ_ONCE for opportunistic checks */
-	WRITE_ONCE(range->flags.__flags, flags.__flags);
+	WRITE_ONCE(svm_pages->flags.__flags, flags.__flags);
 
 	drm_gpusvm_notifier_unlock(gpusvm);
 	kvfree(pfns);
 set_seqno:
-	range->notifier_seq = hmm_range.notifier_seq;
+	svm_pages->notifier_seq = hmm_range.notifier_seq;
 
 	return 0;
 
 err_unmap:
-	__drm_gpusvm_range_unmap_pages(gpusvm, range, num_dma_mapped);
+	__drm_gpusvm_unmap_pages(gpusvm, svm_pages, num_dma_mapped);
 	drm_gpusvm_notifier_unlock(gpusvm);
 err_free:
 	kvfree(pfns);
···
 		goto retry;
 	return err;
 }
+EXPORT_SYMBOL_GPL(drm_gpusvm_get_pages);
+
+/**
+ * drm_gpusvm_range_get_pages() - Get pages for a GPU SVM range
+ * @gpusvm: Pointer to the GPU SVM structure
+ * @range: Pointer to the GPU SVM range structure
+ * @ctx: GPU SVM context
+ *
+ * This function gets pages for a GPU SVM range and ensures they are mapped for
+ * DMA access.
+ *
+ * Return: 0 on success, negative error code on failure.
+ */
+int drm_gpusvm_range_get_pages(struct drm_gpusvm *gpusvm,
+			       struct drm_gpusvm_range *range,
+			       const struct drm_gpusvm_ctx *ctx)
+{
+	return drm_gpusvm_get_pages(gpusvm, &range->pages, gpusvm->mm,
+				    &range->notifier->notifier,
+				    drm_gpusvm_range_start(range),
+				    drm_gpusvm_range_end(range), ctx);
+}
 EXPORT_SYMBOL_GPL(drm_gpusvm_range_get_pages);
 
 /**
+ * drm_gpusvm_unmap_pages() - Unmap GPU svm pages
+ * @gpusvm: Pointer to the GPU SVM structure
+ * @svm_pages: Pointer to the GPU SVM pages structure
+ * @npages: Number of pages in @svm_pages.
+ * @ctx: GPU SVM context
+ *
+ * This function unmaps pages associated with a GPU SVM pages struct. If
+ * @in_notifier is set, it is assumed that gpusvm->notifier_lock is held in
+ * write mode; if it is clear, it acquires gpusvm->notifier_lock in read mode.
+ * Must be called in the invalidate() callback of the corresponding notifier for
+ * IOMMU security model.
+ */
+void drm_gpusvm_unmap_pages(struct drm_gpusvm *gpusvm,
+			    struct drm_gpusvm_pages *svm_pages,
+			    unsigned long npages,
+			    const struct drm_gpusvm_ctx *ctx)
+{
+	if (ctx->in_notifier)
+		lockdep_assert_held_write(&gpusvm->notifier_lock);
+	else
+		drm_gpusvm_notifier_lock(gpusvm);
+
+	__drm_gpusvm_unmap_pages(gpusvm, svm_pages, npages);
+
+	if (!ctx->in_notifier)
+		drm_gpusvm_notifier_unlock(gpusvm);
+}
+EXPORT_SYMBOL_GPL(drm_gpusvm_unmap_pages);
+
+/**
  * drm_gpusvm_range_unmap_pages() - Unmap pages associated with a GPU SVM range
- * drm_gpusvm_range_evict() - Evict GPU SVM range
  * @gpusvm: Pointer to the GPU SVM structure
  * @range: Pointer to the GPU SVM range structure
  * @ctx: GPU SVM context
···
 	unsigned long npages = npages_in_range(drm_gpusvm_range_start(range),
 					       drm_gpusvm_range_end(range));
 
-	if (ctx->in_notifier)
-		lockdep_assert_held_write(&gpusvm->notifier_lock);
-	else
-		drm_gpusvm_notifier_lock(gpusvm);
-
-	__drm_gpusvm_range_unmap_pages(gpusvm, range, npages);
-
-	if (!ctx->in_notifier)
-		drm_gpusvm_notifier_unlock(gpusvm);
+	return drm_gpusvm_unmap_pages(gpusvm, &range->pages, npages, ctx);
 }
 EXPORT_SYMBOL_GPL(drm_gpusvm_range_unmap_pages);
 
···
 {
 	lockdep_assert_held_write(&range->gpusvm->notifier_lock);
 
-	range->flags.unmapped = true;
+	range->pages.flags.unmapped = true;
 	if (drm_gpusvm_range_start(range) < mmu_range->start ||
 	    drm_gpusvm_range_end(range) > mmu_range->end)
-		range->flags.partial_unmap = true;
+		range->pages.flags.partial_unmap = true;
 }
 EXPORT_SYMBOL_GPL(drm_gpusvm_range_set_unmapped);
 

drivers/gpu/drm/xe/Kconfig

···
 	select DRM_TTM
 	select DRM_TTM_HELPER
 	select DRM_EXEC
+	select DRM_GPUSVM if !UML && DEVICE_PRIVATE
 	select DRM_GPUVM
 	select DRM_SCHED
 	select MMU_NOTIFIER
 	select WANT_DEV_COREDUMP
 	select AUXILIARY_BUS
-	select HMM_MIRROR
 	select REGMAP if I2C
 	help
 	  Driver for Intel Xe2 series GPUs and later. Experimental support

drivers/gpu/drm/xe/Kconfig.debug

···
 
 config DRM_XE_USERPTR_INVAL_INJECT
        bool "Inject userptr invalidation -EINVAL errors"
+       depends on DRM_GPUSVM
        default n
        help
          Choose this option when debugging error paths that

drivers/gpu/drm/xe/Makefile

···
 	xe_hw_error.o \
 	xe_hw_fence.o \
 	xe_irq.o \
+	xe_late_bind_fw.o \
 	xe_lrc.o \
 	xe_migrate.o \
 	xe_mmio.o \
···
 	xe_tuning.o \
 	xe_uc.o \
 	xe_uc_fw.o \
+	xe_validation.o \
 	xe_vm.o \
 	xe_vm_madvise.o \
 	xe_vram.o \
···
 	xe_wopcm.o
 
 xe-$(CONFIG_I2C)	+= xe_i2c.o
-xe-$(CONFIG_HMM_MIRROR) += xe_hmm.o
 xe-$(CONFIG_DRM_XE_GPUSVM) += xe_svm.o
+xe-$(CONFIG_DRM_GPUSVM) += xe_userptr.o
 
 # graphics hardware monitoring (HWMON) support
 xe-$(CONFIG_HWMON) += xe_hwmon.o
···
 		xe_gt_stats.o \
 		xe_guc_debugfs.o \
 		xe_huc_debugfs.o \
+		xe_tile_debugfs.o \
 		xe_uc_debugfs.o
 
 	xe-$(CONFIG_PCI_IOV) += xe_gt_sriov_pf_debugfs.o

drivers/gpu/drm/xe/abi/guc_actions_abi.h

···
 	XE_GUC_ACTION_ENTER_S_STATE = 0x501,
 	XE_GUC_ACTION_EXIT_S_STATE = 0x502,
 	XE_GUC_ACTION_GLOBAL_SCHED_POLICY_CHANGE = 0x506,
+	XE_GUC_ACTION_UPDATE_SCHEDULING_POLICIES_KLV = 0x509,
 	XE_GUC_ACTION_SCHED_CONTEXT = 0x1000,
 	XE_GUC_ACTION_SCHED_CONTEXT_MODE_SET = 0x1001,
 	XE_GUC_ACTION_SCHED_CONTEXT_MODE_DONE = 0x1002,
···
 	XE_GUC_ACTION_NOTIFY_FLUSH_LOG_BUFFER_TO_FILE = 0x8003,
 	XE_GUC_ACTION_NOTIFY_CRASH_DUMP_POSTED = 0x8004,
 	XE_GUC_ACTION_NOTIFY_EXCEPTION = 0x8005,
+	XE_GUC_ACTION_TEST_G2G_SEND = 0xF001,
+	XE_GUC_ACTION_TEST_G2G_RECV = 0xF002,
 	XE_GUC_ACTION_LIMIT
 };
 

drivers/gpu/drm/xe/abi/guc_actions_slpc_abi.h

···
 	u8 reserved_mode_definition[4096];
 } __packed;
 
+enum slpc_power_profile {
+	SLPC_POWER_PROFILE_BASE = 0x0,
+	SLPC_POWER_PROFILE_POWER_SAVING = 0x1
+};
+
 /**
  * DOC: SLPC H2G MESSAGE FORMAT
  *

drivers/gpu/drm/xe/abi/guc_klvs_abi.h

···
  *  | 0 | 31:16 | **KEY** - KLV key identifier                                 |
  *  |   |       |   - `GuC Self Config KLVs`_                                  |
  *  |   |       |   - `GuC Opt In Feature KLVs`_                               |
+ *  |   |       |   - `GuC Scheduling Policies KLVs`_                          |
  *  |   |       |   - `GuC VGT Policy KLVs`_                                   |
  *  |   |       |   - `GuC VF Configuration KLVs`_                             |
  *  |   |       |                                                              |
···
 
 #define GUC_KLV_OPT_IN_FEATURE_DYNAMIC_INHIBIT_CONTEXT_SWITCH_KEY 0x4003
 #define GUC_KLV_OPT_IN_FEATURE_DYNAMIC_INHIBIT_CONTEXT_SWITCH_LEN 0u
+
+/**
+ * DOC: GuC Scheduling Policies KLVs
+ *
+ * `GuC KLV`_ keys available for use with UPDATE_SCHEDULING_POLICIES_KLV.
+ *
+ * _`GUC_KLV_SCHEDULING_POLICIES_RENDER_COMPUTE_YIELD` : 0x1001
+ *      Some platforms do not allow concurrent execution of RCS and CCS
+ *      workloads from different address spaces. By default, the GuC prioritizes
+ *      RCS submissions over CCS ones, which can lead to CCS workloads being
+ *      significantly (or completely) starved of execution time. This KLV allows
+ *      the driver to specify a quantum (in ms) and a ratio (percentage value
+ *      between 0 and 100), and the GuC will prioritize the CCS for that
+ *      percentage of each quantum. For example, specifying 100ms and 30% will
+ *      make the GuC prioritize the CCS for 30ms of every 100ms.
+ *      Note that this does not necessarly mean that RCS and CCS engines will
+ *      only be active for their percentage of the quantum, as the restriction
+ *      only kicks in if both classes are fully busy with non-compatible address
+ *      spaces; i.e., if one engine is idle or running the same address space,
+ *      a pending job on the other engine will still be submitted to the HW no
+ *      matter what the ratio is
+ */
+#define GUC_KLV_SCHEDULING_POLICIES_RENDER_COMPUTE_YIELD_KEY	0x1001
+#define GUC_KLV_SCHEDULING_POLICIES_RENDER_COMPUTE_YIELD_LEN	2u
 
 /**
  * DOC: GuC VGT Policy KLVs

drivers/gpu/drm/xe/compat-i915-headers/gem/i915_gem_stolen.h

···
 
 #include "xe_ttm_stolen_mgr.h"
 #include "xe_res_cursor.h"
+#include "xe_validation.h"
 
 struct xe_bo;
 
···
 						       u32 start, u32 end)
 {
 	struct xe_bo *bo;
-	int err;
+	int err = 0;
 	u32 flags = XE_BO_FLAG_PINNED | XE_BO_FLAG_STOLEN;
 
 	if (start < SZ_4K)
···
 		start = ALIGN(start, align);
 	}
 
-	bo = xe_bo_create_locked_range(xe, xe_device_get_root_tile(xe),
-				       NULL, size, start, end,
-				       ttm_bo_type_kernel, flags, 0);
+	bo = xe_bo_create_pin_range_novm(xe, xe_device_get_root_tile(xe),
+					 size, start, end, ttm_bo_type_kernel, flags);
 	if (IS_ERR(bo)) {
 		err = PTR_ERR(bo);
 		bo = NULL;
 		return err;
-	}
-	err = xe_bo_pin(bo);
-	xe_bo_unlock_vm_held(bo);
-
-	if (err) {
-		xe_bo_put(fb->bo);
-		bo = NULL;
 	}
 
 	fb->bo = bo;

drivers/gpu/drm/xe/display/intel_fbdev_fb.c

···
 	obj = ERR_PTR(-ENODEV);
 
 	if (!IS_DGFX(xe) && !XE_GT_WA(xe_root_mmio_gt(xe), 22019338487_display)) {
-		obj = xe_bo_create_pin_map(xe, xe_device_get_root_tile(xe),
-					   NULL, size,
-					   ttm_bo_type_kernel, XE_BO_FLAG_SCANOUT |
-					   XE_BO_FLAG_STOLEN |
-					   XE_BO_FLAG_GGTT);
+		obj = xe_bo_create_pin_map_novm(xe, xe_device_get_root_tile(xe),
+						size,
+						ttm_bo_type_kernel, XE_BO_FLAG_SCANOUT |
+						XE_BO_FLAG_STOLEN |
+						XE_BO_FLAG_GGTT, false);
 		if (!IS_ERR(obj))
 			drm_info(&xe->drm, "Allocated fbdev into stolen\n");
 		else
···
 	}
 
 	if (IS_ERR(obj)) {
-		obj = xe_bo_create_pin_map(xe, xe_device_get_root_tile(xe), NULL, size,
-					   ttm_bo_type_kernel, XE_BO_FLAG_SCANOUT |
-					   XE_BO_FLAG_VRAM_IF_DGFX(xe_device_get_root_tile(xe)) |
-					   XE_BO_FLAG_GGTT);
+		obj = xe_bo_create_pin_map_novm(xe, xe_device_get_root_tile(xe), size,
+						ttm_bo_type_kernel, XE_BO_FLAG_SCANOUT |
+						XE_BO_FLAG_VRAM_IF_DGFX(xe_device_get_root_tile(xe)) |
+						XE_BO_FLAG_GGTT, false);
 	}
 
 	if (IS_ERR(obj)) {

drivers/gpu/drm/xe/display/xe_dsb_buffer.c

···
 		return false;
 
 	/* Set scanout flag for WC mapping */
-	obj = xe_bo_create_pin_map(xe, xe_device_get_root_tile(xe),
-				   NULL, PAGE_ALIGN(size),
-				   ttm_bo_type_kernel,
-				   XE_BO_FLAG_VRAM_IF_DGFX(xe_device_get_root_tile(xe)) |
-				   XE_BO_FLAG_SCANOUT | XE_BO_FLAG_GGTT);
+	obj = xe_bo_create_pin_map_novm(xe, xe_device_get_root_tile(xe),
+					PAGE_ALIGN(size),
+					ttm_bo_type_kernel,
+					XE_BO_FLAG_VRAM_IF_DGFX(xe_device_get_root_tile(xe)) |
+					XE_BO_FLAG_SCANOUT | XE_BO_FLAG_GGTT, false);
 	if (IS_ERR(obj)) {
 		kfree(vma);
 		return false;

drivers/gpu/drm/xe/display/xe_fb_pin.c

···
 				 XE_PAGE_SIZE);
 
 	if (IS_DGFX(xe))
-		dpt = xe_bo_create_pin_map_at_aligned(xe, tile0, NULL,
-						      dpt_size, ~0ull,
-						      ttm_bo_type_kernel,
-						      XE_BO_FLAG_VRAM0 |
-						      XE_BO_FLAG_GGTT |
-						      XE_BO_FLAG_PAGETABLE,
-						      alignment);
+		dpt = xe_bo_create_pin_map_at_novm(xe, tile0,
+						   dpt_size, ~0ull,
+						   ttm_bo_type_kernel,
+						   XE_BO_FLAG_VRAM0 |
+						   XE_BO_FLAG_GGTT |
+						   XE_BO_FLAG_PAGETABLE,
+						   alignment, false);
 	else
-		dpt = xe_bo_create_pin_map_at_aligned(xe, tile0, NULL,
-						      dpt_size,  ~0ull,
-						      ttm_bo_type_kernel,
-						      XE_BO_FLAG_STOLEN |
-						      XE_BO_FLAG_GGTT |
-						      XE_BO_FLAG_PAGETABLE,
-						      alignment);
+		dpt = xe_bo_create_pin_map_at_novm(xe, tile0,
+						   dpt_size,  ~0ull,
+						   ttm_bo_type_kernel,
+						   XE_BO_FLAG_STOLEN |
+						   XE_BO_FLAG_GGTT |
+						   XE_BO_FLAG_PAGETABLE,
+						   alignment, false);
 	if (IS_ERR(dpt))
-		dpt = xe_bo_create_pin_map_at_aligned(xe, tile0, NULL,
-						      dpt_size,  ~0ull,
-						      ttm_bo_type_kernel,
-						      XE_BO_FLAG_SYSTEM |
-						      XE_BO_FLAG_GGTT |
-						      XE_BO_FLAG_PAGETABLE,
-						      alignment);
+		dpt = xe_bo_create_pin_map_at_novm(xe, tile0,
+						   dpt_size,  ~0ull,
+						   ttm_bo_type_kernel,
+						   XE_BO_FLAG_SYSTEM |
+						   XE_BO_FLAG_GGTT |
+						   XE_BO_FLAG_PAGETABLE,
+						   alignment, false);
 	if (IS_ERR(dpt))
 		return PTR_ERR(dpt);
 
···
 	struct i915_vma *vma = kzalloc(sizeof(*vma), GFP_KERNEL);
 	struct drm_gem_object *obj = intel_fb_bo(&fb->base);
 	struct xe_bo *bo = gem_to_xe_bo(obj);
-	int ret;
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
+	int ret = 0;
 
 	if (!vma)
 		return ERR_PTR(-ENODEV);
···
 	 * Pin the framebuffer, we can't use xe_bo_(un)pin functions as the
 	 * assumptions are incorrect for framebuffers
 	 */
-	ret = ttm_bo_reserve(&bo->ttm, false, false, NULL);
-	if (ret)
-		goto err;
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {.interruptible = true},
+			    ret) {
+		ret = drm_exec_lock_obj(&exec, &bo->ttm.base);
+		drm_exec_retry_on_contention(&exec);
+		if (ret)
+			break;
 
-	if (IS_DGFX(xe))
-		ret = xe_bo_migrate(bo, XE_PL_VRAM0);
-	else
-		ret = xe_bo_validate(bo, NULL, true);
-	if (!ret)
-		ttm_bo_pin(&bo->ttm);
-	ttm_bo_unreserve(&bo->ttm);
+		if (IS_DGFX(xe))
+			ret = xe_bo_migrate(bo, XE_PL_VRAM0, NULL, &exec);
+		else
+			ret = xe_bo_validate(bo, NULL, true, &exec);
+		drm_exec_retry_on_contention(&exec);
+		xe_validation_retry_on_oom(&ctx, &ret);
+		if (!ret)
+			ttm_bo_pin(&bo->ttm);
+	}
 	if (ret)
 		goto err;
 

drivers/gpu/drm/xe/display/xe_hdcp_gsc.c

···
 	int ret = 0;
 
 	/* allocate object of two page for HDCP command memory and store it */
-	bo = xe_bo_create_pin_map(xe, xe_device_get_root_tile(xe), NULL, PAGE_SIZE * 2,
-				  ttm_bo_type_kernel,
-				  XE_BO_FLAG_SYSTEM |
-				  XE_BO_FLAG_GGTT);
+	bo = xe_bo_create_pin_map_novm(xe, xe_device_get_root_tile(xe), PAGE_SIZE * 2,
+				       ttm_bo_type_kernel,
+				       XE_BO_FLAG_SYSTEM |
+				       XE_BO_FLAG_GGTT, false);
 
 	if (IS_ERR(bo)) {
 		drm_err(&xe->drm, "Failed to allocate bo for HDCP streaming command!\n");

drivers/gpu/drm/xe/display/xe_plane_initial.c

···
 			page_size);
 	size -= base;
 
-	bo = xe_bo_create_pin_map_at(xe, tile0, NULL, size, phys_base,
-				     ttm_bo_type_kernel, flags);
+	bo = xe_bo_create_pin_map_at_novm(xe, tile0, size, phys_base,
+					  ttm_bo_type_kernel, flags, 0, false);
 	if (IS_ERR(bo)) {
 		drm_dbg(&xe->drm,
 			"Failed to create bo phys_base=%pa size %u with flags %x: %li\n",

drivers/gpu/drm/xe/regs/xe_gt_regs.h

···
 
 #define TDL_CHICKEN				XE_REG_MCR(0xe5f4, XE_REG_OPTION_MASKED)
 #define   QID_WAIT_FOR_THREAD_NOT_RUN_DISABLE	REG_BIT(12)
+#define   EUSTALL_PERF_SAMPLING_DISABLE		REG_BIT(5)
 
 #define LSC_CHICKEN_BIT_0			XE_REG_MCR(0xe7c8)
 #define   DISABLE_D8_D16_COASLESCE		REG_BIT(30)

drivers/gpu/drm/xe/regs/xe_lrc_layout.h

···
 #define INDIRECT_CTX_RING_START_UDW	(0x08 + 1)
 #define INDIRECT_CTX_RING_CTL		(0x0a + 1)
 
-#define CTX_INDIRECT_CTX_OFFSET_MASK	REG_GENMASK(15, 6)
-#define CTX_INDIRECT_CTX_OFFSET_DEFAULT	REG_FIELD_PREP(CTX_INDIRECT_CTX_OFFSET_MASK, 0xd)
-
 #endif

drivers/gpu/drm/xe/tests/xe_bo.c

···
 
 static int ccs_test_migrate(struct xe_tile *tile, struct xe_bo *bo,
 			    bool clear, u64 get_val, u64 assign_val,
-			    struct kunit *test)
+			    struct kunit *test, struct drm_exec *exec)
 {
 	struct dma_fence *fence;
 	struct ttm_tt *ttm;
···
 	u32 offset;
 
 	/* Move bo to VRAM if not already there. */
-	ret = xe_bo_validate(bo, NULL, false);
+	ret = xe_bo_validate(bo, NULL, false, exec);
 	if (ret) {
 		KUNIT_FAIL(test, "Failed to validate bo.\n");
 		return ret;
···
 	}
 
 	/* Evict to system. CCS data should be copied. */
-	ret = xe_bo_evict(bo);
+	ret = xe_bo_evict(bo, exec);
 	if (ret) {
 		KUNIT_FAIL(test, "Failed to evict bo.\n");
 		return ret;
···
 
 	/* TODO: Sanity check */
 	unsigned int bo_flags = XE_BO_FLAG_VRAM_IF_DGFX(tile);
+	struct drm_exec *exec = XE_VALIDATION_OPT_OUT;
 
 	if (IS_DGFX(xe))
 		kunit_info(test, "Testing vram id %u\n", tile->id);
 	else
 		kunit_info(test, "Testing system memory\n");
 
-	bo = xe_bo_create_user(xe, NULL, NULL, SZ_1M, DRM_XE_GEM_CPU_CACHING_WC,
-			       bo_flags);
+	bo = xe_bo_create_user(xe, NULL, SZ_1M, DRM_XE_GEM_CPU_CACHING_WC,
+			       bo_flags, exec);
 	if (IS_ERR(bo)) {
 		KUNIT_FAIL(test, "Failed to create bo.\n");
 		return;
···
 
 	kunit_info(test, "Verifying that CCS data is cleared on creation.\n");
 	ret = ccs_test_migrate(tile, bo, false, 0ULL, 0xdeadbeefdeadbeefULL,
-			       test);
+			       test, exec);
 	if (ret)
 		goto out_unlock;
 
 	kunit_info(test, "Verifying that CCS data survives migration.\n");
 	ret = ccs_test_migrate(tile, bo, false, 0xdeadbeefdeadbeefULL,
-			       0xdeadbeefdeadbeefULL, test);
+			       0xdeadbeefdeadbeefULL, test, exec);
 	if (ret)
 		goto out_unlock;
 
 	kunit_info(test, "Verifying that CCS data can be properly cleared.\n");
-	ret = ccs_test_migrate(tile, bo, true, 0ULL, 0ULL, test);
+	ret = ccs_test_migrate(tile, bo, true, 0ULL, 0ULL, test, exec);
 
 out_unlock:
 	xe_bo_unlock(bo);
···
 	struct xe_bo *bo, *external;
 	unsigned int bo_flags = XE_BO_FLAG_VRAM_IF_DGFX(tile);
 	struct xe_vm *vm = xe_migrate_get_vm(xe_device_get_root_tile(xe)->migrate);
+	struct drm_exec *exec = XE_VALIDATION_OPT_OUT;
 	struct xe_gt *__gt;
 	int err, i, id;
 
···
 
 	for (i = 0; i < 2; ++i) {
 		xe_vm_lock(vm, false);
-		bo = xe_bo_create_user(xe, NULL, vm, 0x10000,
+		bo = xe_bo_create_user(xe, vm, 0x10000,
 				       DRM_XE_GEM_CPU_CACHING_WC,
-				       bo_flags);
+				       bo_flags, exec);
 		xe_vm_unlock(vm);
 		if (IS_ERR(bo)) {
 			KUNIT_FAIL(test, "bo create err=%pe\n", bo);
 			break;
 		}
 
-		external = xe_bo_create_user(xe, NULL, NULL, 0x10000,
+		external = xe_bo_create_user(xe, NULL, 0x10000,
 					     DRM_XE_GEM_CPU_CACHING_WC,
-					     bo_flags);
+					     bo_flags, NULL);
 		if (IS_ERR(external)) {
 			KUNIT_FAIL(test, "external bo create err=%pe\n", external);
 			goto cleanup_bo;
 		}
 
 		xe_bo_lock(external, false);
-		err = xe_bo_pin_external(external, false);
+		err = xe_bo_pin_external(external, false, exec);
 		xe_bo_unlock(external);
 		if (err) {
 			KUNIT_FAIL(test, "external bo pin err=%pe\n",
···
 		if (i) {
 			down_read(&vm->lock);
 			xe_vm_lock(vm, false);
-			err = xe_bo_validate(bo, bo->vm, false);
+			err = xe_bo_validate(bo, bo->vm, false, exec);
 			xe_vm_unlock(vm);
 			up_read(&vm->lock);
 			if (err) {
···
 				goto cleanup_all;
 			}
 			xe_bo_lock(external, false);
-			err = xe_bo_validate(external, NULL, false);
+			err = xe_bo_validate(external, NULL, false, exec);
 			xe_bo_unlock(external);
 			if (err) {
 				KUNIT_FAIL(test, "external bo valid err=%pe\n",
···
 		INIT_LIST_HEAD(&link->link);
 
 		/* We can create bos using WC caching here. But it is slower. */
-		bo = xe_bo_create_user(xe, NULL, NULL, XE_BO_SHRINK_SIZE,
+		bo = xe_bo_create_user(xe, NULL, XE_BO_SHRINK_SIZE,
 				       DRM_XE_GEM_CPU_CACHING_WB,
-				       XE_BO_FLAG_SYSTEM);
+				       XE_BO_FLAG_SYSTEM, NULL);
 		if (IS_ERR(bo)) {
 			if (bo != ERR_PTR(-ENOMEM) && bo != ERR_PTR(-ENOSPC) &&
 			    bo != ERR_PTR(-EINTR) && bo != ERR_PTR(-ERESTARTSYS))

drivers/gpu/drm/xe/tests/xe_dma_buf.c

···
 }
 
 static void check_residency(struct kunit *test, struct xe_bo *exported,
-			    struct xe_bo *imported, struct dma_buf *dmabuf)
+			    struct xe_bo *imported, struct dma_buf *dmabuf,
+			    struct drm_exec *exec)
 {
 	struct dma_buf_test_params *params = to_dma_buf_test_params(test->priv);
 	u32 mem_type;
···
 	 * importer is on a different device. If they're on the same device,
 	 * the exporter and the importer should be the same bo.
 	 */
-	ret = xe_bo_evict(exported);
+	ret = xe_bo_evict(exported, exec);
 	if (ret) {
 		if (ret != -EINTR && ret != -ERESTARTSYS)
 			KUNIT_FAIL(test, "Evicting exporter failed with err=%d.\n",
···
 	}
 
 	/* Re-validate the importer. This should move also exporter in. */
-	ret = xe_bo_validate(imported, NULL, false);
+	ret = xe_bo_validate(imported, NULL, false, exec);
 	if (ret) {
 		if (ret != -EINTR && ret != -ERESTARTSYS)
 			KUNIT_FAIL(test, "Validating importer failed with err=%d.\n",
···
 		size = SZ_64K;
 
 	kunit_info(test, "running %s\n", __func__);
-	bo = xe_bo_create_user(xe, NULL, NULL, size, DRM_XE_GEM_CPU_CACHING_WC,
-			       params->mem_mask);
+	bo = xe_bo_create_user(xe, NULL, size, DRM_XE_GEM_CPU_CACHING_WC,
+			       params->mem_mask, NULL);
 	if (IS_ERR(bo)) {
 		KUNIT_FAIL(test, "xe_bo_create() failed with err=%ld\n",
 			   PTR_ERR(bo));
···
 			KUNIT_FAIL(test,
 				   "xe_gem_prime_import() succeeded when it shouldn't have\n");
 		} else {
+			struct drm_exec *exec = XE_VALIDATION_OPT_OUT;
 			int err;
 
 			/* Is everything where we expect it to be? */
 			xe_bo_lock(import_bo, false);
-			err = xe_bo_validate(import_bo, NULL, false);
+			err = xe_bo_validate(import_bo, NULL, false, exec);
 
 			/* Pinning in VRAM is not allowed. */
 			if (!is_dynamic(params) &&
···
 						  err == -ERESTARTSYS);
 
 			if (!err)
-				check_residency(test, bo, import_bo, dmabuf);
+				check_residency(test, bo, import_bo, dmabuf, exec);
 			xe_bo_unlock(import_bo);
 		}
 		drm_gem_object_put(import);

drivers/gpu/drm/xe/tests/xe_guc_g2g_test.c

···
+// SPDX-License-Identifier: GPL-2.0 AND MIT
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#include <linux/delay.h>
+
+#include <kunit/test.h>
+#include <kunit/visibility.h>
+
+#include "tests/xe_kunit_helpers.h"
+#include "tests/xe_pci_test.h"
+#include "tests/xe_test.h"
+
+#include "xe_bo.h"
+#include "xe_device.h"
+#include "xe_pm.h"
+
+/*
+ * There are different ways to allocate the G2G buffers. The plan for this test
+ * is to make sure that all the possible options work. The particular option
+ * chosen by the driver may vary from one platform to another, it may also change
+ * with time. So to ensure consistency of testing, the relevant driver code is
+ * replicated here to guarantee it won't change without the test being updated
+ * to keep testing the other options.
+ *
+ * In order to test the actual code being used by the driver, there is also the
+ * 'default' scheme. That will use the official driver routines to test whatever
+ * method the driver is using on the current platform at the current time.
+ */
+enum {
+	/* Driver defined allocation scheme */
+	G2G_CTB_TYPE_DEFAULT,
+	/* Single buffer in host memory */
+	G2G_CTB_TYPE_HOST,
+	/* Single buffer in a specific tile, loops across all tiles */
+	G2G_CTB_TYPE_TILE,
+};
+
+/*
+ * Payload is opaque to GuC. So KMD can define any structure or size it wants.
+ */
+struct g2g_test_payload  {
+	u32 tx_dev;
+	u32 tx_tile;
+	u32 rx_dev;
+	u32 rx_tile;
+	u32 seqno;
+};
+
+static void g2g_test_send(struct kunit *test, struct xe_guc *guc,
+			  u32 far_tile, u32 far_dev,
+			  struct g2g_test_payload *payload)
+{
+	struct xe_device *xe = guc_to_xe(guc);
+	struct xe_gt *gt = guc_to_gt(guc);
+	u32 *action, total;
+	size_t payload_len;
+	int ret;
+
+	static_assert(IS_ALIGNED(sizeof(*payload), sizeof(u32)));
+	payload_len = sizeof(*payload) / sizeof(u32);
+
+	total = 4 + payload_len;
+	action = kunit_kmalloc_array(test, total, sizeof(*action), GFP_KERNEL);
+	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, action);
+
+	action[0] = XE_GUC_ACTION_TEST_G2G_SEND;
+	action[1] = far_tile;
+	action[2] = far_dev;
+	action[3] = payload_len;
+	memcpy(action + 4, payload, payload_len * sizeof(u32));
+
+	atomic_inc(&xe->g2g_test_count);
+
+	/*
+	 * Should specify the expected response notification here. Problem is that
+	 * the response will be coming from a different GuC. By the end, it should
+	 * all add up as long as an equal number of messages are sent from each GuC
+	 * and to each GuC. However, in the middle negative reservation space errors
+	 * and such like can occur. Rather than add intrusive changes to the CT layer
+	 * it is simpler to just not bother counting it at all. The system should be
+	 * idle when running the selftest, and the selftest's notification total size
+	 * is well within the G2H allocation size. So there should be no issues with
+	 * needing to block for space, which is all the tracking code is really for.
+	 */
+	ret = xe_guc_ct_send(&guc->ct, action, total, 0, 0);
+	kunit_kfree(test, action);
+	KUNIT_ASSERT_EQ_MSG(test, 0, ret, "G2G send failed: %d [%d:%d -> %d:%d]\n", ret,
+			    gt_to_tile(gt)->id, G2G_DEV(gt), far_tile, far_dev);
+}
+
+/*
+ * NB: Can't use KUNIT_ASSERT and friends in here as this is called asynchronously
+ * from the G2H notification handler. Need that to actually complete rather than
+ * thread-abort in order to keep the rest of the driver alive!
+ */
+int xe_guc_g2g_test_notification(struct xe_guc *guc, u32 *msg, u32 len)
+{
+	struct xe_device *xe = guc_to_xe(guc);
+	struct xe_gt *rx_gt = guc_to_gt(guc), *test_gt, *tx_gt = NULL;
+	u32 tx_tile, tx_dev, rx_tile, rx_dev, idx, got_len;
+	struct g2g_test_payload *payload;
+	size_t payload_len;
+	int ret = 0, i;
+
+	payload_len = sizeof(*payload) / sizeof(u32);
+
+	if (unlikely(len != (G2H_LEN_DW_G2G_NOTIFY_MIN + payload_len))) {
+		xe_gt_err(rx_gt, "G2G test notification invalid length %u", len);
+		ret = -EPROTO;
+		goto done;
+	}
+
+	tx_tile = msg[0];
+	tx_dev = msg[1];
+	got_len = msg[2];
+	payload = (struct g2g_test_payload *)(msg + 3);
+
+	rx_tile = gt_to_tile(rx_gt)->id;
+	rx_dev = G2G_DEV(rx_gt);
+
+	if (got_len != payload_len) {
+		xe_gt_err(rx_gt, "G2G: Invalid payload length: %u vs %zu\n", got_len, payload_len);
+		ret = -EPROTO;
+		goto done;
+	}
+
+	if (payload->tx_dev != tx_dev || payload->tx_tile != tx_tile ||
+	    payload->rx_dev != rx_dev || payload->rx_tile != rx_tile) {
+		xe_gt_err(rx_gt, "G2G: Invalid payload: %d:%d -> %d:%d vs %d:%d -> %d:%d! [%d]\n",
+			  payload->tx_tile, payload->tx_dev, payload->rx_tile, payload->rx_dev,
+			  tx_tile, tx_dev, rx_tile, rx_dev, payload->seqno);
+		ret = -EPROTO;
+		goto done;
+	}
+
+	if (!xe->g2g_test_array) {
+		xe_gt_err(rx_gt, "G2G: Missing test array!\n");
+		ret = -ENOMEM;
+		goto done;
+	}
+
+	for_each_gt(test_gt, xe, i) {
+		if (gt_to_tile(test_gt)->id != tx_tile)
+			continue;
+
+		if (G2G_DEV(test_gt) != tx_dev)
+			continue;
+
+		if (tx_gt) {
+			xe_gt_err(rx_gt, "G2G: Got duplicate TX GTs: %d vs %d for %d:%d!\n",
+				  tx_gt->info.id, test_gt->info.id, tx_tile, tx_dev);
+			ret = -EINVAL;
+			goto done;
+		}
+
+		tx_gt = test_gt;
+	}
+	if (!tx_gt) {
+		xe_gt_err(rx_gt, "G2G: Failed to find a TX GT for %d:%d!\n", tx_tile, tx_dev);
+		ret = -EINVAL;
+		goto done;
+	}
+
+	idx = (tx_gt->info.id * xe->info.gt_count) + rx_gt->info.id;
+
+	if (xe->g2g_test_array[idx] != payload->seqno - 1) {
+		xe_gt_err(rx_gt, "G2G: Seqno mismatch %d vs %d for %d:%d -> %d:%d!\n",
+			  xe->g2g_test_array[idx], payload->seqno - 1,
+			  tx_tile, tx_dev, rx_tile, rx_dev);
+		ret = -EINVAL;
+		goto done;
+	}
+
+	xe->g2g_test_array[idx] = payload->seqno;
+
+done:
+	atomic_dec(&xe->g2g_test_count);
+	return ret;
+}
+
+/*
+ * Send the given seqno from all GuCs to all other GuCs in tile/GT order
+ */
+static void g2g_test_in_order(struct kunit *test, struct xe_device *xe, u32 seqno)
+{
+	struct xe_gt *near_gt, *far_gt;
+	int i, j;
+
+	for_each_gt(near_gt, xe, i) {
+		u32 near_tile = gt_to_tile(near_gt)->id;
+		u32 near_dev = G2G_DEV(near_gt);
+
+		for_each_gt(far_gt, xe, j) {
+			u32 far_tile = gt_to_tile(far_gt)->id;
+			u32 far_dev = G2G_DEV(far_gt);
+			struct g2g_test_payload payload;
+
+			if (far_gt->info.id == near_gt->info.id)
+				continue;
+
+			payload.tx_dev = near_dev;
+			payload.tx_tile = near_tile;
+			payload.rx_dev = far_dev;
+			payload.rx_tile = far_tile;
+			payload.seqno = seqno;
+			g2g_test_send(test, &near_gt->uc.guc, far_tile, far_dev, &payload);
+		}
+	}
+}
+
+#define WAIT_TIME_MS	100
+#define WAIT_COUNT	(1000 / WAIT_TIME_MS)
+
+static void g2g_wait_for_complete(void *_xe)
+{
+	struct xe_device *xe = (struct xe_device *)_xe;
+	struct kunit *test = kunit_get_current_test();
+	int wait = 0;
+
+	/* Wait for all G2H messages to be received */
+	while (atomic_read(&xe->g2g_test_count)) {
+		if (++wait > WAIT_COUNT)
+			break;
+
+		msleep(WAIT_TIME_MS);
+	}
+
+	KUNIT_ASSERT_EQ_MSG(test, 0, atomic_read(&xe->g2g_test_count),
+			    "Timed out waiting for notifications\n");
+	kunit_info(test, "Got all notifications back\n");
+}
+
+#undef WAIT_TIME_MS
+#undef WAIT_COUNT
+
+static void g2g_clean_array(void *_xe)
+{
+	struct xe_device *xe = (struct xe_device *)_xe;
+
+	xe->g2g_test_array = NULL;
+}
+
+#define NUM_LOOPS	16
+
+static void g2g_run_test(struct kunit *test, struct xe_device *xe)
+{
+	u32 seqno, max_array;
+	int ret, i, j;
+
+	max_array = xe->info.gt_count * xe->info.gt_count;
+	xe->g2g_test_array = kunit_kcalloc(test, max_array, sizeof(u32), GFP_KERNEL);
+	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, xe->g2g_test_array);
+
+	ret = kunit_add_action_or_reset(test, g2g_clean_array, xe);
+	KUNIT_ASSERT_EQ_MSG(test, 0, ret, "Failed to register clean up action\n");
+
+	/*
+	 * Send incrementing seqnos from all GuCs to all other GuCs in tile/GT order.
+	 * Tile/GT order doesn't really mean anything to the hardware but it is going
+	 * to be a fixed sequence every time.
+	 *
+	 * Verify that each one comes back having taken the correct route.
+	 */
+	ret = kunit_add_action(test, g2g_wait_for_complete, xe);
+	KUNIT_ASSERT_EQ_MSG(test, 0, ret, "Failed to register clean up action\n");
+	for (seqno = 1; seqno < NUM_LOOPS; seqno++)
+		g2g_test_in_order(test, xe, seqno);
+	seqno--;
+
+	kunit_release_action(test, &g2g_wait_for_complete, xe);
+
+	/* Check for the final seqno in each slot */
+	for (i = 0; i < xe->info.gt_count; i++) {
+		for (j = 0; j < xe->info.gt_count; j++) {
+			u32 idx = (j * xe->info.gt_count) + i;
+
+			if (i == j)
+				KUNIT_ASSERT_EQ_MSG(test, 0, xe->g2g_test_array[idx],
+						    "identity seqno modified: %d for %dx%d!\n",
+						    xe->g2g_test_array[idx], i, j);
+			else
+				KUNIT_ASSERT_EQ_MSG(test, seqno, xe->g2g_test_array[idx],
+						    "invalid seqno: %d vs %d for %dx%d!\n",
+						    xe->g2g_test_array[idx], seqno, i, j);
+		}
+	}
+
+	kunit_kfree(test, xe->g2g_test_array);
+	kunit_release_action(test, &g2g_clean_array, xe);
+
+	kunit_info(test, "Test passed\n");
+}
+
+#undef NUM_LOOPS
+
+static void g2g_ct_stop(struct xe_guc *guc)
+{
+	struct xe_gt *remote_gt, *gt = guc_to_gt(guc);
+	struct xe_device *xe = gt_to_xe(gt);
+	int i, t;
+
+	for_each_gt(remote_gt, xe, i) {
+		u32 tile, dev;
+
+		if (remote_gt->info.id == gt->info.id)
+			continue;
+
+		tile = gt_to_tile(remote_gt)->id;
+		dev = G2G_DEV(remote_gt);
+
+		for (t = 0; t < XE_G2G_TYPE_LIMIT; t++)
+			guc_g2g_deregister(guc, tile, dev, t);
+	}
+}
+
+/* Size of a single allocation that contains all G2G CTBs across all GTs */
+static u32 g2g_ctb_size(struct kunit *test, struct xe_device *xe)
+{
+	unsigned int count = xe->info.gt_count;
+	u32 num_channels = (count * (count - 1)) / 2;
+
+	kunit_info(test, "Size: (%d * %d / 2) * %d * 0x%08X + 0x%08X => 0x%08X [%d]\n",
+		   count, count - 1, XE_G2G_TYPE_LIMIT, G2G_BUFFER_SIZE, G2G_DESC_AREA_SIZE,
+		   num_channels * XE_G2G_TYPE_LIMIT * G2G_BUFFER_SIZE + G2G_DESC_AREA_SIZE,
+		   num_channels * XE_G2G_TYPE_LIMIT);
+
+	return num_channels * XE_G2G_TYPE_LIMIT * G2G_BUFFER_SIZE + G2G_DESC_AREA_SIZE;
+}
+
+/*
+ * Use the driver's regular CTB allocation scheme.
+ */
+static void g2g_alloc_default(struct kunit *test, struct xe_device *xe)
+{
+	struct xe_gt *gt;
+	int i;
+
+	kunit_info(test, "Default [tiles = %d, GTs = %d]\n",
+		   xe->info.tile_count, xe->info.gt_count);
+
+	for_each_gt(gt, xe, i) {
+		struct xe_guc *guc = &gt->uc.guc;
+		int ret;
+
+		ret = guc_g2g_alloc(guc);
+		KUNIT_ASSERT_EQ_MSG(test, 0, ret, "G2G alloc failed: %pe", ERR_PTR(ret));
+		continue;
+	}
+}
+
+static void g2g_distribute(struct kunit *test, struct xe_device *xe, struct xe_bo *bo)
+{
+	struct xe_gt *root_gt, *gt;
+	int i;
+
+	root_gt = xe_device_get_gt(xe, 0);
+	root_gt->uc.guc.g2g.bo = bo;
+	root_gt->uc.guc.g2g.owned = true;
+	kunit_info(test, "[%d.%d] Assigned 0x%p\n", gt_to_tile(root_gt)->id, root_gt->info.id, bo);
+
+	for_each_gt(gt, xe, i) {
+		if (gt->info.id != 0) {
+			gt->uc.guc.g2g.owned = false;
+			gt->uc.guc.g2g.bo = xe_bo_get(bo);
+			kunit_info(test, "[%d.%d] Pinned 0x%p\n",
+				   gt_to_tile(gt)->id, gt->info.id, gt->uc.guc.g2g.bo);
+		}
+
+		KUNIT_ASSERT_NOT_ERR_OR_NULL(test, gt->uc.guc.g2g.bo);
+	}
+}
+
+/*
+ * Allocate a single blob on the host and split between all G2G CTBs.
+ */
+static void g2g_alloc_host(struct kunit *test, struct xe_device *xe)
+{
+	struct xe_bo *bo;
+	u32 g2g_size;
+
+	kunit_info(test, "Host [tiles = %d, GTs = %d]\n", xe->info.tile_count, xe->info.gt_count);
+
+	g2g_size = g2g_ctb_size(test, xe);
+	bo = xe_managed_bo_create_pin_map(xe, xe_device_get_root_tile(xe), g2g_size,
+					  XE_BO_FLAG_SYSTEM |
+					  XE_BO_FLAG_GGTT |
+					  XE_BO_FLAG_GGTT_ALL |
+					  XE_BO_FLAG_GGTT_INVALIDATE);
+	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, bo);
+	kunit_info(test, "[HST] G2G buffer create: 0x%p\n", bo);
+
+	xe_map_memset(xe, &bo->vmap, 0, 0, g2g_size);
+
+	g2g_distribute(test, xe, bo);
+}
+
+/*
+ * Allocate a single blob on the given tile and split between all G2G CTBs.
+ */
+static void g2g_alloc_tile(struct kunit *test, struct xe_device *xe, struct xe_tile *tile)
+{
+	struct xe_bo *bo;
+	u32 g2g_size;
+
+	KUNIT_ASSERT_TRUE(test, IS_DGFX(xe));
+	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, tile);
+
+	kunit_info(test, "Tile %d [tiles = %d, GTs = %d]\n",
+		   tile->id, xe->info.tile_count, xe->info.gt_count);
+
+	g2g_size = g2g_ctb_size(test, xe);
+	bo = xe_managed_bo_create_pin_map(xe, tile, g2g_size,
+					  XE_BO_FLAG_VRAM_IF_DGFX(tile) |
+					  XE_BO_FLAG_GGTT |
+					  XE_BO_FLAG_GGTT_ALL |
+					  XE_BO_FLAG_GGTT_INVALIDATE);
+	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, bo);
+	kunit_info(test, "[%d.*] G2G buffer create: 0x%p\n", tile->id, bo);
+
+	xe_map_memset(xe, &bo->vmap, 0, 0, g2g_size);
+
+	g2g_distribute(test, xe, bo);
+}
+
+static void g2g_free(struct kunit *test, struct xe_device *xe)
+{
+	struct xe_gt *gt;
+	struct xe_bo *bo;
+	int i;
+
+	for_each_gt(gt, xe, i) {
+		bo = gt->uc.guc.g2g.bo;
+		if (!bo)
+			continue;
+
+		if (gt->uc.guc.g2g.owned) {
+			xe_managed_bo_unpin_map_no_vm(bo);
+			kunit_info(test, "[%d.%d] Unmapped 0x%p\n",
+				   gt_to_tile(gt)->id, gt->info.id, bo);
+		} else {
+			xe_bo_put(bo);
+			kunit_info(test, "[%d.%d] Unpinned 0x%p\n",
+				   gt_to_tile(gt)->id, gt->info.id, bo);
+		}
+
+		gt->uc.guc.g2g.bo = NULL;
+	}
+}
+
+static void g2g_stop(struct kunit *test, struct xe_device *xe)
+{
+	struct xe_gt *gt;
+	int i;
+
+	for_each_gt(gt, xe, i) {
+		struct xe_guc *guc = &gt->uc.guc;
+
+		if (!guc->g2g.bo)
+			continue;
+
+		g2g_ct_stop(guc);
+	}
+
+	g2g_free(test, xe);
+}
+
+/*
+ * Generate a unique id for each bi-directional CTB for each pair of
+ * near and far tiles/devices. The id can then be used as an index into
+ * a single allocation that is sub-divided into multiple CTBs.
+ *
+ * For example, with two devices per tile and two tiles, the table should
+ * look like:
+ *           Far <tile>.<dev>
+ *         0.0   0.1   1.0   1.1
+ * N 0.0  --/-- 00/01 02/03 04/05
+ * e 0.1  01/00 --/-- 06/07 08/09
+ * a 1.0  03/02 07/06 --/-- 10/11
+ * r 1.1  05/04 09/08 11/10 --/--
+ *
+ * Where each entry is Rx/Tx channel id.
+ *
+ * So GuC #3 (tile 1, dev 1) talking to GuC #2 (tile 1, dev 0) would
+ * be reading from channel #11 and writing to channel #10. Whereas,
+ * GuC #2 talking to GuC #3 would be read on #10 and write to #11.
+ */
+static int g2g_slot_flat(u32 near_tile, u32 near_dev, u32 far_tile, u32 far_dev,
+			 u32 type, u32 max_inst, bool have_dev)
+{
+	u32 near = near_tile, far = far_tile;
+	u32 idx = 0, x, y, direction;
+	int i;
+
+	if (have_dev) {
+		near = (near << 1) | near_dev;
+		far = (far << 1) | far_dev;
+	}
+
+	/* No need to send to one's self */
+	if (far == near)
+		return -1;
+
+	if (far > near) {
+		/* Top right table half */
+		x = far;
+		y = near;
+
+		/* T/R is 'forwards' direction */
+		direction = type;
+	} else {
+		/* Bottom left table half */
+		x = near;
+		y = far;
+
+		/* B/L is 'backwards' direction */
+		direction = (1 - type);
+	}
+
+	/* Count the rows prior to the target */
+	for (i = y; i > 0; i--)
+		idx += max_inst - i;
+
+	/* Count this row up to the target */
+	idx += (x - 1 - y);
+
+	/* Slots are in Rx/Tx pairs */
+	idx *= 2;
+
+	/* Pick Rx/Tx direction */
+	idx += direction;
+
+	return idx;
+}
+
+static int g2g_register_flat(struct xe_guc *guc, u32 far_tile, u32 far_dev, u32 type, bool have_dev)
+{
+	struct xe_gt *gt = guc_to_gt(guc);
+	struct xe_device *xe = gt_to_xe(gt);
+	u32 near_tile = gt_to_tile(gt)->id;
+	u32 near_dev = G2G_DEV(gt);
+	u32 max = xe->info.gt_count;
+	int idx;
+	u32 base, desc, buf;
+
+	if (!guc->g2g.bo)
+		return -ENODEV;
+
+	idx = g2g_slot_flat(near_tile, near_dev, far_tile, far_dev, type, max, have_dev);
+	xe_assert(xe, idx >= 0);
+
+	base = guc_bo_ggtt_addr(guc, guc->g2g.bo);
+	desc = base + idx * G2G_DESC_SIZE;
+	buf = base + idx * G2G_BUFFER_SIZE + G2G_DESC_AREA_SIZE;
+
+	xe_assert(xe, (desc - base + G2G_DESC_SIZE) <= G2G_DESC_AREA_SIZE);
+	xe_assert(xe, (buf - base + G2G_BUFFER_SIZE) <= xe_bo_size(guc->g2g.bo));
+
+	return guc_action_register_g2g_buffer(guc, type, far_tile, far_dev,
+					      desc, buf, G2G_BUFFER_SIZE);
+}
+
+static void g2g_start(struct kunit *test, struct xe_guc *guc)
+{
+	struct xe_gt *remote_gt, *gt = guc_to_gt(guc);
+	struct xe_device *xe = gt_to_xe(gt);
+	unsigned int i;
+	int t, ret;
+	bool have_dev;
+
+	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, guc->g2g.bo);
+
+	/* GuC interface will need extending if more GT device types are ever created. */
+	KUNIT_ASSERT_TRUE(test,
+			  (gt->info.type == XE_GT_TYPE_MAIN) ||
+			  (gt->info.type == XE_GT_TYPE_MEDIA));
+
+	/* Channel numbering depends on whether there are multiple GTs per tile */
+	have_dev = xe->info.gt_count > xe->info.tile_count;
+
+	for_each_gt(remote_gt, xe, i) {
+		u32 tile, dev;
+
+		if (remote_gt->info.id == gt->info.id)
+			continue;
+
+		tile = gt_to_tile(remote_gt)->id;
+		dev = G2G_DEV(remote_gt);
+
+		for (t = 0; t < XE_G2G_TYPE_LIMIT; t++) {
+			ret = g2g_register_flat(guc, tile, dev, t, have_dev);
+			KUNIT_ASSERT_EQ_MSG(test, 0, ret, "G2G register failed: %pe", ERR_PTR(ret));
+		}
+	}
+}
+
+static void g2g_reinit(struct kunit *test, struct xe_device *xe, int ctb_type, struct xe_tile *tile)
+{
+	struct xe_gt *gt;
+	int i, found = 0;
+
+	g2g_stop(test, xe);
+
+	for_each_gt(gt, xe, i) {
+		struct xe_guc *guc = &gt->uc.guc;
+
+		KUNIT_ASSERT_NULL(test, guc->g2g.bo);
+	}
+
+	switch (ctb_type) {
+	case G2G_CTB_TYPE_DEFAULT:
+		g2g_alloc_default(test, xe);
+		break;
+
+	case G2G_CTB_TYPE_HOST:
+		g2g_alloc_host(test, xe);
+		break;
+
+	case G2G_CTB_TYPE_TILE:
+		g2g_alloc_tile(test, xe, tile);
+		break;
+
+	default:
+		KUNIT_ASSERT_TRUE(test, false);
+	}
+
+	for_each_gt(gt, xe, i) {
+		struct xe_guc *guc = &gt->uc.guc;
+
+		if (!guc->g2g.bo)
+			continue;
+
+		if (ctb_type == G2G_CTB_TYPE_DEFAULT)
+			guc_g2g_start(guc);
+		else
+			g2g_start(test, guc);
+		found++;
+	}
+
+	KUNIT_ASSERT_GT_MSG(test, found, 1, "insufficient G2G channels running: %d", found);
+
+	kunit_info(test, "Testing across %d GTs\n", found);
+}
+
+static void g2g_recreate_ctb(void *_xe)
+{
+	struct xe_device *xe = (struct xe_device *)_xe;
+	struct kunit *test = kunit_get_current_test();
+
+	g2g_stop(test, xe);
+
+	if (xe_guc_g2g_wanted(xe))
+		g2g_reinit(test, xe, G2G_CTB_TYPE_DEFAULT, NULL);
+}
+
+static void g2g_pm_runtime_put(void *_xe)
+{
+	struct xe_device *xe = (struct xe_device *)_xe;
+
+	xe_pm_runtime_put(xe);
+}
+
+static void g2g_pm_runtime_get(struct kunit *test)
+{
+	struct xe_device *xe = test->priv;
+	int ret;
+
+	xe_pm_runtime_get(xe);
+	ret = kunit_add_action_or_reset(test, g2g_pm_runtime_put, xe);
+	KUNIT_ASSERT_EQ_MSG(test, 0, ret, "Failed to register runtime PM action\n");
+}
+
+static void g2g_check_skip(struct kunit *test)
+{
+	struct xe_device *xe = test->priv;
+	struct xe_gt *gt;
+	int i;
+
+	if (IS_SRIOV_VF(xe))
+		kunit_skip(test, "not supported from a VF");
+
+	if (xe->info.gt_count <= 1)
+		kunit_skip(test, "not enough GTs");
+
+	for_each_gt(gt, xe, i) {
+		struct xe_guc *guc = &gt->uc.guc;
+
+		if (guc->fw.build_type == CSS_UKERNEL_INFO_BUILDTYPE_PROD)
+			kunit_skip(test,
+				   "G2G test interface not available in production firmware builds\n");
+	}
+}
+
+/*
+ * Simple test that does not try to recreate the CTBs.
+ * Requires that the platform already enables G2G comms
+ * but has no risk of leaving the system in a broken state
+ * afterwards.
+ */
+static void xe_live_guc_g2g_kunit_default(struct kunit *test)
+{
+	struct xe_device *xe = test->priv;
+
+	if (!xe_guc_g2g_wanted(xe))
+		kunit_skip(test, "G2G not enabled");
+
+	g2g_check_skip(test);
+
+	g2g_pm_runtime_get(test);
+
+	kunit_info(test, "Testing default CTBs\n");
+	g2g_run_test(test, xe);
+
+	kunit_release_action(test, &g2g_pm_runtime_put, xe);
+}
+
+/*
+ * More complex test that re-creates the CTBs in various location to
+ * test access to each location from each GuC. Can be run even on
+ * systems that do not enable G2G by default. On the other hand,
+ * because it recreates the CTBs, if something goes wrong it could
+ * leave the system with broken G2G comms.
+ */
+static void xe_live_guc_g2g_kunit_allmem(struct kunit *test)
+{
+	struct xe_device *xe = test->priv;
+	int ret;
+
+	g2g_check_skip(test);
+
+	g2g_pm_runtime_get(test);
+
+	/* Make sure to leave the system as we found it */
+	ret = kunit_add_action_or_reset(test, g2g_recreate_ctb, xe);
+	KUNIT_ASSERT_EQ_MSG(test, 0, ret, "Failed to register CTB re-creation action\n");
+
+	kunit_info(test, "Testing CTB type 'default'...\n");
+	g2g_reinit(test, xe, G2G_CTB_TYPE_DEFAULT, NULL);
+	g2g_run_test(test, xe);
+
+	kunit_info(test, "Testing CTB type 'host'...\n");
+	g2g_reinit(test, xe, G2G_CTB_TYPE_HOST, NULL);
+	g2g_run_test(test, xe);
+
+	if (IS_DGFX(xe)) {
+		struct xe_tile *tile;
+		int id;
+
+		for_each_tile(tile, xe, id) {
+			kunit_info(test, "Testing CTB type 'tile: #%d'...\n", id);
+
+			g2g_reinit(test, xe, G2G_CTB_TYPE_TILE, tile);
+			g2g_run_test(test, xe);
+		}
+	} else {
+		kunit_info(test, "Skipping local memory on integrated platform\n");
+	}
+
+	kunit_release_action(test, g2g_recreate_ctb, xe);
+	kunit_release_action(test, g2g_pm_runtime_put, xe);
+}
+
+static struct kunit_case xe_guc_g2g_tests[] = {
+	KUNIT_CASE_PARAM(xe_live_guc_g2g_kunit_default, xe_pci_live_device_gen_param),
+	KUNIT_CASE_PARAM(xe_live_guc_g2g_kunit_allmem, xe_pci_live_device_gen_param),
+	{}
+};
+
+VISIBLE_IF_KUNIT
+struct kunit_suite xe_guc_g2g_test_suite = {
+	.name = "xe_guc_g2g",
+	.test_cases = xe_guc_g2g_tests,
+	.init = xe_kunit_helper_xe_device_live_test_init,
+};
+EXPORT_SYMBOL_IF_KUNIT(xe_guc_g2g_test_suite);

drivers/gpu/drm/xe/tests/xe_live_test_mod.c

···
 extern struct kunit_suite xe_dma_buf_test_suite;
 extern struct kunit_suite xe_migrate_test_suite;
 extern struct kunit_suite xe_mocs_test_suite;
+extern struct kunit_suite xe_guc_g2g_test_suite;
 
 kunit_test_suite(xe_bo_test_suite);
 kunit_test_suite(xe_bo_shrink_test_suite);
 kunit_test_suite(xe_dma_buf_test_suite);
 kunit_test_suite(xe_migrate_test_suite);
 kunit_test_suite(xe_mocs_test_suite);
+kunit_test_suite(xe_guc_g2g_test_suite);
 
 MODULE_AUTHOR("Intel Corporation");
 MODULE_LICENSE("GPL");

drivers/gpu/drm/xe/tests/xe_migrate.c

···
 		} } while (0)
 
 static void test_copy(struct xe_migrate *m, struct xe_bo *bo,
-		      struct kunit *test, u32 region)
+		      struct kunit *test, u32 region, struct drm_exec *exec)
 {
 	struct xe_device *xe = tile_to_xe(m->tile);
 	u64 retval, expected = 0;
···
 						   ttm_bo_type_kernel,
 						   region |
 						   XE_BO_FLAG_NEEDS_CPU_ACCESS |
-						   XE_BO_FLAG_PINNED);
+						   XE_BO_FLAG_PINNED,
+						   exec);
 	if (IS_ERR(remote)) {
 		KUNIT_FAIL(test, "Failed to allocate remote bo for %s: %pe\n",
 			   str, remote);
 		return;
 	}
 
-	err = xe_bo_validate(remote, NULL, false);
+	err = xe_bo_validate(remote, NULL, false, exec);
 	if (err) {
 		KUNIT_FAIL(test, "Failed to validate system bo for %s: %i\n",
 			   str, err);
···
 }
 
 static void test_copy_sysmem(struct xe_migrate *m, struct xe_bo *bo,
-			     struct kunit *test)
+			     struct drm_exec *exec, struct kunit *test)
 {
-	test_copy(m, bo, test, XE_BO_FLAG_SYSTEM);
+	test_copy(m, bo, test, XE_BO_FLAG_SYSTEM, exec);
 }
 
 static void test_copy_vram(struct xe_migrate *m, struct xe_bo *bo,
-			   struct kunit *test)
+			   struct drm_exec *exec, struct kunit *test)
 {
 	u32 region;
 
···
 		region = XE_BO_FLAG_VRAM1;
 	else
 		region = XE_BO_FLAG_VRAM0;
-	test_copy(m, bo, test, region);
+	test_copy(m, bo, test, region, exec);
 }
 
-static void xe_migrate_sanity_test(struct xe_migrate *m, struct kunit *test)
+static void xe_migrate_sanity_test(struct xe_migrate *m, struct kunit *test,
+				   struct drm_exec *exec)
 {
 	struct xe_tile *tile = m->tile;
 	struct xe_device *xe = tile_to_xe(tile);
···
 
 	big = xe_bo_create_pin_map(xe, tile, m->q->vm, SZ_4M,
 				   ttm_bo_type_kernel,
-				   XE_BO_FLAG_VRAM_IF_DGFX(tile));
+				   XE_BO_FLAG_VRAM_IF_DGFX(tile),
+				   exec);
 	if (IS_ERR(big)) {
 		KUNIT_FAIL(test, "Failed to allocate bo: %li\n", PTR_ERR(big));
 		goto vunmap;
···
 
 	pt = xe_bo_create_pin_map(xe, tile, m->q->vm, XE_PAGE_SIZE,
 				  ttm_bo_type_kernel,
-				  XE_BO_FLAG_VRAM_IF_DGFX(tile));
+				  XE_BO_FLAG_VRAM_IF_DGFX(tile),
+				  exec);
 	if (IS_ERR(pt)) {
 		KUNIT_FAIL(test, "Failed to allocate fake pt: %li\n",
 			   PTR_ERR(pt));
···
 	tiny = xe_bo_create_pin_map(xe, tile, m->q->vm,
 				    2 * SZ_4K,
 				    ttm_bo_type_kernel,
-				    XE_BO_FLAG_VRAM_IF_DGFX(tile));
+				    XE_BO_FLAG_VRAM_IF_DGFX(tile),
+				    exec);
 	if (IS_ERR(tiny)) {
 		KUNIT_FAIL(test, "Failed to allocate tiny fake pt: %li\n",
 			   PTR_ERR(tiny));
···
 	check(retval, expected, "Command clear small last value", test);
 
 	kunit_info(test, "Copying small buffer object to system\n");
-	test_copy_sysmem(m, tiny, test);
+	test_copy_sysmem(m, tiny, exec, test);
 	if (xe->info.tile_count > 1) {
 		kunit_info(test, "Copying small buffer object to other vram\n");
-		test_copy_vram(m, tiny, test);
+		test_copy_vram(m, tiny, exec, test);
 	}
 
 	/* Clear a big bo */
···
 	check(retval, expected, "Command clear big last value", test);
 
 	kunit_info(test, "Copying big buffer object to system\n");
-	test_copy_sysmem(m, big, test);
+	test_copy_sysmem(m, big, exec, test);
 	if (xe->info.tile_count > 1) {
 		kunit_info(test, "Copying big buffer object to other vram\n");
-		test_copy_vram(m, big, test);
+		test_copy_vram(m, big, exec, test);
 	}
 
 out:
···
 
 	for_each_tile(tile, xe, id) {
 		struct xe_migrate *m = tile->migrate;
+		struct drm_exec *exec = XE_VALIDATION_OPT_OUT;
 
 		kunit_info(test, "Testing tile id %d.\n", id);
 		xe_vm_lock(m->q->vm, false);
-		xe_migrate_sanity_test(m, test);
+		xe_migrate_sanity_test(m, test, exec);
 		xe_vm_unlock(m->q->vm);
 	}
 
···
 
 static void test_migrate(struct xe_device *xe, struct xe_tile *tile,
 			 struct xe_bo *sys_bo, struct xe_bo *vram_bo, struct xe_bo *ccs_bo,
-			 struct kunit *test)
+			 struct drm_exec *exec, struct kunit *test)
 {
 	struct dma_fence *fence;
 	u64 expected, retval;
···
 	dma_fence_put(fence);
 
 	kunit_info(test, "Evict vram buffer object\n");
-	ret = xe_bo_evict(vram_bo);
+	ret = xe_bo_evict(vram_bo, exec);
 	if (ret) {
 		KUNIT_FAIL(test, "Failed to evict bo.\n");
 		return;
···
 	dma_fence_put(fence);
 
 	kunit_info(test, "Restore vram buffer object\n");
-	ret = xe_bo_validate(vram_bo, NULL, false);
+	ret = xe_bo_validate(vram_bo, NULL, false, exec);
 	if (ret) {
 		KUNIT_FAIL(test, "Failed to validate vram bo for: %li\n", ret);
 		return;
···
 {
 	struct xe_bo *sys_bo, *vram_bo = NULL, *ccs_bo = NULL;
 	unsigned int bo_flags = XE_BO_FLAG_VRAM_IF_DGFX(tile);
+	struct drm_exec *exec;
 	long ret;
 
-	sys_bo = xe_bo_create_user(xe, NULL, NULL, SZ_4M,
+	sys_bo = xe_bo_create_user(xe, NULL, SZ_4M,
 				   DRM_XE_GEM_CPU_CACHING_WC,
 				   XE_BO_FLAG_SYSTEM |
 				   XE_BO_FLAG_NEEDS_CPU_ACCESS |
-				   XE_BO_FLAG_PINNED);
+				   XE_BO_FLAG_PINNED, NULL);
 
 	if (IS_ERR(sys_bo)) {
 		KUNIT_FAIL(test, "xe_bo_create() failed with err=%ld\n",
···
 		return;
 	}
 
+	exec = XE_VALIDATION_OPT_OUT;
 	xe_bo_lock(sys_bo, false);
-	ret = xe_bo_validate(sys_bo, NULL, false);
+	ret = xe_bo_validate(sys_bo, NULL, false, exec);
 	if (ret) {
 		KUNIT_FAIL(test, "Failed to validate system bo for: %li\n", ret);
 		goto free_sysbo;
···
 	}
 	xe_bo_unlock(sys_bo);
 
-	ccs_bo = xe_bo_create_user(xe, NULL, NULL, SZ_4M,
+	ccs_bo = xe_bo_create_user(xe, NULL, SZ_4M,
 				   DRM_XE_GEM_CPU_CACHING_WC,
 				   bo_flags | XE_BO_FLAG_NEEDS_CPU_ACCESS |
-				   XE_BO_FLAG_PINNED);
+				   XE_BO_FLAG_PINNED, NULL);
 
 	if (IS_ERR(ccs_bo)) {
 		KUNIT_FAIL(test, "xe_bo_create() failed with err=%ld\n",
···
 	}
 
 	xe_bo_lock(ccs_bo, false);
-	ret = xe_bo_validate(ccs_bo, NULL, false);
+	ret = xe_bo_validate(ccs_bo, NULL, false, exec);
 	if (ret) {
 		KUNIT_FAIL(test, "Failed to validate system bo for: %li\n", ret);
 		goto free_ccsbo;
···
 	}
 	xe_bo_unlock(ccs_bo);
 
-	vram_bo = xe_bo_create_user(xe, NULL, NULL, SZ_4M,
+	vram_bo = xe_bo_create_user(xe, NULL, SZ_4M,
 				    DRM_XE_GEM_CPU_CACHING_WC,
 				    bo_flags | XE_BO_FLAG_NEEDS_CPU_ACCESS |
-				    XE_BO_FLAG_PINNED);
+				    XE_BO_FLAG_PINNED, NULL);
 	if (IS_ERR(vram_bo)) {
 		KUNIT_FAIL(test, "xe_bo_create() failed with err=%ld\n",
 			   PTR_ERR(vram_bo));
···
 	}
 
 	xe_bo_lock(vram_bo, false);
-	ret = xe_bo_validate(vram_bo, NULL, false);
+	ret = xe_bo_validate(vram_bo, NULL, false, exec);
 	if (ret) {
 		KUNIT_FAIL(test, "Failed to validate vram bo for: %li\n", ret);
 		goto free_vrambo;
···
 	}
 
 	test_clear(xe, tile, sys_bo, vram_bo, test);
-	test_migrate(xe, tile, sys_bo, vram_bo, ccs_bo, test);
+	test_migrate(xe, tile, sys_bo, vram_bo, ccs_bo, exec, test);
 	xe_bo_unlock(vram_bo);
 
 	xe_bo_lock(vram_bo, false);

drivers/gpu/drm/xe/tests/xe_pci.c

···
 #include <kunit/test-bug.h>
 #include <kunit/visibility.h>
 
+#define PLATFORM_CASE(platform__, graphics_step__)					\
+	{										\
+		.platform = XE_ ## platform__,						\
+		.subplatform = XE_SUBPLATFORM_NONE,					\
+		.step = { .graphics = STEP_ ## graphics_step__ }			\
+	}
+
+#define SUBPLATFORM_CASE(platform__, subplatform__, graphics_step__)			\
+	{										\
+		.platform = XE_ ## platform__,						\
+		.subplatform = XE_SUBPLATFORM_ ## platform__ ## _ ## subplatform__,	\
+		.step = { .graphics = STEP_ ## graphics_step__ }			\
+	}
+
+#define GMDID_CASE(platform__, graphics_verx100__, graphics_step__,			\
+		   media_verx100__, media_step__)					\
+	{										\
+		.platform = XE_ ## platform__,						\
+		.subplatform = XE_SUBPLATFORM_NONE,					\
+		.graphics_verx100 = graphics_verx100__,					\
+		.media_verx100 = media_verx100__,					\
+		.step = { .graphics = STEP_ ## graphics_step__,				\
+			   .media = STEP_ ## media_step__ }				\
+	}
+
+static const struct xe_pci_fake_data cases[] = {
+	PLATFORM_CASE(TIGERLAKE, B0),
+	PLATFORM_CASE(DG1, A0),
+	PLATFORM_CASE(DG1, B0),
+	PLATFORM_CASE(ALDERLAKE_S, A0),
+	PLATFORM_CASE(ALDERLAKE_S, B0),
+	PLATFORM_CASE(ALDERLAKE_S, C0),
+	PLATFORM_CASE(ALDERLAKE_S, D0),
+	PLATFORM_CASE(ALDERLAKE_P, A0),
+	PLATFORM_CASE(ALDERLAKE_P, B0),
+	PLATFORM_CASE(ALDERLAKE_P, C0),
+	SUBPLATFORM_CASE(ALDERLAKE_S, RPLS, D0),
+	SUBPLATFORM_CASE(ALDERLAKE_P, RPLU, E0),
+	SUBPLATFORM_CASE(DG2, G10, C0),
+	SUBPLATFORM_CASE(DG2, G11, B1),
+	SUBPLATFORM_CASE(DG2, G12, A1),
+	GMDID_CASE(METEORLAKE, 1270, A0, 1300, A0),
+	GMDID_CASE(METEORLAKE, 1271, A0, 1300, A0),
+	GMDID_CASE(METEORLAKE, 1274, A0, 1300, A0),
+	GMDID_CASE(LUNARLAKE, 2004, A0, 2000, A0),
+	GMDID_CASE(LUNARLAKE, 2004, B0, 2000, A0),
+	GMDID_CASE(BATTLEMAGE, 2001, A0, 1301, A1),
+	GMDID_CASE(PANTHERLAKE, 3000, A0, 3000, A0),
+};
+
+KUNIT_ARRAY_PARAM(platform, cases, xe_pci_fake_data_desc);
+
+/**
+ * xe_pci_fake_data_gen_params - Generate struct xe_pci_fake_data parameters
+ * @prev: the pointer to the previous parameter to iterate from or NULL
+ * @desc: output buffer with minimum size of KUNIT_PARAM_DESC_SIZE
+ *
+ * This function prepares struct xe_pci_fake_data parameter.
+ *
+ * To be used only as a parameter generator function in &KUNIT_CASE_PARAM.
+ *
+ * Return: pointer to the next parameter or NULL if no more parameters
+ */
+const void *xe_pci_fake_data_gen_params(const void *prev, char *desc)
+{
+	return platform_gen_params(prev, desc);
+}
+EXPORT_SYMBOL_IF_KUNIT(xe_pci_fake_data_gen_params);
+
+static const struct xe_device_desc *lookup_desc(enum xe_platform p)
+{
+	const struct xe_device_desc *desc;
+	const struct pci_device_id *ids;
+
+	for (ids = pciidlist; ids->driver_data; ids++) {
+		desc = (const void *)ids->driver_data;
+		if (desc->platform == p)
+			return desc;
+	}
+	return NULL;
+}
+
+static const struct xe_subplatform_desc *lookup_sub_desc(enum xe_platform p, enum xe_subplatform s)
+{
+	const struct xe_device_desc *desc = lookup_desc(p);
+	const struct xe_subplatform_desc *spd;
+
+	if (desc && desc->subplatforms)
+		for (spd = desc->subplatforms; spd->subplatform; spd++)
+			if (spd->subplatform == s)
+				return spd;
+	return NULL;
+}
+
+static const char *lookup_platform_name(enum xe_platform p)
+{
+	const struct xe_device_desc *desc = lookup_desc(p);
+
+	return desc ? desc->platform_name : "INVALID";
+}
+
+static const char *__lookup_subplatform_name(enum xe_platform p, enum xe_subplatform s)
+{
+	const struct xe_subplatform_desc *desc = lookup_sub_desc(p, s);
+
+	return desc ? desc->name : "INVALID";
+}
+
+static const char *lookup_subplatform_name(enum xe_platform p, enum xe_subplatform s)
+{
+	return s == XE_SUBPLATFORM_NONE ? "" : __lookup_subplatform_name(p, s);
+}
+
+static const char *subplatform_prefix(enum xe_subplatform s)
+{
+	return s == XE_SUBPLATFORM_NONE ? "" : " ";
+}
+
+static const char *step_prefix(enum xe_step step)
+{
+	return step == STEP_NONE ? "" : " ";
+}
+
+static const char *step_name(enum xe_step step)
+{
+	return step == STEP_NONE ? "" : xe_step_name(step);
+}
+
+static const char *sriov_prefix(enum xe_sriov_mode mode)
+{
+	return mode <= XE_SRIOV_MODE_NONE ? "" : " ";
+}
+
+static const char *sriov_name(enum xe_sriov_mode mode)
+{
+	return mode <= XE_SRIOV_MODE_NONE ? "" : xe_sriov_mode_to_string(mode);
+}
+
+static const char *lookup_graphics_name(unsigned int verx100)
+{
+	const struct xe_ip *ip = find_graphics_ip(verx100);
+
+	return ip ? ip->name : "";
+}
+
+static const char *lookup_media_name(unsigned int verx100)
+{
+	const struct xe_ip *ip = find_media_ip(verx100);
+
+	return ip ? ip->name : "";
+}
+
+/**
+ * xe_pci_fake_data_desc - Describe struct xe_pci_fake_data parameter
+ * @param: the &struct xe_pci_fake_data parameter to describe
+ * @desc: output buffer with minimum size of KUNIT_PARAM_DESC_SIZE
+ *
+ * This function prepares description of the struct xe_pci_fake_data parameter.
+ *
+ * It is tailored for use in parameterized KUnit tests where parameter generator
+ * is based on the struct xe_pci_fake_data arrays.
+ */
+void xe_pci_fake_data_desc(const struct xe_pci_fake_data *param, char *desc)
+{
+	if (param->graphics_verx100 || param->media_verx100)
+		snprintf(desc, KUNIT_PARAM_DESC_SIZE, "%s%s%s %u.%02u(%s)%s%s %u.%02u(%s)%s%s%s%s",
+			 lookup_platform_name(param->platform),
+			 subplatform_prefix(param->subplatform),
+			 lookup_subplatform_name(param->platform, param->subplatform),
+			 param->graphics_verx100 / 100, param->graphics_verx100 % 100,
+			 lookup_graphics_name(param->graphics_verx100),
+			 step_prefix(param->step.graphics), step_name(param->step.graphics),
+			 param->media_verx100 / 100, param->media_verx100 % 100,
+			 lookup_media_name(param->media_verx100),
+			 step_prefix(param->step.media), step_name(param->step.media),
+			 sriov_prefix(param->sriov_mode), sriov_name(param->sriov_mode));
+	else
+		snprintf(desc, KUNIT_PARAM_DESC_SIZE, "%s%s%s%s%s%s%s",
+			 lookup_platform_name(param->platform),
+			 subplatform_prefix(param->subplatform),
+			 lookup_subplatform_name(param->platform, param->subplatform),
+			 step_prefix(param->step.graphics), step_name(param->step.graphics),
+			 sriov_prefix(param->sriov_mode), sriov_name(param->sriov_mode));
+}
+EXPORT_SYMBOL_IF_KUNIT(xe_pci_fake_data_desc);
+
 static void xe_ip_kunit_desc(const struct xe_ip *param, char *desc)
 {
 	snprintf(desc, KUNIT_PARAM_DESC_SIZE, "%u.%02u %s",
 		 param->verx100 / 100, param->verx100 % 100, param->name);
 }
+
+/*
+ * Pre-GMDID Graphics and Media IPs definitions.
+ *
+ * Mimic the way GMDID IPs are declared so the same
+ * param generator can be used for both
+ */
+static const struct xe_ip pre_gmdid_graphics_ips[] = {
+	graphics_ip_xelp,
+	graphics_ip_xelpp,
+	graphics_ip_xehpg,
+	graphics_ip_xehpc,
+};
+
+static const struct xe_ip pre_gmdid_media_ips[] = {
+	media_ip_xem,
+	media_ip_xehpm,
+};
+
+KUNIT_ARRAY_PARAM(pre_gmdid_graphics_ip, pre_gmdid_graphics_ips, xe_ip_kunit_desc);
+KUNIT_ARRAY_PARAM(pre_gmdid_media_ip, pre_gmdid_media_ips, xe_ip_kunit_desc);
 
 KUNIT_ARRAY_PARAM(graphics_ip, graphics_ips, xe_ip_kunit_desc);
 KUNIT_ARRAY_PARAM(media_ip, media_ips, xe_ip_kunit_desc);
···
  */
 const void *xe_pci_graphics_ip_gen_param(const void *prev, char *desc)
 {
+	const void *next = pre_gmdid_graphics_ip_gen_params(prev, desc);
+
+	if (next)
+		return next;
+	if (is_insidevar(prev, pre_gmdid_graphics_ips))
+		prev = NULL;
+
 	return graphics_ip_gen_params(prev, desc);
 }
 EXPORT_SYMBOL_IF_KUNIT(xe_pci_graphics_ip_gen_param);
···
  */
 const void *xe_pci_media_ip_gen_param(const void *prev, char *desc)
 {
+	const void *next = pre_gmdid_media_ip_gen_params(prev, desc);
+
+	if (next)
+		return next;
+	if (is_insidevar(prev, pre_gmdid_media_ips))
+		prev = NULL;
+
 	return media_ip_gen_params(prev, desc);
 }
 EXPORT_SYMBOL_IF_KUNIT(xe_pci_media_ip_gen_param);
···
 
 	if (type == GMDID_MEDIA) {
 		*ver = data->media_verx100;
-		*revid = xe_step_to_gmdid(data->media_step);
+		*revid = xe_step_to_gmdid(data->step.media);
 	} else {
 		*ver = data->graphics_verx100;
-		*revid = xe_step_to_gmdid(data->graphics_step);
+		*revid = xe_step_to_gmdid(data->step.graphics);
 	}
 }
 

drivers/gpu/drm/xe/tests/xe_pci_test.h

···
 
 #include "xe_platform_types.h"
 #include "xe_sriov_types.h"
+#include "xe_step_types.h"
 
 struct xe_device;
 
···
 	enum xe_sriov_mode sriov_mode;
 	enum xe_platform platform;
 	enum xe_subplatform subplatform;
+	struct xe_step_info step;
 	u32 graphics_verx100;
 	u32 media_verx100;
-	u32 graphics_step;
-	u32 media_step;
 };
 
 int xe_pci_fake_device_init(struct xe_device *xe);
+const void *xe_pci_fake_data_gen_params(const void *prev, char *desc);
+void xe_pci_fake_data_desc(const struct xe_pci_fake_data *param, char *desc);
 
 const void *xe_pci_graphics_ip_gen_param(const void *prev, char *desc);
 const void *xe_pci_media_ip_gen_param(const void *prev, char *desc);

drivers/gpu/drm/xe/tests/xe_wa_test.c

···
 #include "xe_tuning.h"
 #include "xe_wa.h"
 
-struct platform_test_case {
-	const char *name;
-	enum xe_platform platform;
-	enum xe_subplatform subplatform;
-	u32 graphics_verx100;
-	u32 media_verx100;
-	struct xe_step_info step;
-};
-
-#define PLATFORM_CASE(platform__, graphics_step__)				\
-	{									\
-		.name = #platform__ " (" #graphics_step__ ")",			\
-		.platform = XE_ ## platform__,					\
-		.subplatform = XE_SUBPLATFORM_NONE,				\
-		.step = { .graphics = STEP_ ## graphics_step__ }		\
-	}
-
-
-#define SUBPLATFORM_CASE(platform__, subplatform__, graphics_step__)			\
-	{										\
-		.name = #platform__ "_" #subplatform__ " (" #graphics_step__ ")",	\
-		.platform = XE_ ## platform__,						\
-		.subplatform = XE_SUBPLATFORM_ ## platform__ ## _ ## subplatform__,	\
-		.step = { .graphics = STEP_ ## graphics_step__ }			\
-	}
-
-#define GMDID_CASE(platform__, graphics_verx100__, graphics_step__,		\
-		   media_verx100__, media_step__)				\
-	{									\
-		.name = #platform__ " (g:" #graphics_step__ ", m:" #media_step__ ")",\
-		.platform = XE_ ## platform__,					\
-		.subplatform = XE_SUBPLATFORM_NONE,				\
-		.graphics_verx100 = graphics_verx100__,				\
-		.media_verx100 = media_verx100__,				\
-		.step = { .graphics = STEP_ ## graphics_step__,			\
-			   .media = STEP_ ## media_step__ }			\
-	}
-
-static const struct platform_test_case cases[] = {
-	PLATFORM_CASE(TIGERLAKE, B0),
-	PLATFORM_CASE(DG1, A0),
-	PLATFORM_CASE(DG1, B0),
-	PLATFORM_CASE(ALDERLAKE_S, A0),
-	PLATFORM_CASE(ALDERLAKE_S, B0),
-	PLATFORM_CASE(ALDERLAKE_S, C0),
-	PLATFORM_CASE(ALDERLAKE_S, D0),
-	PLATFORM_CASE(ALDERLAKE_P, A0),
-	PLATFORM_CASE(ALDERLAKE_P, B0),
-	PLATFORM_CASE(ALDERLAKE_P, C0),
-	SUBPLATFORM_CASE(ALDERLAKE_S, RPLS, D0),
-	SUBPLATFORM_CASE(ALDERLAKE_P, RPLU, E0),
-	SUBPLATFORM_CASE(DG2, G10, C0),
-	SUBPLATFORM_CASE(DG2, G11, B1),
-	SUBPLATFORM_CASE(DG2, G12, A1),
-	GMDID_CASE(METEORLAKE, 1270, A0, 1300, A0),
-	GMDID_CASE(METEORLAKE, 1271, A0, 1300, A0),
-	GMDID_CASE(METEORLAKE, 1274, A0, 1300, A0),
-	GMDID_CASE(LUNARLAKE, 2004, A0, 2000, A0),
-	GMDID_CASE(LUNARLAKE, 2004, B0, 2000, A0),
-	GMDID_CASE(BATTLEMAGE, 2001, A0, 1301, A1),
-	GMDID_CASE(PANTHERLAKE, 3000, A0, 3000, A0),
-};
-
-static void platform_desc(const struct platform_test_case *t, char *desc)
-{
-	strscpy(desc, t->name, KUNIT_PARAM_DESC_SIZE);
-}
-
-KUNIT_ARRAY_PARAM(platform, cases, platform_desc);
-
 static int xe_wa_test_init(struct kunit *test)
 {
-	const struct platform_test_case *param = test->param_value;
-	struct xe_pci_fake_data data = {
-		.platform = param->platform,
-		.subplatform = param->subplatform,
-		.graphics_verx100 = param->graphics_verx100,
-		.media_verx100 = param->media_verx100,
-		.graphics_step = param->step.graphics,
-		.media_step = param->step.media,
-	};
+	const struct xe_pci_fake_data *param = test->param_value;
+	struct xe_pci_fake_data data = *param;
 	struct xe_device *xe;
 	struct device *dev;
 	int ret;
···
 	return 0;
 }
 
-static void xe_wa_test_exit(struct kunit *test)
-{
-	struct xe_device *xe = test->priv;
-
-	drm_kunit_helper_free_device(test, xe->drm.dev);
-}
-
 static void xe_wa_gt(struct kunit *test)
 {
 	struct xe_device *xe = test->priv;
···
 }
 
 static struct kunit_case xe_wa_tests[] = {
-	KUNIT_CASE_PARAM(xe_wa_gt, platform_gen_params),
+	KUNIT_CASE_PARAM(xe_wa_gt, xe_pci_fake_data_gen_params),
 	{}
 };
 
 static struct kunit_suite xe_rtp_test_suite = {
 	.name = "xe_wa",
 	.init = xe_wa_test_init,
-	.exit = xe_wa_test_exit,
 	.test_cases = xe_wa_tests,
 };
 

drivers/gpu/drm/xe/xe_bb.c

···
 			    enum xe_sriov_vf_ccs_rw_ctxs ctx_id)
 {
 	struct xe_bb *bb = kmalloc(sizeof(*bb), GFP_KERNEL);
-	struct xe_tile *tile = gt_to_tile(gt);
+	struct xe_device *xe = gt_to_xe(gt);
 	struct xe_sa_manager *bb_pool;
 	int err;
 
···
 	 * So, this extra DW acts as a guard here.
 	 */
 
-	bb_pool = tile->sriov.vf.ccs[ctx_id].mem.ccs_bb_pool;
+	bb_pool = xe->sriov.vf.ccs.contexts[ctx_id].mem.ccs_bb_pool;
 	bb->bo = xe_sa_bo_new(bb_pool, 4 * (dwords + 1));
 
 	if (IS_ERR(bb->bo)) {

drivers/gpu/drm/xe/xe_bo.c

···
 	 * CCS meta data is migrated from TT -> SMEM. So, let us detach the
 	 * BBs from BO as it is no longer needed.
 	 */
-	if (IS_VF_CCS_BB_VALID(xe, bo) && old_mem_type == XE_PL_TT &&
+	if (IS_VF_CCS_READY(xe) && old_mem_type == XE_PL_TT &&
 	    new_mem->mem_type == XE_PL_SYSTEM)
 		xe_sriov_vf_ccs_detach_bo(bo);
 
-	if (IS_SRIOV_VF(xe) &&
+	if (IS_VF_CCS_READY(xe) &&
 	    ((move_lacks_source && new_mem->mem_type == XE_PL_TT) ||
 	     (old_mem_type == XE_PL_SYSTEM && new_mem->mem_type == XE_PL_TT)) &&
 	    handle_system_ccs)
···
 		if (timeout < 0)
 			ret = timeout;
 
-		if (IS_VF_CCS_BB_VALID(xe, bo))
+		if (IS_VF_CCS_READY(xe))
 			xe_sriov_vf_ccs_detach_bo(bo);
 
 		xe_tt_unmap_sg(xe, ttm_bo->ttm);
···
 int xe_bo_notifier_prepare_pinned(struct xe_bo *bo)
 {
 	struct xe_device *xe = ttm_to_xe_device(bo->ttm.bdev);
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
 	struct xe_bo *backup;
 	int ret = 0;
 
-	xe_bo_lock(bo, false);
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {.exclusive = true}, ret) {
+		ret = drm_exec_lock_obj(&exec, &bo->ttm.base);
+		drm_exec_retry_on_contention(&exec);
+		xe_assert(xe, !ret);
+		xe_assert(xe, !bo->backup_obj);
 
-	xe_assert(xe, !bo->backup_obj);
+		/*
+		 * Since this is called from the PM notifier we might have raced with
+		 * someone unpinning this after we dropped the pinned list lock and
+		 * grabbing the above bo lock.
+		 */
+		if (!xe_bo_is_pinned(bo))
+			break;
 
-	/*
-	 * Since this is called from the PM notifier we might have raced with
-	 * someone unpinning this after we dropped the pinned list lock and
-	 * grabbing the above bo lock.
-	 */
-	if (!xe_bo_is_pinned(bo))
-		goto out_unlock_bo;
+		if (!xe_bo_is_vram(bo))
+			break;
 
-	if (!xe_bo_is_vram(bo))
-		goto out_unlock_bo;
+		if (bo->flags & XE_BO_FLAG_PINNED_NORESTORE)
+			break;
 
-	if (bo->flags & XE_BO_FLAG_PINNED_NORESTORE)
-		goto out_unlock_bo;
+		backup = xe_bo_init_locked(xe, NULL, NULL, bo->ttm.base.resv, NULL, xe_bo_size(bo),
+					   DRM_XE_GEM_CPU_CACHING_WB, ttm_bo_type_kernel,
+					   XE_BO_FLAG_SYSTEM | XE_BO_FLAG_NEEDS_CPU_ACCESS |
+					   XE_BO_FLAG_PINNED, &exec);
+		if (IS_ERR(backup)) {
+			drm_exec_retry_on_contention(&exec);
+			ret = PTR_ERR(backup);
+			xe_validation_retry_on_oom(&ctx, &ret);
+			break;
+		}
 
-	backup = ___xe_bo_create_locked(xe, NULL, NULL, bo->ttm.base.resv, NULL, xe_bo_size(bo),
-					DRM_XE_GEM_CPU_CACHING_WB, ttm_bo_type_kernel,
-					XE_BO_FLAG_SYSTEM | XE_BO_FLAG_NEEDS_CPU_ACCESS |
-					XE_BO_FLAG_PINNED);
-	if (IS_ERR(backup)) {
-		ret = PTR_ERR(backup);
-		goto out_unlock_bo;
+		backup->parent_obj = xe_bo_get(bo); /* Released by bo_destroy */
+		ttm_bo_pin(&backup->ttm);
+		bo->backup_obj = backup;
 	}
 
-	backup->parent_obj = xe_bo_get(bo); /* Released by bo_destroy */
-	ttm_bo_pin(&backup->ttm);
-	bo->backup_obj = backup;
-
-out_unlock_bo:
-	xe_bo_unlock(bo);
 	return ret;
 }
 
···
 	return 0;
 }
 
+static int xe_bo_evict_pinned_copy(struct xe_bo *bo, struct xe_bo *backup)
+{
+	struct xe_device *xe = xe_bo_device(bo);
+	bool unmap = false;
+	int ret = 0;
+
+	if (xe_bo_is_user(bo) || (bo->flags & XE_BO_FLAG_PINNED_LATE_RESTORE)) {
+		struct xe_migrate *migrate;
+		struct dma_fence *fence;
+
+		if (bo->tile)
+			migrate = bo->tile->migrate;
+		else
+			migrate = mem_type_to_migrate(xe, bo->ttm.resource->mem_type);
+
+		xe_assert(xe, bo->ttm.base.resv == backup->ttm.base.resv);
+		ret = dma_resv_reserve_fences(bo->ttm.base.resv, 1);
+		if (ret)
+			goto out_backup;
+
+		fence = xe_migrate_copy(migrate, bo, backup, bo->ttm.resource,
+					backup->ttm.resource, false);
+		if (IS_ERR(fence)) {
+			ret = PTR_ERR(fence);
+			goto out_backup;
+		}
+
+		dma_resv_add_fence(bo->ttm.base.resv, fence,
+				   DMA_RESV_USAGE_KERNEL);
+		dma_fence_put(fence);
+	} else {
+		ret = xe_bo_vmap(backup);
+		if (ret)
+			goto out_backup;
+
+		if (iosys_map_is_null(&bo->vmap)) {
+			ret = xe_bo_vmap(bo);
+			if (ret)
+				goto out_vunmap;
+			unmap = true;
+		}
+
+		xe_map_memcpy_from(xe, backup->vmap.vaddr, &bo->vmap, 0,
+				   xe_bo_size(bo));
+	}
+
+	if (!bo->backup_obj)
+		bo->backup_obj = backup;
+out_vunmap:
+	xe_bo_vunmap(backup);
+out_backup:
+	if (unmap)
+		xe_bo_vunmap(bo);
+
+	return ret;
+}
+
 /**
  * xe_bo_evict_pinned() - Evict a pinned VRAM object to system memory
  * @bo: The buffer object to move.
···
 int xe_bo_evict_pinned(struct xe_bo *bo)
 {
 	struct xe_device *xe = ttm_to_xe_device(bo->ttm.bdev);
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
 	struct xe_bo *backup = bo->backup_obj;
 	bool backup_created = false;
-	bool unmap = false;
 	int ret = 0;
 
-	xe_bo_lock(bo, false);
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {.exclusive = true}, ret) {
+		ret = drm_exec_lock_obj(&exec, &bo->ttm.base);
+		drm_exec_retry_on_contention(&exec);
+		xe_assert(xe, !ret);
 
-	if (WARN_ON(!bo->ttm.resource)) {
-		ret = -EINVAL;
-		goto out_unlock_bo;
-	}
-
-	if (WARN_ON(!xe_bo_is_pinned(bo))) {
-		ret = -EINVAL;
-		goto out_unlock_bo;
-	}
-
-	if (!xe_bo_is_vram(bo))
-		goto out_unlock_bo;
-
-	if (bo->flags & XE_BO_FLAG_PINNED_NORESTORE)
-		goto out_unlock_bo;
-
-	if (!backup) {
-		backup = ___xe_bo_create_locked(xe, NULL, NULL, bo->ttm.base.resv,
-						NULL, xe_bo_size(bo),
-						DRM_XE_GEM_CPU_CACHING_WB, ttm_bo_type_kernel,
-						XE_BO_FLAG_SYSTEM | XE_BO_FLAG_NEEDS_CPU_ACCESS |
-						XE_BO_FLAG_PINNED);
-		if (IS_ERR(backup)) {
-			ret = PTR_ERR(backup);
-			goto out_unlock_bo;
-		}
-		backup->parent_obj = xe_bo_get(bo); /* Released by bo_destroy */
-		backup_created = true;
-	}
-
-	if (xe_bo_is_user(bo) || (bo->flags & XE_BO_FLAG_PINNED_LATE_RESTORE)) {
-		struct xe_migrate *migrate;
-		struct dma_fence *fence;
-
-		if (bo->tile)
-			migrate = bo->tile->migrate;
-		else
-			migrate = mem_type_to_migrate(xe, bo->ttm.resource->mem_type);
-
-		ret = dma_resv_reserve_fences(bo->ttm.base.resv, 1);
-		if (ret)
-			goto out_backup;
-
-		ret = dma_resv_reserve_fences(backup->ttm.base.resv, 1);
-		if (ret)
-			goto out_backup;
-
-		fence = xe_migrate_copy(migrate, bo, backup, bo->ttm.resource,
-					backup->ttm.resource, false);
-		if (IS_ERR(fence)) {
-			ret = PTR_ERR(fence);
-			goto out_backup;
+		if (WARN_ON(!bo->ttm.resource)) {
+			ret = -EINVAL;
+			break;
 		}
 
-		dma_resv_add_fence(bo->ttm.base.resv, fence,
-				   DMA_RESV_USAGE_KERNEL);
-		dma_resv_add_fence(backup->ttm.base.resv, fence,
-				   DMA_RESV_USAGE_KERNEL);
-		dma_fence_put(fence);
-	} else {
-		ret = xe_bo_vmap(backup);
-		if (ret)
-			goto out_backup;
-
-		if (iosys_map_is_null(&bo->vmap)) {
-			ret = xe_bo_vmap(bo);
-			if (ret)
-				goto out_backup;
-			unmap = true;
+		if (WARN_ON(!xe_bo_is_pinned(bo))) {
+			ret = -EINVAL;
+			break;
 		}
 
-		xe_map_memcpy_from(xe, backup->vmap.vaddr, &bo->vmap, 0,
-				   xe_bo_size(bo));
+		if (!xe_bo_is_vram(bo))
+			break;
+
+		if (bo->flags & XE_BO_FLAG_PINNED_NORESTORE)
+			break;
+
+		if (!backup) {
+			backup = xe_bo_init_locked(xe, NULL, NULL, bo->ttm.base.resv, NULL,
+						   xe_bo_size(bo),
+						   DRM_XE_GEM_CPU_CACHING_WB, ttm_bo_type_kernel,
+						   XE_BO_FLAG_SYSTEM | XE_BO_FLAG_NEEDS_CPU_ACCESS |
+						   XE_BO_FLAG_PINNED, &exec);
+			if (IS_ERR(backup)) {
+				drm_exec_retry_on_contention(&exec);
+				ret = PTR_ERR(backup);
+				xe_validation_retry_on_oom(&ctx, &ret);
+				break;
+			}
+			backup->parent_obj = xe_bo_get(bo); /* Released by bo_destroy */
+			backup_created = true;
+		}
+
+		ret = xe_bo_evict_pinned_copy(bo, backup);
 	}
 
-	if (!bo->backup_obj)
-		bo->backup_obj = backup;
-
-out_backup:
-	xe_bo_vunmap(backup);
 	if (ret && backup_created)
 		xe_bo_put(backup);
-out_unlock_bo:
-	if (unmap)
-		xe_bo_vunmap(bo);
-	xe_bo_unlock(bo);
+
 	return ret;
 }
 
···
 		if (ret)
 			goto out_unlock_bo;
 
-		ret = dma_resv_reserve_fences(backup->ttm.base.resv, 1);
-		if (ret)
-			goto out_unlock_bo;
-
 		fence = xe_migrate_copy(migrate, backup, bo,
 					backup->ttm.resource, bo->ttm.resource,
 					false);
···
 		}
 
 		dma_resv_add_fence(bo->ttm.base.resv, fence,
-				   DMA_RESV_USAGE_KERNEL);
-		dma_resv_add_fence(backup->ttm.base.resv, fence,
 				   DMA_RESV_USAGE_KERNEL);
 		dma_fence_put(fence);
 	} else {
···
 	if (!xe_bo_is_xe_bo(ttm_bo))
 		return;
 
-	if (IS_VF_CCS_BB_VALID(ttm_to_xe_device(ttm_bo->bdev), bo))
+	if (IS_VF_CCS_READY(ttm_to_xe_device(ttm_bo->bdev)))
 		xe_sriov_vf_ccs_detach_bo(bo);
 
 	/*
···
 	       bo->attr.atomic_access == DRM_XE_ATOMIC_CPU;
 }
 
-static vm_fault_t xe_gem_fault(struct vm_fault *vmf)
+/* Populate the bo if swapped out, or migrate if the access mode requires that. */
+static int xe_bo_fault_migrate(struct xe_bo *bo, struct ttm_operation_ctx *ctx,
+			       struct drm_exec *exec)
+{
+	struct ttm_buffer_object *tbo = &bo->ttm;
+	int err = 0;
+
+	if (ttm_manager_type(tbo->bdev, tbo->resource->mem_type)->use_tt) {
+		xe_assert(xe_bo_device(bo),
+			  dma_resv_test_signaled(tbo->base.resv, DMA_RESV_USAGE_KERNEL) ||
+			  (tbo->ttm && ttm_tt_is_populated(tbo->ttm)));
+		err = ttm_bo_populate(&bo->ttm, ctx);
+	} else if (should_migrate_to_smem(bo)) {
+		xe_assert(xe_bo_device(bo), bo->flags & XE_BO_FLAG_SYSTEM);
+		err = xe_bo_migrate(bo, XE_PL_TT, ctx, exec);
+	}
+
+	return err;
+}
+
+/* Call into TTM to populate PTEs, and register bo for PTE removal on runtime suspend. */
+static vm_fault_t __xe_bo_cpu_fault(struct vm_fault *vmf, struct xe_device *xe, struct xe_bo *bo)
+{
+	vm_fault_t ret;
+
+	trace_xe_bo_cpu_fault(bo);
+
+	ret = ttm_bo_vm_fault_reserved(vmf, vmf->vma->vm_page_prot,
+				       TTM_BO_VM_NUM_PREFAULT);
+	/*
+	 * When TTM is actually called to insert PTEs, ensure no blocking conditions
+	 * remain, in which case TTM may drop locks and return VM_FAULT_RETRY.
+	 */
+	xe_assert(xe, ret != VM_FAULT_RETRY);
+
+	if (ret == VM_FAULT_NOPAGE &&
+	    mem_type_is_vram(bo->ttm.resource->mem_type)) {
+		mutex_lock(&xe->mem_access.vram_userfault.lock);
+		if (list_empty(&bo->vram_userfault_link))
+			list_add(&bo->vram_userfault_link,
+				 &xe->mem_access.vram_userfault.list);
+		mutex_unlock(&xe->mem_access.vram_userfault.lock);
+	}
+
+	return ret;
+}
+
+static vm_fault_t xe_err_to_fault_t(int err)
+{
+	switch (err) {
+	case 0:
+	case -EINTR:
+	case -ERESTARTSYS:
+	case -EAGAIN:
+		return VM_FAULT_NOPAGE;
+	case -ENOMEM:
+	case -ENOSPC:
+		return VM_FAULT_OOM;
+	default:
+		break;
+	}
+	return VM_FAULT_SIGBUS;
+}
+
+static bool xe_ttm_bo_is_imported(struct ttm_buffer_object *tbo)
+{
+	dma_resv_assert_held(tbo->base.resv);
+
+	return tbo->ttm &&
+		(tbo->ttm->page_flags & (TTM_TT_FLAG_EXTERNAL | TTM_TT_FLAG_EXTERNAL_MAPPABLE)) ==
+		TTM_TT_FLAG_EXTERNAL;
+}
+
+static vm_fault_t xe_bo_cpu_fault_fastpath(struct vm_fault *vmf, struct xe_device *xe,
+					   struct xe_bo *bo, bool needs_rpm)
+{
+	struct ttm_buffer_object *tbo = &bo->ttm;
+	vm_fault_t ret = VM_FAULT_RETRY;
+	struct xe_validation_ctx ctx;
+	struct ttm_operation_ctx tctx = {
+		.interruptible = true,
+		.no_wait_gpu = true,
+		.gfp_retry_mayfail = true,
+
+	};
+	int err;
+
+	if (needs_rpm && !xe_pm_runtime_get_if_active(xe))
+		return VM_FAULT_RETRY;
+
+	err = xe_validation_ctx_init(&ctx, &xe->val, NULL,
+				     (struct xe_val_flags) {
+					     .interruptible = true,
+					     .no_block = true
+				     });
+	if (err)
+		goto out_pm;
+
+	if (!dma_resv_trylock(tbo->base.resv))
+		goto out_validation;
+
+	if (xe_ttm_bo_is_imported(tbo)) {
+		ret = VM_FAULT_SIGBUS;
+		drm_dbg(&xe->drm, "CPU trying to access an imported buffer object.\n");
+		goto out_unlock;
+	}
+
+	err = xe_bo_fault_migrate(bo, &tctx, NULL);
+	if (err) {
+		/* Return VM_FAULT_RETRY on these errors. */
+		if (err != -ENOMEM && err != -ENOSPC && err != -EBUSY)
+			ret = xe_err_to_fault_t(err);
+		goto out_unlock;
+	}
+
+	if (dma_resv_test_signaled(bo->ttm.base.resv, DMA_RESV_USAGE_KERNEL))
+		ret = __xe_bo_cpu_fault(vmf, xe, bo);
+
+out_unlock:
+	dma_resv_unlock(tbo->base.resv);
+out_validation:
+	xe_validation_ctx_fini(&ctx);
+out_pm:
+	if (needs_rpm)
+		xe_pm_runtime_put(xe);
+
+	return ret;
+}
+
+static vm_fault_t xe_bo_cpu_fault(struct vm_fault *vmf)
 {
 	struct ttm_buffer_object *tbo = vmf->vma->vm_private_data;
 	struct drm_device *ddev = tbo->base.dev;
 	struct xe_device *xe = to_xe_device(ddev);
 	struct xe_bo *bo = ttm_to_xe_bo(tbo);
 	bool needs_rpm = bo->flags & XE_BO_FLAG_VRAM_MASK;
+	bool retry_after_wait = false;
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
 	vm_fault_t ret;
-	int idx, r = 0;
+	int err = 0;
+	int idx;
+
+	if (!drm_dev_enter(&xe->drm, &idx))
+		return ttm_bo_vm_dummy_page(vmf, vmf->vma->vm_page_prot);
+
+	ret = xe_bo_cpu_fault_fastpath(vmf, xe, bo, needs_rpm);
+	if (ret != VM_FAULT_RETRY)
+		goto out;
+
+	if (fault_flag_allow_retry_first(vmf->flags)) {
+		if (vmf->flags & FAULT_FLAG_RETRY_NOWAIT)
+			goto out;
+		retry_after_wait = true;
+		xe_bo_get(bo);
+		mmap_read_unlock(vmf->vma->vm_mm);
+	} else {
+		ret = VM_FAULT_NOPAGE;
+	}
+
+	/*
+	 * The fastpath failed and we were not required to return and retry immediately.
+	 * We're now running in one of two modes:
+	 *
+	 * 1) retry_after_wait == true: The mmap_read_lock() is dropped, and we're trying
+	 * to resolve blocking waits. But we can't resolve the fault since the
+	 * mmap_read_lock() is dropped. After retrying the fault, the aim is that the fastpath
+	 * should succeed. But it may fail since we drop the bo lock.
+	 *
+	 * 2) retry_after_wait == false: The fastpath failed, typically even after
+	 * a retry. Do whatever's necessary to resolve the fault.
+	 *
+	 * This construct is recommended to avoid excessive waits under the mmap_lock.
+	 */
 
 	if (needs_rpm)
 		xe_pm_runtime_get(xe);
 
-	ret = ttm_bo_vm_reserve(tbo, vmf);
-	if (ret)
-		goto out;
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {.interruptible = true},
+			    err) {
+		struct ttm_operation_ctx tctx = {
+			.interruptible = true,
+			.no_wait_gpu = false,
+			.gfp_retry_mayfail = retry_after_wait,
+		};
+		long lerr;
 
-	if (drm_dev_enter(ddev, &idx)) {
-		trace_xe_bo_cpu_fault(bo);
+		err = drm_exec_lock_obj(&exec, &tbo->base);
+		drm_exec_retry_on_contention(&exec);
+		if (err)
+			break;
 
-		if (should_migrate_to_smem(bo)) {
-			xe_assert(xe, bo->flags & XE_BO_FLAG_SYSTEM);
-
-			r = xe_bo_migrate(bo, XE_PL_TT);
-			if (r == -EBUSY || r == -ERESTARTSYS || r == -EINTR)
-				ret = VM_FAULT_NOPAGE;
-			else if (r)
-				ret = VM_FAULT_SIGBUS;
+		if (xe_ttm_bo_is_imported(tbo)) {
+			err = -EFAULT;
+			drm_dbg(&xe->drm, "CPU trying to access an imported buffer object.\n");
+			break;
 		}
-		if (!ret)
-			ret = ttm_bo_vm_fault_reserved(vmf,
-						       vmf->vma->vm_page_prot,
-						       TTM_BO_VM_NUM_PREFAULT);
-		drm_dev_exit(idx);
 
-		if (ret == VM_FAULT_RETRY &&
-		    !(vmf->flags & FAULT_FLAG_RETRY_NOWAIT))
-			goto out;
-
-		/*
-		 * ttm_bo_vm_reserve() already has dma_resv_lock.
-		 */
-		if (ret == VM_FAULT_NOPAGE &&
-		    mem_type_is_vram(tbo->resource->mem_type)) {
-			mutex_lock(&xe->mem_access.vram_userfault.lock);
-			if (list_empty(&bo->vram_userfault_link))
-				list_add(&bo->vram_userfault_link,
-					 &xe->mem_access.vram_userfault.list);
-			mutex_unlock(&xe->mem_access.vram_userfault.lock);
+		err = xe_bo_fault_migrate(bo, &tctx, &exec);
+		if (err) {
+			drm_exec_retry_on_contention(&exec);
+			xe_validation_retry_on_oom(&ctx, &err);
+			break;
 		}
-	} else {
-		ret = ttm_bo_vm_dummy_page(vmf, vmf->vma->vm_page_prot);
+
+		lerr = dma_resv_wait_timeout(tbo->base.resv,
+					     DMA_RESV_USAGE_KERNEL, true,
+					     MAX_SCHEDULE_TIMEOUT);
+		if (lerr < 0) {
+			err = lerr;
+			break;
+		}
+
+		if (!retry_after_wait)
+			ret = __xe_bo_cpu_fault(vmf, xe, bo);
 	}
+	/* if retry_after_wait == true, we *must* return VM_FAULT_RETRY. */
+	if (err && !retry_after_wait)
+		ret = xe_err_to_fault_t(err);
 
-	dma_resv_unlock(tbo->base.resv);
-out:
 	if (needs_rpm)
 		xe_pm_runtime_put(xe);
+
+	if (retry_after_wait)
+		xe_bo_put(bo);
+out:
+	drm_dev_exit(idx);
 
 	return ret;
 }
···
 }
 
 static const struct vm_operations_struct xe_gem_vm_ops = {
-	.fault = xe_gem_fault,
+	.fault = xe_bo_cpu_fault,
 	.open = ttm_bo_vm_open,
 	.close = ttm_bo_vm_close,
 	.access = xe_bo_vm_access,
···
 	kfree(bo);
 }
 
-struct xe_bo *___xe_bo_create_locked(struct xe_device *xe, struct xe_bo *bo,
-				     struct xe_tile *tile, struct dma_resv *resv,
-				     struct ttm_lru_bulk_move *bulk, size_t size,
-				     u16 cpu_caching, enum ttm_bo_type type,
-				     u32 flags)
+/**
+ * xe_bo_init_locked() - Initialize or create an xe_bo.
+ * @xe: The xe device.
+ * @bo: An already allocated buffer object or NULL
+ * if the function should allocate a new one.
+ * @tile: The tile to select for migration of this bo, and the tile used for
+ * GGTT binding if any. Only to be non-NULL for ttm_bo_type_kernel bos.
+ * @resv: Pointer to a locked shared reservation object to use fo this bo,
+ * or NULL for the xe_bo to use its own.
+ * @bulk: The bulk move to use for LRU bumping, or NULL for external bos.
+ * @size: The storage size to use for the bo.
+ * @cpu_caching: The cpu caching used for system memory backing store.
+ * @type: The TTM buffer object type.
+ * @flags: XE_BO_FLAG_ flags.
+ * @exec: The drm_exec transaction to use for exhaustive eviction.
+ *
+ * Initialize or create an xe buffer object. On failure, any allocated buffer
+ * object passed in @bo will have been unreferenced.
+ *
+ * Return: The buffer object on success. Negative error pointer on failure.
+ */
+struct xe_bo *xe_bo_init_locked(struct xe_device *xe, struct xe_bo *bo,
+				struct xe_tile *tile, struct dma_resv *resv,
+				struct ttm_lru_bulk_move *bulk, size_t size,
+				u16 cpu_caching, enum ttm_bo_type type,
+				u32 flags, struct drm_exec *exec)
 {
 	struct ttm_operation_ctx ctx = {
 		.interruptible = true,
···
 		ctx.resv = resv;
 	}
 
+	xe_validation_assert_exec(xe, exec, &bo->ttm.base);
 	if (!(flags & XE_BO_FLAG_FIXED_PLACEMENT)) {
 		err = __xe_bo_placement_for_flags(xe, bo, bo->flags);
 		if (WARN_ON(err)) {
···
 		      struct xe_tile *tile, struct xe_vm *vm,
 		      size_t size, u64 start, u64 end,
 		      u16 cpu_caching, enum ttm_bo_type type, u32 flags,
-		      u64 alignment)
+		      u64 alignment, struct drm_exec *exec)
 {
 	struct xe_bo *bo = NULL;
 	int err;
···
 		}
 	}
 
-	bo = ___xe_bo_create_locked(xe, bo, tile, vm ? xe_vm_resv(vm) : NULL,
-				    vm && !xe_vm_in_fault_mode(vm) &&
-				    flags & XE_BO_FLAG_USER ?
-				    &vm->lru_bulk_move : NULL, size,
-				    cpu_caching, type, flags);
+	bo = xe_bo_init_locked(xe, bo, tile, vm ? xe_vm_resv(vm) : NULL,
+			       vm && !xe_vm_in_fault_mode(vm) &&
+			       flags & XE_BO_FLAG_USER ?
+			       &vm->lru_bulk_move : NULL, size,
+			       cpu_caching, type, flags, exec);
 	if (IS_ERR(bo))
 		return bo;
 
···
 
 			if (flags & XE_BO_FLAG_FIXED_PLACEMENT) {
 				err = xe_ggtt_insert_bo_at(t->mem.ggtt, bo,
-							   start + xe_bo_size(bo), U64_MAX);
+							   start + xe_bo_size(bo), U64_MAX,
+							   exec);
 			} else {
-				err = xe_ggtt_insert_bo(t->mem.ggtt, bo);
+				err = xe_ggtt_insert_bo(t->mem.ggtt, bo, exec);
 			}
 			if (err)
 				goto err_unlock_put_bo;
···
 	return ERR_PTR(err);
 }
 
-struct xe_bo *
-xe_bo_create_locked_range(struct xe_device *xe,
-			  struct xe_tile *tile, struct xe_vm *vm,
-			  size_t size, u64 start, u64 end,
-			  enum ttm_bo_type type, u32 flags, u64 alignment)
-{
-	return __xe_bo_create_locked(xe, tile, vm, size, start, end, 0, type,
-				     flags, alignment);
-}
-
+/**
+ * xe_bo_create_locked() - Create a BO
+ * @xe: The xe device.
+ * @tile: The tile to select for migration of this bo, and the tile used for
+ * GGTT binding if any. Only to be non-NULL for ttm_bo_type_kernel bos.
+ * @vm: The local vm or NULL for external objects.
+ * @size: The storage size to use for the bo.
+ * @type: The TTM buffer object type.
+ * @flags: XE_BO_FLAG_ flags.
+ * @exec: The drm_exec transaction to use for exhaustive eviction.
+ *
+ * Create a locked xe BO with no range- nor alignment restrictions.
+ *
+ * Return: The buffer object on success. Negative error pointer on failure.
+ */
 struct xe_bo *xe_bo_create_locked(struct xe_device *xe, struct xe_tile *tile,
 				  struct xe_vm *vm, size_t size,
-				  enum ttm_bo_type type, u32 flags)
+				  enum ttm_bo_type type, u32 flags,
+				  struct drm_exec *exec)
 {
 	return __xe_bo_create_locked(xe, tile, vm, size, 0, ~0ULL, 0, type,
-				     flags, 0);
+				     flags, 0, exec);
 }
 
-struct xe_bo *xe_bo_create_user(struct xe_device *xe, struct xe_tile *tile,
+static struct xe_bo *xe_bo_create_novm(struct xe_device *xe, struct xe_tile *tile,
+				       size_t size, u16 cpu_caching,
+				       enum ttm_bo_type type, u32 flags,
+				       u64 alignment, bool intr)
+{
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
+	struct xe_bo *bo;
+	int ret = 0;
+
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {.interruptible = intr},
+			    ret) {
+		bo = __xe_bo_create_locked(xe, tile, NULL, size, 0, ~0ULL,
+					   cpu_caching, type, flags, alignment, &exec);
+		drm_exec_retry_on_contention(&exec);
+		if (IS_ERR(bo)) {
+			ret = PTR_ERR(bo);
+			xe_validation_retry_on_oom(&ctx, &ret);
+		} else {
+			xe_bo_unlock(bo);
+		}
+	}
+
+	return ret ? ERR_PTR(ret) : bo;
+}
+
+/**
+ * xe_bo_create_user() - Create a user BO
+ * @xe: The xe device.
+ * @vm: The local vm or NULL for external objects.
+ * @size: The storage size to use for the bo.
+ * @cpu_caching: The caching mode to be used for system backing store.
+ * @flags: XE_BO_FLAG_ flags.
+ * @exec: The drm_exec transaction to use for exhaustive eviction, or NULL
+ * if such a transaction should be initiated by the call.
+ *
+ * Create a bo on behalf of user-space.
+ *
+ * Return: The buffer object on success. Negative error pointer on failure.
+ */
+struct xe_bo *xe_bo_create_user(struct xe_device *xe,
 				struct xe_vm *vm, size_t size,
 				u16 cpu_caching,
-				u32 flags)
+				u32 flags, struct drm_exec *exec)
 {
-	struct xe_bo *bo = __xe_bo_create_locked(xe, tile, vm, size, 0, ~0ULL,
-						 cpu_caching, ttm_bo_type_device,
-						 flags | XE_BO_FLAG_USER, 0);
-	if (!IS_ERR(bo))
-		xe_bo_unlock_vm_held(bo);
+	struct xe_bo *bo;
+
+	flags |= XE_BO_FLAG_USER;
+
+	if (vm || exec) {
+		xe_assert(xe, exec);
+		bo = __xe_bo_create_locked(xe, NULL, vm, size, 0, ~0ULL,
+					   cpu_caching, ttm_bo_type_device,
+					   flags, 0, exec);
+		if (!IS_ERR(bo))
+			xe_bo_unlock_vm_held(bo);
+	} else {
+		bo = xe_bo_create_novm(xe, NULL, size, cpu_caching,
+				       ttm_bo_type_device, flags, 0, true);
+	}
 
 	return bo;
 }
 
-struct xe_bo *xe_bo_create(struct xe_device *xe, struct xe_tile *tile,
-			   struct xe_vm *vm, size_t size,
-			   enum ttm_bo_type type, u32 flags)
+/**
+ * xe_bo_create_pin_range_novm() - Create and pin a BO with range options.
+ * @xe: The xe device.
+ * @tile: The tile to select for migration of this bo, and the tile used for
+ * GGTT binding if any. Only to be non-NULL for ttm_bo_type_kernel bos.
+ * @size: The storage size to use for the bo.
+ * @start: Start of fixed VRAM range or 0.
+ * @end: End of fixed VRAM range or ~0ULL.
+ * @type: The TTM buffer object type.
+ * @flags: XE_BO_FLAG_ flags.
+ *
+ * Create an Xe BO with range- and options. If @start and @end indicate
+ * a fixed VRAM range, this must be a ttm_bo_type_kernel bo with VRAM placement
+ * only.
+ *
+ * Return: The buffer object on success. Negative error pointer on failure.
+ */
+struct xe_bo *xe_bo_create_pin_range_novm(struct xe_device *xe, struct xe_tile *tile,
+					  size_t size, u64 start, u64 end,
+					  enum ttm_bo_type type, u32 flags)
 {
-	struct xe_bo *bo = xe_bo_create_locked(xe, tile, vm, size, type, flags);
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
+	struct xe_bo *bo;
+	int err = 0;
 
-	if (!IS_ERR(bo))
-		xe_bo_unlock_vm_held(bo);
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {}, err) {
+		bo = __xe_bo_create_locked(xe, tile, NULL, size, start, end,
+					   0, type, flags, 0, &exec);
+		if (IS_ERR(bo)) {
+			drm_exec_retry_on_contention(&exec);
+			err = PTR_ERR(bo);
+			xe_validation_retry_on_oom(&ctx, &err);
+			break;
+		}
 
-	return bo;
+		err = xe_bo_pin(bo, &exec);
+		xe_bo_unlock(bo);
+		if (err) {
+			xe_bo_put(bo);
+			drm_exec_retry_on_contention(&exec);
+			xe_validation_retry_on_oom(&ctx, &err);
+			break;
+		}
+	}
+
+	return err ? ERR_PTR(err) : bo;
 }
 
-struct xe_bo *xe_bo_create_pin_map_at(struct xe_device *xe, struct xe_tile *tile,
-				      struct xe_vm *vm,
-				      size_t size, u64 offset,
-				      enum ttm_bo_type type, u32 flags)
-{
-	return xe_bo_create_pin_map_at_aligned(xe, tile, vm, size, offset,
-					       type, flags, 0);
-}
-
-struct xe_bo *xe_bo_create_pin_map_at_aligned(struct xe_device *xe,
-					      struct xe_tile *tile,
-					      struct xe_vm *vm,
-					      size_t size, u64 offset,
-					      enum ttm_bo_type type, u32 flags,
-					      u64 alignment)
+static struct xe_bo *xe_bo_create_pin_map_at_aligned(struct xe_device *xe,
+						     struct xe_tile *tile,
+						     struct xe_vm *vm,
+						     size_t size, u64 offset,
+						     enum ttm_bo_type type, u32 flags,
+						     u64 alignment, struct drm_exec *exec)
 {
 	struct xe_bo *bo;
 	int err;
 	u64 start = offset == ~0ull ? 0 : offset;
-	u64 end = offset == ~0ull ? offset : start + size;
+	u64 end = offset == ~0ull ? ~0ull : start + size;
 
 	if (flags & XE_BO_FLAG_STOLEN &&
 	    xe_ttm_stolen_cpu_access_needs_ggtt(xe))
 		flags |= XE_BO_FLAG_GGTT;
 
-	bo = xe_bo_create_locked_range(xe, tile, vm, size, start, end, type,
-				       flags | XE_BO_FLAG_NEEDS_CPU_ACCESS | XE_BO_FLAG_PINNED,
-				       alignment);
+	bo = __xe_bo_create_locked(xe, tile, vm, size, start, end, 0, type,
+				   flags | XE_BO_FLAG_NEEDS_CPU_ACCESS | XE_BO_FLAG_PINNED,
+				   alignment, exec);
 	if (IS_ERR(bo))
 		return bo;
 
-	err = xe_bo_pin(bo);
+	err = xe_bo_pin(bo, exec);
 	if (err)
 		goto err_put;
 
···
 	return ERR_PTR(err);
 }
 
+/**
+ * xe_bo_create_pin_map_at_novm() - Create pinned and mapped bo at optional VRAM offset
+ * @xe: The xe device.
+ * @tile: The tile to select for migration of this bo, and the tile used for
+ * GGTT binding if any. Only to be non-NULL for ttm_bo_type_kernel bos.
+ * @size: The storage size to use for the bo.
+ * @offset: Optional VRAM offset or %~0ull for don't care.
+ * @type: The TTM buffer object type.
+ * @flags: XE_BO_FLAG_ flags.
+ * @alignment: GGTT alignment.
+ * @intr: Whether to execute any waits for backing store interruptible.
+ *
+ * Create a pinned and optionally mapped bo with VRAM offset and GGTT alignment
+ * options. The bo will be external and not associated with a VM.
+ *
+ * Return: The buffer object on success. Negative error pointer on failure.
+ * In particular, the function may return ERR_PTR(%-EINTR) if @intr was set
+ * to true on entry.
+ */
+struct xe_bo *
+xe_bo_create_pin_map_at_novm(struct xe_device *xe, struct xe_tile *tile,
+			     size_t size, u64 offset, enum ttm_bo_type type, u32 flags,
+			     u64 alignment, bool intr)
+{
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
+	struct xe_bo *bo;
+	int ret = 0;
+
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {.interruptible = intr},
+			    ret) {
+		bo = xe_bo_create_pin_map_at_aligned(xe, tile, NULL, size, offset,
+						     type, flags, alignment, &exec);
+		if (IS_ERR(bo)) {
+			drm_exec_retry_on_contention(&exec);
+			ret = PTR_ERR(bo);
+			xe_validation_retry_on_oom(&ctx, &ret);
+		}
+	}
+
+	return ret ? ERR_PTR(ret) : bo;
+}
+
+/**
+ * xe_bo_create_pin_map() - Create pinned and mapped bo
+ * @xe: The xe device.
+ * @tile: The tile to select for migration of this bo, and the tile used for
+ * @vm: The vm to associate the buffer object with. The vm's resv must be locked
+ * with the transaction represented by @exec.
+ * GGTT binding if any. Only to be non-NULL for ttm_bo_type_kernel bos.
+ * @size: The storage size to use for the bo.
+ * @type: The TTM buffer object type.
+ * @flags: XE_BO_FLAG_ flags.
+ * @exec: The drm_exec transaction to use for exhaustive eviction, and
+ * previously used for locking @vm's resv.
+ *
+ * Create a pinned and mapped bo. The bo will be external and not associated
+ * with a VM.
+ *
+ * Return: The buffer object on success. Negative error pointer on failure.
+ * In particular, the function may return ERR_PTR(%-EINTR) if @exec was
+ * configured for interruptible locking.
+ */
 struct xe_bo *xe_bo_create_pin_map(struct xe_device *xe, struct xe_tile *tile,
 				   struct xe_vm *vm, size_t size,
-				   enum ttm_bo_type type, u32 flags)
+				   enum ttm_bo_type type, u32 flags,
+				   struct drm_exec *exec)
 {
-	return xe_bo_create_pin_map_at(xe, tile, vm, size, ~0ull, type, flags);
+	return xe_bo_create_pin_map_at_aligned(xe, tile, vm, size, ~0ull, type, flags,
+					       0, exec);
+}
+
+/**
+ * xe_bo_create_pin_map_novm() - Create pinned and mapped bo
+ * @xe: The xe device.
+ * @tile: The tile to select for migration of this bo, and the tile used for
+ * GGTT binding if any. Only to be non-NULL for ttm_bo_type_kernel bos.
+ * @size: The storage size to use for the bo.
+ * @type: The TTM buffer object type.
+ * @flags: XE_BO_FLAG_ flags.
+ * @intr: Whether to execut any waits for backing store interruptible.
+ *
+ * Create a pinned and mapped bo. The bo will be external and not associated
+ * with a VM.
+ *
+ * Return: The buffer object on success. Negative error pointer on failure.
+ * In particular, the function may return ERR_PTR(%-EINTR) if @intr was set
+ * to true on entry.
+ */
+struct xe_bo *xe_bo_create_pin_map_novm(struct xe_device *xe, struct xe_tile *tile,
+					size_t size, enum ttm_bo_type type, u32 flags,
+					bool intr)
+{
+	return xe_bo_create_pin_map_at_novm(xe, tile, size, ~0ull, type, flags, 0, intr);
 }
 
 static void __xe_bo_unpin_map_no_vm(void *arg)
···
 	int ret;
 
 	KUNIT_STATIC_STUB_REDIRECT(xe_managed_bo_create_pin_map, xe, tile, size, flags);
-
-	bo = xe_bo_create_pin_map(xe, tile, NULL, size, ttm_bo_type_kernel, flags);
+	bo = xe_bo_create_pin_map_novm(xe, tile, size, ttm_bo_type_kernel, flags, true);
 	if (IS_ERR(bo))
 		return bo;
 
···
 		return ERR_PTR(ret);
 
 	return bo;
+}
+
+void xe_managed_bo_unpin_map_no_vm(struct xe_bo *bo)
+{
+	devm_release_action(xe_bo_device(bo)->drm.dev, __xe_bo_unpin_map_no_vm, bo);
 }
 
 struct xe_bo *xe_managed_bo_create_from_data(struct xe_device *xe, struct xe_tile *tile,
···
  * xe_bo_pin_external - pin an external BO
  * @bo: buffer object to be pinned
  * @in_place: Pin in current placement, don't attempt to migrate.
+ * @exec: The drm_exec transaction to use for exhaustive eviction.
  *
  * Pin an external (not tied to a VM, can be exported via dma-buf / prime FD)
  * BO. Unique call compared to xe_bo_pin as this function has it own set of
···
  *
  * Returns 0 for success, negative error code otherwise.
  */
-int xe_bo_pin_external(struct xe_bo *bo, bool in_place)
+int xe_bo_pin_external(struct xe_bo *bo, bool in_place, struct drm_exec *exec)
 {
 	struct xe_device *xe = xe_bo_device(bo);
 	int err;
···
 
 	if (!xe_bo_is_pinned(bo)) {
 		if (!in_place) {
-			err = xe_bo_validate(bo, NULL, false);
+			err = xe_bo_validate(bo, NULL, false, exec);
 			if (err)
 				return err;
 		}
···
 	return 0;
 }
 
-int xe_bo_pin(struct xe_bo *bo)
+/**
+ * xe_bo_pin() - Pin a kernel bo after potentially migrating it
+ * @bo: The kernel bo to pin.
+ * @exec: The drm_exec transaction to use for exhaustive eviction.
+ *
+ * Attempts to migrate a bo to @bo->placement. If that succeeds,
+ * pins the bo.
+ *
+ * Return: %0 on success, negative error code on migration failure.
+ */
+int xe_bo_pin(struct xe_bo *bo, struct drm_exec *exec)
 {
 	struct ttm_place *place = &bo->placements[0];
 	struct xe_device *xe = xe_bo_device(bo);
···
 	/* We only expect at most 1 pin */
 	xe_assert(xe, !xe_bo_is_pinned(bo));
 
-	err = xe_bo_validate(bo, NULL, false);
+	err = xe_bo_validate(bo, NULL, false, exec);
 	if (err)
 		return err;
 
···
  *      NULL. Used together with @allow_res_evict.
  * @allow_res_evict: Whether it's allowed to evict bos sharing @vm's
  *                   reservation object.
+ * @exec: The drm_exec transaction to use for exhaustive eviction.
  *
  * Make sure the bo is in allowed placement, migrating it if necessary. If
  * needed, other bos will be evicted. If bos selected for eviction shares
···
  * Return: 0 on success, negative error code on failure. May return
  * -EINTR or -ERESTARTSYS if internal waits are interrupted by a signal.
  */
-int xe_bo_validate(struct xe_bo *bo, struct xe_vm *vm, bool allow_res_evict)
+int xe_bo_validate(struct xe_bo *bo, struct xe_vm *vm, bool allow_res_evict,
+		   struct drm_exec *exec)
 {
 	struct ttm_operation_ctx ctx = {
 		.interruptible = true,
···
 
 	xe_vm_set_validating(vm, allow_res_evict);
 	trace_xe_bo_validate(bo);
+	xe_validation_assert_exec(xe_bo_device(bo), exec, &bo->ttm.base);
 	ret = ttm_bo_validate(&bo->ttm, &bo->placement, &ctx);
 	xe_vm_clear_validating(vm, allow_res_evict);
 
···
 	struct xe_device *xe = to_xe_device(dev);
 	struct xe_file *xef = to_xe_file(file);
 	struct drm_xe_gem_create *args = data;
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
 	struct xe_vm *vm = NULL;
-	ktime_t end = 0;
 	struct xe_bo *bo;
 	unsigned int bo_flags;
 	u32 handle;
···
 			return -ENOENT;
 	}
 
-retry:
-	if (vm) {
-		err = xe_vm_lock(vm, true);
-		if (err)
-			goto out_vm;
+	err = 0;
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {.interruptible = true},
+			    err) {
+		if (vm) {
+			err = xe_vm_drm_exec_lock(vm, &exec);
+			drm_exec_retry_on_contention(&exec);
+			if (err)
+				break;
+		}
+		bo = xe_bo_create_user(xe, vm, args->size, args->cpu_caching,
+				       bo_flags, &exec);
+		drm_exec_retry_on_contention(&exec);
+		if (IS_ERR(bo)) {
+			err = PTR_ERR(bo);
+			xe_validation_retry_on_oom(&ctx, &err);
+			break;
+		}
 	}
-
-	bo = xe_bo_create_user(xe, NULL, vm, args->size, args->cpu_caching,
-			       bo_flags);
-
-	if (vm)
-		xe_vm_unlock(vm);
-
-	if (IS_ERR(bo)) {
-		err = PTR_ERR(bo);
-		if (xe_vm_validate_should_retry(NULL, err, &end))
-			goto retry;
+	if (err)
 		goto out_vm;
-	}
 
 	if (args->extensions) {
 		err = gem_create_user_extensions(xe, bo, args->extensions, 0);
···
  * xe_bo_migrate - Migrate an object to the desired region id
  * @bo: The buffer object to migrate.
  * @mem_type: The TTM region type to migrate to.
+ * @tctx: A pointer to a struct ttm_operation_ctx or NULL if
+ * a default interruptibe ctx is to be used.
+ * @exec: The drm_exec transaction to use for exhaustive eviction.
  *
  * Attempt to migrate the buffer object to the desired memory region. The
  * buffer object may not be pinned, and must be locked.
···
  * Return: 0 on success. Negative error code on failure. In particular may
  * return -EINTR or -ERESTARTSYS if signal pending.
  */
-int xe_bo_migrate(struct xe_bo *bo, u32 mem_type)
+int xe_bo_migrate(struct xe_bo *bo, u32 mem_type, struct ttm_operation_ctx *tctx,
+		  struct drm_exec *exec)
 {
 	struct xe_device *xe = ttm_to_xe_device(bo->ttm.bdev);
 	struct ttm_operation_ctx ctx = {
···
 	struct ttm_place requested;
 
 	xe_bo_assert_held(bo);
+	tctx = tctx ? tctx : &ctx;
 
 	if (bo->ttm.resource->mem_type == mem_type)
 		return 0;
···
 		add_vram(xe, bo, &requested, bo->flags, mem_type, &c);
 	}
 
-	return ttm_bo_validate(&bo->ttm, &placement, &ctx);
+	if (!tctx->no_wait_gpu)
+		xe_validation_assert_exec(xe_bo_device(bo), exec, &bo->ttm.base);
+	return ttm_bo_validate(&bo->ttm, &placement, tctx);
 }
 
 /**
  * xe_bo_evict - Evict an object to evict placement
  * @bo: The buffer object to migrate.
+ * @exec: The drm_exec transaction to use for exhaustive eviction.
  *
  * On successful completion, the object memory will be moved to evict
  * placement. This function blocks until the object has been fully moved.
  *
  * Return: 0 on success. Negative error code on failure.
  */
-int xe_bo_evict(struct xe_bo *bo)
+int xe_bo_evict(struct xe_bo *bo, struct drm_exec *exec)
 {
 	struct ttm_operation_ctx ctx = {
 		.interruptible = false,
···
 	args->size = ALIGN(mul_u32_u32(args->pitch, args->height),
 			   page_size);
 
-	bo = xe_bo_create_user(xe, NULL, NULL, args->size,
+	bo = xe_bo_create_user(xe, NULL, args->size,
 			       DRM_XE_GEM_CPU_CACHING_WC,
 			       XE_BO_FLAG_VRAM_IF_DGFX(xe_device_get_root_tile(xe)) |
 			       XE_BO_FLAG_SCANOUT |
-			       XE_BO_FLAG_NEEDS_CPU_ACCESS);
+			       XE_BO_FLAG_NEEDS_CPU_ACCESS, NULL);
 	if (IS_ERR(bo))
 		return PTR_ERR(bo);
 

drivers/gpu/drm/xe/xe_bo.h

···
 
 #include "xe_bo_types.h"
 #include "xe_macros.h"
+#include "xe_validation.h"
 #include "xe_vm_types.h"
 #include "xe_vm.h"
 #include "xe_vram_types.h"
···
 struct xe_bo *xe_bo_alloc(void);
 void xe_bo_free(struct xe_bo *bo);
 
-struct xe_bo *___xe_bo_create_locked(struct xe_device *xe, struct xe_bo *bo,
-				     struct xe_tile *tile, struct dma_resv *resv,
-				     struct ttm_lru_bulk_move *bulk, size_t size,
-				     u16 cpu_caching, enum ttm_bo_type type,
-				     u32 flags);
-struct xe_bo *
-xe_bo_create_locked_range(struct xe_device *xe,
-			  struct xe_tile *tile, struct xe_vm *vm,
-			  size_t size, u64 start, u64 end,
-			  enum ttm_bo_type type, u32 flags, u64 alignment);
+struct xe_bo *xe_bo_init_locked(struct xe_device *xe, struct xe_bo *bo,
+				struct xe_tile *tile, struct dma_resv *resv,
+				struct ttm_lru_bulk_move *bulk, size_t size,
+				u16 cpu_caching, enum ttm_bo_type type,
+				u32 flags, struct drm_exec *exec);
 struct xe_bo *xe_bo_create_locked(struct xe_device *xe, struct xe_tile *tile,
 				  struct xe_vm *vm, size_t size,
-				  enum ttm_bo_type type, u32 flags);
-struct xe_bo *xe_bo_create(struct xe_device *xe, struct xe_tile *tile,
-			   struct xe_vm *vm, size_t size,
-			   enum ttm_bo_type type, u32 flags);
-struct xe_bo *xe_bo_create_user(struct xe_device *xe, struct xe_tile *tile,
-				struct xe_vm *vm, size_t size,
-				u16 cpu_caching,
-				u32 flags);
+				  enum ttm_bo_type type, u32 flags,
+				  struct drm_exec *exec);
+struct xe_bo *xe_bo_create_user(struct xe_device *xe, struct xe_vm *vm, size_t size,
+				u16 cpu_caching, u32 flags, struct drm_exec *exec);
 struct xe_bo *xe_bo_create_pin_map(struct xe_device *xe, struct xe_tile *tile,
 				   struct xe_vm *vm, size_t size,
-				   enum ttm_bo_type type, u32 flags);
-struct xe_bo *xe_bo_create_pin_map_at(struct xe_device *xe, struct xe_tile *tile,
-				      struct xe_vm *vm, size_t size, u64 offset,
-				      enum ttm_bo_type type, u32 flags);
-struct xe_bo *xe_bo_create_pin_map_at_aligned(struct xe_device *xe,
-					      struct xe_tile *tile,
-					      struct xe_vm *vm,
-					      size_t size, u64 offset,
-					      enum ttm_bo_type type, u32 flags,
-					      u64 alignment);
+				   enum ttm_bo_type type, u32 flags,
+				   struct drm_exec *exec);
+struct xe_bo *xe_bo_create_pin_map_novm(struct xe_device *xe, struct xe_tile *tile,
+					size_t size, enum ttm_bo_type type, u32 flags,
+					bool intr);
+struct xe_bo *xe_bo_create_pin_range_novm(struct xe_device *xe, struct xe_tile *tile,
+					  size_t size, u64 start, u64 end,
+					  enum ttm_bo_type type, u32 flags);
+struct xe_bo *
+xe_bo_create_pin_map_at_novm(struct xe_device *xe, struct xe_tile *tile,
+			     size_t size, u64 offset, enum ttm_bo_type type,
+			     u32 flags, u64 alignment, bool intr);
 struct xe_bo *xe_managed_bo_create_pin_map(struct xe_device *xe, struct xe_tile *tile,
 					   size_t size, u32 flags);
+void xe_managed_bo_unpin_map_no_vm(struct xe_bo *bo);
 struct xe_bo *xe_managed_bo_create_from_data(struct xe_device *xe, struct xe_tile *tile,
 					     const void *data, size_t size, u32 flags);
 int xe_managed_bo_reinit_in_vram(struct xe_device *xe, struct xe_tile *tile, struct xe_bo **src);
···
 	}
 }
 
-int xe_bo_pin_external(struct xe_bo *bo, bool in_place);
-int xe_bo_pin(struct xe_bo *bo);
+int xe_bo_pin_external(struct xe_bo *bo, bool in_place, struct drm_exec *exec);
+int xe_bo_pin(struct xe_bo *bo, struct drm_exec *exec);
 void xe_bo_unpin_external(struct xe_bo *bo);
 void xe_bo_unpin(struct xe_bo *bo);
-int xe_bo_validate(struct xe_bo *bo, struct xe_vm *vm, bool allow_res_evict);
+int xe_bo_validate(struct xe_bo *bo, struct xe_vm *vm, bool allow_res_evict,
+		   struct drm_exec *exec);
 
 static inline bool xe_bo_is_pinned(struct xe_bo *bo)
 {
···
 
 bool xe_bo_can_migrate(struct xe_bo *bo, u32 mem_type);
 
-int xe_bo_migrate(struct xe_bo *bo, u32 mem_type);
-int xe_bo_evict(struct xe_bo *bo);
+int xe_bo_migrate(struct xe_bo *bo, u32 mem_type, struct ttm_operation_ctx *ctc,
+		  struct drm_exec *exec);
+int xe_bo_evict(struct xe_bo *bo, struct drm_exec *exec);
 
 int xe_bo_evict_pinned(struct xe_bo *bo);
 int xe_bo_notifier_prepare_pinned(struct xe_bo *bo);
···
 static inline size_t xe_bo_ccs_pages_start(struct xe_bo *bo)
 {
 	return PAGE_ALIGN(xe_bo_size(bo));
+}
+
+/**
+ * xe_bo_has_valid_ccs_bb - Check if CCS's BBs were setup for the BO.
+ * @bo: the &xe_bo to check
+ *
+ * The CCS's BBs should only be setup by the driver VF, but it is safe
+ * to call this function also by non-VF driver.
+ *
+ * Return: true iff the CCS's BBs are setup, false otherwise.
+ */
+static inline bool xe_bo_has_valid_ccs_bb(struct xe_bo *bo)
+{
+	return bo->bb_ccs[XE_SRIOV_VF_CCS_READ_CTX] &&
+	       bo->bb_ccs[XE_SRIOV_VF_CCS_WRITE_CTX];
 }
 
 static inline bool xe_bo_has_pages(struct xe_bo *bo)

drivers/gpu/drm/xe/xe_bo_types.h

···
 /* TODO: To be selected with VM_MADVISE */
 #define	XE_BO_PRIORITY_NORMAL	1
 
-/** @xe_bo: XE buffer object */
+/**
+ * struct xe_bo - Xe buffer object
+ */
 struct xe_bo {
 	/** @ttm: TTM base buffer object */
 	struct ttm_buffer_object ttm;
···
 	struct xe_ggtt_node *ggtt_node[XE_MAX_TILES_PER_DEVICE];
 	/** @vmap: iosys map of this buffer */
 	struct iosys_map vmap;
-	/** @ttm_kmap: TTM bo kmap object for internal use only. Keep off. */
+	/** @kmap: TTM bo kmap object for internal use only. Keep off. */
 	struct ttm_bo_kmap_obj kmap;
 	/** @pinned_link: link to present / evicted list of pinned BO */
 	struct list_head pinned_link;
···
 	/** @created: Whether the bo has passed initial creation */
 	bool created;
 
-	/** @ccs_cleared */
+	/** @ccs_cleared: true means that CCS region of BO is already cleared */
 	bool ccs_cleared;
 
-	/** @bb_ccs_rw: BB instructions of CCS read/write. Valid only for VF */
+	/** @bb_ccs: BB instructions of CCS read/write. Valid only for VF */
 	struct xe_bb *bb_ccs[XE_SRIOV_VF_CCS_CTX_COUNT];
 
 	/**
···
 	struct drm_pagemap_devmem devmem_allocation;
 
 	/** @vram_userfault_link: Link into @mem_access.vram_userfault.list */
-		struct list_head vram_userfault_link;
+	struct list_head vram_userfault_link;
 
-	/** @min_align: minimum alignment needed for this BO if different
+	/**
+	 * @min_align: minimum alignment needed for this BO if different
 	 * from default
 	 */
 	u64 min_align;

drivers/gpu/drm/xe/xe_configfs.c

···
  */
 
 #include <linux/bitops.h>
+#include <linux/ctype.h>
 #include <linux/configfs.h>
 #include <linux/cleanup.h>
 #include <linux/find.h>
···
 #include <linux/pci.h>
 #include <linux/string.h>
 
+#include "instructions/xe_mi_commands.h"
 #include "xe_configfs.h"
 #include "xe_hw_engine_types.h"
 #include "xe_module.h"
···
  * DOC: Xe Configfs
  *
  * Overview
- * =========
+ * ========
  *
  * Configfs is a filesystem-based manager of kernel objects. XE KMD registers a
  * configfs subsystem called ``xe`` that creates a directory in the mounted
···
  *
  * To create a device, the ``xe`` module should already be loaded, but some
  * attributes can only be set before binding the device. It can be accomplished
- * by blocking the driver autoprobe:
+ * by blocking the driver autoprobe::
  *
  *	# echo 0 > /sys/bus/pci/drivers_autoprobe
  *	# modprobe xe
···
  *
  * This attribute can only be set before binding to the device.
  *
+ * Context restore BB
+ * ------------------
+ *
+ * Allow to execute a batch buffer during any context switches. When the
+ * GPU is restoring the context, it executes additional commands. It's useful
+ * for testing additional workarounds and validating certain HW behaviors: it's
+ * not intended for normal execution and will taint the kernel with TAINT_TEST
+ * when used.
+ *
+ * Currently this is implemented only for post and mid context restore.
+ * Examples:
+ *
+ * #. Execute a LRI command to write 0xDEADBEEF to register 0x4f10 after the
+ *    normal context restore::
+ *
+ *	# echo 'rcs cmd 11000001 4F100 DEADBEEF' \
+ *		> /sys/kernel/config/xe/0000:03:00.0/ctx_restore_post_bb
+ *
+ * #. Execute a LRI command to write 0xDEADBEEF to register 0x4f10 at the
+ *    beginning of the context restore::
+ *
+ *	# echo 'rcs cmd 11000001 4F100 DEADBEEF' \
+ *		> /sys/kernel/config/xe/0000:03:00.0/ctx_restore_mid_bb
+
+ * #. Load certain values in a couple of registers (it can be used as a simpler
+ *    alternative to the `cmd`) action::
+ *
+ *	# cat > /sys/kernel/config/xe/0000:03:00.0/ctx_restore_post_bb <<EOF
+ *	rcs reg 4F100 DEADBEEF
+ *	rcs reg 4F104 FFFFFFFF
+ *	EOF
+ *
+ *    .. note::
+ *
+ *       When using multiple lines, make sure to use a command that is
+ *       implemented with a single write syscall, like HEREDOC.
+ *
+ * These attributes can only be set before binding to the device.
+ *
  * Remove devices
  * ==============
  *
···
  *	# rmdir /sys/kernel/config/xe/0000:03:00.0/
  */
 
+/* Similar to struct xe_bb, but not tied to HW (yet) */
+struct wa_bb {
+	u32 *cs;
+	u32 len; /* in dwords */
+};
+
 struct xe_config_group_device {
 	struct config_group group;
 
 	struct xe_config_device {
 		u64 engines_allowed;
+		struct wa_bb ctx_restore_post_bb[XE_ENGINE_CLASS_MAX];
+		struct wa_bb ctx_restore_mid_bb[XE_ENGINE_CLASS_MAX];
 		bool survivability_mode;
 		bool enable_psmi;
 	} config;
 
 	/* protects attributes */
 	struct mutex lock;
+	/* matching descriptor */
+	const struct xe_device_desc *desc;
 };
 
 static const struct xe_config_device device_defaults = {
···
 struct engine_info {
 	const char *cls;
 	u64 mask;
+	enum xe_engine_class engine_class;
 };
 
 /* Some helpful macros to aid on the sizing of buffer allocation when parsing */
···
 #define MAX_ENGINE_INSTANCE_CHARS 2
 
 static const struct engine_info engine_info[] = {
-	{ .cls = "rcs", .mask = XE_HW_ENGINE_RCS_MASK },
-	{ .cls = "bcs", .mask = XE_HW_ENGINE_BCS_MASK },
-	{ .cls = "vcs", .mask = XE_HW_ENGINE_VCS_MASK },
-	{ .cls = "vecs", .mask = XE_HW_ENGINE_VECS_MASK },
-	{ .cls = "ccs", .mask = XE_HW_ENGINE_CCS_MASK },
-	{ .cls = "gsccs", .mask = XE_HW_ENGINE_GSCCS_MASK },
+	{ .cls = "rcs", .mask = XE_HW_ENGINE_RCS_MASK, .engine_class = XE_ENGINE_CLASS_RENDER },
+	{ .cls = "bcs", .mask = XE_HW_ENGINE_BCS_MASK, .engine_class = XE_ENGINE_CLASS_COPY },
+	{ .cls = "vcs", .mask = XE_HW_ENGINE_VCS_MASK, .engine_class = XE_ENGINE_CLASS_VIDEO_DECODE },
+	{ .cls = "vecs", .mask = XE_HW_ENGINE_VECS_MASK, .engine_class = XE_ENGINE_CLASS_VIDEO_ENHANCE },
+	{ .cls = "ccs", .mask = XE_HW_ENGINE_CCS_MASK, .engine_class = XE_ENGINE_CLASS_COMPUTE },
+	{ .cls = "gsccs", .mask = XE_HW_ENGINE_GSCCS_MASK, .engine_class = XE_ENGINE_CLASS_OTHER },
 };
 
 static struct xe_config_group_device *to_xe_config_group_device(struct config_item *item)
···
 	return p - page;
 }
 
-static bool lookup_engine_mask(const char *pattern, u64 *mask)
+/*
+ * Lookup engine_info. If @mask is not NULL, reduce the mask according to the
+ * instance in @pattern.
+ *
+ * Examples of inputs:
+ * - lookup_engine_info("rcs0", &mask): return "rcs" entry from @engine_info and
+ *   mask == BIT_ULL(XE_HW_ENGINE_RCS0)
+ * - lookup_engine_info("rcs*", &mask): return "rcs" entry from @engine_info and
+ *   mask == XE_HW_ENGINE_RCS_MASK
+ * - lookup_engine_info("rcs", NULL): return "rcs" entry from @engine_info
+ */
+static const struct engine_info *lookup_engine_info(const char *pattern, u64 *mask)
 {
 	for (size_t i = 0; i < ARRAY_SIZE(engine_info); i++) {
 		u8 instance;
···
 			continue;
 
 		pattern += strlen(engine_info[i].cls);
+		if (!mask && !*pattern)
+			return &engine_info[i];
 
 		if (!strcmp(pattern, "*")) {
 			*mask = engine_info[i].mask;
-			return true;
+			return &engine_info[i];
 		}
 
 		if (kstrtou8(pattern, 10, &instance))
-			return false;
+			return NULL;
 
 		bit = __ffs64(engine_info[i].mask) + instance;
 		if (bit >= fls64(engine_info[i].mask))
-			return false;
+			return NULL;
 
 		*mask = BIT_ULL(bit);
-		return true;
+		return &engine_info[i];
 	}
 
-	return false;
+	return NULL;
+}
+
+static int parse_engine(const char *s, const char *end_chars, u64 *mask,
+			const struct engine_info **pinfo)
+{
+	char buf[MAX_ENGINE_CLASS_CHARS + MAX_ENGINE_INSTANCE_CHARS + 1];
+	const struct engine_info *info;
+	size_t len;
+
+	len = strcspn(s, end_chars);
+	if (len >= sizeof(buf))
+		return -EINVAL;
+
+	memcpy(buf, s, len);
+	buf[len] = '\0';
+
+	info = lookup_engine_info(buf, mask);
+	if (!info)
+		return -ENOENT;
+
+	if (pinfo)
+		*pinfo = info;
+
+	return len;
 }
 
 static ssize_t engines_allowed_store(struct config_item *item, const char *page,
 				     size_t len)
 {
 	struct xe_config_group_device *dev = to_xe_config_group_device(item);
-	size_t patternlen, p;
+	ssize_t patternlen, p;
 	u64 mask, val = 0;
 
 	for (p = 0; p < len; p += patternlen + 1) {
-		char buf[MAX_ENGINE_CLASS_CHARS + MAX_ENGINE_INSTANCE_CHARS + 1];
-
-		patternlen = strcspn(page + p, ",\n");
-		if (patternlen >= sizeof(buf))
-			return -EINVAL;
-
-		memcpy(buf, page + p, patternlen);
-		buf[patternlen] = '\0';
-
-		if (!lookup_engine_mask(buf, &mask))
+		patternlen = parse_engine(page + p, ",\n", &mask, NULL);
+		if (patternlen < 0)
 			return -EINVAL;
 
 		val |= mask;
···
 	return len;
 }
 
+static bool wa_bb_read_advance(bool dereference, char **p,
+			       const char *append, size_t len,
+			       size_t *max_size)
+{
+	if (dereference) {
+		if (len >= *max_size)
+			return false;
+		*max_size -= len;
+		if (append)
+			memcpy(*p, append, len);
+	}
+
+	*p += len;
+
+	return true;
+}
+
+static ssize_t wa_bb_show(struct xe_config_group_device *dev,
+			  struct wa_bb wa_bb[static XE_ENGINE_CLASS_MAX],
+			  char *data, size_t sz)
+{
+	char *p = data;
+
+	guard(mutex)(&dev->lock);
+
+	for (size_t i = 0; i < ARRAY_SIZE(engine_info); i++) {
+		enum xe_engine_class ec = engine_info[i].engine_class;
+		size_t len;
+
+		if (!wa_bb[ec].len)
+			continue;
+
+		len = snprintf(p, sz, "%s:", engine_info[i].cls);
+		if (!wa_bb_read_advance(data, &p, NULL, len, &sz))
+			return -ENOBUFS;
+
+		for (size_t j = 0; j < wa_bb[ec].len; j++) {
+			len = snprintf(p, sz, " %08x", wa_bb[ec].cs[j]);
+			if (!wa_bb_read_advance(data, &p, NULL, len, &sz))
+				return -ENOBUFS;
+		}
+
+		if (!wa_bb_read_advance(data, &p, "\n", 1, &sz))
+			return -ENOBUFS;
+	}
+
+	if (!wa_bb_read_advance(data, &p, "", 1, &sz))
+		return -ENOBUFS;
+
+	/* Reserve one more to match check for '\0' */
+	if (!data)
+		p++;
+
+	return p - data;
+}
+
+static ssize_t ctx_restore_mid_bb_show(struct config_item *item, char *page)
+{
+	struct xe_config_group_device *dev = to_xe_config_group_device(item);
+
+	return wa_bb_show(dev, dev->config.ctx_restore_mid_bb, page, SZ_4K);
+}
+
+static ssize_t ctx_restore_post_bb_show(struct config_item *item, char *page)
+{
+	struct xe_config_group_device *dev = to_xe_config_group_device(item);
+
+	return wa_bb_show(dev, dev->config.ctx_restore_post_bb, page, SZ_4K);
+}
+
+static void wa_bb_append(struct wa_bb *wa_bb, u32 val)
+{
+	if (wa_bb->cs)
+		wa_bb->cs[wa_bb->len] = val;
+
+	wa_bb->len++;
+}
+
+static ssize_t parse_hex(const char *line, u32 *pval)
+{
+	char numstr[12];
+	const char *p;
+	ssize_t numlen;
+
+	p = line + strspn(line, " \t");
+	if (!*p || *p == '\n')
+		return 0;
+
+	numlen = strcspn(p, " \t\n");
+	if (!numlen || numlen >= sizeof(numstr) - 1)
+		return -EINVAL;
+
+	memcpy(numstr, p, numlen);
+	numstr[numlen] = '\0';
+	p += numlen;
+
+	if (kstrtou32(numstr, 16, pval))
+		return -EINVAL;
+
+	return p - line;
+}
+
+/*
+ * Parse lines with the format
+ *
+ *	<engine-class> cmd <u32> <u32...>
+ *	<engine-class> reg <u32_addr> <u32_val>
+ *
+ * and optionally save them in @wa_bb[i].cs is non-NULL.
+ *
+ * Return the number of dwords parsed.
+ */
+static ssize_t parse_wa_bb_lines(const char *lines,
+				 struct wa_bb wa_bb[static XE_ENGINE_CLASS_MAX])
+{
+	ssize_t dwords = 0, ret;
+	const char *p;
+
+	for (p = lines; *p; p++) {
+		const struct engine_info *info = NULL;
+		u32 val, val2;
+
+		/* Also allow empty lines */
+		p += strspn(p, " \t\n");
+		if (!*p)
+			break;
+
+		ret = parse_engine(p, " \t\n", NULL, &info);
+		if (ret < 0)
+			return ret;
+
+		p += ret;
+		p += strspn(p, " \t");
+
+		if (str_has_prefix(p, "cmd")) {
+			for (p += strlen("cmd"); *p;) {
+				ret = parse_hex(p, &val);
+				if (ret < 0)
+					return -EINVAL;
+				if (!ret)
+					break;
+
+				p += ret;
+				dwords++;
+				wa_bb_append(&wa_bb[info->engine_class], val);
+			}
+		} else if (str_has_prefix(p, "reg")) {
+			p += strlen("reg");
+			ret = parse_hex(p, &val);
+			if (ret <= 0)
+				return -EINVAL;
+
+			p += ret;
+			ret = parse_hex(p, &val2);
+			if (ret <= 0)
+				return -EINVAL;
+
+			p += ret;
+			dwords += 3;
+			wa_bb_append(&wa_bb[info->engine_class],
+				     MI_LOAD_REGISTER_IMM | MI_LRI_NUM_REGS(1));
+			wa_bb_append(&wa_bb[info->engine_class], val);
+			wa_bb_append(&wa_bb[info->engine_class], val2);
+		} else {
+			return -EINVAL;
+		}
+	}
+
+	return dwords;
+}
+
+static ssize_t wa_bb_store(struct wa_bb wa_bb[static XE_ENGINE_CLASS_MAX],
+			   struct xe_config_group_device *dev,
+			   const char *page, size_t len)
+{
+	/* tmp_wa_bb must match wa_bb's size */
+	struct wa_bb tmp_wa_bb[XE_ENGINE_CLASS_MAX] = { };
+	ssize_t count, class;
+	u32 *tmp;
+
+	/* 1. Count dwords - wa_bb[i].cs is NULL for all classes */
+	count = parse_wa_bb_lines(page, tmp_wa_bb);
+	if (count < 0)
+		return count;
+
+	guard(mutex)(&dev->lock);
+
+	if (is_bound(dev))
+		return -EBUSY;
+
+	/*
+	 * 2. Allocate a u32 array and set the pointers to the right positions
+	 * according to the length of each class' wa_bb
+	 */
+	tmp = krealloc(wa_bb[0].cs, count * sizeof(u32), GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
+
+	if (!count) {
+		memset(wa_bb, 0, sizeof(tmp_wa_bb));
+		return len;
+	}
+
+	for (class = 0, count = 0; class < XE_ENGINE_CLASS_MAX; ++class) {
+		tmp_wa_bb[class].cs = tmp + count;
+		count += tmp_wa_bb[class].len;
+		tmp_wa_bb[class].len = 0;
+	}
+
+	/* 3. Parse wa_bb lines again, this time saving the values */
+	count = parse_wa_bb_lines(page, tmp_wa_bb);
+	if (count < 0)
+		return count;
+
+	memcpy(wa_bb, tmp_wa_bb, sizeof(tmp_wa_bb));
+
+	return len;
+}
+
+static ssize_t ctx_restore_mid_bb_store(struct config_item *item,
+					const char *data, size_t sz)
+{
+	struct xe_config_group_device *dev = to_xe_config_group_device(item);
+
+	return wa_bb_store(dev->config.ctx_restore_mid_bb, dev, data, sz);
+}
+
+static ssize_t ctx_restore_post_bb_store(struct config_item *item,
+					 const char *data, size_t sz)
+{
+	struct xe_config_group_device *dev = to_xe_config_group_device(item);
+
+	return wa_bb_store(dev->config.ctx_restore_post_bb, dev, data, sz);
+}
+
+CONFIGFS_ATTR(, ctx_restore_mid_bb);
+CONFIGFS_ATTR(, ctx_restore_post_bb);
 CONFIGFS_ATTR(, enable_psmi);
 CONFIGFS_ATTR(, engines_allowed);
 CONFIGFS_ATTR(, survivability_mode);
 
 static struct configfs_attribute *xe_config_device_attrs[] = {
+	&attr_ctx_restore_mid_bb,
+	&attr_ctx_restore_post_bb,
 	&attr_enable_psmi,
 	&attr_engines_allowed,
 	&attr_survivability_mode,
···
 	struct xe_config_group_device *dev = to_xe_config_group_device(item);
 
 	mutex_destroy(&dev->lock);
+
+	kfree(dev->config.ctx_restore_post_bb[0].cs);
 	kfree(dev);
 }
 
···
 	.release	= xe_config_device_release,
 };
 
+static bool xe_config_device_is_visible(struct config_item *item,
+					struct configfs_attribute *attr, int n)
+{
+	struct xe_config_group_device *dev = to_xe_config_group_device(item);
+
+	if (attr == &attr_survivability_mode) {
+		if (!dev->desc->is_dgfx || dev->desc->platform < XE_BATTLEMAGE)
+			return false;
+	}
+
+	return true;
+}
+
+static struct configfs_group_operations xe_config_device_group_ops = {
+	.is_visible	= xe_config_device_is_visible,
+};
+
 static const struct config_item_type xe_config_device_type = {
 	.ct_item_ops	= &xe_config_device_ops,
+	.ct_group_ops	= &xe_config_device_group_ops,
 	.ct_attrs	= xe_config_device_attrs,
 	.ct_owner	= THIS_MODULE,
 };
···
 	if (!dev)
 		return ERR_PTR(-ENOMEM);
 
+	dev->desc = match;
 	set_device_defaults(&dev->config);
 
 	config_group_init_type_name(&dev->group, name, &xe_config_device_type);
···
 	return &dev->group;
 }
 
-static struct configfs_group_operations xe_config_device_group_ops = {
+static struct configfs_group_operations xe_config_group_ops = {
 	.make_group	= xe_config_make_device_group,
 };
 
 static const struct config_item_type xe_configfs_type = {
-	.ct_group_ops	= &xe_config_device_group_ops,
+	.ct_group_ops	= &xe_config_group_ops,
 	.ct_owner	= THIS_MODULE,
 };
 
···
 }
 
 /**
- * xe_configfs_clear_survivability_mode - clear configfs survivability mode
- * @pdev: pci device
- */
-void xe_configfs_clear_survivability_mode(struct pci_dev *pdev)
-{
-	struct xe_config_group_device *dev = find_xe_config_group_device(pdev);
-
-	if (!dev)
-		return;
-
-	guard(mutex)(&dev->lock);
-	dev->config.survivability_mode = 0;
-
-	config_group_put(&dev->group);
-}
-
-/**
  * xe_configfs_get_engines_allowed - get engine allowed mask from configfs
  * @pdev: pci device
  *
···
 		return false;
 
 	ret = dev->config.enable_psmi;
-	config_item_put(&dev->group.cg_item);
+	config_group_put(&dev->group);
 
 	return ret;
+}
+
+/**
+ * xe_configfs_get_ctx_restore_mid_bb - get configfs ctx_restore_mid_bb setting
+ * @pdev: pci device
+ * @class: hw engine class
+ * @cs: pointer to the bb to use - only valid during probe
+ *
+ * Return: Number of dwords used in the mid_ctx_restore setting in configfs
+ */
+u32 xe_configfs_get_ctx_restore_mid_bb(struct pci_dev *pdev,
+				       enum xe_engine_class class,
+				       const u32 **cs)
+{
+	struct xe_config_group_device *dev = find_xe_config_group_device(pdev);
+	u32 len;
+
+	if (!dev)
+		return 0;
+
+	if (cs)
+		*cs = dev->config.ctx_restore_mid_bb[class].cs;
+
+	len = dev->config.ctx_restore_mid_bb[class].len;
+	config_group_put(&dev->group);
+
+	return len;
+}
+
+/**
+ * xe_configfs_get_ctx_restore_post_bb - get configfs ctx_restore_post_bb setting
+ * @pdev: pci device
+ * @class: hw engine class
+ * @cs: pointer to the bb to use - only valid during probe
+ *
+ * Return: Number of dwords used in the post_ctx_restore setting in configfs
+ */
+u32 xe_configfs_get_ctx_restore_post_bb(struct pci_dev *pdev,
+					enum xe_engine_class class,
+					const u32 **cs)
+{
+	struct xe_config_group_device *dev = find_xe_config_group_device(pdev);
+	u32 len;
+
+	if (!dev)
+		return 0;
+
+	*cs = dev->config.ctx_restore_post_bb[class].cs;
+	len = dev->config.ctx_restore_post_bb[class].len;
+	config_group_put(&dev->group);
+
+	return len;
 }
 
 int __init xe_configfs_init(void)
···
 	return 0;
 }
 
-void __exit xe_configfs_exit(void)
+void xe_configfs_exit(void)
 {
 	configfs_unregister_subsystem(&xe_configfs);
 	mutex_destroy(&xe_configfs.su_mutex);

drivers/gpu/drm/xe/xe_configfs.h

···
 #include <linux/limits.h>
 #include <linux/types.h>
 
+#include <xe_hw_engine_types.h>
+
 struct pci_dev;
 
 #if IS_ENABLED(CONFIG_CONFIGFS_FS)
···
 void xe_configfs_exit(void);
 void xe_configfs_check_device(struct pci_dev *pdev);
 bool xe_configfs_get_survivability_mode(struct pci_dev *pdev);
-void xe_configfs_clear_survivability_mode(struct pci_dev *pdev);
 u64 xe_configfs_get_engines_allowed(struct pci_dev *pdev);
 bool xe_configfs_get_psmi_enabled(struct pci_dev *pdev);
+u32 xe_configfs_get_ctx_restore_mid_bb(struct pci_dev *pdev, enum xe_engine_class,
+				       const u32 **cs);
+u32 xe_configfs_get_ctx_restore_post_bb(struct pci_dev *pdev, enum xe_engine_class,
+					const u32 **cs);
 #else
 static inline int xe_configfs_init(void) { return 0; }
 static inline void xe_configfs_exit(void) { }
 static inline void xe_configfs_check_device(struct pci_dev *pdev) { }
 static inline bool xe_configfs_get_survivability_mode(struct pci_dev *pdev) { return false; }
-static inline void xe_configfs_clear_survivability_mode(struct pci_dev *pdev) { }
 static inline u64 xe_configfs_get_engines_allowed(struct pci_dev *pdev) { return U64_MAX; }
 static inline bool xe_configfs_get_psmi_enabled(struct pci_dev *pdev) { return false; }
+static inline u32 xe_configfs_get_ctx_restore_mid_bb(struct pci_dev *pdev, enum xe_engine_class,
+						     const u32 **cs) { return 0; }
+static inline u32 xe_configfs_get_ctx_restore_post_bb(struct pci_dev *pdev, enum xe_engine_class,
+						      const u32 **cs) { return 0; }
 #endif
 
 #endif

drivers/gpu/drm/xe/xe_debugfs.c

···
 #include "xe_pxp_debugfs.h"
 #include "xe_sriov.h"
 #include "xe_sriov_pf.h"
+#include "xe_sriov_vf.h"
 #include "xe_step.h"
+#include "xe_tile_debugfs.h"
 #include "xe_wa.h"
 #include "xe_vsec.h"
 
···
 DECLARE_FAULT_ATTR(inject_csc_hw_error);
 
 static void read_residency_counter(struct xe_device *xe, struct xe_mmio *mmio,
-				   u32 offset, char *name, struct drm_printer *p)
+				   u32 offset, const char *name, struct drm_printer *p)
 {
 	u64 residency = 0;
 	int ret;
···
 	p = drm_seq_file_printer(m);
 	xe_pm_runtime_get(xe);
 	mmio = xe_root_tile_mmio(xe);
-	struct {
+	static const struct {
 		u32 offset;
-		char *name;
+		const char *name;
 	} residencies[] = {
 		{BMG_G2_RESIDENCY_OFFSET, "Package G2"},
 		{BMG_G6_RESIDENCY_OFFSET, "Package G6"},
···
 	xe_pm_runtime_get(xe);
 	mmio = xe_root_tile_mmio(xe);
 
-	struct {
+	static const struct {
 		u32 offset;
-		char *name;
+		const char *name;
 	} residencies[] = {
 		{BMG_PCIE_LINK_L0_RESIDENCY_OFFSET, "PCIE LINK L0 RESIDENCY"},
 		{BMG_PCIE_LINK_L1_RESIDENCY_OFFSET, "PCIE LINK L1 RESIDENCY"},
···
 	.write = atomic_svm_timeslice_ms_set,
 };
 
-static void create_tile_debugfs(struct xe_tile *tile, struct dentry *root)
+static ssize_t disable_late_binding_show(struct file *f, char __user *ubuf,
+					 size_t size, loff_t *pos)
 {
-	char name[8];
+	struct xe_device *xe = file_inode(f)->i_private;
+	struct xe_late_bind *late_bind = &xe->late_bind;
+	char buf[32];
+	int len;
 
-	snprintf(name, sizeof(name), "tile%u", tile->id);
-	tile->debugfs = debugfs_create_dir(name, root);
-	if (IS_ERR(tile->debugfs))
-		return;
+	len = scnprintf(buf, sizeof(buf), "%d\n", late_bind->disable);
 
-	/*
-	 * Store the xe_tile pointer as private data of the tile/ directory
-	 * node so other tile specific attributes under that directory may
-	 * refer to it by looking at its parent node private data.
-	 */
-	tile->debugfs->d_inode->i_private = tile;
+	return simple_read_from_buffer(ubuf, size, pos, buf, len);
 }
+
+static ssize_t disable_late_binding_set(struct file *f, const char __user *ubuf,
+					size_t size, loff_t *pos)
+{
+	struct xe_device *xe = file_inode(f)->i_private;
+	struct xe_late_bind *late_bind = &xe->late_bind;
+	u32 uval;
+	ssize_t ret;
+
+	ret = kstrtouint_from_user(ubuf, size, sizeof(uval), &uval);
+	if (ret)
+		return ret;
+
+	if (uval > 1)
+		return -EINVAL;
+
+	late_bind->disable = !!uval;
+	return size;
+}
+
+static const struct file_operations disable_late_binding_fops = {
+	.owner = THIS_MODULE,
+	.read = disable_late_binding_show,
+	.write = disable_late_binding_set,
+};
 
 void xe_debugfs_register(struct xe_device *xe)
 {
···
 				 ARRAY_SIZE(debugfs_list),
 				 root, minor);
 
-	if (xe->info.platform == XE_BATTLEMAGE) {
+	if (xe->info.platform == XE_BATTLEMAGE && !IS_SRIOV_VF(xe)) {
 		drm_debugfs_create_files(debugfs_residencies,
 					 ARRAY_SIZE(debugfs_residencies),
 					 root, minor);
···
 
 	debugfs_create_file("atomic_svm_timeslice_ms", 0600, root, xe,
 			    &atomic_svm_timeslice_ms_fops);
+
+	debugfs_create_file("disable_late_binding", 0600, root, xe,
+			    &disable_late_binding_fops);
 
 	for (mem_type = XE_PL_VRAM0; mem_type <= XE_PL_VRAM1; ++mem_type) {
 		man = ttm_manager_type(bdev, mem_type);
···
 		ttm_resource_manager_create_debugfs(man, root, "stolen_mm");
 
 	for_each_tile(tile, xe, tile_id)
-		create_tile_debugfs(tile, root);
+		xe_tile_debugfs_register(tile);
 
 	for_each_gt(gt, xe, id)
 		xe_gt_debugfs_register(gt);
···
 
 	if (IS_SRIOV_PF(xe))
 		xe_sriov_pf_debugfs_register(xe, root);
+	else if (IS_SRIOV_VF(xe))
+		xe_sriov_vf_debugfs_register(xe, root);
 }

drivers/gpu/drm/xe/xe_device.c

···
 #include "xe_hwmon.h"
 #include "xe_i2c.h"
 #include "xe_irq.h"
+#include "xe_late_bind_fw.h"
 #include "xe_mmio.h"
 #include "xe_module.h"
 #include "xe_nvm.h"
···
 	if (err)
 		goto err;
 
+	xe_validation_device_init(&xe->val);
+
 	init_waitqueue_head(&xe->ufence_wq);
 
 	init_rwsem(&xe->usm.lock);
···
  * re-init and saving/restoring (or re-populating) the wiped memory. Since we
  * perform the FLR as the very last action before releasing access to the HW
  * during the driver release flow, we don't attempt recovery at all, because
- * if/when a new instance of i915 is bound to the device it will do a full
+ * if/when a new instance of Xe is bound to the device it will do a full
  * re-init anyway.
  */
 static void __xe_driver_flr(struct xe_device *xe)
···
 	if (err)
 		return err;
 
+	err = xe_late_bind_init(&xe->late_bind);
+	if (err)
+		return err;
+
 	err = xe_oa_init(xe);
 	if (err)
 		return err;
···
 
 	xe_vsec_init(xe);
 
-	err = xe_sriov_late_init(xe);
+	err = xe_sriov_init_late(xe);
 	if (err)
 		goto err_unregister_display;
 

drivers/gpu/drm/xe/xe_device_sysfs.c

···
 
 static DEVICE_ATTR_RW(vram_d3cold_threshold);
 
+static struct attribute *vram_attrs[] = {
+	&dev_attr_vram_d3cold_threshold.attr,
+	NULL
+};
+
+static const struct attribute_group vram_attr_group = {
+	.attrs = vram_attrs,
+};
+
 static ssize_t
 lb_fan_control_version_show(struct device *dev, struct device_attribute *attr, char *buf)
 {
···
 }
 static DEVICE_ATTR_ADMIN_RO(lb_voltage_regulator_version);
 
-static int late_bind_create_files(struct device *dev)
+static struct attribute *late_bind_attrs[] = {
+	&dev_attr_lb_fan_control_version.attr,
+	&dev_attr_lb_voltage_regulator_version.attr,
+	NULL
+};
+
+static umode_t late_bind_attr_is_visible(struct kobject *kobj,
+					 struct attribute *attr, int n)
 {
+	struct device *dev = kobj_to_dev(kobj);
 	struct xe_device *xe = pdev_to_xe_device(to_pci_dev(dev));
 	struct xe_tile *root = xe_device_get_root_tile(xe);
 	u32 cap = 0;
···
 
 	ret = xe_pcode_read(root, PCODE_MBOX(PCODE_LATE_BINDING, GET_CAPABILITY_STATUS, 0),
 			    &cap, NULL);
-	if (ret) {
-		if (ret == -ENXIO) {
-			drm_dbg(&xe->drm, "Late binding not supported by firmware\n");
-			ret = 0;
-		}
-		goto out;
-	}
-
-	if (REG_FIELD_GET(V1_FAN_SUPPORTED, cap)) {
-		ret = sysfs_create_file(&dev->kobj, &dev_attr_lb_fan_control_version.attr);
-		if (ret)
-			goto out;
-	}
-
-	if (REG_FIELD_GET(VR_PARAMS_SUPPORTED, cap))
-		ret = sysfs_create_file(&dev->kobj, &dev_attr_lb_voltage_regulator_version.attr);
-out:
 	xe_pm_runtime_put(xe);
-
-	return ret;
-}
-
-static void late_bind_remove_files(struct device *dev)
-{
-	struct xe_device *xe = pdev_to_xe_device(to_pci_dev(dev));
-	struct xe_tile *root = xe_device_get_root_tile(xe);
-	u32 cap = 0;
-	int ret;
-
-	xe_pm_runtime_get(xe);
-
-	ret = xe_pcode_read(root, PCODE_MBOX(PCODE_LATE_BINDING, GET_CAPABILITY_STATUS, 0),
-			    &cap, NULL);
 	if (ret)
-		goto out;
+		return 0;
 
-	if (REG_FIELD_GET(V1_FAN_SUPPORTED, cap))
-		sysfs_remove_file(&dev->kobj, &dev_attr_lb_fan_control_version.attr);
+	if (attr == &dev_attr_lb_fan_control_version.attr &&
+	    REG_FIELD_GET(V1_FAN_SUPPORTED, cap))
+		return attr->mode;
+	if (attr == &dev_attr_lb_voltage_regulator_version.attr &&
+	    REG_FIELD_GET(VR_PARAMS_SUPPORTED, cap))
+		return attr->mode;
 
-	if (REG_FIELD_GET(VR_PARAMS_SUPPORTED, cap))
-		sysfs_remove_file(&dev->kobj, &dev_attr_lb_voltage_regulator_version.attr);
-out:
-	xe_pm_runtime_put(xe);
+	return 0;
 }
+
+static const struct attribute_group late_bind_attr_group = {
+	.attrs = late_bind_attrs,
+	.is_visible = late_bind_attr_is_visible,
+};
 
 /**
  * DOC: PCIe Gen5 Limitations
···
 }
 static DEVICE_ATTR_ADMIN_RO(auto_link_downgrade_status);
 
-static const struct attribute *auto_link_downgrade_attrs[] = {
+static struct attribute *auto_link_downgrade_attrs[] = {
 	&dev_attr_auto_link_downgrade_capable.attr,
 	&dev_attr_auto_link_downgrade_status.attr,
 	NULL
 };
 
-static void xe_device_sysfs_fini(void *arg)
-{
-	struct xe_device *xe = arg;
-
-	if (xe->d3cold.capable)
-		sysfs_remove_file(&xe->drm.dev->kobj, &dev_attr_vram_d3cold_threshold.attr);
-
-	if (xe->info.platform == XE_BATTLEMAGE) {
-		sysfs_remove_files(&xe->drm.dev->kobj, auto_link_downgrade_attrs);
-		late_bind_remove_files(xe->drm.dev);
-	}
-}
+static const struct attribute_group auto_link_downgrade_attr_group = {
+	.attrs = auto_link_downgrade_attrs,
+};
 
 int xe_device_sysfs_init(struct xe_device *xe)
 {
···
 	int ret;
 
 	if (xe->d3cold.capable) {
-		ret = sysfs_create_file(&dev->kobj, &dev_attr_vram_d3cold_threshold.attr);
+		ret = devm_device_add_group(dev, &vram_attr_group);
 		if (ret)
 			return ret;
 	}
 
-	if (xe->info.platform == XE_BATTLEMAGE) {
-		ret = sysfs_create_files(&dev->kobj, auto_link_downgrade_attrs);
+	if (xe->info.platform == XE_BATTLEMAGE && !IS_SRIOV_VF(xe)) {
+		ret = devm_device_add_group(dev, &auto_link_downgrade_attr_group);
 		if (ret)
 			return ret;
 
-		ret = late_bind_create_files(dev);
+		ret = devm_device_add_group(dev, &late_bind_attr_group);
 		if (ret)
 			return ret;
 	}
 
-	return devm_add_action_or_reset(dev, xe_device_sysfs_fini, xe);
+	return 0;
 }

drivers/gpu/drm/xe/xe_device_types.h

···
 
 #include "xe_devcoredump_types.h"
 #include "xe_heci_gsc.h"
+#include "xe_late_bind_fw_types.h"
 #include "xe_lmtt_types.h"
 #include "xe_memirq_types.h"
 #include "xe_oa_types.h"
···
 #include "xe_sriov_vf_ccs_types.h"
 #include "xe_step_types.h"
 #include "xe_survivability_mode_types.h"
+#include "xe_validation.h"
 
 #if IS_ENABLED(CONFIG_DRM_XE_DEBUG)
 #define TEST_VM_OPS_ERROR
···
 		struct {
 			/** @sriov.vf.ggtt_balloon: GGTT regions excluded from use. */
 			struct xe_ggtt_node *ggtt_balloon[2];
-
-			/** @sriov.vf.ccs: CCS read and write contexts for VF. */
-			struct xe_tile_vf_ccs ccs[XE_SRIOV_VF_CCS_CTX_COUNT];
 		} vf;
 	} sriov;
 
···
 		u8 has_heci_cscfi:1;
 		/** @info.has_heci_gscfi: device has heci gscfi */
 		u8 has_heci_gscfi:1;
+		/** @info.has_late_bind: Device has firmware late binding support */
+		u8 has_late_bind:1;
 		/** @info.has_llc: Device has a shared CPU+GPU last level cache */
 		u8 has_llc:1;
 		/** @info.has_mbx_power_limits: Device has support to manage power limits using
···
 	/** @nvm: discrete graphics non-volatile memory */
 	struct intel_dg_nvm_dev *nvm;
 
+	/** @late_bind: xe mei late bind interface */
+	struct xe_late_bind late_bind;
+
 	/** @oa: oa observation subsystem */
 	struct xe_oa oa;
 
···
 	 */
 	atomic64_t global_total_pages;
 #endif
+	/** @val: The domain for exhaustive eviction, which is currently per device. */
+	struct xe_validation_device val;
 
 	/** @psmi: GPU debugging via additional validation HW */
 	struct {
···
 		/** @psmi.region_mask: Mask of valid memory regions */
 		u8 region_mask;
 	} psmi;
+
+#if IS_ENABLED(CONFIG_DRM_XE_KUNIT_TEST)
+	/** @g2g_test_array: for testing G2G communications */
+	u32 *g2g_test_array;
+	/** @g2g_test_count: for testing G2G communications */
+	atomic_t g2g_test_count;
+#endif
 
 	/* private: */
 

drivers/gpu/drm/xe/xe_dma_buf.c

···
 	struct drm_gem_object *obj = attach->dmabuf->priv;
 	struct xe_bo *bo = gem_to_xe_bo(obj);
 	struct xe_device *xe = xe_bo_device(bo);
+	struct drm_exec *exec = XE_VALIDATION_UNSUPPORTED;
 	int ret;
 
 	/*
···
 		return -EINVAL;
 	}
 
-	ret = xe_bo_migrate(bo, XE_PL_TT);
+	ret = xe_bo_migrate(bo, XE_PL_TT, NULL, exec);
 	if (ret) {
 		if (ret != -EINTR && ret != -ERESTARTSYS)
 			drm_dbg(&xe->drm,
···
 		return ret;
 	}
 
-	ret = xe_bo_pin_external(bo, true);
+	ret = xe_bo_pin_external(bo, true, exec);
 	xe_assert(xe, !ret);
 
 	return 0;
···
 	struct dma_buf *dma_buf = attach->dmabuf;
 	struct drm_gem_object *obj = dma_buf->priv;
 	struct xe_bo *bo = gem_to_xe_bo(obj);
+	struct drm_exec *exec = XE_VALIDATION_UNSUPPORTED;
 	struct sg_table *sgt;
 	int r = 0;
 
···
 
 	if (!xe_bo_is_pinned(bo)) {
 		if (!attach->peer2peer)
-			r = xe_bo_migrate(bo, XE_PL_TT);
+			r = xe_bo_migrate(bo, XE_PL_TT, NULL, exec);
 		else
-			r = xe_bo_validate(bo, NULL, false);
+			r = xe_bo_validate(bo, NULL, false, exec);
 		if (r)
 			return ERR_PTR(r);
 	}
···
 	struct xe_bo *bo = gem_to_xe_bo(obj);
 	bool reads =  (direction == DMA_BIDIRECTIONAL ||
 		       direction == DMA_FROM_DEVICE);
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
+	int ret = 0;
 
 	if (!reads)
 		return 0;
 
 	/* Can we do interruptible lock here? */
-	xe_bo_lock(bo, false);
-	(void)xe_bo_migrate(bo, XE_PL_TT);
-	xe_bo_unlock(bo);
+	xe_validation_guard(&ctx, &xe_bo_device(bo)->val, &exec, (struct xe_val_flags) {}, ret) {
+		ret = drm_exec_lock_obj(&exec, &bo->ttm.base);
+		drm_exec_retry_on_contention(&exec);
+		if (ret)
+			break;
 
+		ret = xe_bo_migrate(bo, XE_PL_TT, NULL, &exec);
+		drm_exec_retry_on_contention(&exec);
+		xe_validation_retry_on_oom(&ctx, &ret);
+	}
+
+	/* If we failed, cpu-access takes place in current placement. */
 	return 0;
 }
 
···
 {
 	struct dma_resv *resv = dma_buf->resv;
 	struct xe_device *xe = to_xe_device(dev);
+	struct xe_validation_ctx ctx;
+	struct drm_gem_object *dummy_obj;
+	struct drm_exec exec;
 	struct xe_bo *bo;
-	int ret;
+	int ret = 0;
 
-	dma_resv_lock(resv, NULL);
-	bo = ___xe_bo_create_locked(xe, storage, NULL, resv, NULL, dma_buf->size,
-				    0, /* Will require 1way or 2way for vm_bind */
-				    ttm_bo_type_sg, XE_BO_FLAG_SYSTEM);
-	if (IS_ERR(bo)) {
-		ret = PTR_ERR(bo);
-		goto error;
+	dummy_obj = drm_gpuvm_resv_object_alloc(&xe->drm);
+	if (!dummy_obj)
+		return ERR_PTR(-ENOMEM);
+
+	dummy_obj->resv = resv;
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {}, ret) {
+		ret = drm_exec_lock_obj(&exec, dummy_obj);
+		drm_exec_retry_on_contention(&exec);
+		if (ret)
+			break;
+
+		bo = xe_bo_init_locked(xe, storage, NULL, resv, NULL, dma_buf->size,
+				       0, /* Will require 1way or 2way for vm_bind */
+				       ttm_bo_type_sg, XE_BO_FLAG_SYSTEM, &exec);
+		drm_exec_retry_on_contention(&exec);
+		if (IS_ERR(bo)) {
+			ret = PTR_ERR(bo);
+			xe_validation_retry_on_oom(&ctx, &ret);
+			break;
+		}
 	}
-	dma_resv_unlock(resv);
+	drm_gem_object_put(dummy_obj);
 
-	return &bo->ttm.base;
-
-error:
-	dma_resv_unlock(resv);
-	return ERR_PTR(ret);
+	return ret ? ERR_PTR(ret) : &bo->ttm.base;
 }
 
 static void xe_dma_buf_move_notify(struct dma_buf_attachment *attach)
 {
 	struct drm_gem_object *obj = attach->importer_priv;
 	struct xe_bo *bo = gem_to_xe_bo(obj);
+	struct drm_exec *exec = XE_VALIDATION_UNSUPPORTED;
 
-	XE_WARN_ON(xe_bo_evict(bo));
+	XE_WARN_ON(xe_bo_evict(bo, exec));
 }
 
 static const struct dma_buf_attach_ops xe_dma_buf_attach_ops = {

drivers/gpu/drm/xe/xe_eu_stall.c

···
 
 	size = stream->per_xecore_buf_size * last_xecore;
 
-	bo = xe_bo_create_pin_map_at_aligned(tile->xe, tile, NULL,
-					     size, ~0ull, ttm_bo_type_kernel,
-					     XE_BO_FLAG_SYSTEM | XE_BO_FLAG_GGTT, SZ_64);
+	bo = xe_bo_create_pin_map_at_novm(tile->xe, tile, size, ~0ull, ttm_bo_type_kernel,
+					  XE_BO_FLAG_SYSTEM | XE_BO_FLAG_GGTT, SZ_64, false);
 	if (IS_ERR(bo)) {
 		kfree(stream->xecore_buf);
 		return PTR_ERR(bo);

drivers/gpu/drm/xe/xe_exec.c

···
 #include "xe_ring_ops_types.h"
 #include "xe_sched_job.h"
 #include "xe_sync.h"
+#include "xe_svm.h"
 #include "xe_vm.h"
 
 /**
···
 static int xe_exec_fn(struct drm_gpuvm_exec *vm_exec)
 {
 	struct xe_vm *vm = container_of(vm_exec->vm, struct xe_vm, gpuvm);
+	int ret;
 
 	/* The fence slot added here is intended for the exec sched job. */
-	return xe_vm_validate_rebind(vm, &vm_exec->exec, 1);
+	xe_vm_set_validation_exec(vm, &vm_exec->exec);
+	ret = xe_vm_validate_rebind(vm, &vm_exec->exec, 1);
+	xe_vm_set_validation_exec(vm, NULL);
+	return ret;
 }
 
 int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
···
 	struct drm_gpuvm_exec vm_exec = {.extra.fn = xe_exec_fn};
 	struct drm_exec *exec = &vm_exec.exec;
 	u32 i, num_syncs, num_ufence = 0;
+	struct xe_validation_ctx ctx;
 	struct xe_sched_job *job;
 	struct xe_vm *vm;
 	bool write_locked, skip_retry = false;
-	ktime_t end = 0;
 	int err = 0;
 	struct xe_hw_engine_group *group;
 	enum xe_hw_engine_group_execution_mode mode, previous_mode;
···
 	if (err)
 		goto err_unlock_list;
 
-	vm_exec.vm = &vm->gpuvm;
-	vm_exec.flags = DRM_EXEC_INTERRUPTIBLE_WAIT;
-	if (xe_vm_in_lr_mode(vm)) {
-		drm_exec_init(exec, vm_exec.flags, 0);
-	} else {
-		err = drm_gpuvm_exec_lock(&vm_exec);
-		if (err) {
-			if (xe_vm_validate_should_retry(exec, err, &end))
-				err = -EAGAIN;
+	if (!xe_vm_in_lr_mode(vm)) {
+		vm_exec.vm = &vm->gpuvm;
+		vm_exec.flags = DRM_EXEC_INTERRUPTIBLE_WAIT;
+		err = xe_validation_exec_lock(&ctx, &vm_exec, &xe->val);
+		if (err)
 			goto err_unlock_list;
-		}
 	}
 
 	if (xe_vm_is_closed_or_banned(q->vm)) {
···
 		if (err)
 			goto err_put_job;
 
-		err = down_read_interruptible(&vm->userptr.notifier_lock);
+		err = xe_svm_notifier_lock_interruptible(vm);
 		if (err)
 			goto err_put_job;
 
···
 
 err_repin:
 	if (!xe_vm_in_lr_mode(vm))
-		up_read(&vm->userptr.notifier_lock);
+		xe_svm_notifier_unlock(vm);
 err_put_job:
 	if (err)
 		xe_sched_job_put(job);
 err_exec:
-	drm_exec_fini(exec);
+	if (!xe_vm_in_lr_mode(vm))
+		xe_validation_ctx_fini(&ctx);
 err_unlock_list:
 	up_read(&vm->lock);
 	if (err == -EAGAIN && !skip_retry)

drivers/gpu/drm/xe/xe_exec_queue.c

···
 	return err;
 }
 
+static void __xe_exec_queue_fini(struct xe_exec_queue *q)
+{
+	int i;
+
+	q->ops->fini(q);
+
+	for (i = 0; i < q->width; ++i)
+		xe_lrc_put(q->lrc[i]);
+}
+
 struct xe_exec_queue *xe_exec_queue_create(struct xe_device *xe, struct xe_vm *vm,
 					   u32 logical_mask, u16 width,
 					   struct xe_hw_engine *hwe, u32 flags,
···
 	if (xe_exec_queue_uses_pxp(q)) {
 		err = xe_pxp_exec_queue_add(xe->pxp, q);
 		if (err)
-			goto err_post_alloc;
+			goto err_post_init;
 	}
 
 	return q;
 
+err_post_init:
+	__xe_exec_queue_fini(q);
 err_post_alloc:
 	__xe_exec_queue_free(q);
 	return ERR_PTR(err);
···
 			xe_exec_queue_put(eq);
 	}
 
-	q->ops->fini(q);
+	q->ops->destroy(q);
 }
 
 void xe_exec_queue_fini(struct xe_exec_queue *q)
 {
-	int i;
-
 	/*
 	 * Before releasing our ref to lrc and xef, accumulate our run ticks
 	 * and wakeup any waiters.
···
 	if (q->xef && atomic_dec_and_test(&q->xef->exec_queue.pending_removal))
 		wake_up_var(&q->xef->exec_queue.pending_removal);
 
-	for (i = 0; i < q->width; ++i)
-		xe_lrc_put(q->lrc[i]);
-
+	__xe_exec_queue_fini(q);
 	__xe_exec_queue_free(q);
 }
 

drivers/gpu/drm/xe/xe_exec_queue_types.h

···
 	int (*init)(struct xe_exec_queue *q);
 	/** @kill: Kill inflight submissions for backend */
 	void (*kill)(struct xe_exec_queue *q);
-	/** @fini: Fini exec queue for submission backend */
+	/** @fini: Undoes the init() for submission backend */
 	void (*fini)(struct xe_exec_queue *q);
+	/**
+	 * @destroy: Destroy exec queue for submission backend. The backend
+	 * function must call xe_exec_queue_fini() (which will in turn call the
+	 * fini() backend function) to ensure the queue is properly cleaned up.
+	 */
+	void (*destroy)(struct xe_exec_queue *q);
 	/** @set_priority: Set priority for exec queue */
 	int (*set_priority)(struct xe_exec_queue *q,
 			    enum xe_exec_queue_priority priority);

drivers/gpu/drm/xe/xe_execlist.c

···
 	return err;
 }
 
-static void execlist_exec_queue_fini_async(struct work_struct *w)
+static void execlist_exec_queue_fini(struct xe_exec_queue *q)
+{
+	struct xe_execlist_exec_queue *exl = q->execlist;
+
+	drm_sched_entity_fini(&exl->entity);
+	drm_sched_fini(&exl->sched);
+
+	kfree(exl);
+}
+
+static void execlist_exec_queue_destroy_async(struct work_struct *w)
 {
 	struct xe_execlist_exec_queue *ee =
-		container_of(w, struct xe_execlist_exec_queue, fini_async);
+		container_of(w, struct xe_execlist_exec_queue, destroy_async);
 	struct xe_exec_queue *q = ee->q;
 	struct xe_execlist_exec_queue *exl = q->execlist;
 	struct xe_device *xe = gt_to_xe(q->gt);
···
 		list_del(&exl->active_link);
 	spin_unlock_irqrestore(&exl->port->lock, flags);
 
-	drm_sched_entity_fini(&exl->entity);
-	drm_sched_fini(&exl->sched);
-	kfree(exl);
-
 	xe_exec_queue_fini(q);
 }
 
···
 	/* NIY */
 }
 
-static void execlist_exec_queue_fini(struct xe_exec_queue *q)
+static void execlist_exec_queue_destroy(struct xe_exec_queue *q)
 {
-	INIT_WORK(&q->execlist->fini_async, execlist_exec_queue_fini_async);
-	queue_work(system_unbound_wq, &q->execlist->fini_async);
+	INIT_WORK(&q->execlist->destroy_async, execlist_exec_queue_destroy_async);
+	queue_work(system_unbound_wq, &q->execlist->destroy_async);
 }
 
 static int execlist_exec_queue_set_priority(struct xe_exec_queue *q,
···
 	.init = execlist_exec_queue_init,
 	.kill = execlist_exec_queue_kill,
 	.fini = execlist_exec_queue_fini,
+	.destroy = execlist_exec_queue_destroy,
 	.set_priority = execlist_exec_queue_set_priority,
 	.set_timeslice = execlist_exec_queue_set_timeslice,
 	.set_preempt_timeout = execlist_exec_queue_set_preempt_timeout,

drivers/gpu/drm/xe/xe_execlist_types.h

···
 
 	bool has_run;
 
-	struct work_struct fini_async;
+	struct work_struct destroy_async;
 
 	enum xe_exec_queue_priority active_priority;
 	struct list_head active_link;

drivers/gpu/drm/xe/xe_ggtt.c

···
 #include "xe_pm.h"
 #include "xe_res_cursor.h"
 #include "xe_sriov.h"
+#include "xe_tile_printk.h"
 #include "xe_tile_sriov_vf.h"
 #include "xe_tlb_inval.h"
 #include "xe_wa.h"
···
 		gsm_size = probe_gsm_size(pdev);
 
 	if (gsm_size == 0) {
-		drm_err(&xe->drm, "Hardware reported no preallocated GSM\n");
+		xe_tile_err(ggtt->tile, "Hardware reported no preallocated GSM\n");
 		return -ENOMEM;
 	}
 
···
 
 	if (IS_ENABLED(CONFIG_DRM_XE_DEBUG)) {
 		string_get_size(node->size, 1, STRING_UNITS_2, buf, sizeof(buf));
-		xe_gt_dbg(ggtt->tile->primary_gt, "GGTT %#llx-%#llx (%s) %s\n",
-			  node->start, node->start + node->size, buf, description);
+		xe_tile_dbg(ggtt->tile, "GGTT %#llx-%#llx (%s) %s\n",
+			    node->start, node->start + node->size, buf, description);
 	}
 }
 
···
 
 	err = drm_mm_reserve_node(&ggtt->mm, &node->base);
 
-	if (xe_gt_WARN(ggtt->tile->primary_gt, err,
-		       "Failed to balloon GGTT %#llx-%#llx (%pe)\n",
-		       node->base.start, node->base.start + node->base.size, ERR_PTR(err)))
+	if (xe_tile_WARN(ggtt->tile, err, "Failed to balloon GGTT %#llx-%#llx (%pe)\n",
+			 node->base.start, node->base.start + node->base.size, ERR_PTR(err)))
 		return err;
 
 	xe_ggtt_dump_node(ggtt, &node->base, "balloon");
···
 }
 
 static int __xe_ggtt_insert_bo_at(struct xe_ggtt *ggtt, struct xe_bo *bo,
-				  u64 start, u64 end)
+				  u64 start, u64 end, struct drm_exec *exec)
 {
 	u64 alignment = bo->min_align > 0 ? bo->min_align : XE_PAGE_SIZE;
 	u8 tile_id = ggtt->tile->id;
···
 		return 0;
 	}
 
-	err = xe_bo_validate(bo, NULL, false);
+	err = xe_bo_validate(bo, NULL, false, exec);
 	if (err)
 		return err;
 
···
  * @bo: the &xe_bo to be inserted
  * @start: address where it will be inserted
  * @end: end of the range where it will be inserted
+ * @exec: The drm_exec transaction to use for exhaustive eviction.
  *
  * Return: 0 on success or a negative error code on failure.
  */
 int xe_ggtt_insert_bo_at(struct xe_ggtt *ggtt, struct xe_bo *bo,
-			 u64 start, u64 end)
+			 u64 start, u64 end, struct drm_exec *exec)
 {
-	return __xe_ggtt_insert_bo_at(ggtt, bo, start, end);
+	return __xe_ggtt_insert_bo_at(ggtt, bo, start, end, exec);
 }
 
 /**
  * xe_ggtt_insert_bo - Insert BO into GGTT
  * @ggtt: the &xe_ggtt where bo will be inserted
  * @bo: the &xe_bo to be inserted
+ * @exec: The drm_exec transaction to use for exhaustive eviction.
  *
  * Return: 0 on success or a negative error code on failure.
  */
-int xe_ggtt_insert_bo(struct xe_ggtt *ggtt, struct xe_bo *bo)
+int xe_ggtt_insert_bo(struct xe_ggtt *ggtt, struct xe_bo *bo,
+		      struct drm_exec *exec)
 {
-	return __xe_ggtt_insert_bo_at(ggtt, bo, 0, U64_MAX);
+	return __xe_ggtt_insert_bo_at(ggtt, bo, 0, U64_MAX, exec);
 }
 
 /**

drivers/gpu/drm/xe/xe_ggtt.h

···
 
 struct drm_printer;
 struct xe_tile;
+struct drm_exec;
 
 struct xe_ggtt *xe_ggtt_alloc(struct xe_tile *tile);
 int xe_ggtt_init_early(struct xe_ggtt *ggtt);
···
 void xe_ggtt_map_bo(struct xe_ggtt *ggtt, struct xe_ggtt_node *node,
 		    struct xe_bo *bo, u16 pat_index);
 void xe_ggtt_map_bo_unlocked(struct xe_ggtt *ggtt, struct xe_bo *bo);
-int xe_ggtt_insert_bo(struct xe_ggtt *ggtt, struct xe_bo *bo);
+int xe_ggtt_insert_bo(struct xe_ggtt *ggtt, struct xe_bo *bo, struct drm_exec *exec);
 int xe_ggtt_insert_bo_at(struct xe_ggtt *ggtt, struct xe_bo *bo,
-			 u64 start, u64 end);
+			 u64 start, u64 end, struct drm_exec *exec);
 void xe_ggtt_remove_bo(struct xe_ggtt *ggtt, struct xe_bo *bo);
 u64 xe_ggtt_largest_hole(struct xe_ggtt *ggtt, u64 alignment, u64 *spare);
 

drivers/gpu/drm/xe/xe_gsc.c

···
 	u64 ggtt_offset;
 	int err;
 
-	bo = xe_bo_create_pin_map(xe, tile, NULL, GSC_VER_PKT_SZ * 2,
-				  ttm_bo_type_kernel,
-				  XE_BO_FLAG_SYSTEM |
-				  XE_BO_FLAG_GGTT);
+	bo = xe_bo_create_pin_map_novm(xe, tile, GSC_VER_PKT_SZ * 2,
+				       ttm_bo_type_kernel,
+				       XE_BO_FLAG_SYSTEM |
+				       XE_BO_FLAG_GGTT, false);
 	if (IS_ERR(bo)) {
 		xe_gt_err(gt, "failed to allocate bo for GSC version query\n");
 		return PTR_ERR(bo);

drivers/gpu/drm/xe/xe_gt.c

···
 	 * FIXME: if xe_uc_sanitize is called here, on TGL driver will not
 	 * reload
 	 */
-	gt->uc.guc.submission_state.enabled = false;
+	xe_guc_submit_disable(&gt->uc.guc);
 }
 
 static void xe_gt_enable_host_l2_vram(struct xe_gt *gt)

drivers/gpu/drm/xe/xe_gt_debugfs.c

···
 #include "xe_reg_whitelist.h"
 #include "xe_sa.h"
 #include "xe_sriov.h"
+#include "xe_sriov_vf_ccs.h"
 #include "xe_tuning.h"
 #include "xe_uc_debugfs.h"
 #include "xe_wa.h"
···
 	xe_pm_runtime_put(gt_to_xe(gt));
 
 	return ret;
-}
-
-static int sa_info(struct xe_gt *gt, struct drm_printer *p)
-{
-	struct xe_tile *tile = gt_to_tile(gt);
-
-	xe_pm_runtime_get(gt_to_xe(gt));
-	drm_suballoc_dump_debug_info(&tile->mem.kernel_bb_pool->base, p,
-				     xe_sa_manager_gpu_addr(tile->mem.kernel_bb_pool));
-	xe_pm_runtime_put(gt_to_xe(gt));
-
-	return 0;
-}
-
-static int sa_info_vf_ccs(struct xe_gt *gt, struct drm_printer *p)
-{
-	struct xe_tile *tile = gt_to_tile(gt);
-	struct xe_sa_manager *bb_pool;
-	enum xe_sriov_vf_ccs_rw_ctxs ctx_id;
-
-	if (!IS_VF_CCS_READY(gt_to_xe(gt)))
-		return 0;
-
-	xe_pm_runtime_get(gt_to_xe(gt));
-
-	for_each_ccs_rw_ctx(ctx_id) {
-		bb_pool = tile->sriov.vf.ccs[ctx_id].mem.ccs_bb_pool;
-		if (!bb_pool)
-			break;
-
-		drm_printf(p, "ccs %s bb suballoc info\n", ctx_id ? "write" : "read");
-		drm_printf(p, "-------------------------\n");
-		drm_suballoc_dump_debug_info(&bb_pool->base, p, xe_sa_manager_gpu_addr(bb_pool));
-		drm_puts(p, "\n");
-	}
-
-	xe_pm_runtime_put(gt_to_xe(gt));
-
-	return 0;
 }
 
 static int topology(struct xe_gt *gt, struct drm_printer *p)
···
  * - without access to the PF specific data
  */
 static const struct drm_info_list vf_safe_debugfs_list[] = {
-	{"sa_info", .show = xe_gt_debugfs_simple_show, .data = sa_info},
 	{"topology", .show = xe_gt_debugfs_simple_show, .data = topology},
 	{"ggtt", .show = xe_gt_debugfs_simple_show, .data = ggtt},
 	{"register-save-restore", .show = xe_gt_debugfs_simple_show, .data = register_save_restore},
···
 	{"default_lrc_bcs", .show = xe_gt_debugfs_simple_show, .data = bcs_default_lrc},
 	{"default_lrc_vcs", .show = xe_gt_debugfs_simple_show, .data = vcs_default_lrc},
 	{"default_lrc_vecs", .show = xe_gt_debugfs_simple_show, .data = vecs_default_lrc},
-	{"stats", .show = xe_gt_debugfs_simple_show, .data = xe_gt_stats_print_info},
 	{"hwconfig", .show = xe_gt_debugfs_simple_show, .data = hwconfig},
-};
-
-/*
- * only for GT debugfs files which are valid on VF. Not valid on PF.
- */
-static const struct drm_info_list vf_only_debugfs_list[] = {
-	{"sa_info_vf_ccs", .show = xe_gt_debugfs_simple_show, .data = sa_info_vf_ccs},
 };
 
 /* everything else should be added here */
···
 		call(gt);
 	return count;
 }
+
+static ssize_t stats_write(struct file *file, const char __user *userbuf,
+			   size_t count, loff_t *ppos)
+{
+	struct seq_file *s = file->private_data;
+	struct xe_gt *gt = s->private;
+
+	return write_to_gt_call(userbuf, count, ppos, xe_gt_stats_clear, gt);
+}
+
+static int stats_show(struct seq_file *s, void *unused)
+{
+	struct drm_printer p = drm_seq_file_printer(s);
+	struct xe_gt *gt = s->private;
+
+	return xe_gt_stats_print_info(gt, &p);
+}
+DEFINE_SHOW_STORE_ATTRIBUTE(stats);
 
 static void force_reset(struct xe_gt *gt)
 {
···
 	root->d_inode->i_private = gt;
 
 	/* VF safe */
+	debugfs_create_file("stats", 0600, root, gt, &stats_fops);
 	debugfs_create_file("force_reset", 0600, root, gt, &force_reset_fops);
 	debugfs_create_file("force_reset_sync", 0600, root, gt, &force_reset_sync_fops);
 
···
 		drm_debugfs_create_files(pf_only_debugfs_list,
 					 ARRAY_SIZE(pf_only_debugfs_list),
 					 root, minor);
-	else
-		drm_debugfs_create_files(vf_only_debugfs_list,
-					 ARRAY_SIZE(vf_only_debugfs_list),
-					 root, minor);
-
 
 	xe_uc_debugfs_register(&gt->uc, root);
 

drivers/gpu/drm/xe/xe_gt_freq.c

···
 }
 static struct kobj_attribute attr_max_freq = __ATTR_RW(max_freq);
 
+static ssize_t power_profile_show(struct kobject *kobj,
+				  struct kobj_attribute *attr,
+				  char *buff)
+{
+	struct device *dev = kobj_to_dev(kobj);
+
+	xe_guc_pc_get_power_profile(dev_to_pc(dev), buff);
+
+	return strlen(buff);
+}
+
+static ssize_t power_profile_store(struct kobject *kobj,
+				   struct kobj_attribute *attr,
+				   const char *buff, size_t count)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct xe_guc_pc *pc = dev_to_pc(dev);
+	int err;
+
+	xe_pm_runtime_get(dev_to_xe(dev));
+	err = xe_guc_pc_set_power_profile(pc, buff);
+	xe_pm_runtime_put(dev_to_xe(dev));
+
+	return err ?: count;
+}
+static struct kobj_attribute attr_power_profile = __ATTR_RW(power_profile);
+
 static const struct attribute *freq_attrs[] = {
 	&attr_act_freq.attr,
 	&attr_cur_freq.attr,
···
 	&attr_rpn_freq.attr,
 	&attr_min_freq.attr,
 	&attr_max_freq.attr,
+	&attr_power_profile.attr,
 	NULL
 };
 

drivers/gpu/drm/xe/xe_gt_mcr.c

···
  * @group: pointer to storage for steering group ID
  * @instance: pointer to storage for steering instance ID
  */
-void xe_gt_mcr_get_dss_steering(struct xe_gt *gt, unsigned int dss, u16 *group, u16 *instance)
+void xe_gt_mcr_get_dss_steering(const struct xe_gt *gt, unsigned int dss, u16 *group, u16 *instance)
 {
 	xe_gt_assert(gt, dss < XE_MAX_DSS_FUSE_BITS);
 

drivers/gpu/drm/xe/xe_gt_mcr.h

···
 					  u8 *group, u8 *instance);
 
 void xe_gt_mcr_steering_dump(struct xe_gt *gt, struct drm_printer *p);
-void xe_gt_mcr_get_dss_steering(struct xe_gt *gt, unsigned int dss, u16 *group, u16 *instance);
+void xe_gt_mcr_get_dss_steering(const struct xe_gt *gt,
+				unsigned int dss, u16 *group, u16 *instance);
 u32 xe_gt_mcr_steering_info_to_dss_id(struct xe_gt *gt, u16 group, u16 instance);
 
 /*

drivers/gpu/drm/xe/xe_gt_pagefault.c

···
 	if (!bo)
 		return 0;
 
-	err = need_vram_move ? xe_bo_migrate(bo, vram->placement) :
-			       xe_bo_validate(bo, vm, true);
-
-	return err;
+	return need_vram_move ? xe_bo_migrate(bo, vram->placement, NULL, exec) :
+		xe_bo_validate(bo, vm, true, exec);
 }
 
 static int handle_vma_pagefault(struct xe_gt *gt, struct xe_vma *vma,
···
 {
 	struct xe_vm *vm = xe_vma_vm(vma);
 	struct xe_tile *tile = gt_to_tile(gt);
+	struct xe_validation_ctx ctx;
 	struct drm_exec exec;
 	struct dma_fence *fence;
-	ktime_t end = 0;
 	int err, needs_vram;
 
 	lockdep_assert_held_write(&vm->lock);
···
 	}
 
 	/* Lock VM and BOs dma-resv */
-	drm_exec_init(&exec, 0, 0);
+	xe_validation_ctx_init(&ctx, &vm->xe->val, &exec, (struct xe_val_flags) {});
 	drm_exec_until_all_locked(&exec) {
 		err = xe_pf_begin(&exec, vma, needs_vram == 1, tile->mem.vram);
 		drm_exec_retry_on_contention(&exec);
-		if (xe_vm_validate_should_retry(&exec, err, &end))
-			err = -EAGAIN;
+		xe_validation_retry_on_oom(&ctx, &err);
 		if (err)
 			goto unlock_dma_resv;
 
 		/* Bind VMA only to the GT that has faulted */
 		trace_xe_vma_pf_bind(vma);
+		xe_vm_set_validation_exec(vm, &exec);
 		fence = xe_vma_rebind(vm, vma, BIT(tile->id));
+		xe_vm_set_validation_exec(vm, NULL);
 		if (IS_ERR(fence)) {
 			err = PTR_ERR(fence);
-			if (xe_vm_validate_should_retry(&exec, err, &end))
-				err = -EAGAIN;
+			xe_validation_retry_on_oom(&ctx, &err);
 			goto unlock_dma_resv;
 		}
 	}
···
 	dma_fence_put(fence);
 
 unlock_dma_resv:
-	drm_exec_fini(&exec);
+	xe_validation_ctx_fini(&ctx);
 	if (err == -EAGAIN)
 		goto retry_userptr;
 
···
 {
 	struct xe_device *xe = gt_to_xe(gt);
 	struct xe_tile *tile = gt_to_tile(gt);
+	struct xe_validation_ctx ctx;
 	struct drm_exec exec;
 	struct xe_vm *vm;
 	struct xe_vma *vma;
···
 		goto unlock_vm;
 
 	/* Lock VM and BOs dma-resv */
-	drm_exec_init(&exec, 0, 0);
+	xe_validation_ctx_init(&ctx, &vm->xe->val, &exec, (struct xe_val_flags) {});
 	drm_exec_until_all_locked(&exec) {
 		ret = xe_pf_begin(&exec, vma, IS_DGFX(vm->xe), tile->mem.vram);
 		drm_exec_retry_on_contention(&exec);
-		if (ret)
-			break;
+		xe_validation_retry_on_oom(&ctx, &ret);
 	}
 
-	drm_exec_fini(&exec);
+	xe_validation_ctx_fini(&ctx);
 unlock_vm:
 	up_read(&vm->lock);
 	xe_vm_put(vm);

drivers/gpu/drm/xe/xe_gt_printk.h

···
 #ifndef _XE_GT_PRINTK_H_
 #define _XE_GT_PRINTK_H_
 
-#include <drm/drm_print.h>
-
 #include "xe_gt_types.h"
+#include "xe_tile_printk.h"
+
+#define __XE_GT_PRINTK_FMT(_gt, _fmt, _args...)	"GT%u: " _fmt, (_gt)->info.id, ##_args
 
 #define xe_gt_printk(_gt, _level, _fmt, ...) \
-	drm_##_level(&gt_to_xe(_gt)->drm, "GT%u: " _fmt, (_gt)->info.id, ##__VA_ARGS__)
+	xe_tile_printk((_gt)->tile, _level, __XE_GT_PRINTK_FMT((_gt), _fmt, ##__VA_ARGS__))
+
+#define xe_gt_err(_gt, _fmt, ...) \
+	xe_gt_printk((_gt), err, _fmt, ##__VA_ARGS__)
 
 #define xe_gt_err_once(_gt, _fmt, ...) \
 	xe_gt_printk((_gt), err_once, _fmt, ##__VA_ARGS__)
 
-#define xe_gt_err(_gt, _fmt, ...) \
-	xe_gt_printk((_gt), err, _fmt, ##__VA_ARGS__)
+#define xe_gt_err_ratelimited(_gt, _fmt, ...) \
+	xe_gt_printk((_gt), err_ratelimited, _fmt, ##__VA_ARGS__)
 
 #define xe_gt_warn(_gt, _fmt, ...) \
 	xe_gt_printk((_gt), warn, _fmt, ##__VA_ARGS__)
···
 #define xe_gt_dbg(_gt, _fmt, ...) \
 	xe_gt_printk((_gt), dbg, _fmt, ##__VA_ARGS__)
 
-#define xe_gt_err_ratelimited(_gt, _fmt, ...) \
-	xe_gt_printk((_gt), err_ratelimited, _fmt, ##__VA_ARGS__)
+#define xe_gt_WARN_type(_gt, _type, _condition, _fmt, ...) \
+	xe_tile_WARN##_type((_gt)->tile, _condition, _fmt, ## __VA_ARGS__)
 
 #define xe_gt_WARN(_gt, _condition, _fmt, ...) \
-	drm_WARN(&gt_to_xe(_gt)->drm, _condition, "GT%u: " _fmt, (_gt)->info.id, ##__VA_ARGS__)
+	xe_gt_WARN_type((_gt),, _condition, __XE_GT_PRINTK_FMT((_gt), _fmt, ##__VA_ARGS__))
 
 #define xe_gt_WARN_ONCE(_gt, _condition, _fmt, ...) \
-	drm_WARN_ONCE(&gt_to_xe(_gt)->drm, _condition, "GT%u: " _fmt, (_gt)->info.id, ##__VA_ARGS__)
+	xe_gt_WARN_type((_gt), _ONCE, _condition, __XE_GT_PRINTK_FMT((_gt), _fmt, ##__VA_ARGS__))
 
 #define xe_gt_WARN_ON(_gt, _condition) \
-	xe_gt_WARN((_gt), _condition, "%s(%s)", "gt_WARN_ON", __stringify(_condition))
+	xe_gt_WARN((_gt), _condition, "%s(%s)", "WARN_ON", __stringify(_condition))
 
 #define xe_gt_WARN_ON_ONCE(_gt, _condition) \
-	xe_gt_WARN_ONCE((_gt), _condition, "%s(%s)", "gt_WARN_ON_ONCE", __stringify(_condition))
+	xe_gt_WARN_ONCE((_gt), _condition, "%s(%s)", "WARN_ON_ONCE", __stringify(_condition))
 
 static inline void __xe_gt_printfn_err(struct drm_printer *p, struct va_format *vaf)
 {
···
 
 	/*
 	 * The original xe_gt_dbg() callsite annotations are useless here,
-	 * redirect to the tweaked drm_dbg_printer() instead.
+	 * redirect to the tweaked xe_tile_dbg_printer() instead.
 	 */
-	dbg = drm_dbg_printer(&gt_to_xe(gt)->drm, DRM_UT_DRIVER, NULL);
+	dbg = xe_tile_dbg_printer((gt)->tile);
 	dbg.origin = p->origin;
 
-	drm_printf(&dbg, "GT%u: %pV", gt->info.id, vaf);
+	drm_printf(&dbg, __XE_GT_PRINTK_FMT(gt, "%pV", vaf));
 }
 
 /**

drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c

···
 		return 0;
 
 	xe_gt_assert(gt, pf_get_lmem_alignment(gt) == SZ_2M);
-	bo = xe_bo_create_locked(xe, tile, NULL,
-				 ALIGN(size, PAGE_SIZE),
-				 ttm_bo_type_kernel,
-				 XE_BO_FLAG_VRAM_IF_DGFX(tile) |
-				 XE_BO_FLAG_NEEDS_2M |
-				 XE_BO_FLAG_PINNED |
-				 XE_BO_FLAG_PINNED_LATE_RESTORE);
+	bo = xe_bo_create_pin_range_novm(xe, tile,
+					 ALIGN(size, PAGE_SIZE), 0, ~0ull,
+					 ttm_bo_type_kernel,
+					 XE_BO_FLAG_VRAM_IF_DGFX(tile) |
+					 XE_BO_FLAG_NEEDS_2M |
+					 XE_BO_FLAG_PINNED |
+					 XE_BO_FLAG_PINNED_LATE_RESTORE);
 	if (IS_ERR(bo))
 		return PTR_ERR(bo);
-
-	err = xe_bo_pin(bo);
-	xe_bo_unlock(bo);
-	if (unlikely(err)) {
-		xe_bo_put(bo);
-		return err;
-	}
 
 	config->lmem_obj = bo;
 
···
 	u64 fair;
 
 	fair = div_u64(available, num_vfs);
-	fair = rounddown_pow_of_two(fair);	/* XXX: ttm_vram_mgr & drm_buddy limitation */
 	fair = ALIGN_DOWN(fair, alignment);
 #ifdef MAX_FAIR_LMEM
 	fair = min_t(u64, MAX_FAIR_LMEM, fair);

drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.c

···
 	xe_gt_assert(gt, size % sizeof(u32) == 0);
 	xe_gt_assert(gt, size == ndwords * sizeof(u32));
 
-	bo = xe_bo_create_pin_map(xe, tile, NULL,
-				  ALIGN(size, PAGE_SIZE),
-				  ttm_bo_type_kernel,
-				  XE_BO_FLAG_SYSTEM |
-				  XE_BO_FLAG_GGTT |
-				  XE_BO_FLAG_GGTT_INVALIDATE);
+	bo = xe_bo_create_pin_map_novm(xe, tile,
+				       ALIGN(size, PAGE_SIZE),
+				       ttm_bo_type_kernel,
+				       XE_BO_FLAG_SYSTEM |
+				       XE_BO_FLAG_GGTT |
+				       XE_BO_FLAG_GGTT_INVALIDATE, false);
 	if (IS_ERR(bo))
 		return PTR_ERR(bo);
 
···
 	xe_gt_assert(gt, size % sizeof(u32) == 0);
 	xe_gt_assert(gt, size == ndwords * sizeof(u32));
 
-	bo = xe_bo_create_pin_map(xe, tile, NULL,
-				  ALIGN(size, PAGE_SIZE),
-				  ttm_bo_type_kernel,
-				  XE_BO_FLAG_SYSTEM |
-				  XE_BO_FLAG_GGTT |
-				  XE_BO_FLAG_GGTT_INVALIDATE);
+	bo = xe_bo_create_pin_map_novm(xe, tile,
+				       ALIGN(size, PAGE_SIZE),
+				       ttm_bo_type_kernel,
+				       XE_BO_FLAG_SYSTEM |
+				       XE_BO_FLAG_GGTT |
+				       XE_BO_FLAG_GGTT_INVALIDATE, false);
 	if (IS_ERR(bo))
 		return PTR_ERR(bo);
 

drivers/gpu/drm/xe/xe_gt_stats.c

···
 	atomic64_add(incr, &gt->stats.counters[id]);
 }
 
+#define DEF_STAT_STR(ID, name) [XE_GT_STATS_ID_##ID] = name
+
 static const char *const stat_description[__XE_GT_STATS_NUM_IDS] = {
-	"svm_pagefault_count",
-	"tlb_inval_count",
-	"vma_pagefault_count",
-	"vma_pagefault_kb",
+	DEF_STAT_STR(SVM_PAGEFAULT_COUNT, "svm_pagefault_count"),
+	DEF_STAT_STR(TLB_INVAL, "tlb_inval_count"),
+	DEF_STAT_STR(SVM_TLB_INVAL_COUNT, "svm_tlb_inval_count"),
+	DEF_STAT_STR(SVM_TLB_INVAL_US, "svm_tlb_inval_us"),
+	DEF_STAT_STR(VMA_PAGEFAULT_COUNT, "vma_pagefault_count"),
+	DEF_STAT_STR(VMA_PAGEFAULT_KB, "vma_pagefault_kb"),
+	DEF_STAT_STR(SVM_4K_PAGEFAULT_COUNT, "svm_4K_pagefault_count"),
+	DEF_STAT_STR(SVM_64K_PAGEFAULT_COUNT, "svm_64K_pagefault_count"),
+	DEF_STAT_STR(SVM_2M_PAGEFAULT_COUNT, "svm_2M_pagefault_count"),
+	DEF_STAT_STR(SVM_4K_VALID_PAGEFAULT_COUNT, "svm_4K_valid_pagefault_count"),
+	DEF_STAT_STR(SVM_64K_VALID_PAGEFAULT_COUNT, "svm_64K_valid_pagefault_count"),
+	DEF_STAT_STR(SVM_2M_VALID_PAGEFAULT_COUNT, "svm_2M_valid_pagefault_count"),
+	DEF_STAT_STR(SVM_4K_PAGEFAULT_US, "svm_4K_pagefault_us"),
+	DEF_STAT_STR(SVM_64K_PAGEFAULT_US, "svm_64K_pagefault_us"),
+	DEF_STAT_STR(SVM_2M_PAGEFAULT_US, "svm_2M_pagefault_us"),
+	DEF_STAT_STR(SVM_4K_MIGRATE_COUNT, "svm_4K_migrate_count"),
+	DEF_STAT_STR(SVM_64K_MIGRATE_COUNT, "svm_64K_migrate_count"),
+	DEF_STAT_STR(SVM_2M_MIGRATE_COUNT, "svm_2M_migrate_count"),
+	DEF_STAT_STR(SVM_4K_MIGRATE_US, "svm_4K_migrate_us"),
+	DEF_STAT_STR(SVM_64K_MIGRATE_US, "svm_64K_migrate_us"),
+	DEF_STAT_STR(SVM_2M_MIGRATE_US, "svm_2M_migrate_us"),
+	DEF_STAT_STR(SVM_DEVICE_COPY_US, "svm_device_copy_us"),
+	DEF_STAT_STR(SVM_4K_DEVICE_COPY_US, "svm_4K_device_copy_us"),
+	DEF_STAT_STR(SVM_64K_DEVICE_COPY_US, "svm_64K_device_copy_us"),
+	DEF_STAT_STR(SVM_2M_DEVICE_COPY_US, "svm_2M_device_copy_us"),
+	DEF_STAT_STR(SVM_CPU_COPY_US, "svm_cpu_copy_us"),
+	DEF_STAT_STR(SVM_4K_CPU_COPY_US, "svm_4K_cpu_copy_us"),
+	DEF_STAT_STR(SVM_64K_CPU_COPY_US, "svm_64K_cpu_copy_us"),
+	DEF_STAT_STR(SVM_2M_CPU_COPY_US, "svm_2M_cpu_copy_us"),
+	DEF_STAT_STR(SVM_DEVICE_COPY_KB, "svm_device_copy_kb"),
+	DEF_STAT_STR(SVM_CPU_COPY_KB, "svm_cpu_copy_kb"),
+	DEF_STAT_STR(SVM_4K_GET_PAGES_US, "svm_4K_get_pages_us"),
+	DEF_STAT_STR(SVM_64K_GET_PAGES_US, "svm_64K_get_pages_us"),
+	DEF_STAT_STR(SVM_2M_GET_PAGES_US, "svm_2M_get_pages_us"),
+	DEF_STAT_STR(SVM_4K_BIND_US, "svm_4K_bind_us"),
+	DEF_STAT_STR(SVM_64K_BIND_US, "svm_64K_bind_us"),
+	DEF_STAT_STR(SVM_2M_BIND_US, "svm_2M_bind_us"),
 };
 
 /**
···
 			   atomic64_read(&gt->stats.counters[id]));
 
 	return 0;
+}
+
+/**
+ * xe_gt_stats_clear - Clear the GT stats
+ * @gt: GT structure
+ *
+ * This clear (zeros) all the available GT stats.
+ */
+void xe_gt_stats_clear(struct xe_gt *gt)
+{
+	int id;
+
+	for (id = 0; id < ARRAY_SIZE(gt->stats.counters); ++id)
+		atomic64_set(&gt->stats.counters[id], 0);
 }

drivers/gpu/drm/xe/xe_gt_stats.h

···
 
 #ifdef CONFIG_DEBUG_FS
 int xe_gt_stats_print_info(struct xe_gt *gt, struct drm_printer *p);
+void xe_gt_stats_clear(struct xe_gt *gt);
 void xe_gt_stats_incr(struct xe_gt *gt, const enum xe_gt_stats_id id, int incr);
 #else
 static inline void

drivers/gpu/drm/xe/xe_gt_stats_types.h

···
 enum xe_gt_stats_id {
 	XE_GT_STATS_ID_SVM_PAGEFAULT_COUNT,
 	XE_GT_STATS_ID_TLB_INVAL,
+	XE_GT_STATS_ID_SVM_TLB_INVAL_COUNT,
+	XE_GT_STATS_ID_SVM_TLB_INVAL_US,
 	XE_GT_STATS_ID_VMA_PAGEFAULT_COUNT,
 	XE_GT_STATS_ID_VMA_PAGEFAULT_KB,
+	XE_GT_STATS_ID_SVM_4K_PAGEFAULT_COUNT,
+	XE_GT_STATS_ID_SVM_64K_PAGEFAULT_COUNT,
+	XE_GT_STATS_ID_SVM_2M_PAGEFAULT_COUNT,
+	XE_GT_STATS_ID_SVM_4K_VALID_PAGEFAULT_COUNT,
+	XE_GT_STATS_ID_SVM_64K_VALID_PAGEFAULT_COUNT,
+	XE_GT_STATS_ID_SVM_2M_VALID_PAGEFAULT_COUNT,
+	XE_GT_STATS_ID_SVM_4K_PAGEFAULT_US,
+	XE_GT_STATS_ID_SVM_64K_PAGEFAULT_US,
+	XE_GT_STATS_ID_SVM_2M_PAGEFAULT_US,
+	XE_GT_STATS_ID_SVM_4K_MIGRATE_COUNT,
+	XE_GT_STATS_ID_SVM_64K_MIGRATE_COUNT,
+	XE_GT_STATS_ID_SVM_2M_MIGRATE_COUNT,
+	XE_GT_STATS_ID_SVM_4K_MIGRATE_US,
+	XE_GT_STATS_ID_SVM_64K_MIGRATE_US,
+	XE_GT_STATS_ID_SVM_2M_MIGRATE_US,
+	XE_GT_STATS_ID_SVM_DEVICE_COPY_US,
+	XE_GT_STATS_ID_SVM_4K_DEVICE_COPY_US,
+	XE_GT_STATS_ID_SVM_64K_DEVICE_COPY_US,
+	XE_GT_STATS_ID_SVM_2M_DEVICE_COPY_US,
+	XE_GT_STATS_ID_SVM_CPU_COPY_US,
+	XE_GT_STATS_ID_SVM_4K_CPU_COPY_US,
+	XE_GT_STATS_ID_SVM_64K_CPU_COPY_US,
+	XE_GT_STATS_ID_SVM_2M_CPU_COPY_US,
+	XE_GT_STATS_ID_SVM_DEVICE_COPY_KB,
+	XE_GT_STATS_ID_SVM_CPU_COPY_KB,
+	XE_GT_STATS_ID_SVM_4K_GET_PAGES_US,
+	XE_GT_STATS_ID_SVM_64K_GET_PAGES_US,
+	XE_GT_STATS_ID_SVM_2M_GET_PAGES_US,
+	XE_GT_STATS_ID_SVM_4K_BIND_US,
+	XE_GT_STATS_ID_SVM_64K_BIND_US,
+	XE_GT_STATS_ID_SVM_2M_BIND_US,
 	/* must be the last entry */
 	__XE_GT_STATS_NUM_IDS,
 };

drivers/gpu/drm/xe/xe_gt_topology.c

···
 #include "regs/xe_gt_regs.h"
 #include "xe_assert.h"
 #include "xe_gt.h"
+#include "xe_gt_mcr.h"
 #include "xe_gt_printk.h"
 #include "xe_mmio.h"
 #include "xe_wa.h"
···
 	}
 }
 
+bool xe_gt_topology_report_l3(struct xe_gt *gt)
+{
+	/*
+	 * No known userspace needs/uses the L3 bank mask reported by
+	 * the media GT, and the hardware itself is known to report bogus
+	 * values on several platforms.  Only report L3 bank mask as part
+	 * of the media GT's topology on pre-Xe3 platforms since that's
+	 * already part of our ABI.
+	 */
+	if (xe_gt_is_media_type(gt) && MEDIA_VER(gt_to_xe(gt)) >= 30)
+		return false;
+
+	return true;
+}
+
 static void
 load_l3_bank_mask(struct xe_gt *gt, xe_l3_bank_mask_t l3_bank_mask)
 {
···
 	struct xe_mmio *mmio = &gt->mmio;
 	u32 fuse3 = xe_mmio_read32(mmio, MIRROR_FUSE3);
 
-	/*
-	 * PTL platforms with media version 30.00 do not provide proper values
-	 * for the media GT's L3 bank registers.  Skip the readout since we
-	 * don't have any way to obtain real values.
-	 *
-	 * This may get re-described as an official workaround in the future,
-	 * but there's no tracking number assigned yet so we use a custom
-	 * OOB workaround descriptor.
-	 */
-	if (XE_GT_WA(gt, no_media_l3))
+	if (!xe_gt_topology_report_l3(gt))
 		return;
 
 	if (GRAPHICS_VER(xe) >= 30) {
···
 	drm_printf(p, "EU type:             %s\n",
 		   eu_type_to_str(gt->fuse_topo.eu_type));
 
-	drm_printf(p, "L3 bank mask:        %*pb\n", XE_MAX_L3_BANK_MASK_BITS,
-		   gt->fuse_topo.l3_bank_mask);
+	if (xe_gt_topology_report_l3(gt))
+		drm_printf(p, "L3 bank mask:        %*pb\n", XE_MAX_L3_BANK_MASK_BITS,
+			   gt->fuse_topo.l3_bank_mask);
 }
 
 /*
···
 bool xe_gt_has_compute_dss(struct xe_gt *gt, unsigned int dss)
 {
 	return test_bit(dss, gt->fuse_topo.c_dss_mask);
+}
+
+bool xe_gt_has_discontiguous_dss_groups(const struct xe_gt *gt)
+{
+	unsigned int xecore;
+	int last_group = -1;
+	u16 group, instance;
+
+	for_each_dss_steering(xecore, gt, group, instance) {
+		if (last_group != group) {
+			if (group - last_group > 1)
+				return true;
+			last_group = group;
+		}
+	}
+	return false;
 }

drivers/gpu/drm/xe/xe_gt_topology.h

···
 bool xe_gt_has_geometry_dss(struct xe_gt *gt, unsigned int dss);
 bool xe_gt_has_compute_dss(struct xe_gt *gt, unsigned int dss);
 
+bool xe_gt_has_discontiguous_dss_groups(const struct xe_gt *gt);
+
+bool xe_gt_topology_report_l3(struct xe_gt *gt);
+
 #endif /* _XE_GT_TOPOLOGY_H_ */

drivers/gpu/drm/xe/xe_guc.c

···
 	if (!GUC_LOG_LEVEL_IS_VERBOSE(level))
 		flags |= GUC_LOG_DISABLED;
 	else
-		flags |= GUC_LOG_LEVEL_TO_VERBOSITY(level) <<
-			 GUC_LOG_VERBOSITY_SHIFT;
+		flags |= FIELD_PREP(GUC_LOG_VERBOSITY, GUC_LOG_LEVEL_TO_VERBOSITY(level));
 
 	return flags;
 }
···
 	BUILD_BUG_ON(!CAPTURE_BUFFER_SIZE);
 	BUILD_BUG_ON(!IS_ALIGNED(CAPTURE_BUFFER_SIZE, CAPTURE_UNIT));
 
-	BUILD_BUG_ON((CRASH_BUFFER_SIZE / LOG_UNIT - 1) >
-			(GUC_LOG_CRASH_MASK >> GUC_LOG_CRASH_SHIFT));
-	BUILD_BUG_ON((DEBUG_BUFFER_SIZE / LOG_UNIT - 1) >
-			(GUC_LOG_DEBUG_MASK >> GUC_LOG_DEBUG_SHIFT));
-	BUILD_BUG_ON((CAPTURE_BUFFER_SIZE / CAPTURE_UNIT - 1) >
-			(GUC_LOG_CAPTURE_MASK >> GUC_LOG_CAPTURE_SHIFT));
-
 	flags = GUC_LOG_VALID |
 		GUC_LOG_NOTIFY_ON_HALF_FULL |
 		CAPTURE_FLAG |
 		LOG_FLAG |
-		((CRASH_BUFFER_SIZE / LOG_UNIT - 1) << GUC_LOG_CRASH_SHIFT) |
-		((DEBUG_BUFFER_SIZE / LOG_UNIT - 1) << GUC_LOG_DEBUG_SHIFT) |
-		((CAPTURE_BUFFER_SIZE / CAPTURE_UNIT - 1) <<
-		 GUC_LOG_CAPTURE_SHIFT) |
-		(offset << GUC_LOG_BUF_ADDR_SHIFT);
+		FIELD_PREP(GUC_LOG_CRASH, CRASH_BUFFER_SIZE / LOG_UNIT - 1) |
+		FIELD_PREP(GUC_LOG_DEBUG, DEBUG_BUFFER_SIZE / LOG_UNIT - 1) |
+		FIELD_PREP(GUC_LOG_CAPTURE, CAPTURE_BUFFER_SIZE / CAPTURE_UNIT - 1) |
+		FIELD_PREP(GUC_LOG_BUF_ADDR, offset);
 
 	#undef LOG_UNIT
 	#undef LOG_FLAG
···
 static u32 guc_ctl_ads_flags(struct xe_guc *guc)
 {
 	u32 ads = guc_bo_ggtt_addr(guc, guc->ads.bo) >> PAGE_SHIFT;
-	u32 flags = ads << GUC_ADS_ADDR_SHIFT;
+	u32 flags = FIELD_PREP(GUC_ADS_ADDR, ads);
 
 	return flags;
 }
···
 	if (ret)
 		return ret;
 
-	ret = xe_managed_bo_reinit_in_vram(xe, tile, &guc->ct.bo);
-	if (ret)
-		return ret;
-
 	return 0;
 }
 
···
 	if (ret)
 		return ret;
 
+	ret = xe_guc_ct_init_post_hwconfig(&guc->ct);
+	if (ret)
+		return ret;
+
 	guc_init_params_post_hwconfig(guc);
 
 	ret = xe_guc_submit_init(guc, ~0);
···
 			return ret;
 	}
 
-	guc->submission_state.enabled = true;
-
-	return 0;
+	return xe_guc_submit_enable(guc);
 }
 
 int xe_guc_reset(struct xe_guc *guc)
···
 #endif
 #define GUC_LOAD_TIME_WARN_MS      200
 
-static void guc_wait_ucode(struct xe_guc *guc)
+static int guc_wait_ucode(struct xe_guc *guc)
 {
 	struct xe_gt *gt = guc_to_gt(guc);
 	struct xe_mmio *mmio = &gt->mmio;
···
 			break;
 		}
 
-		xe_device_declare_wedged(gt_to_xe(gt));
+		return -EPROTO;
 	} else if (delta_ms > GUC_LOAD_TIME_WARN_MS) {
 		xe_gt_warn(gt, "excessive init time: %lldms! [status = 0x%08X, timeouts = %d]\n",
 			   delta_ms, status, count);
···
 			  delta_ms, xe_guc_pc_get_act_freq(guc_pc), guc_pc_get_cur_freq(guc_pc),
 			  before_freq, status, count);
 	}
+
+	return 0;
 }
+ALLOW_ERROR_INJECTION(guc_wait_ucode, ERRNO);
 
 static int __xe_guc_upload(struct xe_guc *guc)
 {
···
 		goto out;
 
 	/* Wait for authentication */
-	guc_wait_ucode(guc);
+	ret = guc_wait_ucode(guc);
+	if (ret)
+		goto out;
 
 	xe_uc_fw_change_status(&guc->fw, XE_UC_FIRMWARE_RUNNING);
 	return 0;
 
 out:
 	xe_uc_fw_change_status(&guc->fw, XE_UC_FIRMWARE_LOAD_FAIL);
-	return 0	/* FIXME: ret, don't want to stop load currently */;
+	return ret;
 }
 
 static int vf_guc_min_load_for_hwconfig(struct xe_guc *guc)
···
 {
 	xe_uc_fw_sanitize(&guc->fw);
 	xe_guc_ct_disable(&guc->ct);
-	guc->submission_state.enabled = false;
+	xe_guc_submit_disable(guc);
 }
 
 int xe_guc_reset_prepare(struct xe_guc *guc)
···
 	xe_guc_ct_stop(&guc->ct);
 	xe_guc_submit_wedge(guc);
 }
+
+#if IS_ENABLED(CONFIG_DRM_XE_KUNIT_TEST)
+#include "tests/xe_guc_g2g_test.c"
+#endif

drivers/gpu/drm/xe/xe_guc.h

···
 int xe_guc_start(struct xe_guc *guc);
 void xe_guc_declare_wedged(struct xe_guc *guc);
 
+#if IS_ENABLED(CONFIG_DRM_XE_KUNIT_TEST)
+int xe_guc_g2g_test_notification(struct xe_guc *guc, u32 *payload, u32 len);
+#endif
+
 static inline u16 xe_engine_class_to_guc_class(enum xe_engine_class class)
 {
 	switch (class) {

drivers/gpu/drm/xe/xe_guc_ads.c

···
 	if (XE_GT_WA(gt, 13011645652)) {
 		u32 data = 0xC40;
 
-		guc_waklv_enable(ads, &data, sizeof(data) / sizeof(u32), &offset, &remain,
+		guc_waklv_enable(ads, &data, 1, &offset, &remain,
 				 GUC_WA_KLV_NP_RD_WRITE_TO_CLEAR_RCSM_AT_CGP_LATE_RESTORE);
 	}
 
···
 			0x0,
 			0xF,
 		};
-		guc_waklv_enable(ads, data, sizeof(data) / sizeof(u32), &offset, &remain,
+		guc_waklv_enable(ads, data, ARRAY_SIZE(data), &offset, &remain,
 				 GUC_WA_KLV_RESTORE_UNSAVED_MEDIA_CONTROL_REG);
 	}
 

drivers/gpu/drm/xe/xe_guc_ct.c

···
 static void g2h_worker_func(struct work_struct *w);
 static void safe_mode_worker_func(struct work_struct *w);
 static void ct_exit_safe_mode(struct xe_guc_ct *ct);
+static void guc_ct_change_state(struct xe_guc_ct *ct,
+				enum xe_guc_ct_state state);
 
 #if IS_ENABLED(CONFIG_DRM_XE_DEBUG)
 enum {
···
 }
 ALLOW_ERROR_INJECTION(xe_guc_ct_init_noalloc, ERRNO); /* See xe_pci_probe() */
 
+static void guc_action_disable_ct(void *arg)
+{
+	struct xe_guc_ct *ct = arg;
+
+	guc_ct_change_state(ct, XE_GUC_CT_STATE_DISABLED);
+}
+
 int xe_guc_ct_init(struct xe_guc_ct *ct)
 {
 	struct xe_device *xe = ct_to_xe(ct);
···
 		return PTR_ERR(bo);
 
 	ct->bo = bo;
-	return 0;
+
+	return devm_add_action_or_reset(xe->drm.dev, guc_action_disable_ct, ct);
 }
 ALLOW_ERROR_INJECTION(xe_guc_ct_init, ERRNO); /* See xe_pci_probe() */
+
+/**
+ * xe_guc_ct_init_post_hwconfig - Reinitialize the GuC CTB in VRAM
+ * @ct: the &xe_guc_ct
+ *
+ * Allocate a new BO in VRAM and free the previous BO that was allocated
+ * in system memory (SMEM). Applicable only for DGFX products.
+ *
+ * Return: 0 on success, or a negative errno on failure.
+ */
+int xe_guc_ct_init_post_hwconfig(struct xe_guc_ct *ct)
+{
+	struct xe_device *xe = ct_to_xe(ct);
+	struct xe_gt *gt = ct_to_gt(ct);
+	struct xe_tile *tile = gt_to_tile(gt);
+	int ret;
+
+	xe_assert(xe, !xe_guc_ct_enabled(ct));
+
+	if (IS_DGFX(xe)) {
+		ret = xe_managed_bo_reinit_in_vram(xe, tile, &ct->bo);
+		if (ret)
+			return ret;
+	}
+
+	devm_remove_action(xe->drm.dev, guc_action_disable_ct, ct);
+	return devm_add_action_or_reset(xe->drm.dev, guc_action_disable_ct, ct);
+}
 
 #define desc_read(xe_, guc_ctb__, field_)			\
 	xe_map_rd_field(xe_, &guc_ctb__->desc, 0,		\
···
 	return true;
 }
 
+#define GUC_SEND_RETRY_LIMIT	50
+#define GUC_SEND_RETRY_MSLEEP	5
+
 static int guc_ct_send_recv(struct xe_guc_ct *ct, const u32 *action, u32 len,
 			    u32 *response_buffer, bool no_fail)
 {
 	struct xe_gt *gt = ct_to_gt(ct);
 	struct g2h_fence g2h_fence;
+	unsigned int retries = 0;
 	int ret = 0;
 
 	/*
···
 		xe_gt_dbg(gt, "H2G action %#x retrying: reason %#x\n",
 			  action[0], g2h_fence.reason);
 		mutex_unlock(&ct->lock);
+		if (++retries > GUC_SEND_RETRY_LIMIT) {
+			xe_gt_err(gt, "H2G action %#x reached retry limit=%u, aborting\n",
+				  action[0], GUC_SEND_RETRY_LIMIT);
+			return -ELOOP;
+		}
+		msleep(GUC_SEND_RETRY_MSLEEP * retries);
 		goto retry;
 	}
 	if (g2h_fence.fail) {
···
 	case XE_GUC_ACTION_NOTIFY_EXCEPTION:
 		ret = guc_crash_process_msg(ct, action);
 		break;
+#if IS_ENABLED(CONFIG_DRM_XE_KUNIT_TEST)
+	case XE_GUC_ACTION_TEST_G2G_RECV:
+		ret = xe_guc_g2g_test_notification(guc, payload, adj_len);
+		break;
+#endif
 	default:
 		xe_gt_err(gt, "unexpected G2H action 0x%04x\n", action);
 	}

drivers/gpu/drm/xe/xe_guc_ct.h

···
 
 int xe_guc_ct_init_noalloc(struct xe_guc_ct *ct);
 int xe_guc_ct_init(struct xe_guc_ct *ct);
+int xe_guc_ct_init_post_hwconfig(struct xe_guc_ct *ct);
 int xe_guc_ct_enable(struct xe_guc_ct *ct);
 void xe_guc_ct_disable(struct xe_guc_ct *ct);
 void xe_guc_ct_stop(struct xe_guc_ct *ct);

drivers/gpu/drm/xe/xe_guc_engine_activity.c

···
 	struct xe_tile *tile = gt_to_tile(gt);
 	struct xe_bo *bo, *metadata_bo;
 
-	metadata_bo = xe_bo_create_pin_map(gt_to_xe(gt), tile, NULL, PAGE_ALIGN(metadata_size),
-					   ttm_bo_type_kernel, XE_BO_FLAG_SYSTEM |
-					   XE_BO_FLAG_GGTT | XE_BO_FLAG_GGTT_INVALIDATE);
+	metadata_bo = xe_bo_create_pin_map_novm(gt_to_xe(gt), tile, PAGE_ALIGN(metadata_size),
+						ttm_bo_type_kernel, XE_BO_FLAG_SYSTEM |
+						XE_BO_FLAG_GGTT | XE_BO_FLAG_GGTT_INVALIDATE,
+						false);
 
 	if (IS_ERR(metadata_bo))
 		return PTR_ERR(metadata_bo);
 
-	bo = xe_bo_create_pin_map(gt_to_xe(gt), tile, NULL, PAGE_ALIGN(size),
-				  ttm_bo_type_kernel, XE_BO_FLAG_VRAM_IF_DGFX(tile) |
-				  XE_BO_FLAG_GGTT | XE_BO_FLAG_GGTT_INVALIDATE);
+	bo = xe_bo_create_pin_map_novm(gt_to_xe(gt), tile, PAGE_ALIGN(size),
+				       ttm_bo_type_kernel, XE_BO_FLAG_VRAM_IF_DGFX(tile) |
+				       XE_BO_FLAG_GGTT | XE_BO_FLAG_GGTT_INVALIDATE, false);
 
 	if (IS_ERR(bo)) {
 		xe_bo_unpin_map_no_vm(metadata_bo);

drivers/gpu/drm/xe/xe_guc_exec_queue_types.h

···
 	struct xe_sched_msg static_msgs[MAX_STATIC_MSG_TYPE];
 	/** @lr_tdr: long running TDR worker */
 	struct work_struct lr_tdr;
-	/** @fini_async: do final fini async from this worker */
-	struct work_struct fini_async;
+	/** @destroy_async: do final destroy async from this worker */
+	struct work_struct destroy_async;
 	/** @resume_time: time of last resume */
 	u64 resume_time;
 	/** @state: GuC specific state for this xe_exec_queue */

drivers/gpu/drm/xe/xe_guc_fwif.h

···
 #define G2H_LEN_DW_SCHED_CONTEXT_MODE_SET	4
 #define G2H_LEN_DW_DEREGISTER_CONTEXT		3
 #define G2H_LEN_DW_TLB_INVALIDATE		3
+#define G2H_LEN_DW_G2G_NOTIFY_MIN		3
 
 #define GUC_ID_MAX			65535
 #define GUC_ID_UNKNOWN			0xffffffff
···
 	u32 hwlrca_hi;
 };
 #define CONTEXT_REGISTRATION_FLAG_KMD	BIT(0)
+#define CONTEXT_REGISTRATION_FLAG_TYPE	GENMASK(2, 1)
 
 /* 32-bit KLV structure as used by policy updates and others */
 struct guc_klv_generic_dw_t {
···
 #define   GUC_LOG_NOTIFY_ON_HALF_FULL	BIT(1)
 #define   GUC_LOG_CAPTURE_ALLOC_UNITS	BIT(2)
 #define   GUC_LOG_LOG_ALLOC_UNITS	BIT(3)
-#define   GUC_LOG_CRASH_SHIFT		4
-#define   GUC_LOG_CRASH_MASK		(0x3 << GUC_LOG_CRASH_SHIFT)
-#define   GUC_LOG_DEBUG_SHIFT		6
-#define   GUC_LOG_DEBUG_MASK	        (0xF << GUC_LOG_DEBUG_SHIFT)
-#define   GUC_LOG_CAPTURE_SHIFT		10
-#define   GUC_LOG_CAPTURE_MASK	        (0x3 << GUC_LOG_CAPTURE_SHIFT)
-#define   GUC_LOG_BUF_ADDR_SHIFT	12
+#define   GUC_LOG_CRASH			REG_GENMASK(5, 4)
+#define   GUC_LOG_DEBUG			REG_GENMASK(9, 6)
+#define   GUC_LOG_CAPTURE		REG_GENMASK(11, 10)
+#define   GUC_LOG_BUF_ADDR		REG_GENMASK(31, 12)
 
 #define GUC_CTL_WA			1
 #define   GUC_WA_GAM_CREDITS		BIT(10)
···
 #define   GUC_CTL_DISABLE_SCHEDULER	BIT(14)
 
 #define GUC_CTL_DEBUG			3
-#define   GUC_LOG_VERBOSITY_SHIFT	0
-#define   GUC_LOG_VERBOSITY_LOW		(0 << GUC_LOG_VERBOSITY_SHIFT)
-#define   GUC_LOG_VERBOSITY_MED		(1 << GUC_LOG_VERBOSITY_SHIFT)
-#define   GUC_LOG_VERBOSITY_HIGH	(2 << GUC_LOG_VERBOSITY_SHIFT)
-#define   GUC_LOG_VERBOSITY_ULTRA	(3 << GUC_LOG_VERBOSITY_SHIFT)
-#define	  GUC_LOG_VERBOSITY_MIN		0
+#define   GUC_LOG_VERBOSITY		REG_GENMASK(1, 0)
 #define	  GUC_LOG_VERBOSITY_MAX		3
-#define	  GUC_LOG_VERBOSITY_MASK	0x0000000f
-#define	  GUC_LOG_DESTINATION_MASK	(3 << 4)
-#define   GUC_LOG_DISABLED		(1 << 6)
-#define   GUC_PROFILE_ENABLED		(1 << 7)
+#define	  GUC_LOG_DESTINATION		REG_GENMASK(5, 4)
+#define   GUC_LOG_DISABLED		BIT(6)
+#define   GUC_PROFILE_ENABLED		BIT(7)
 
 #define GUC_CTL_ADS			4
-#define   GUC_ADS_ADDR_SHIFT		1
-#define   GUC_ADS_ADDR_MASK		(0xFFFFF << GUC_ADS_ADDR_SHIFT)
+#define   GUC_ADS_ADDR			REG_GENMASK(21, 1)
 
 #define GUC_CTL_DEVID			5
 

drivers/gpu/drm/xe/xe_guc_log.h

···
 #define DEBUG_BUFFER_SIZE       SZ_8M
 #define CAPTURE_BUFFER_SIZE     SZ_2M
 #else
-#define CRASH_BUFFER_SIZE	SZ_8K
+#define CRASH_BUFFER_SIZE	SZ_16K
 #define DEBUG_BUFFER_SIZE	SZ_64K
 #define CAPTURE_BUFFER_SIZE	SZ_1M
 #endif

drivers/gpu/drm/xe/xe_guc_pc.c

···
  * Xe driver enables SLPC with all of its defaults features and frequency
  * selection, which varies per platform.
  *
+ * Power profiles add another level of control to SLPC. When power saving
+ * profile is chosen, SLPC will use conservative thresholds to ramp frequency,
+ * thus saving power. Base profile is default and ensures balanced performance
+ * for any workload.
+ *
  * Render-C States:
  * ================
  *
···
 	return ret;
 }
 
+static const char *power_profile_to_string(struct xe_guc_pc *pc)
+{
+	switch (pc->power_profile) {
+	case SLPC_POWER_PROFILE_BASE:
+		return "base";
+	case SLPC_POWER_PROFILE_POWER_SAVING:
+		return "power_saving";
+	default:
+		return "invalid";
+	}
+}
+
+void xe_guc_pc_get_power_profile(struct xe_guc_pc *pc, char *profile)
+{
+	switch (pc->power_profile) {
+	case SLPC_POWER_PROFILE_BASE:
+		sprintf(profile, "[%s]    %s\n", "base", "power_saving");
+		break;
+	case SLPC_POWER_PROFILE_POWER_SAVING:
+		sprintf(profile, "%s    [%s]\n", "base", "power_saving");
+		break;
+	default:
+		sprintf(profile, "invalid");
+	}
+}
+
+int xe_guc_pc_set_power_profile(struct xe_guc_pc *pc, const char *buf)
+{
+	int ret = 0;
+	u32 val;
+
+	if (strncmp("base", buf, strlen("base")) == 0)
+		val = SLPC_POWER_PROFILE_BASE;
+	else if (strncmp("power_saving", buf, strlen("power_saving")) == 0)
+		val = SLPC_POWER_PROFILE_POWER_SAVING;
+	else
+		return -EINVAL;
+
+	guard(mutex)(&pc->freq_lock);
+	xe_pm_runtime_get_noresume(pc_to_xe(pc));
+
+	ret = pc_action_set_param(pc,
+				  SLPC_PARAM_POWER_PROFILE,
+				  val);
+	if (ret)
+		xe_gt_err_once(pc_to_gt(pc), "Failed to set power profile to %d: %pe\n",
+			       val, ERR_PTR(ret));
+	else
+		pc->power_profile = val;
+
+	xe_pm_runtime_put(pc_to_xe(pc));
+
+	return ret;
+}
+
 /**
  * xe_guc_pc_start - Start GuC's Power Conservation component
  * @pc: Xe_GuC_PC instance
···
 	/* Enable SLPC Optimized Strategy for compute */
 	ret = pc_action_set_strategy(pc, SLPC_OPTIMIZED_STRATEGY_COMPUTE);
 
+	/* Set cached value of power_profile */
+	ret = xe_guc_pc_set_power_profile(pc, power_profile_to_string(pc));
+	if (unlikely(ret))
+		xe_gt_err(gt, "Failed to set SLPC power profile: %pe\n", ERR_PTR(ret));
+
 out:
 	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	return ret;
···
 		return PTR_ERR(bo);
 
 	pc->bo = bo;
+
+	pc->power_profile = SLPC_POWER_PROFILE_BASE;
 
 	return devm_add_action_or_reset(xe->drm.dev, xe_guc_pc_fini_hw, pc);
 }

drivers/gpu/drm/xe/xe_guc_pc.h

···
 int xe_guc_pc_set_min_freq(struct xe_guc_pc *pc, u32 freq);
 int xe_guc_pc_get_max_freq(struct xe_guc_pc *pc, u32 *freq);
 int xe_guc_pc_set_max_freq(struct xe_guc_pc *pc, u32 freq);
+int xe_guc_pc_set_power_profile(struct xe_guc_pc *pc, const char *buf);
+void xe_guc_pc_get_power_profile(struct xe_guc_pc *pc, char *profile);
 
 enum xe_gt_idle_state xe_guc_pc_c_status(struct xe_guc_pc *pc);
 u64 xe_guc_pc_rc6_residency(struct xe_guc_pc *pc);

drivers/gpu/drm/xe/xe_guc_pc_types.h

···
 	struct mutex freq_lock;
 	/** @freq_ready: Only handle freq changes, if they are really ready */
 	bool freq_ready;
+	/** @power_profile: Base or power_saving profile */
+	u32 power_profile;
 };
 
 #endif	/* _XE_GUC_PC_TYPES_H_ */

drivers/gpu/drm/xe/xe_guc_submit.c

···
 #include "xe_guc_ct.h"
 #include "xe_guc_exec_queue_types.h"
 #include "xe_guc_id_mgr.h"
+#include "xe_guc_klv_helpers.h"
 #include "xe_guc_submit_types.h"
 #include "xe_hw_engine.h"
 #include "xe_hw_fence.h"
···
 	return drmm_add_action_or_reset(&xe->drm, guc_submit_fini, guc);
 }
 
+/*
+ * Given that we want to guarantee enough RCS throughput to avoid missing
+ * frames, we set the yield policy to 20% of each 80ms interval.
+ */
+#define RC_YIELD_DURATION	80	/* in ms */
+#define RC_YIELD_RATIO		20	/* in percent */
+static u32 *emit_render_compute_yield_klv(u32 *emit)
+{
+	*emit++ = PREP_GUC_KLV_TAG(SCHEDULING_POLICIES_RENDER_COMPUTE_YIELD);
+	*emit++ = RC_YIELD_DURATION;
+	*emit++ = RC_YIELD_RATIO;
+
+	return emit;
+}
+
+#define SCHEDULING_POLICY_MAX_DWORDS 16
+static int guc_init_global_schedule_policy(struct xe_guc *guc)
+{
+	u32 data[SCHEDULING_POLICY_MAX_DWORDS];
+	u32 *emit = data;
+	u32 count = 0;
+	int ret;
+
+	if (GUC_SUBMIT_VER(guc) < MAKE_GUC_VER(1, 1, 0))
+		return 0;
+
+	*emit++ = XE_GUC_ACTION_UPDATE_SCHEDULING_POLICIES_KLV;
+
+	if (CCS_MASK(guc_to_gt(guc)))
+		emit = emit_render_compute_yield_klv(emit);
+
+	count = emit - data;
+	if (count > 1) {
+		xe_assert(guc_to_xe(guc), count <= SCHEDULING_POLICY_MAX_DWORDS);
+
+		ret = xe_guc_ct_send_block(&guc->ct, data, count);
+		if (ret < 0) {
+			xe_gt_err(guc_to_gt(guc),
+				  "failed to enable GuC scheduling policies: %pe\n",
+				  ERR_PTR(ret));
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+int xe_guc_submit_enable(struct xe_guc *guc)
+{
+	int ret;
+
+	ret = guc_init_global_schedule_policy(guc);
+	if (ret)
+		return ret;
+
+	guc->submission_state.enabled = true;
+
+	return 0;
+}
+
+void xe_guc_submit_disable(struct xe_guc *guc)
+{
+	guc->submission_state.enabled = false;
+}
+
 static void __release_guc_id(struct xe_guc *guc, struct xe_exec_queue *q, u32 xa_count)
 {
 	int i;
···
 	info.engine_submit_mask = q->logical_mask;
 	info.hwlrca_lo = lower_32_bits(xe_lrc_descriptor(lrc));
 	info.hwlrca_hi = upper_32_bits(xe_lrc_descriptor(lrc));
-	info.flags = CONTEXT_REGISTRATION_FLAG_KMD;
-
-	if (ctx_type != GUC_CONTEXT_NORMAL)
-		info.flags |= BIT(ctx_type);
+	info.flags = CONTEXT_REGISTRATION_FLAG_KMD |
+		FIELD_PREP(CONTEXT_REGISTRATION_FLAG_TYPE, ctx_type);
 
 	if (xe_exec_queue_is_parallel(q)) {
 		u64 ggtt_addr = xe_lrc_parallel_ggtt_addr(lrc);
···
 	return DRM_GPU_SCHED_STAT_NO_HANG;
 }
 
-static void __guc_exec_queue_fini_async(struct work_struct *w)
+static void guc_exec_queue_fini(struct xe_exec_queue *q)
 {
-	struct xe_guc_exec_queue *ge =
-		container_of(w, struct xe_guc_exec_queue, fini_async);
-	struct xe_exec_queue *q = ge->q;
+	struct xe_guc_exec_queue *ge = q->guc;
 	struct xe_guc *guc = exec_queue_to_guc(q);
 
-	xe_pm_runtime_get(guc_to_xe(guc));
-	trace_xe_exec_queue_destroy(q);
-
 	release_guc_id(guc, q);
-	if (xe_exec_queue_is_lr(q))
-		cancel_work_sync(&ge->lr_tdr);
-	/* Confirm no work left behind accessing device structures */
-	cancel_delayed_work_sync(&ge->sched.base.work_tdr);
 	xe_sched_entity_fini(&ge->entity);
 	xe_sched_fini(&ge->sched);
 
···
 	 * (timeline name).
 	 */
 	kfree_rcu(ge, rcu);
+}
+
+static void __guc_exec_queue_destroy_async(struct work_struct *w)
+{
+	struct xe_guc_exec_queue *ge =
+		container_of(w, struct xe_guc_exec_queue, destroy_async);
+	struct xe_exec_queue *q = ge->q;
+	struct xe_guc *guc = exec_queue_to_guc(q);
+
+	xe_pm_runtime_get(guc_to_xe(guc));
+	trace_xe_exec_queue_destroy(q);
+
+	if (xe_exec_queue_is_lr(q))
+		cancel_work_sync(&ge->lr_tdr);
+	/* Confirm no work left behind accessing device structures */
+	cancel_delayed_work_sync(&ge->sched.base.work_tdr);
+
 	xe_exec_queue_fini(q);
+
 	xe_pm_runtime_put(guc_to_xe(guc));
 }
 
-static void guc_exec_queue_fini_async(struct xe_exec_queue *q)
+static void guc_exec_queue_destroy_async(struct xe_exec_queue *q)
 {
 	struct xe_guc *guc = exec_queue_to_guc(q);
 	struct xe_device *xe = guc_to_xe(guc);
 
-	INIT_WORK(&q->guc->fini_async, __guc_exec_queue_fini_async);
+	INIT_WORK(&q->guc->destroy_async, __guc_exec_queue_destroy_async);
 
 	/* We must block on kernel engines so slabs are empty on driver unload */
 	if (q->flags & EXEC_QUEUE_FLAG_PERMANENT || exec_queue_wedged(q))
-		__guc_exec_queue_fini_async(&q->guc->fini_async);
+		__guc_exec_queue_destroy_async(&q->guc->destroy_async);
 	else
-		queue_work(xe->destroy_wq, &q->guc->fini_async);
+		queue_work(xe->destroy_wq, &q->guc->destroy_async);
 }
 
-static void __guc_exec_queue_fini(struct xe_guc *guc, struct xe_exec_queue *q)
+static void __guc_exec_queue_destroy(struct xe_guc *guc, struct xe_exec_queue *q)
 {
 	/*
 	 * Might be done from within the GPU scheduler, need to do async as we
···
 	 * this we and don't really care when everything is fini'd, just that it
 	 * is.
 	 */
-	guc_exec_queue_fini_async(q);
+	guc_exec_queue_destroy_async(q);
 }
 
 static void __guc_exec_queue_process_msg_cleanup(struct xe_sched_msg *msg)
···
 	if (exec_queue_registered(q))
 		disable_scheduling_deregister(guc, q);
 	else
-		__guc_exec_queue_fini(guc, q);
+		__guc_exec_queue_destroy(guc, q);
 }
 
 static bool guc_exec_queue_allowed_to_change_state(struct xe_exec_queue *q)
···
 #define STATIC_MSG_CLEANUP	0
 #define STATIC_MSG_SUSPEND	1
 #define STATIC_MSG_RESUME	2
-static void guc_exec_queue_fini(struct xe_exec_queue *q)
+static void guc_exec_queue_destroy(struct xe_exec_queue *q)
 {
 	struct xe_sched_msg *msg = q->guc->static_msgs + STATIC_MSG_CLEANUP;
 
 	if (!(q->flags & EXEC_QUEUE_FLAG_PERMANENT) && !exec_queue_wedged(q))
 		guc_exec_queue_add_msg(q, msg, CLEANUP);
 	else
-		__guc_exec_queue_fini(exec_queue_to_guc(q), q);
+		__guc_exec_queue_destroy(exec_queue_to_guc(q), q);
 }
 
 static int guc_exec_queue_set_priority(struct xe_exec_queue *q,
···
 	.init = guc_exec_queue_init,
 	.kill = guc_exec_queue_kill,
 	.fini = guc_exec_queue_fini,
+	.destroy = guc_exec_queue_destroy,
 	.set_priority = guc_exec_queue_set_priority,
 	.set_timeslice = guc_exec_queue_set_timeslice,
 	.set_preempt_timeout = guc_exec_queue_set_preempt_timeout,
···
 		if (exec_queue_extra_ref(q) || xe_exec_queue_is_lr(q))
 			xe_exec_queue_put(q);
 		else if (exec_queue_destroyed(q))
-			__guc_exec_queue_fini(guc, q);
+			__guc_exec_queue_destroy(guc, q);
 	}
 	if (q->guc->suspend_pending) {
 		set_exec_queue_suspended(q);
···
 
 	q = xa_load(&guc->submission_state.exec_queue_lookup, guc_id);
 	if (unlikely(!q)) {
-		xe_gt_err(gt, "Not engine present for guc_id %u\n", guc_id);
+		xe_gt_err(gt, "No exec queue found for guc_id %u\n", guc_id);
 		return NULL;
 	}
 
···
 	if (exec_queue_extra_ref(q) || xe_exec_queue_is_lr(q))
 		xe_exec_queue_put(q);
 	else
-		__guc_exec_queue_fini(guc, q);
+		__guc_exec_queue_destroy(guc, q);
 }
 
 int xe_guc_deregister_done_handler(struct xe_guc *guc, u32 *msg, u32 len)
···
 }
 
 /**
- * xe_guc_register_exec_queue - Register exec queue for a given context type.
+ * xe_guc_register_vf_exec_queue - Register exec queue for a given context type.
  * @q: Execution queue
  * @ctx_type: Type of the context
  *
···
  *
  * Returns - None.
  */
-void xe_guc_register_exec_queue(struct xe_exec_queue *q, int ctx_type)
+void xe_guc_register_vf_exec_queue(struct xe_exec_queue *q, int ctx_type)
 {
 	struct xe_guc *guc = exec_queue_to_guc(q);
 	struct xe_device *xe = guc_to_xe(guc);
+	struct xe_gt *gt = guc_to_gt(guc);
 
-	xe_assert(xe, IS_SRIOV_VF(xe));
-	xe_assert(xe, !IS_DGFX(xe));
-	xe_assert(xe, (ctx_type > GUC_CONTEXT_NORMAL &&
-		       ctx_type < GUC_CONTEXT_COUNT));
+	xe_gt_assert(gt, IS_SRIOV_VF(xe));
+	xe_gt_assert(gt, !IS_DGFX(xe));
+	xe_gt_assert(gt, ctx_type == GUC_CONTEXT_COMPRESSION_SAVE ||
+		     ctx_type == GUC_CONTEXT_COMPRESSION_RESTORE);
+	xe_gt_assert(gt, GUC_SUBMIT_VER(guc) >= MAKE_GUC_VER(1, 23, 0));
 
 	register_exec_queue(q, ctx_type);
 	enable_scheduling(q);

drivers/gpu/drm/xe/xe_guc_submit.h

···
 struct xe_guc;
 
 int xe_guc_submit_init(struct xe_guc *guc, unsigned int num_ids);
+int xe_guc_submit_enable(struct xe_guc *guc);
+void xe_guc_submit_disable(struct xe_guc *guc);
 
 int xe_guc_submit_reset_prepare(struct xe_guc *guc);
 void xe_guc_submit_reset_wait(struct xe_guc *guc);
···
 void
 xe_guc_exec_queue_snapshot_free(struct xe_guc_submit_exec_queue_snapshot *snapshot);
 void xe_guc_submit_print(struct xe_guc *guc, struct drm_printer *p);
-void xe_guc_register_exec_queue(struct xe_exec_queue *q, int ctx_type);
+void xe_guc_register_vf_exec_queue(struct xe_exec_queue *q, int ctx_type);
 
 int xe_guc_contexts_hwsp_rebase(struct xe_guc *guc, void *scratch);
 

drivers/gpu/drm/xe/xe_hmm.c

···
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2024 Intel Corporation
- */
-
-#include <linux/scatterlist.h>
-#include <linux/mmu_notifier.h>
-#include <linux/dma-mapping.h>
-#include <linux/memremap.h>
-#include <linux/swap.h>
-#include <linux/hmm.h>
-#include <linux/mm.h>
-#include "xe_hmm.h"
-#include "xe_vm.h"
-#include "xe_bo.h"
-
-static u64 xe_npages_in_range(unsigned long start, unsigned long end)
-{
-	return (end - start) >> PAGE_SHIFT;
-}
-
-static int xe_alloc_sg(struct xe_device *xe, struct sg_table *st,
-		       struct hmm_range *range, struct rw_semaphore *notifier_sem)
-{
-	unsigned long i, npages, hmm_pfn;
-	unsigned long num_chunks = 0;
-	int ret;
-
-	/* HMM docs says this is needed. */
-	ret = down_read_interruptible(notifier_sem);
-	if (ret)
-		return ret;
-
-	if (mmu_interval_read_retry(range->notifier, range->notifier_seq)) {
-		up_read(notifier_sem);
-		return -EAGAIN;
-	}
-
-	npages = xe_npages_in_range(range->start, range->end);
-	for (i = 0; i < npages;) {
-		unsigned long len;
-
-		hmm_pfn = range->hmm_pfns[i];
-		xe_assert(xe, hmm_pfn & HMM_PFN_VALID);
-
-		len = 1UL << hmm_pfn_to_map_order(hmm_pfn);
-
-		/* If order > 0 the page may extend beyond range->start */
-		len -= (hmm_pfn & ~HMM_PFN_FLAGS) & (len - 1);
-		i += len;
-		num_chunks++;
-	}
-	up_read(notifier_sem);
-
-	return sg_alloc_table(st, num_chunks, GFP_KERNEL);
-}
-
-/**
- * xe_build_sg() - build a scatter gather table for all the physical pages/pfn
- * in a hmm_range. dma-map pages if necessary. dma-address is save in sg table
- * and will be used to program GPU page table later.
- * @xe: the xe device who will access the dma-address in sg table
- * @range: the hmm range that we build the sg table from. range->hmm_pfns[]
- * has the pfn numbers of pages that back up this hmm address range.
- * @st: pointer to the sg table.
- * @notifier_sem: The xe notifier lock.
- * @write: whether we write to this range. This decides dma map direction
- * for system pages. If write we map it bi-diretional; otherwise
- * DMA_TO_DEVICE
- *
- * All the contiguous pfns will be collapsed into one entry in
- * the scatter gather table. This is for the purpose of efficiently
- * programming GPU page table.
- *
- * The dma_address in the sg table will later be used by GPU to
- * access memory. So if the memory is system memory, we need to
- * do a dma-mapping so it can be accessed by GPU/DMA.
- *
- * FIXME: This function currently only support pages in system
- * memory. If the memory is GPU local memory (of the GPU who
- * is going to access memory), we need gpu dpa (device physical
- * address), and there is no need of dma-mapping. This is TBD.
- *
- * FIXME: dma-mapping for peer gpu device to access remote gpu's
- * memory. Add this when you support p2p
- *
- * This function allocates the storage of the sg table. It is
- * caller's responsibility to free it calling sg_free_table.
- *
- * Returns 0 if successful; -ENOMEM if fails to allocate memory
- */
-static int xe_build_sg(struct xe_device *xe, struct hmm_range *range,
-		       struct sg_table *st,
-		       struct rw_semaphore *notifier_sem,
-		       bool write)
-{
-	unsigned long npages = xe_npages_in_range(range->start, range->end);
-	struct device *dev = xe->drm.dev;
-	struct scatterlist *sgl;
-	struct page *page;
-	unsigned long i, j;
-
-	lockdep_assert_held(notifier_sem);
-
-	i = 0;
-	for_each_sg(st->sgl, sgl, st->nents, j) {
-		unsigned long hmm_pfn, size;
-
-		hmm_pfn = range->hmm_pfns[i];
-		page = hmm_pfn_to_page(hmm_pfn);
-		xe_assert(xe, !is_device_private_page(page));
-
-		size = 1UL << hmm_pfn_to_map_order(hmm_pfn);
-		size -= page_to_pfn(page) & (size - 1);
-		i += size;
-
-		if (unlikely(j == st->nents - 1)) {
-			xe_assert(xe, i >= npages);
-			if (i > npages)
-				size -= (i - npages);
-
-			sg_mark_end(sgl);
-		} else {
-			xe_assert(xe, i < npages);
-		}
-
-		sg_set_page(sgl, page, size << PAGE_SHIFT, 0);
-	}
-
-	return dma_map_sgtable(dev, st, write ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE,
-			       DMA_ATTR_SKIP_CPU_SYNC | DMA_ATTR_NO_KERNEL_MAPPING);
-}
-
-static void xe_hmm_userptr_set_mapped(struct xe_userptr_vma *uvma)
-{
-	struct xe_userptr *userptr = &uvma->userptr;
-	struct xe_vm *vm = xe_vma_vm(&uvma->vma);
-
-	lockdep_assert_held_write(&vm->lock);
-	lockdep_assert_held(&vm->userptr.notifier_lock);
-
-	mutex_lock(&userptr->unmap_mutex);
-	xe_assert(vm->xe, !userptr->mapped);
-	userptr->mapped = true;
-	mutex_unlock(&userptr->unmap_mutex);
-}
-
-void xe_hmm_userptr_unmap(struct xe_userptr_vma *uvma)
-{
-	struct xe_userptr *userptr = &uvma->userptr;
-	struct xe_vma *vma = &uvma->vma;
-	bool write = !xe_vma_read_only(vma);
-	struct xe_vm *vm = xe_vma_vm(vma);
-	struct xe_device *xe = vm->xe;
-
-	if (!lockdep_is_held_type(&vm->userptr.notifier_lock, 0) &&
-	    !lockdep_is_held_type(&vm->lock, 0) &&
-	    !(vma->gpuva.flags & XE_VMA_DESTROYED)) {
-		/* Don't unmap in exec critical section. */
-		xe_vm_assert_held(vm);
-		/* Don't unmap while mapping the sg. */
-		lockdep_assert_held(&vm->lock);
-	}
-
-	mutex_lock(&userptr->unmap_mutex);
-	if (userptr->sg && userptr->mapped)
-		dma_unmap_sgtable(xe->drm.dev, userptr->sg,
-				  write ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE, 0);
-	userptr->mapped = false;
-	mutex_unlock(&userptr->unmap_mutex);
-}
-
-/**
- * xe_hmm_userptr_free_sg() - Free the scatter gather table of userptr
- * @uvma: the userptr vma which hold the scatter gather table
- *
- * With function xe_userptr_populate_range, we allocate storage of
- * the userptr sg table. This is a helper function to free this
- * sg table, and dma unmap the address in the table.
- */
-void xe_hmm_userptr_free_sg(struct xe_userptr_vma *uvma)
-{
-	struct xe_userptr *userptr = &uvma->userptr;
-
-	xe_assert(xe_vma_vm(&uvma->vma)->xe, userptr->sg);
-	xe_hmm_userptr_unmap(uvma);
-	sg_free_table(userptr->sg);
-	userptr->sg = NULL;
-}
-
-/**
- * xe_hmm_userptr_populate_range() - Populate physical pages of a virtual
- * address range
- *
- * @uvma: userptr vma which has information of the range to populate.
- * @is_mm_mmap_locked: True if mmap_read_lock is already acquired by caller.
- *
- * This function populate the physical pages of a virtual
- * address range. The populated physical pages is saved in
- * userptr's sg table. It is similar to get_user_pages but call
- * hmm_range_fault.
- *
- * This function also read mmu notifier sequence # (
- * mmu_interval_read_begin), for the purpose of later
- * comparison (through mmu_interval_read_retry).
- *
- * This must be called with mmap read or write lock held.
- *
- * This function allocates the storage of the userptr sg table.
- * It is caller's responsibility to free it calling sg_free_table.
- *
- * returns: 0 for success; negative error no on failure
- */
-int xe_hmm_userptr_populate_range(struct xe_userptr_vma *uvma,
-				  bool is_mm_mmap_locked)
-{
-	unsigned long timeout =
-		jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
-	unsigned long *pfns;
-	struct xe_userptr *userptr;
-	struct xe_vma *vma = &uvma->vma;
-	u64 userptr_start = xe_vma_userptr(vma);
-	u64 userptr_end = userptr_start + xe_vma_size(vma);
-	struct xe_vm *vm = xe_vma_vm(vma);
-	struct hmm_range hmm_range = {
-		.pfn_flags_mask = 0, /* ignore pfns */
-		.default_flags = HMM_PFN_REQ_FAULT,
-		.start = userptr_start,
-		.end = userptr_end,
-		.notifier = &uvma->userptr.notifier,
-		.dev_private_owner = vm->xe,
-	};
-	bool write = !xe_vma_read_only(vma);
-	unsigned long notifier_seq;
-	u64 npages;
-	int ret;
-
-	userptr = &uvma->userptr;
-
-	if (is_mm_mmap_locked)
-		mmap_assert_locked(userptr->notifier.mm);
-
-	if (vma->gpuva.flags & XE_VMA_DESTROYED)
-		return 0;
-
-	notifier_seq = mmu_interval_read_begin(&userptr->notifier);
-	if (notifier_seq == userptr->notifier_seq)
-		return 0;
-
-	if (userptr->sg)
-		xe_hmm_userptr_free_sg(uvma);
-
-	npages = xe_npages_in_range(userptr_start, userptr_end);
-	pfns = kvmalloc_array(npages, sizeof(*pfns), GFP_KERNEL);
-	if (unlikely(!pfns))
-		return -ENOMEM;
-
-	if (write)
-		hmm_range.default_flags |= HMM_PFN_REQ_WRITE;
-
-	if (!mmget_not_zero(userptr->notifier.mm)) {
-		ret = -EFAULT;
-		goto free_pfns;
-	}
-
-	hmm_range.hmm_pfns = pfns;
-
-	while (true) {
-		hmm_range.notifier_seq = mmu_interval_read_begin(&userptr->notifier);
-
-		if (!is_mm_mmap_locked)
-			mmap_read_lock(userptr->notifier.mm);
-
-		ret = hmm_range_fault(&hmm_range);
-
-		if (!is_mm_mmap_locked)
-			mmap_read_unlock(userptr->notifier.mm);
-
-		if (ret == -EBUSY) {
-			if (time_after(jiffies, timeout))
-				break;
-
-			continue;
-		}
-		break;
-	}
-
-	mmput(userptr->notifier.mm);
-
-	if (ret)
-		goto free_pfns;
-
-	ret = xe_alloc_sg(vm->xe, &userptr->sgt, &hmm_range, &vm->userptr.notifier_lock);
-	if (ret)
-		goto free_pfns;
-
-	ret = down_read_interruptible(&vm->userptr.notifier_lock);
-	if (ret)
-		goto free_st;
-
-	if (mmu_interval_read_retry(hmm_range.notifier, hmm_range.notifier_seq)) {
-		ret = -EAGAIN;
-		goto out_unlock;
-	}
-
-	ret = xe_build_sg(vm->xe, &hmm_range, &userptr->sgt,
-			  &vm->userptr.notifier_lock, write);
-	if (ret)
-		goto out_unlock;
-
-	userptr->sg = &userptr->sgt;
-	xe_hmm_userptr_set_mapped(uvma);
-	userptr->notifier_seq = hmm_range.notifier_seq;
-	up_read(&vm->userptr.notifier_lock);
-	kvfree(pfns);
-	return 0;
-
-out_unlock:
-	up_read(&vm->userptr.notifier_lock);
-free_st:
-	sg_free_table(&userptr->sgt);
-free_pfns:
-	kvfree(pfns);
-	return ret;
-}

drivers/gpu/drm/xe/xe_hmm.h

···
-/* SPDX-License-Identifier: MIT
- *
- * Copyright © 2024 Intel Corporation
- */
-
-#ifndef _XE_HMM_H_
-#define _XE_HMM_H_
-
-#include <linux/types.h>
-
-struct xe_userptr_vma;
-
-int xe_hmm_userptr_populate_range(struct xe_userptr_vma *uvma, bool is_mm_mmap_locked);
-
-void xe_hmm_userptr_free_sg(struct xe_userptr_vma *uvma);
-
-void xe_hmm_userptr_unmap(struct xe_userptr_vma *uvma);
-#endif

drivers/gpu/drm/xe/xe_hwmon.c

···
  */
 static void xe_hwmon_power_max_read(struct xe_hwmon *hwmon, u32 attr, int channel, long *value)
 {
-	u64 reg_val = 0, min, max;
+	u32 reg_val = 0;
 	struct xe_device *xe = hwmon->xe;
 	struct xe_reg rapl_limit, pkg_power_sku;
 	struct xe_mmio *mmio = xe_root_tile_mmio(xe);
···
 	mutex_lock(&hwmon->hwmon_lock);
 
 	if (hwmon->xe->info.has_mbx_power_limits) {
-		xe_hwmon_pcode_read_power_limit(hwmon, attr, channel, (u32 *)&reg_val);
+		xe_hwmon_pcode_read_power_limit(hwmon, attr, channel, &reg_val);
 	} else {
 		rapl_limit = xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, channel);
 		pkg_power_sku = xe_hwmon_get_reg(hwmon, REG_PKG_POWER_SKU, channel);
···
 	/* Check if PL limits are disabled. */
 	if (!(reg_val & PWR_LIM_EN)) {
 		*value = PL_DISABLE;
-		drm_info(&hwmon->xe->drm, "%s disabled for channel %d, val 0x%016llx\n",
+		drm_info(&hwmon->xe->drm, "%s disabled for channel %d, val 0x%08x\n",
 			 PWR_ATTR_TO_STR(attr), channel, reg_val);
 		goto unlock;
 	}
 
 	reg_val = REG_FIELD_GET(PWR_LIM_VAL, reg_val);
-	*value = mul_u64_u32_shr(reg_val, SF_POWER, hwmon->scl_shift_power);
+	*value = mul_u32_u32(reg_val, SF_POWER) >> hwmon->scl_shift_power;
 
 	/* For platforms with mailbox power limit support clamping would be done by pcode. */
 	if (!hwmon->xe->info.has_mbx_power_limits) {
-		reg_val = xe_mmio_read64_2x32(mmio, pkg_power_sku);
-		min = REG_FIELD_GET(PKG_MIN_PWR, reg_val);
-		max = REG_FIELD_GET(PKG_MAX_PWR, reg_val);
+		u64 pkg_pwr, min, max;
+
+		pkg_pwr = xe_mmio_read64_2x32(mmio, pkg_power_sku);
+		min = REG_FIELD_GET(PKG_MIN_PWR, pkg_pwr);
+		max = REG_FIELD_GET(PKG_MAX_PWR, pkg_pwr);
 		min = mul_u64_u32_shr(min, SF_POWER, hwmon->scl_shift_power);
 		max = mul_u64_u32_shr(max, SF_POWER, hwmon->scl_shift_power);
 		if (min && max)
···
 {
 	struct xe_hwmon *hwmon = dev_get_drvdata(dev);
 	struct xe_mmio *mmio = xe_root_tile_mmio(hwmon->xe);
-	u32 x, y, x_w = 2; /* 2 bits */
-	u64 r, tau4, out;
+	u32 reg_val, x, y, x_w = 2; /* 2 bits */
+	u64 tau4, out;
 	int channel = (to_sensor_dev_attr(attr)->index % 2) ? CHANNEL_PKG : CHANNEL_CARD;
 	u32 power_attr = (to_sensor_dev_attr(attr)->index > 1) ? PL2_HWMON_ATTR : PL1_HWMON_ATTR;
 
···
 	mutex_lock(&hwmon->hwmon_lock);
 
 	if (hwmon->xe->info.has_mbx_power_limits) {
-		ret = xe_hwmon_pcode_read_power_limit(hwmon, power_attr, channel, (u32 *)&r);
+		ret = xe_hwmon_pcode_read_power_limit(hwmon, power_attr, channel, &reg_val);
 		if (ret) {
 			drm_err(&hwmon->xe->drm,
-				"power interval read fail, ch %d, attr %d, r 0%llx, ret %d\n",
-				channel, power_attr, r, ret);
-			r = 0;
+				"power interval read fail, ch %d, attr %d, val 0x%08x, ret %d\n",
+				channel, power_attr, reg_val, ret);
+			reg_val = 0;
 		}
 	} else {
-		r = xe_mmio_read32(mmio, xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, channel));
+		reg_val = xe_mmio_read32(mmio, xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT,
+								channel));
 	}
 
 	mutex_unlock(&hwmon->hwmon_lock);
 
 	xe_pm_runtime_put(hwmon->xe);
 
-	x = REG_FIELD_GET(PWR_LIM_TIME_X, r);
-	y = REG_FIELD_GET(PWR_LIM_TIME_Y, r);
+	x = REG_FIELD_GET(PWR_LIM_TIME_X, reg_val);
+	y = REG_FIELD_GET(PWR_LIM_TIME_Y, reg_val);
 
 	/*
 	 * tau = (1 + (x / 4)) * power(2,y), x = bits(23:22), y = bits(21:17)
···
 			xe_hwmon_fan_input_read(hwmon, channel, &fan_speed);
 }
 
-static void xe_hwmon_mutex_destroy(void *arg)
-{
-	struct xe_hwmon *hwmon = arg;
-
-	mutex_destroy(&hwmon->hwmon_lock);
-}
-
 int xe_hwmon_register(struct xe_device *xe)
 {
 	struct device *dev = xe->drm.dev;
···
 	if (!hwmon)
 		return -ENOMEM;
 
-	mutex_init(&hwmon->hwmon_lock);
-	ret = devm_add_action_or_reset(dev, xe_hwmon_mutex_destroy, hwmon);
+	ret = devm_mutex_init(dev, &hwmon->hwmon_lock);
 	if (ret)
 		return ret;
 

drivers/gpu/drm/xe/xe_i2c.c

···
 		return;
 
 	if (d3cold)
-		xe_mmio_rmw32(mmio, I2C_CONFIG_CMD, 0, PCI_COMMAND_MEMORY);
+		xe_mmio_rmw32(mmio, I2C_CONFIG_CMD, 0, PCI_COMMAND_MEMORY | PCI_COMMAND_MASTER);
 
 	xe_mmio_rmw32(mmio, I2C_CONFIG_PMCSR, PCI_PM_CTRL_STATE_MASK, (__force u32)PCI_D0);
 	drm_dbg(&xe->drm, "pmcsr: 0x%08x\n", xe_mmio_read32(mmio, I2C_CONFIG_PMCSR));

drivers/gpu/drm/xe/xe_late_bind_fw.c

···
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#include <linux/component.h>
+#include <linux/delay.h>
+#include <linux/firmware.h>
+
+#include <drm/drm_managed.h>
+#include <drm/intel/i915_component.h>
+#include <drm/intel/intel_lb_mei_interface.h>
+#include <drm/drm_print.h>
+
+#include "xe_device.h"
+#include "xe_late_bind_fw.h"
+#include "xe_pcode.h"
+#include "xe_pcode_api.h"
+#include "xe_pm.h"
+
+/*
+ * The component should load quite quickly in most cases, but it could take
+ * a bit. Using a very big timeout just to cover the worst case scenario
+ */
+#define LB_INIT_TIMEOUT_MS 20000
+
+/*
+ * Retry interval set to 6 seconds, in steps of 200 ms, to allow time for
+ * other OS components to release the MEI CL handle
+ */
+#define LB_FW_LOAD_RETRY_MAXCOUNT 30
+#define LB_FW_LOAD_RETRY_PAUSE_MS 200
+
+static const u32 fw_id_to_type[] = {
+		[XE_LB_FW_FAN_CONTROL] = INTEL_LB_TYPE_FAN_CONTROL,
+	};
+
+static const char * const fw_id_to_name[] = {
+		[XE_LB_FW_FAN_CONTROL] = "fan_control",
+	};
+
+static struct xe_device *
+late_bind_to_xe(struct xe_late_bind *late_bind)
+{
+	return container_of(late_bind, struct xe_device, late_bind);
+}
+
+static struct xe_device *
+late_bind_fw_to_xe(struct xe_late_bind_fw *lb_fw)
+{
+	return container_of(lb_fw, struct xe_device, late_bind.late_bind_fw[lb_fw->id]);
+}
+
+/* Refer to the "Late Bind based Firmware Layout" documentation entry for details */
+static int parse_cpd_header(struct xe_late_bind_fw *lb_fw,
+			    const void *data, size_t size, const char *manifest_entry)
+{
+	struct xe_device *xe = late_bind_fw_to_xe(lb_fw);
+	const struct gsc_cpd_header_v2 *header = data;
+	const struct gsc_manifest_header *manifest;
+	const struct gsc_cpd_entry *entry;
+	size_t min_size = sizeof(*header);
+	u32 offset;
+	int i;
+
+	/* manifest_entry is mandatory */
+	xe_assert(xe, manifest_entry);
+
+	if (size < min_size || header->header_marker != GSC_CPD_HEADER_MARKER)
+		return -ENOENT;
+
+	if (header->header_length < sizeof(struct gsc_cpd_header_v2)) {
+		drm_err(&xe->drm, "%s late binding fw: Invalid CPD header length %u!\n",
+			fw_id_to_name[lb_fw->id], header->header_length);
+		return -EINVAL;
+	}
+
+	min_size = header->header_length + sizeof(struct gsc_cpd_entry) * header->num_of_entries;
+	if (size < min_size) {
+		drm_err(&xe->drm, "%s late binding fw: too small! %zu < %zu\n",
+			fw_id_to_name[lb_fw->id], size, min_size);
+		return -ENODATA;
+	}
+
+	/* Look for the manifest first */
+	entry = (void *)header + header->header_length;
+	for (i = 0; i < header->num_of_entries; i++, entry++)
+		if (strcmp(entry->name, manifest_entry) == 0)
+			offset = entry->offset & GSC_CPD_ENTRY_OFFSET_MASK;
+
+	if (!offset) {
+		drm_err(&xe->drm, "%s late binding fw: Failed to find manifest_entry\n",
+			fw_id_to_name[lb_fw->id]);
+		return -ENODATA;
+	}
+
+	min_size = offset + sizeof(struct gsc_manifest_header);
+	if (size < min_size) {
+		drm_err(&xe->drm, "%s late binding fw: too small! %zu < %zu\n",
+			fw_id_to_name[lb_fw->id], size, min_size);
+		return -ENODATA;
+	}
+
+	manifest = data + offset;
+
+	lb_fw->version = manifest->fw_version;
+
+	return 0;
+}
+
+/* Refer to the "Late Bind based Firmware Layout" documentation entry for details */
+static int parse_lb_layout(struct xe_late_bind_fw *lb_fw,
+			   const void *data, size_t size, const char *fpt_entry)
+{
+	struct xe_device *xe = late_bind_fw_to_xe(lb_fw);
+	const struct csc_fpt_header *header = data;
+	const struct csc_fpt_entry *entry;
+	size_t min_size = sizeof(*header);
+	u32 offset;
+	int i;
+
+	/* fpt_entry is mandatory */
+	xe_assert(xe, fpt_entry);
+
+	if (size < min_size || header->header_marker != CSC_FPT_HEADER_MARKER)
+		return -ENOENT;
+
+	if (header->header_length < sizeof(struct csc_fpt_header)) {
+		drm_err(&xe->drm, "%s late binding fw: Invalid FPT header length %u!\n",
+			fw_id_to_name[lb_fw->id], header->header_length);
+		return -EINVAL;
+	}
+
+	min_size = header->header_length + sizeof(struct csc_fpt_entry) * header->num_of_entries;
+	if (size < min_size) {
+		drm_err(&xe->drm, "%s late binding fw: too small! %zu < %zu\n",
+			fw_id_to_name[lb_fw->id], size, min_size);
+		return -ENODATA;
+	}
+
+	/* Look for the cpd header first */
+	entry = (void *)header + header->header_length;
+	for (i = 0; i < header->num_of_entries; i++, entry++)
+		if (strcmp(entry->name, fpt_entry) == 0)
+			offset = entry->offset;
+
+	if (!offset) {
+		drm_err(&xe->drm, "%s late binding fw: Failed to find fpt_entry\n",
+			fw_id_to_name[lb_fw->id]);
+		return -ENODATA;
+	}
+
+	min_size = offset + sizeof(struct gsc_cpd_header_v2);
+	if (size < min_size) {
+		drm_err(&xe->drm, "%s late binding fw: too small! %zu < %zu\n",
+			fw_id_to_name[lb_fw->id], size, min_size);
+		return -ENODATA;
+	}
+
+	return parse_cpd_header(lb_fw, data + offset, size - offset, "LTES.man");
+}
+
+static const char *xe_late_bind_parse_status(uint32_t status)
+{
+	switch (status) {
+	case INTEL_LB_STATUS_SUCCESS:
+		return "success";
+	case INTEL_LB_STATUS_4ID_MISMATCH:
+		return "4Id Mismatch";
+	case INTEL_LB_STATUS_ARB_FAILURE:
+		return "ARB Failure";
+	case INTEL_LB_STATUS_GENERAL_ERROR:
+		return "General Error";
+	case INTEL_LB_STATUS_INVALID_PARAMS:
+		return "Invalid Params";
+	case INTEL_LB_STATUS_INVALID_SIGNATURE:
+		return "Invalid Signature";
+	case INTEL_LB_STATUS_INVALID_PAYLOAD:
+		return "Invalid Payload";
+	case INTEL_LB_STATUS_TIMEOUT:
+		return "Timeout";
+	default:
+		return "Unknown error";
+	}
+}
+
+static int xe_late_bind_fw_num_fans(struct xe_late_bind *late_bind)
+{
+	struct xe_device *xe = late_bind_to_xe(late_bind);
+	struct xe_tile *root_tile = xe_device_get_root_tile(xe);
+	u32 uval;
+
+	if (!xe_pcode_read(root_tile,
+			   PCODE_MBOX(FAN_SPEED_CONTROL, FSC_READ_NUM_FANS, 0), &uval, NULL))
+		return uval;
+	else
+		return 0;
+}
+
+void xe_late_bind_wait_for_worker_completion(struct xe_late_bind *late_bind)
+{
+	struct xe_device *xe = late_bind_to_xe(late_bind);
+	struct xe_late_bind_fw *lbfw;
+	int fw_id;
+
+	for (fw_id = 0; fw_id < XE_LB_FW_MAX_ID; fw_id++) {
+		lbfw = &late_bind->late_bind_fw[fw_id];
+		if (lbfw->payload && late_bind->wq) {
+			drm_dbg(&xe->drm, "Flush work: load %s firmware\n",
+				fw_id_to_name[lbfw->id]);
+			flush_work(&lbfw->work);
+		}
+	}
+}
+
+static void xe_late_bind_work(struct work_struct *work)
+{
+	struct xe_late_bind_fw *lbfw = container_of(work, struct xe_late_bind_fw, work);
+	struct xe_late_bind *late_bind = container_of(lbfw, struct xe_late_bind,
+						      late_bind_fw[lbfw->id]);
+	struct xe_device *xe = late_bind_to_xe(late_bind);
+	int retry = LB_FW_LOAD_RETRY_MAXCOUNT;
+	int ret;
+	int slept;
+
+	xe_device_assert_mem_access(xe);
+
+	/* we can queue this before the component is bound */
+	for (slept = 0; slept < LB_INIT_TIMEOUT_MS; slept += 100) {
+		if (late_bind->component.ops)
+			break;
+		msleep(100);
+	}
+
+	if (!late_bind->component.ops) {
+		drm_err(&xe->drm, "Late bind component not bound\n");
+		/* Do not re-attempt fw load */
+		drmm_kfree(&xe->drm, (void *)lbfw->payload);
+		lbfw->payload = NULL;
+		goto out;
+	}
+
+	drm_dbg(&xe->drm, "Load %s firmware\n", fw_id_to_name[lbfw->id]);
+
+	do {
+		ret = late_bind->component.ops->push_payload(late_bind->component.mei_dev,
+							     lbfw->type,
+							     lbfw->flags,
+							     lbfw->payload,
+							     lbfw->payload_size);
+		if (!ret)
+			break;
+		msleep(LB_FW_LOAD_RETRY_PAUSE_MS);
+	} while (--retry && ret == -EBUSY);
+
+	if (!ret) {
+		drm_dbg(&xe->drm, "Load %s firmware successful\n",
+			fw_id_to_name[lbfw->id]);
+		goto out;
+	}
+
+	if (ret > 0)
+		drm_err(&xe->drm, "Load %s firmware failed with err %d, %s\n",
+			fw_id_to_name[lbfw->id], ret, xe_late_bind_parse_status(ret));
+	else
+		drm_err(&xe->drm, "Load %s firmware failed with err %d",
+			fw_id_to_name[lbfw->id], ret);
+	/* Do not re-attempt fw load */
+	drmm_kfree(&xe->drm, (void *)lbfw->payload);
+	lbfw->payload = NULL;
+
+out:
+	xe_pm_runtime_put(xe);
+}
+
+int xe_late_bind_fw_load(struct xe_late_bind *late_bind)
+{
+	struct xe_device *xe = late_bind_to_xe(late_bind);
+	struct xe_late_bind_fw *lbfw;
+	int fw_id;
+
+	if (!late_bind->component_added)
+		return -ENODEV;
+
+	if (late_bind->disable)
+		return 0;
+
+	for (fw_id = 0; fw_id < XE_LB_FW_MAX_ID; fw_id++) {
+		lbfw = &late_bind->late_bind_fw[fw_id];
+		if (lbfw->payload) {
+			xe_pm_runtime_get_noresume(xe);
+			queue_work(late_bind->wq, &lbfw->work);
+		}
+	}
+	return 0;
+}
+
+static int __xe_late_bind_fw_init(struct xe_late_bind *late_bind, u32 fw_id)
+{
+	struct xe_device *xe = late_bind_to_xe(late_bind);
+	struct pci_dev *pdev = to_pci_dev(xe->drm.dev);
+	struct xe_late_bind_fw *lb_fw;
+	const struct firmware *fw;
+	u32 num_fans;
+	int ret;
+
+	if (fw_id >= XE_LB_FW_MAX_ID)
+		return -EINVAL;
+
+	lb_fw = &late_bind->late_bind_fw[fw_id];
+
+	lb_fw->id = fw_id;
+	lb_fw->type = fw_id_to_type[lb_fw->id];
+	lb_fw->flags &= ~INTEL_LB_FLAG_IS_PERSISTENT;
+
+	if (lb_fw->type == INTEL_LB_TYPE_FAN_CONTROL) {
+		num_fans = xe_late_bind_fw_num_fans(late_bind);
+		drm_dbg(&xe->drm, "Number of Fans: %d\n", num_fans);
+		if (!num_fans)
+			return 0;
+	}
+
+	snprintf(lb_fw->blob_path, sizeof(lb_fw->blob_path), "xe/%s_8086_%04x_%04x_%04x.bin",
+		 fw_id_to_name[lb_fw->id], pdev->device,
+		 pdev->subsystem_vendor, pdev->subsystem_device);
+
+	drm_dbg(&xe->drm, "Request late binding firmware %s\n", lb_fw->blob_path);
+	ret = firmware_request_nowarn(&fw, lb_fw->blob_path, xe->drm.dev);
+	if (ret) {
+		drm_dbg(&xe->drm, "%s late binding fw not available for current device",
+			fw_id_to_name[lb_fw->id]);
+		return 0;
+	}
+
+	if (fw->size > XE_LB_MAX_PAYLOAD_SIZE) {
+		drm_err(&xe->drm, "Firmware %s size %zu is larger than max pay load size %u\n",
+			lb_fw->blob_path, fw->size, XE_LB_MAX_PAYLOAD_SIZE);
+		release_firmware(fw);
+		return -ENODATA;
+	}
+
+	ret = parse_lb_layout(lb_fw, fw->data, fw->size, "LTES");
+	if (ret)
+		return ret;
+
+	lb_fw->payload_size = fw->size;
+	lb_fw->payload = drmm_kzalloc(&xe->drm, lb_fw->payload_size, GFP_KERNEL);
+	if (!lb_fw->payload) {
+		release_firmware(fw);
+		return -ENOMEM;
+	}
+
+	drm_info(&xe->drm, "Using %s firmware from %s version %u.%u.%u.%u\n",
+		 fw_id_to_name[lb_fw->id], lb_fw->blob_path,
+		 lb_fw->version.major, lb_fw->version.minor,
+		 lb_fw->version.hotfix, lb_fw->version.build);
+
+	memcpy((void *)lb_fw->payload, fw->data, lb_fw->payload_size);
+	release_firmware(fw);
+	INIT_WORK(&lb_fw->work, xe_late_bind_work);
+
+	return 0;
+}
+
+static int xe_late_bind_fw_init(struct xe_late_bind *late_bind)
+{
+	int ret;
+	int fw_id;
+
+	late_bind->wq = alloc_ordered_workqueue("late-bind-ordered-wq", 0);
+	if (!late_bind->wq)
+		return -ENOMEM;
+
+	for (fw_id = 0; fw_id < XE_LB_FW_MAX_ID; fw_id++) {
+		ret = __xe_late_bind_fw_init(late_bind, fw_id);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static int xe_late_bind_component_bind(struct device *xe_kdev,
+				       struct device *mei_kdev, void *data)
+{
+	struct xe_device *xe = kdev_to_xe_device(xe_kdev);
+	struct xe_late_bind *late_bind = &xe->late_bind;
+
+	late_bind->component.ops = data;
+	late_bind->component.mei_dev = mei_kdev;
+
+	return 0;
+}
+
+static void xe_late_bind_component_unbind(struct device *xe_kdev,
+					  struct device *mei_kdev, void *data)
+{
+	struct xe_device *xe = kdev_to_xe_device(xe_kdev);
+	struct xe_late_bind *late_bind = &xe->late_bind;
+
+	xe_late_bind_wait_for_worker_completion(late_bind);
+
+	late_bind->component.ops = NULL;
+}
+
+static const struct component_ops xe_late_bind_component_ops = {
+	.bind   = xe_late_bind_component_bind,
+	.unbind = xe_late_bind_component_unbind,
+};
+
+static void xe_late_bind_remove(void *arg)
+{
+	struct xe_late_bind *late_bind = arg;
+	struct xe_device *xe = late_bind_to_xe(late_bind);
+
+	xe_late_bind_wait_for_worker_completion(late_bind);
+
+	late_bind->component_added = false;
+
+	component_del(xe->drm.dev, &xe_late_bind_component_ops);
+	if (late_bind->wq) {
+		destroy_workqueue(late_bind->wq);
+		late_bind->wq = NULL;
+	}
+}
+
+/**
+ * xe_late_bind_init() - add xe mei late binding component
+ * @late_bind: pointer to late bind structure.
+ *
+ * Return: 0 if the initialization was successful, a negative errno otherwise.
+ */
+int xe_late_bind_init(struct xe_late_bind *late_bind)
+{
+	struct xe_device *xe = late_bind_to_xe(late_bind);
+	int err;
+
+	if (!xe->info.has_late_bind)
+		return 0;
+
+	if (!IS_ENABLED(CONFIG_INTEL_MEI_LB) || !IS_ENABLED(CONFIG_INTEL_MEI_GSC)) {
+		drm_info(&xe->drm, "Can't init xe mei late bind missing mei component\n");
+		return 0;
+	}
+
+	err = component_add_typed(xe->drm.dev, &xe_late_bind_component_ops,
+				  INTEL_COMPONENT_LB);
+	if (err < 0) {
+		drm_err(&xe->drm, "Failed to add mei late bind component (%pe)\n", ERR_PTR(err));
+		return err;
+	}
+
+	late_bind->component_added = true;
+
+	err = devm_add_action_or_reset(xe->drm.dev, xe_late_bind_remove, late_bind);
+	if (err)
+		return err;
+
+	err = xe_late_bind_fw_init(late_bind);
+	if (err)
+		return err;
+
+	return xe_late_bind_fw_load(late_bind);
+}

drivers/gpu/drm/xe/xe_late_bind_fw.h

···
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#ifndef _XE_LATE_BIND_FW_H_
+#define _XE_LATE_BIND_FW_H_
+
+#include <linux/types.h>
+
+struct xe_late_bind;
+
+int xe_late_bind_init(struct xe_late_bind *late_bind);
+int xe_late_bind_fw_load(struct xe_late_bind *late_bind);
+void xe_late_bind_wait_for_worker_completion(struct xe_late_bind *late_bind);
+
+#endif

drivers/gpu/drm/xe/xe_late_bind_fw_types.h

···
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#ifndef _XE_LATE_BIND_TYPES_H_
+#define _XE_LATE_BIND_TYPES_H_
+
+#include <linux/iosys-map.h>
+#include <linux/mutex.h>
+#include <linux/types.h>
+#include <linux/workqueue.h>
+#include "xe_uc_fw_abi.h"
+
+#define XE_LB_MAX_PAYLOAD_SIZE SZ_4K
+
+/**
+ * xe_late_bind_fw_id - enum to determine late binding fw index
+ */
+enum xe_late_bind_fw_id {
+	XE_LB_FW_FAN_CONTROL = 0,
+	XE_LB_FW_MAX_ID
+};
+
+/**
+ * struct xe_late_bind_fw
+ */
+struct xe_late_bind_fw {
+	/** @id: firmware index */
+	u32 id;
+	/** @blob_path: firmware binary path */
+	char blob_path[PATH_MAX];
+	/** @type: firmware type */
+	u32  type;
+	/** @flags: firmware flags */
+	u32  flags;
+	/** @payload: to store the late binding blob */
+	const u8  *payload;
+	/** @payload_size: late binding blob payload_size */
+	size_t payload_size;
+	/** @work: worker to upload latebind blob */
+	struct work_struct work;
+	/** @version: late binding blob manifest version */
+	struct gsc_version version;
+};
+
+/**
+ * struct xe_late_bind_component - Late Binding services component
+ * @mei_dev: device that provide Late Binding service.
+ * @ops: Ops implemented by Late Binding driver, used by Xe driver.
+ *
+ * Communication between Xe and MEI drivers for Late Binding services
+ */
+struct xe_late_bind_component {
+	struct device *mei_dev;
+	const struct intel_lb_component_ops *ops;
+};
+
+/**
+ * struct xe_late_bind
+ */
+struct xe_late_bind {
+	/** @component: struct for communication with mei component */
+	struct xe_late_bind_component component;
+	/** @late_bind_fw: late binding firmware array */
+	struct xe_late_bind_fw late_bind_fw[XE_LB_FW_MAX_ID];
+	/** @wq: workqueue to submit request to download late bind blob */
+	struct workqueue_struct *wq;
+	/** @component_added: whether the component has been added */
+	bool component_added;
+	/** @disable: to block late binding reload during pm resume flow*/
+	bool disable;
+};
+
+#endif

drivers/gpu/drm/xe/xe_lmtt.c

···
 		goto out;
 	}
 
-	bo = xe_bo_create_pin_map(lmtt_to_xe(lmtt), lmtt_to_tile(lmtt), NULL,
-				  PAGE_ALIGN(lmtt->ops->lmtt_pte_size(level) *
-					     lmtt->ops->lmtt_pte_num(level)),
-				  ttm_bo_type_kernel,
-				  XE_BO_FLAG_VRAM_IF_DGFX(lmtt_to_tile(lmtt)) |
-				  XE_BO_FLAG_NEEDS_64K);
+	bo = xe_bo_create_pin_map_novm(lmtt_to_xe(lmtt), lmtt_to_tile(lmtt),
+				       PAGE_ALIGN(lmtt->ops->lmtt_pte_size(level) *
+						  lmtt->ops->lmtt_pte_num(level)),
+				       ttm_bo_type_kernel,
+				       XE_BO_FLAG_VRAM_IF_DGFX(lmtt_to_tile(lmtt)) |
+				       XE_BO_FLAG_NEEDS_64K, false);
 	if (IS_ERR(bo)) {
 		err = PTR_ERR(bo);
 		goto out_free_pt;

drivers/gpu/drm/xe/xe_lrc.c

···
 #include <generated/xe_wa_oob.h>
 
 #include <linux/ascii85.h>
+#include <linux/panic.h>
 
 #include "instructions/xe_mi_commands.h"
 #include "instructions/xe_gfxpipe_commands.h"
···
 #include "regs/xe_lrc_layout.h"
 #include "xe_bb.h"
 #include "xe_bo.h"
+#include "xe_configfs.h"
 #include "xe_device.h"
 #include "xe_drm_client.h"
 #include "xe_exec_queue_types.h"
···
 static bool
 gt_engine_needs_indirect_ctx(struct xe_gt *gt, enum xe_engine_class class)
 {
+	struct xe_device *xe = gt_to_xe(gt);
+
 	if (XE_GT_WA(gt, 16010904313) &&
 	    (class == XE_ENGINE_CLASS_RENDER ||
 	     class == XE_ENGINE_CLASS_COMPUTE))
+		return true;
+
+	if (xe_configfs_get_ctx_restore_mid_bb(to_pci_dev(xe->drm.dev),
+					       class, NULL))
 		return true;
 
 	return false;
···
 	return cmd - batch;
 }
 
+static ssize_t setup_configfs_post_ctx_restore_bb(struct xe_lrc *lrc,
+						  struct xe_hw_engine *hwe,
+						  u32 *batch, size_t max_len)
+{
+	struct xe_device *xe = gt_to_xe(lrc->gt);
+	const u32 *user_batch;
+	u32 *cmd = batch;
+	u32 count;
+
+	count = xe_configfs_get_ctx_restore_post_bb(to_pci_dev(xe->drm.dev),
+						    hwe->class, &user_batch);
+	if (!count)
+		return 0;
+
+	if (count > max_len)
+		return -ENOSPC;
+
+	/*
+	 * This should be used only for tests and validation. Taint the kernel
+	 * as anything could be submitted directly in context switches
+	 */
+	add_taint(TAINT_TEST, LOCKDEP_STILL_OK);
+
+	memcpy(cmd, user_batch, count * sizeof(u32));
+	cmd += count;
+
+	return cmd - batch;
+}
+
+static ssize_t setup_configfs_mid_ctx_restore_bb(struct xe_lrc *lrc,
+						 struct xe_hw_engine *hwe,
+						 u32 *batch, size_t max_len)
+{
+	struct xe_device *xe = gt_to_xe(lrc->gt);
+	const u32 *user_batch;
+	u32 *cmd = batch;
+	u32 count;
+
+	count = xe_configfs_get_ctx_restore_mid_bb(to_pci_dev(xe->drm.dev),
+						   hwe->class, &user_batch);
+	if (!count)
+		return 0;
+
+	if (count > max_len)
+		return -ENOSPC;
+
+	/*
+	 * This should be used only for tests and validation. Taint the kernel
+	 * as anything could be submitted directly in context switches
+	 */
+	add_taint(TAINT_TEST, LOCKDEP_STILL_OK);
+
+	memcpy(cmd, user_batch, count * sizeof(u32));
+	cmd += count;
+
+	return cmd - batch;
+}
+
 static ssize_t setup_invalidate_state_cache_wa(struct xe_lrc *lrc,
 					       struct xe_hw_engine *hwe,
 					       u32 *batch, size_t max_len)
···
 		{ .setup = setup_timestamp_wa },
 		{ .setup = setup_invalidate_state_cache_wa },
 		{ .setup = setup_utilization_wa },
+		{ .setup = setup_configfs_post_ctx_restore_bb },
 	};
 	struct bo_setup_state state = {
 		.lrc = lrc,
···
 static int
 setup_indirect_ctx(struct xe_lrc *lrc, struct xe_hw_engine *hwe)
 {
-	static struct bo_setup rcs_funcs[] = {
+	static const struct bo_setup rcs_funcs[] = {
 		{ .setup = setup_timestamp_wa },
+		{ .setup = setup_configfs_mid_ctx_restore_bb },
+	};
+	static const struct bo_setup xcs_funcs[] = {
+		{ .setup = setup_configfs_mid_ctx_restore_bb },
 	};
 	struct bo_setup_state state = {
 		.lrc = lrc,
···
 	    hwe->class == XE_ENGINE_CLASS_COMPUTE) {
 		state.funcs = rcs_funcs;
 		state.num_funcs = ARRAY_SIZE(rcs_funcs);
+	} else {
+		state.funcs = xcs_funcs;
+		state.num_funcs = ARRAY_SIZE(xcs_funcs);
 	}
 
 	if (xe_gt_WARN_ON(lrc->gt, !state.funcs))
···
 	finish_bo(&state);
 	kfree(state.buffer);
 
+	/*
+	 * Enable INDIRECT_CTX leaving INDIRECT_CTX_OFFSET at its default: it
+	 * varies per engine class, but the default is good enough
+	 */
 	xe_lrc_write_ctx_reg(lrc,
 			     CTX_CS_INDIRECT_CTX,
 			     (xe_bo_ggtt_addr(lrc->bo) + state.offset) |
 			     /* Size in CLs. */
 			     (state.written * sizeof(u32) / 64));
-	xe_lrc_write_ctx_reg(lrc,
-			     CTX_CS_INDIRECT_CTX_OFFSET,
-			     CTX_INDIRECT_CTX_OFFSET_DEFAULT);
 
 	return 0;
 }
···
 	if (vm && vm->xef) /* userspace */
 		bo_flags |= XE_BO_FLAG_PINNED_LATE_RESTORE;
 
-	lrc->bo = xe_bo_create_pin_map(xe, tile, NULL, bo_size,
-				       ttm_bo_type_kernel,
-				       bo_flags);
+	lrc->bo = xe_bo_create_pin_map_novm(xe, tile,
+					    bo_size,
+					    ttm_bo_type_kernel,
+					    bo_flags, false);
 	if (IS_ERR(lrc->bo))
 		return PTR_ERR(lrc->bo);
 

drivers/gpu/drm/xe/xe_migrate.c

···
 #include "xe_sched_job.h"
 #include "xe_sync.h"
 #include "xe_trace_bo.h"
+#include "xe_validation.h"
 #include "xe_vm.h"
 #include "xe_vram.h"
 
···
 }
 
 static int xe_migrate_prepare_vm(struct xe_tile *tile, struct xe_migrate *m,
-				 struct xe_vm *vm)
+				 struct xe_vm *vm, struct drm_exec *exec)
 {
 	struct xe_device *xe = tile_to_xe(tile);
 	u16 pat_index = xe->pat.idx[XE_CACHE_WB];
···
 				  num_entries * XE_PAGE_SIZE,
 				  ttm_bo_type_kernel,
 				  XE_BO_FLAG_VRAM_IF_DGFX(tile) |
-				  XE_BO_FLAG_PAGETABLE);
+				  XE_BO_FLAG_PAGETABLE, exec);
 	if (IS_ERR(bo))
 		return PTR_ERR(bo);
 
···
 	return m;
 }
 
+static int xe_migrate_lock_prepare_vm(struct xe_tile *tile, struct xe_migrate *m, struct xe_vm *vm)
+{
+	struct xe_device *xe = tile_to_xe(tile);
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
+	int err = 0;
+
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {}, err) {
+		err = xe_vm_drm_exec_lock(vm, &exec);
+		drm_exec_retry_on_contention(&exec);
+		err = xe_migrate_prepare_vm(tile, m, vm, &exec);
+		drm_exec_retry_on_contention(&exec);
+		xe_validation_retry_on_oom(&ctx, &err);
+	}
+
+	return err;
+}
+
 /**
  * xe_migrate_init() - Initialize a migrate context
  * @m: The migration context
···
 	if (IS_ERR(vm))
 		return PTR_ERR(vm);
 
-	xe_vm_lock(vm, false);
-	err = xe_migrate_prepare_vm(tile, m, vm);
-	xe_vm_unlock(vm);
+	err = xe_migrate_lock_prepare_vm(tile, m, vm);
 	if (err)
-		goto err_out;
+		return err;
 
 	if (xe->info.has_usm) {
 		struct xe_hw_engine *hwe = xe_gt_hw_engine(primary_gt,
···
 		batch_size += pte_update_size(m, pte_flags, src, &src_it, &src_L0,
 					      &src_L0_ofs, &src_L0_pt, 0, 0,
 					      avail_pts);
-
-		pte_flags = dst_is_vram ? PTE_UPDATE_FLAG_IS_VRAM : 0;
-		batch_size += pte_update_size(m, pte_flags, dst, &dst_it, &src_L0,
-					      &dst_L0_ofs, &dst_L0_pt, 0,
-					      avail_pts, avail_pts);
+		if (copy_only_ccs) {
+			dst_L0_ofs = src_L0_ofs;
+		} else {
+			pte_flags = dst_is_vram ? PTE_UPDATE_FLAG_IS_VRAM : 0;
+			batch_size += pte_update_size(m, pte_flags, dst,
+						      &dst_it, &src_L0,
+						      &dst_L0_ofs, &dst_L0_pt,
+						      0, avail_pts, avail_pts);
+		}
 
 		if (copy_system_ccs) {
 			xe_assert(xe, type_device);
···
 
 		if (dst_is_vram && xe_migrate_allow_identity(src_L0, &dst_it))
 			xe_res_next(&dst_it, src_L0);
-		else
+		else if (!copy_only_ccs)
 			emit_pte(m, bb, dst_L0_pt, dst_is_vram, copy_system_ccs,
 				 &dst_it, src_L0, dst);
 
···
 		if (!fence) {
 			err = xe_sched_job_add_deps(job, src_bo->ttm.base.resv,
 						    DMA_RESV_USAGE_BOOKKEEP);
-			if (!err && src_bo != dst_bo)
+			if (!err && src_bo->ttm.base.resv != dst_bo->ttm.base.resv)
 				err = xe_sched_job_add_deps(job, dst_bo->ttm.base.resv,
 							    DMA_RESV_USAGE_BOOKKEEP);
 			if (err)

drivers/gpu/drm/xe/xe_nvm.c

···
 
 static void xe_nvm_release_dev(struct device *dev)
 {
+	struct auxiliary_device *aux = container_of(dev, struct auxiliary_device, dev);
+	struct intel_dg_nvm_dev *nvm = container_of(aux, struct intel_dg_nvm_dev, aux_dev);
+
+	kfree(nvm);
 }
 
 static bool xe_nvm_non_posted_erase(struct xe_device *xe)
···
 
 	auxiliary_device_delete(&nvm->aux_dev);
 	auxiliary_device_uninit(&nvm->aux_dev);
-	kfree(nvm);
 	xe->nvm = NULL;
 }

drivers/gpu/drm/xe/xe_oa.c

···
 {
 	struct xe_bo *bo;
 
-	bo = xe_bo_create_pin_map(stream->oa->xe, stream->gt->tile, NULL,
-				  size, ttm_bo_type_kernel,
-				  XE_BO_FLAG_SYSTEM | XE_BO_FLAG_GGTT);
+	bo = xe_bo_create_pin_map_novm(stream->oa->xe, stream->gt->tile,
+				       size, ttm_bo_type_kernel,
+				       XE_BO_FLAG_SYSTEM | XE_BO_FLAG_GGTT, false);
 	if (IS_ERR(bo))
 		return PTR_ERR(bo);
 

drivers/gpu/drm/xe/xe_pci.c

···
 	.has_mbx_power_limits = true,
 	.has_gsc_nvm = 1,
 	.has_heci_cscfi = 1,
+	.has_late_bind = true,
 	.has_sriov = true,
 	.max_gt_per_tile = 2,
 	.needs_scratch = true,
···
 	*revid = REG_FIELD_GET(GMD_ID_REVID, val);
 }
 
+static const struct xe_ip *find_graphics_ip(unsigned int verx100)
+{
+	KUNIT_STATIC_STUB_REDIRECT(find_graphics_ip, verx100);
+
+	for (int i = 0; i < ARRAY_SIZE(graphics_ips); i++)
+		if (graphics_ips[i].verx100 == verx100)
+			return &graphics_ips[i];
+	return NULL;
+}
+
+static const struct xe_ip *find_media_ip(unsigned int verx100)
+{
+	KUNIT_STATIC_STUB_REDIRECT(find_media_ip, verx100);
+
+	for (int i = 0; i < ARRAY_SIZE(media_ips); i++)
+		if (media_ips[i].verx100 == verx100)
+			return &media_ips[i];
+	return NULL;
+}
+
 /*
  * Read IP version from hardware and select graphics/media IP descriptors
  * based on the result.
···
 
 	read_gmdid(xe, GMDID_GRAPHICS, &ver, graphics_revid);
 
-	for (int i = 0; i < ARRAY_SIZE(graphics_ips); i++) {
-		if (ver == graphics_ips[i].verx100) {
-			*graphics_ip = &graphics_ips[i];
-
-			break;
-		}
-	}
-
+	*graphics_ip = find_graphics_ip(ver);
 	if (!*graphics_ip) {
 		drm_err(&xe->drm, "Hardware reports unknown graphics version %u.%02u\n",
 			ver / 100, ver % 100);
···
 	if (ver == 0)
 		return;
 
-	for (int i = 0; i < ARRAY_SIZE(media_ips); i++) {
-		if (ver == media_ips[i].verx100) {
-			*media_ip = &media_ips[i];
-
-			break;
-		}
-	}
-
+	*media_ip = find_media_ip(ver);
 	if (!*media_ip) {
 		drm_err(&xe->drm, "Hardware reports unknown media version %u.%02u\n",
 			ver / 100, ver % 100);
···
 	xe->info.has_gsc_nvm = desc->has_gsc_nvm;
 	xe->info.has_heci_gscfi = desc->has_heci_gscfi;
 	xe->info.has_heci_cscfi = desc->has_heci_cscfi;
+	xe->info.has_late_bind = desc->has_late_bind;
 	xe->info.has_llc = desc->has_llc;
 	xe->info.has_pxp = desc->has_pxp;
 	xe->info.has_sriov = desc->has_sriov;

drivers/gpu/drm/xe/xe_pci_types.h

···
 	u8 has_gsc_nvm:1;
 	u8 has_heci_gscfi:1;
 	u8 has_heci_cscfi:1;
+	u8 has_late_bind:1;
 	u8 has_llc:1;
 	u8 has_mbx_power_limits:1;
 	u8 has_pxp:1;

drivers/gpu/drm/xe/xe_pm.c

···
 #include "xe_gt_idle.h"
 #include "xe_i2c.h"
 #include "xe_irq.h"
+#include "xe_late_bind_fw.h"
 #include "xe_pcode.h"
 #include "xe_pxp.h"
 #include "xe_sriov_vf_ccs.h"
···
 	if (err)
 		goto err;
 
+	xe_late_bind_wait_for_worker_completion(&xe->late_bind);
+
 	for_each_gt(gt, xe, id)
 		xe_gt_suspend_prepare(gt);
 
···
 
 	xe_pxp_pm_resume(xe->pxp);
 
-	if (IS_SRIOV_VF(xe))
+	if (IS_VF_CCS_READY(xe))
 		xe_sriov_vf_ccs_register_context(xe);
+
+	xe_late_bind_fw_load(&xe->late_bind);
 
 	drm_dbg(&xe->drm, "Device resumed\n");
 	return 0;
···
 
 	xe_pxp_pm_resume(xe->pxp);
 
-	if (IS_SRIOV_VF(xe))
+	if (IS_VF_CCS_READY(xe))
 		xe_sriov_vf_ccs_register_context(xe);
+
+	if (xe->d3cold.allowed)
+		xe_late_bind_fw_load(&xe->late_bind);
 
 out:
 	xe_rpm_lockmap_release(xe);

drivers/gpu/drm/xe/xe_printk.h

···
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#ifndef _XE_PRINTK_H_
+#define _XE_PRINTK_H_
+
+#include <drm/drm_print.h>
+
+#include "xe_device_types.h"
+
+#define __XE_PRINTK_FMT(_xe, _fmt, _args...)	_fmt, ##_args
+
+#define xe_printk(_xe, _level, _fmt, ...) \
+	drm_##_level(&(_xe)->drm, __XE_PRINTK_FMT((_xe), _fmt, ## __VA_ARGS__))
+
+#define xe_err(_xe, _fmt, ...) \
+	xe_printk((_xe), err, _fmt, ##__VA_ARGS__)
+
+#define xe_err_once(_xe, _fmt, ...) \
+	xe_printk((_xe), err_once, _fmt, ##__VA_ARGS__)
+
+#define xe_err_ratelimited(_xe, _fmt, ...) \
+	xe_printk((_xe), err_ratelimited, _fmt, ##__VA_ARGS__)
+
+#define xe_warn(_xe, _fmt, ...) \
+	xe_printk((_xe), warn, _fmt, ##__VA_ARGS__)
+
+#define xe_notice(_xe, _fmt, ...) \
+	xe_printk((_xe), notice, _fmt, ##__VA_ARGS__)
+
+#define xe_info(_xe, _fmt, ...) \
+	xe_printk((_xe), info, _fmt, ##__VA_ARGS__)
+
+#define xe_dbg(_xe, _fmt, ...) \
+	xe_printk((_xe), dbg, _fmt, ##__VA_ARGS__)
+
+#define xe_WARN_type(_xe, _type, _condition, _fmt, ...) \
+	drm_WARN##_type(&(_xe)->drm, _condition, _fmt, ## __VA_ARGS__)
+
+#define xe_WARN(_xe, _condition, _fmt, ...) \
+	xe_WARN_type((_xe),, _condition, __XE_PRINTK_FMT((_xe), _fmt, ## __VA_ARGS__))
+
+#define xe_WARN_ONCE(_xe, _condition, _fmt, ...) \
+	xe_WARN_type((_xe), _ONCE, _condition, __XE_PRINTK_FMT((_xe), _fmt, ## __VA_ARGS__))
+
+#define xe_WARN_ON(_xe, _condition) \
+	xe_WARN((_xe), _condition, "%s(%s)", "WARN_ON", __stringify(_condition))
+
+#define xe_WARN_ON_ONCE(_xe, _condition) \
+	xe_WARN_ONCE((_xe), _condition, "%s(%s)", "WARN_ON_ONCE", __stringify(_condition))
+
+static inline void __xe_printfn_err(struct drm_printer *p, struct va_format *vaf)
+{
+	struct xe_device *xe = p->arg;
+
+	xe_err(xe, "%pV", vaf);
+}
+
+static inline void __xe_printfn_info(struct drm_printer *p, struct va_format *vaf)
+{
+	struct xe_device *xe = p->arg;
+
+	xe_info(xe, "%pV", vaf);
+}
+
+static inline void __xe_printfn_dbg(struct drm_printer *p, struct va_format *vaf)
+{
+	struct xe_device *xe = p->arg;
+	struct drm_printer ddp;
+
+	/*
+	 * The original xe_dbg() callsite annotations are useless here,
+	 * redirect to the tweaked drm_dbg_printer() instead.
+	 */
+	ddp = drm_dbg_printer(&xe->drm, DRM_UT_DRIVER, NULL);
+	ddp.origin = p->origin;
+
+	drm_printf(&ddp, __XE_PRINTK_FMT(xe, "%pV", vaf));
+}
+
+/**
+ * xe_err_printer - Construct a &drm_printer that outputs to xe_err()
+ * @xe: the &xe_device pointer to use in xe_err()
+ *
+ * Return: The &drm_printer object.
+ */
+static inline struct drm_printer xe_err_printer(struct xe_device *xe)
+{
+	struct drm_printer p = {
+		.printfn = __xe_printfn_err,
+		.arg = xe,
+	};
+	return p;
+}
+
+/**
+ * xe_info_printer - Construct a &drm_printer that outputs to xe_info()
+ * @xe: the &xe_device pointer to use in xe_info()
+ *
+ * Return: The &drm_printer object.
+ */
+static inline struct drm_printer xe_info_printer(struct xe_device *xe)
+{
+	struct drm_printer p = {
+		.printfn = __xe_printfn_info,
+		.arg = xe,
+	};
+	return p;
+}
+
+/**
+ * xe_dbg_printer - Construct a &drm_printer that outputs like xe_dbg()
+ * @xe: the &xe_device pointer to use in xe_dbg()
+ *
+ * Return: The &drm_printer object.
+ */
+static inline struct drm_printer xe_dbg_printer(struct xe_device *xe)
+{
+	struct drm_printer p = {
+		.printfn = __xe_printfn_dbg,
+		.arg = xe,
+		.origin = (const void *)_THIS_IP_,
+	};
+	return p;
+}
+
+#endif

drivers/gpu/drm/xe/xe_psmi.c

···
 static struct xe_bo *psmi_alloc_object(struct xe_device *xe,
 				       unsigned int id, size_t bo_size)
 {
-	struct xe_bo *bo = NULL;
 	struct xe_tile *tile;
-	int err;
 
 	if (!id || !bo_size)
 		return NULL;
···
 	tile = &xe->tiles[id - 1];
 
 	/* VRAM: Allocate GEM object for the capture buffer */
-	bo = xe_bo_create_locked(xe, tile, NULL, bo_size,
-				 ttm_bo_type_kernel,
-				 XE_BO_FLAG_VRAM_IF_DGFX(tile) |
-				 XE_BO_FLAG_PINNED |
-				 XE_BO_FLAG_PINNED_LATE_RESTORE |
-				 XE_BO_FLAG_NEEDS_CPU_ACCESS);
-
-	if (!IS_ERR(bo)) {
-		/* Buffer written by HW, ensure stays resident */
-		err = xe_bo_pin(bo);
-		if (err)
-			bo = ERR_PTR(err);
-		xe_bo_unlock(bo);
-	}
-
-	return bo;
+	return xe_bo_create_pin_range_novm(xe, tile, bo_size, 0, ~0ull,
+					   ttm_bo_type_kernel,
+					   XE_BO_FLAG_VRAM_IF_DGFX(tile) |
+					   XE_BO_FLAG_PINNED |
+					   XE_BO_FLAG_PINNED_LATE_RESTORE |
+					   XE_BO_FLAG_NEEDS_CPU_ACCESS);
 }
 
 /*

drivers/gpu/drm/xe/xe_pt.c

···
 #include "xe_drm_client.h"
 #include "xe_exec_queue.h"
 #include "xe_gt.h"
-#include "xe_tlb_inval_job.h"
 #include "xe_migrate.h"
 #include "xe_pt_types.h"
 #include "xe_pt_walk.h"
 #include "xe_res_cursor.h"
 #include "xe_sched_job.h"
-#include "xe_sync.h"
 #include "xe_svm.h"
+#include "xe_sync.h"
 #include "xe_tlb_inval_job.h"
 #include "xe_trace.h"
 #include "xe_ttm_stolen_mgr.h"
+#include "xe_userptr.h"
 #include "xe_vm.h"
 
 struct xe_pt_dir {
···
  * @vm: The vm to create for.
  * @tile: The tile to create for.
  * @level: The page-table level.
+ * @exec: The drm_exec object used to lock the vm.
  *
  * Allocate and initialize a single struct xe_pt metadata structure. Also
  * create the corresponding page-table bo, but don't initialize it. If the
···
  * error.
  */
 struct xe_pt *xe_pt_create(struct xe_vm *vm, struct xe_tile *tile,
-			   unsigned int level)
+			   unsigned int level, struct drm_exec *exec)
 {
 	struct xe_pt *pt;
 	struct xe_bo *bo;
···
 		bo_flags |= XE_BO_FLAG_PINNED_LATE_RESTORE;
 
 	pt->level = level;
+
+	drm_WARN_ON(&vm->xe->drm, IS_ERR_OR_NULL(exec));
 	bo = xe_bo_create_pin_map(vm->xe, tile, vm, SZ_4K,
 				  ttm_bo_type_kernel,
-				  bo_flags);
+				  bo_flags, exec);
 	if (IS_ERR(bo)) {
 		err = PTR_ERR(bo);
 		goto err_kfree;
···
 	if (covers || !*child) {
 		u64 flags = 0;
 
-		xe_child = xe_pt_create(xe_walk->vm, xe_walk->tile, level - 1);
+		xe_child = xe_pt_create(xe_walk->vm, xe_walk->tile, level - 1,
+					xe_vm_validation_exec(vm));
 		if (IS_ERR(xe_child))
 			return PTR_ERR(xe_child);
 
···
 			return -EAGAIN;
 		}
 		if (xe_svm_range_has_dma_mapping(range)) {
-			xe_res_first_dma(range->base.dma_addr, 0,
+			xe_res_first_dma(range->base.pages.dma_addr, 0,
 					 range->base.itree.last + 1 - range->base.itree.start,
 					 &curs);
 			xe_svm_range_debug(range, "BIND PREPARE - MIXED");
···
 
 	if (!xe_vma_is_null(vma) && !range) {
 		if (xe_vma_is_userptr(vma))
-			xe_res_first_sg(to_userptr_vma(vma)->userptr.sg, 0,
-					xe_vma_size(vma), &curs);
+			xe_res_first_dma(to_userptr_vma(vma)->userptr.pages.dma_addr, 0,
+					 xe_vma_size(vma), &curs);
 		else if (xe_bo_is_vram(bo) || xe_bo_is_stolen(bo))
 			xe_res_first(bo->ttm.resource, xe_vma_bo_offset(vma),
 				     xe_vma_size(vma), &curs);
···
 	if (xe_vma_bo(vma))
 		xe_bo_assert_held(xe_vma_bo(vma));
 	else if (xe_vma_is_userptr(vma))
-		lockdep_assert_held(&xe_vma_vm(vma)->userptr.notifier_lock);
+		lockdep_assert_held(&xe_vma_vm(vma)->svm.gpusvm.notifier_lock);
 
 	if (!(pt_mask & BIT(tile->id)))
 		return false;
···
 	xe_pt_commit_prepare_locks_assert(vma);
 
 	if (xe_vma_is_userptr(vma))
-		lockdep_assert_held_read(&vm->userptr.notifier_lock);
+		xe_svm_assert_held_read(vm);
 }
 
 static void xe_pt_commit(struct xe_vma *vma,
···
 				     pt_update_ops, rftree);
 }
 
+#if IS_ENABLED(CONFIG_DRM_GPUSVM)
 #ifdef CONFIG_DRM_XE_USERPTR_INVAL_INJECT
 
 static bool xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
···
 	struct xe_userptr_vma *uvma;
 	unsigned long notifier_seq;
 
-	lockdep_assert_held_read(&vm->userptr.notifier_lock);
+	xe_svm_assert_held_read(vm);
 
 	if (!xe_vma_is_userptr(vma))
 		return 0;
···
 	if (xe_pt_userptr_inject_eagain(uvma))
 		xe_vma_userptr_force_invalidate(uvma);
 
-	notifier_seq = uvma->userptr.notifier_seq;
+	notifier_seq = uvma->userptr.pages.notifier_seq;
 
 	if (!mmu_interval_read_retry(&uvma->userptr.notifier,
 				     notifier_seq))
···
 	return 0;
 }
 
-static int op_check_userptr(struct xe_vm *vm, struct xe_vma_op *op,
-			    struct xe_vm_pgtable_update_ops *pt_update)
+static int op_check_svm_userptr(struct xe_vm *vm, struct xe_vma_op *op,
+				struct xe_vm_pgtable_update_ops *pt_update)
 {
 	int err = 0;
 
-	lockdep_assert_held_read(&vm->userptr.notifier_lock);
+	xe_svm_assert_held_read(vm);
 
 	switch (op->base.op) {
 	case DRM_GPUVA_OP_MAP:
···
 	case DRM_GPUVA_OP_UNMAP:
 		break;
 	case DRM_GPUVA_OP_PREFETCH:
-		err = vma_check_userptr(vm, gpuva_to_vma(op->base.prefetch.va),
-					pt_update);
+		if (xe_vma_is_cpu_addr_mirror(gpuva_to_vma(op->base.prefetch.va))) {
+			struct xe_svm_range *range = op->map_range.range;
+			unsigned long i;
+
+			xe_assert(vm->xe,
+				  xe_vma_is_cpu_addr_mirror(gpuva_to_vma(op->base.prefetch.va)));
+			xa_for_each(&op->prefetch_range.range, i, range) {
+				xe_svm_range_debug(range, "PRE-COMMIT");
+
+				if (!xe_svm_range_pages_valid(range)) {
+					xe_svm_range_debug(range, "PRE-COMMIT - RETRY");
+					return -ENODATA;
+				}
+			}
+		} else {
+			err = vma_check_userptr(vm, gpuva_to_vma(op->base.prefetch.va), pt_update);
+		}
 		break;
+#if IS_ENABLED(CONFIG_DRM_XE_GPUSVM)
+	case DRM_GPUVA_OP_DRIVER:
+		if (op->subop == XE_VMA_SUBOP_MAP_RANGE) {
+			struct xe_svm_range *range = op->map_range.range;
+
+			xe_assert(vm->xe, xe_vma_is_cpu_addr_mirror(op->map_range.vma));
+
+			xe_svm_range_debug(range, "PRE-COMMIT");
+
+			if (!xe_svm_range_pages_valid(range)) {
+				xe_svm_range_debug(range, "PRE-COMMIT - RETRY");
+				return -EAGAIN;
+			}
+		}
+		break;
+#endif
 	default:
 		drm_warn(&vm->xe->drm, "NOT POSSIBLE");
 	}
···
 	return err;
 }
 
-static int xe_pt_userptr_pre_commit(struct xe_migrate_pt_update *pt_update)
+static int xe_pt_svm_userptr_pre_commit(struct xe_migrate_pt_update *pt_update)
 {
 	struct xe_vm *vm = pt_update->vops->vm;
 	struct xe_vma_ops *vops = pt_update->vops;
···
 	if (err)
 		return err;
 
-	down_read(&vm->userptr.notifier_lock);
+	xe_svm_notifier_lock(vm);
 
 	list_for_each_entry(op, &vops->list, link) {
-		err = op_check_userptr(vm, op, pt_update_ops);
+		err = op_check_svm_userptr(vm, op, pt_update_ops);
 		if (err) {
-			up_read(&vm->userptr.notifier_lock);
+			xe_svm_notifier_unlock(vm);
 			break;
 		}
 	}
 
 	return err;
-}
-
-#if IS_ENABLED(CONFIG_DRM_XE_GPUSVM)
-static int xe_pt_svm_pre_commit(struct xe_migrate_pt_update *pt_update)
-{
-	struct xe_vm *vm = pt_update->vops->vm;
-	struct xe_vma_ops *vops = pt_update->vops;
-	struct xe_vma_op *op;
-	unsigned long i;
-	int err;
-
-	err = xe_pt_pre_commit(pt_update);
-	if (err)
-		return err;
-
-	xe_svm_notifier_lock(vm);
-
-	list_for_each_entry(op, &vops->list, link) {
-		struct xe_svm_range *range = NULL;
-
-		if (op->subop == XE_VMA_SUBOP_UNMAP_RANGE)
-			continue;
-
-		if (op->base.op == DRM_GPUVA_OP_PREFETCH) {
-			xe_assert(vm->xe,
-				  xe_vma_is_cpu_addr_mirror(gpuva_to_vma(op->base.prefetch.va)));
-			xa_for_each(&op->prefetch_range.range, i, range) {
-				xe_svm_range_debug(range, "PRE-COMMIT");
-
-				if (!xe_svm_range_pages_valid(range)) {
-					xe_svm_range_debug(range, "PRE-COMMIT - RETRY");
-					xe_svm_notifier_unlock(vm);
-					return -ENODATA;
-				}
-			}
-		} else {
-			xe_assert(vm->xe, xe_vma_is_cpu_addr_mirror(op->map_range.vma));
-			xe_assert(vm->xe, op->subop == XE_VMA_SUBOP_MAP_RANGE);
-			range = op->map_range.range;
-
-			xe_svm_range_debug(range, "PRE-COMMIT");
-
-			if (!xe_svm_range_pages_valid(range)) {
-				xe_svm_range_debug(range, "PRE-COMMIT - RETRY");
-				xe_svm_notifier_unlock(vm);
-				return -EAGAIN;
-			}
-		}
-	}
-
-	return 0;
 }
 #endif
 
···
 						 xe_vma_start(vma),
 						 xe_vma_end(vma));
 		++pt_update_ops->current_op;
-		pt_update_ops->needs_userptr_lock |= xe_vma_is_userptr(vma);
+		pt_update_ops->needs_svm_lock |= xe_vma_is_userptr(vma);
 
 		/*
 		 * If rebind, we have to invalidate TLB on !LR vms to invalidate
···
 	xe_pt_update_ops_rfence_interval(pt_update_ops, xe_vma_start(vma),
 					 xe_vma_end(vma));
 	++pt_update_ops->current_op;
-	pt_update_ops->needs_userptr_lock |= xe_vma_is_userptr(vma);
+	pt_update_ops->needs_svm_lock |= xe_vma_is_userptr(vma);
 	pt_update_ops->needs_invalidation = true;
 
 	xe_pt_commit_prepare_unbind(vma, pt_op->entries, pt_op->num_entries);
···
 			   vma->tile_invalidated & ~BIT(tile->id));
 	vma->tile_staged &= ~BIT(tile->id);
 	if (xe_vma_is_userptr(vma)) {
-		lockdep_assert_held_read(&vm->userptr.notifier_lock);
+		xe_svm_assert_held_read(vm);
 		to_userptr_vma(vma)->userptr.initial_bind = true;
 	}
 
···
 	if (!vma->tile_present) {
 		list_del_init(&vma->combined_links.rebind);
 		if (xe_vma_is_userptr(vma)) {
-			lockdep_assert_held_read(&vm->userptr.notifier_lock);
+			xe_svm_assert_held_read(vm);
 
 			spin_lock(&vm->userptr.invalidated_lock);
 			list_del_init(&to_userptr_vma(vma)->userptr.invalidate_link);
···
 	.pre_commit = xe_pt_pre_commit,
 };
 
-static const struct xe_migrate_pt_update_ops userptr_migrate_ops = {
+#if IS_ENABLED(CONFIG_DRM_GPUSVM)
+static const struct xe_migrate_pt_update_ops svm_userptr_migrate_ops = {
 	.populate = xe_vm_populate_pgtable,
 	.clear = xe_migrate_clear_pgtable_callback,
-	.pre_commit = xe_pt_userptr_pre_commit,
-};
-
-#if IS_ENABLED(CONFIG_DRM_XE_GPUSVM)
-static const struct xe_migrate_pt_update_ops svm_migrate_ops = {
-	.populate = xe_vm_populate_pgtable,
-	.clear = xe_migrate_clear_pgtable_callback,
-	.pre_commit = xe_pt_svm_pre_commit,
+	.pre_commit = xe_pt_svm_userptr_pre_commit,
 };
 #else
-static const struct xe_migrate_pt_update_ops svm_migrate_ops;
+static const struct xe_migrate_pt_update_ops svm_userptr_migrate_ops;
 #endif
 
 static struct xe_dep_scheduler *to_dep_scheduler(struct xe_exec_queue *q,
···
 	int err = 0, i;
 	struct xe_migrate_pt_update update = {
 		.ops = pt_update_ops->needs_svm_lock ?
-			&svm_migrate_ops :
-			pt_update_ops->needs_userptr_lock ?
-			&userptr_migrate_ops :
+			&svm_userptr_migrate_ops :
 			&migrate_ops,
 		.vops = vops,
 		.tile_id = tile->id,
···
 
 	if (pt_update_ops->needs_svm_lock)
 		xe_svm_notifier_unlock(vm);
-	if (pt_update_ops->needs_userptr_lock)
-		up_read(&vm->userptr.notifier_lock);
 
 	xe_tlb_inval_job_put(mjob);
 	xe_tlb_inval_job_put(ijob);

drivers/gpu/drm/xe/xe_pt.h

···
 #include "xe_pt_types.h"
 
 struct dma_fence;
+struct drm_exec;
 struct xe_bo;
 struct xe_device;
 struct xe_exec_queue;
···
 unsigned int xe_pt_shift(unsigned int level);
 
 struct xe_pt *xe_pt_create(struct xe_vm *vm, struct xe_tile *tile,
-			   unsigned int level);
+			   unsigned int level, struct drm_exec *exec);
 
 void xe_pt_populate_empty(struct xe_tile *tile, struct xe_vm *vm,
 			  struct xe_pt *pt);

drivers/gpu/drm/xe/xe_pt_types.h

···
 	u32 current_op;
 	/** @needs_svm_lock: Needs SVM lock */
 	bool needs_svm_lock;
-	/** @needs_userptr_lock: Needs userptr lock */
-	bool needs_userptr_lock;
 	/** @needs_invalidation: Needs invalidation */
 	bool needs_invalidation;
 	/**

drivers/gpu/drm/xe/xe_pxp.c

···
 
 	return ret;
 }
+ALLOW_ERROR_INJECTION(xe_pxp_exec_queue_add, ERRNO);
 
 static void __pxp_exec_queue_remove(struct xe_pxp *pxp, struct xe_exec_queue *q, bool lock)
 {

drivers/gpu/drm/xe/xe_pxp_submit.c

···
 	 * Each termination is 16 DWORDS, so 4K is enough to contain a
 	 * termination for each sessions.
 	 */
-	bo = xe_bo_create_pin_map(xe, tile, NULL, SZ_4K, ttm_bo_type_kernel,
-				  XE_BO_FLAG_SYSTEM | XE_BO_FLAG_PINNED | XE_BO_FLAG_GGTT);
+	bo = xe_bo_create_pin_map_novm(xe, tile, SZ_4K, ttm_bo_type_kernel,
+				       XE_BO_FLAG_SYSTEM | XE_BO_FLAG_PINNED | XE_BO_FLAG_GGTT,
+				       false);
 	if (IS_ERR(bo)) {
 		err = PTR_ERR(bo);
 		goto out_queue;
···
 {
 	struct xe_tile *tile = gt_to_tile(gt);
 	struct xe_device *xe = tile_to_xe(tile);
+	struct xe_validation_ctx ctx;
 	struct xe_hw_engine *hwe;
+	struct drm_exec exec;
 	struct xe_vm *vm;
 	struct xe_bo *bo;
 	struct xe_exec_queue *q;
···
 		return PTR_ERR(vm);
 
 	/* We allocate a single object for the batch and the in/out memory */
-	xe_vm_lock(vm, false);
-	bo = xe_bo_create_pin_map(xe, tile, vm, PXP_BB_SIZE + inout_size * 2,
-				  ttm_bo_type_kernel,
-				  XE_BO_FLAG_SYSTEM | XE_BO_FLAG_PINNED | XE_BO_FLAG_NEEDS_UC);
-	xe_vm_unlock(vm);
-	if (IS_ERR(bo)) {
-		err = PTR_ERR(bo);
-		goto vm_out;
+
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags){}, err) {
+		err = xe_vm_drm_exec_lock(vm, &exec);
+		drm_exec_retry_on_contention(&exec);
+		if (err)
+			break;
+
+		bo = xe_bo_create_pin_map(xe, tile, vm, PXP_BB_SIZE + inout_size * 2,
+					  ttm_bo_type_kernel,
+					  XE_BO_FLAG_SYSTEM | XE_BO_FLAG_PINNED |
+					  XE_BO_FLAG_NEEDS_UC, &exec);
+		drm_exec_retry_on_contention(&exec);
+		if (IS_ERR(bo)) {
+			err = PTR_ERR(bo);
+			xe_validation_retry_on_oom(&ctx, &err);
+			break;
+		}
 	}
+	if (err)
+		goto vm_out;
 
 	fence = xe_vm_bind_kernel_bo(vm, bo, NULL, 0, XE_CACHE_WB);
 	if (IS_ERR(fence)) {

drivers/gpu/drm/xe/xe_query.c

···
 #include "xe_force_wake.h"
 #include "xe_ggtt.h"
 #include "xe_gt.h"
+#include "xe_gt_topology.h"
 #include "xe_guc_hwconfig.h"
 #include "xe_macros.h"
 #include "xe_mmio.h"
···
 			sizeof_field(struct xe_gt, fuse_topo.eu_mask_per_dss);
 
 		/* L3bank mask may not be available for some GTs */
-		if (!XE_GT_WA(gt, no_media_l3))
+		if (xe_gt_topology_report_l3(gt))
 			query_size += sizeof(struct drm_xe_query_topology_mask) +
 				sizeof_field(struct xe_gt, fuse_topo.l3_bank_mask);
 	}
···
 		 * mask, then it's better to omit L3 from the query rather than
 		 * reporting bogus or zeroed information to userspace.
 		 */
-		if (!XE_GT_WA(gt, no_media_l3)) {
+		if (xe_gt_topology_report_l3(gt)) {
 			topo.type = DRM_XE_TOPO_L3_BANK;
 			err = copy_mask(&query_ptr, &topo, gt->fuse_topo.l3_bank_mask,
 					sizeof(gt->fuse_topo.l3_bank_mask));

drivers/gpu/drm/xe/xe_rtp.c

···
 {
 	return xe_configfs_get_psmi_enabled(to_pci_dev(gt_to_xe(gt)->drm.dev));
 }
+
+bool xe_rtp_match_gt_has_discontiguous_dss_groups(const struct xe_gt *gt,
+						  const struct xe_hw_engine *hwe)
+{
+	return xe_gt_has_discontiguous_dss_groups(gt);
+}

drivers/gpu/drm/xe/xe_rtp.h

···
 bool xe_rtp_match_psmi_enabled(const struct xe_gt *gt,
 			       const struct xe_hw_engine *hwe);
 
+bool xe_rtp_match_gt_has_discontiguous_dss_groups(const struct xe_gt *gt,
+						  const struct xe_hw_engine *hwe);
+
 #endif

drivers/gpu/drm/xe/xe_sriov.c

···
 }
 
 /**
- * xe_sriov_late_init() - SR-IOV late initialization functions.
+ * xe_sriov_init_late() - SR-IOV late initialization functions.
  * @xe: the &xe_device to initialize
- *
- * On VF this function will initialize code for CCS migration.
  *
  * Return: 0 on success or a negative error code on failure.
  */
-int xe_sriov_late_init(struct xe_device *xe)
+int xe_sriov_init_late(struct xe_device *xe)
 {
-	int err = 0;
+	if (IS_SRIOV_VF(xe))
+		return xe_sriov_vf_init_late(xe);
 
-	if (IS_VF_CCS_INIT_NEEDED(xe))
-		err = xe_sriov_vf_ccs_init(xe);
-
-	return err;
+	return 0;
 }

drivers/gpu/drm/xe/xe_sriov.h

···
 void xe_sriov_probe_early(struct xe_device *xe);
 void xe_sriov_print_info(struct xe_device *xe, struct drm_printer *p);
 int xe_sriov_init(struct xe_device *xe);
-int xe_sriov_late_init(struct xe_device *xe);
+int xe_sriov_init_late(struct xe_device *xe);
 
 static inline enum xe_sriov_mode xe_device_sriov_mode(const struct xe_device *xe)
 {

drivers/gpu/drm/xe/xe_sriov_vf.c

···
  * Copyright © 2023-2024 Intel Corporation
  */
 
+#include <drm/drm_debugfs.h>
 #include <drm/drm_managed.h>
 
 #include "xe_assert.h"
···
 #include "xe_gt.h"
 #include "xe_gt_sriov_printk.h"
 #include "xe_gt_sriov_vf.h"
+#include "xe_guc.h"
 #include "xe_guc_ct.h"
 #include "xe_guc_submit.h"
 #include "xe_irq.h"
···
 #include "xe_sriov.h"
 #include "xe_sriov_printk.h"
 #include "xe_sriov_vf.h"
+#include "xe_sriov_vf_ccs.h"
 #include "xe_tile_sriov_vf.h"
 
 /**
···
  *      |                               |                               |
  */
 
-static bool vf_migration_supported(struct xe_device *xe)
+/**
+ * xe_sriov_vf_migration_supported - Report whether SR-IOV VF migration is
+ * supported or not.
+ * @xe: the &xe_device to check
+ *
+ * Returns: true if VF migration is supported, false otherwise.
+ */
+bool xe_sriov_vf_migration_supported(struct xe_device *xe)
+{
+	xe_assert(xe, IS_SRIOV_VF(xe));
+	return xe->sriov.vf.migration.enabled;
+}
+
+static void vf_disable_migration(struct xe_device *xe, const char *fmt, ...)
+{
+	struct va_format vaf;
+	va_list va_args;
+
+	xe_assert(xe, IS_SRIOV_VF(xe));
+
+	va_start(va_args, fmt);
+	vaf.fmt = fmt;
+	vaf.va  = &va_args;
+	xe_sriov_notice(xe, "migration disabled: %pV\n", &vaf);
+	va_end(va_args);
+
+	xe->sriov.vf.migration.enabled = false;
+}
+
+static void migration_worker_func(struct work_struct *w);
+
+static void vf_migration_init_early(struct xe_device *xe)
 {
 	/*
 	 * TODO: Add conditions to allow specific platforms, when they're
 	 * supported at production quality.
 	 */
-	return IS_ENABLED(CONFIG_DRM_XE_DEBUG);
-}
+	if (!IS_ENABLED(CONFIG_DRM_XE_DEBUG))
+		return vf_disable_migration(xe,
+					    "experimental feature not available on production builds");
 
-static void migration_worker_func(struct work_struct *w);
+	if (GRAPHICS_VER(xe) < 20)
+		return vf_disable_migration(xe, "requires gfx version >= 20, but only %u found",
+					    GRAPHICS_VER(xe));
+
+	if (!IS_DGFX(xe)) {
+		struct xe_uc_fw_version guc_version;
+
+		xe_gt_sriov_vf_guc_versions(xe_device_get_gt(xe, 0), NULL, &guc_version);
+		if (MAKE_GUC_VER_STRUCT(guc_version) < MAKE_GUC_VER(1, 23, 0))
+			return vf_disable_migration(xe,
+						    "CCS migration requires GuC ABI >= 1.23 but only %u.%u found",
+						    guc_version.major, guc_version.minor);
+	}
+
+	INIT_WORK(&xe->sriov.vf.migration.worker, migration_worker_func);
+
+	xe->sriov.vf.migration.enabled = true;
+	xe_sriov_dbg(xe, "migration support enabled\n");
+}
 
 /**
  * xe_sriov_vf_init_early - Initialize SR-IOV VF specific data.
···
  */
 void xe_sriov_vf_init_early(struct xe_device *xe)
 {
-	INIT_WORK(&xe->sriov.vf.migration.worker, migration_worker_func);
-
-	if (!vf_migration_supported(xe))
-		xe_sriov_info(xe, "migration not supported by this module version\n");
+	vf_migration_init_early(xe);
 }
 
 /**
···
 	xe_pm_runtime_get(xe);
 	vf_post_migration_shutdown(xe);
 
-	if (!vf_migration_supported(xe)) {
-		xe_sriov_err(xe, "migration not supported by this module version\n");
+	if (!xe_sriov_vf_migration_supported(xe)) {
+		xe_sriov_err(xe, "migration is not supported\n");
 		err = -ENOTRECOVERABLE;
 		goto fail;
 	}
···
 	started = queue_work(xe->sriov.wq, &xe->sriov.vf.migration.worker);
 	drm_info(&xe->drm, "VF migration recovery %s\n", started ?
 		 "scheduled" : "already in progress");
+}
+
+/**
+ * xe_sriov_vf_init_late() - SR-IOV VF late initialization functions.
+ * @xe: the &xe_device to initialize
+ *
+ * This function initializes code for CCS migration.
+ *
+ * Return: 0 on success or a negative error code on failure.
+ */
+int xe_sriov_vf_init_late(struct xe_device *xe)
+{
+	int err = 0;
+
+	if (xe_sriov_vf_migration_supported(xe))
+		err = xe_sriov_vf_ccs_init(xe);
+
+	return err;
+}
+
+static int sa_info_vf_ccs(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = m->private;
+	struct xe_device *xe = to_xe_device(node->minor->dev);
+	struct drm_printer p = drm_seq_file_printer(m);
+
+	xe_sriov_vf_ccs_print(xe, &p);
+	return 0;
+}
+
+static const struct drm_info_list debugfs_list[] = {
+	{ .name = "sa_info_vf_ccs", .show = sa_info_vf_ccs },
+};
+
+/**
+ * xe_sriov_vf_debugfs_register - Register VF debugfs attributes.
+ * @xe: the &xe_device
+ * @root: the root &dentry
+ *
+ * Prepare debugfs attributes exposed by the VF.
+ */
+void xe_sriov_vf_debugfs_register(struct xe_device *xe, struct dentry *root)
+{
+	drm_debugfs_create_files(debugfs_list, ARRAY_SIZE(debugfs_list),
+				 root, xe->drm.primary);
 }

drivers/gpu/drm/xe/xe_sriov_vf.h

···
 #ifndef _XE_SRIOV_VF_H_
 #define _XE_SRIOV_VF_H_
 
+#include <linux/types.h>
+
+struct dentry;
 struct xe_device;
 
 void xe_sriov_vf_init_early(struct xe_device *xe);
+int xe_sriov_vf_init_late(struct xe_device *xe);
 void xe_sriov_vf_start_migration_recovery(struct xe_device *xe);
+bool xe_sriov_vf_migration_supported(struct xe_device *xe);
+void xe_sriov_vf_debugfs_register(struct xe_device *xe, struct dentry *root);
 
 #endif

drivers/gpu/drm/xe/xe_sriov_vf_ccs.c

···
 #include "xe_guc_submit.h"
 #include "xe_lrc.h"
 #include "xe_migrate.h"
+#include "xe_pm.h"
 #include "xe_sa.h"
 #include "xe_sriov_printk.h"
+#include "xe_sriov_vf.h"
 #include "xe_sriov_vf_ccs.h"
 #include "xe_sriov_vf_ccs_types.h"
 
···
 	return round_up(bb_pool_size * 2, SZ_1M);
 }
 
-static int alloc_bb_pool(struct xe_tile *tile, struct xe_tile_vf_ccs *ctx)
+static int alloc_bb_pool(struct xe_tile *tile, struct xe_sriov_vf_ccs_ctx *ctx)
 {
 	struct xe_device *xe = tile_to_xe(tile);
 	struct xe_sa_manager *sa_manager;
···
 	return 0;
 }
 
-static void ccs_rw_update_ring(struct xe_tile_vf_ccs *ctx)
+static void ccs_rw_update_ring(struct xe_sriov_vf_ccs_ctx *ctx)
 {
 	u64 addr = xe_sa_manager_gpu_addr(ctx->mem.ccs_bb_pool);
 	struct xe_lrc *lrc = xe_exec_queue_lrc(ctx->mig_q);
···
 	xe_lrc_set_ring_tail(lrc, lrc->ring.tail);
 }
 
-static int register_save_restore_context(struct xe_tile_vf_ccs *ctx)
+static int register_save_restore_context(struct xe_sriov_vf_ccs_ctx *ctx)
 {
-	int err = -EINVAL;
 	int ctx_type;
 
 	switch (ctx->ctx_id) {
···
 		ctx_type = GUC_CONTEXT_COMPRESSION_RESTORE;
 		break;
 	default:
-		return err;
+		return -EINVAL;
 	}
 
-	xe_guc_register_exec_queue(ctx->mig_q, ctx_type);
+	xe_guc_register_vf_exec_queue(ctx->mig_q, ctx_type);
 	return 0;
 }
 
···
  */
 int xe_sriov_vf_ccs_register_context(struct xe_device *xe)
 {
-	struct xe_tile *tile = xe_device_get_root_tile(xe);
 	enum xe_sriov_vf_ccs_rw_ctxs ctx_id;
-	struct xe_tile_vf_ccs *ctx;
+	struct xe_sriov_vf_ccs_ctx *ctx;
 	int err;
 
-	if (!IS_VF_CCS_READY(xe))
-		return 0;
+	xe_assert(xe, IS_VF_CCS_READY(xe));
 
 	for_each_ccs_rw_ctx(ctx_id) {
-		ctx = &tile->sriov.vf.ccs[ctx_id];
+		ctx = &xe->sriov.vf.ccs.contexts[ctx_id];
 		err = register_save_restore_context(ctx);
 		if (err)
 			return err;
···
 
 static void xe_sriov_vf_ccs_fini(void *arg)
 {
-	struct xe_tile_vf_ccs *ctx = arg;
+	struct xe_sriov_vf_ccs_ctx *ctx = arg;
 	struct xe_lrc *lrc = xe_exec_queue_lrc(ctx->mig_q);
 
 	/*
···
 {
 	struct xe_tile *tile = xe_device_get_root_tile(xe);
 	enum xe_sriov_vf_ccs_rw_ctxs ctx_id;
-	struct xe_tile_vf_ccs *ctx;
+	struct xe_sriov_vf_ccs_ctx *ctx;
 	struct xe_exec_queue *q;
 	u32 flags;
 	int err;
 
 	xe_assert(xe, IS_SRIOV_VF(xe));
-	xe_assert(xe, !IS_DGFX(xe));
-	xe_assert(xe, xe_device_has_flat_ccs(xe));
+	xe_assert(xe, xe_sriov_vf_migration_supported(xe));
+
+	if (IS_DGFX(xe) || !xe_device_has_flat_ccs(xe))
+		return 0;
 
 	for_each_ccs_rw_ctx(ctx_id) {
-		ctx = &tile->sriov.vf.ccs[ctx_id];
+		ctx = &xe->sriov.vf.ccs.contexts[ctx_id];
 		ctx->ctx_id = ctx_id;
 
 		flags = EXEC_QUEUE_FLAG_KERNEL |
···
 {
 	struct xe_device *xe = xe_bo_device(bo);
 	enum xe_sriov_vf_ccs_rw_ctxs ctx_id;
-	struct xe_tile_vf_ccs *ctx;
+	struct xe_sriov_vf_ccs_ctx *ctx;
 	struct xe_tile *tile;
 	struct xe_bb *bb;
 	int err = 0;
 
-	if (!IS_VF_CCS_READY(xe))
-		return 0;
+	xe_assert(xe, IS_VF_CCS_READY(xe));
 
 	tile = xe_device_get_root_tile(xe);
 
···
 		/* bb should be NULL here. Assert if not NULL */
 		xe_assert(xe, !bb);
 
-		ctx = &tile->sriov.vf.ccs[ctx_id];
+		ctx = &xe->sriov.vf.ccs.contexts[ctx_id];
 		err = xe_migrate_ccs_rw_copy(tile, ctx->mig_q, bo, ctx_id);
 	}
 	return err;
···
 	enum xe_sriov_vf_ccs_rw_ctxs ctx_id;
 	struct xe_bb *bb;
 
-	if (!IS_VF_CCS_READY(xe))
+	xe_assert(xe, IS_VF_CCS_READY(xe));
+
+	if (!xe_bo_has_valid_ccs_bb(bo))
 		return 0;
 
 	for_each_ccs_rw_ctx(ctx_id) {
···
 		bo->bb_ccs[ctx_id] = NULL;
 	}
 	return 0;
+}
+
+/**
+ * xe_sriov_vf_ccs_print - Print VF CCS details.
+ * @xe: the &xe_device
+ * @p: the &drm_printer
+ *
+ * This function is for VF use only.
+ */
+void xe_sriov_vf_ccs_print(struct xe_device *xe, struct drm_printer *p)
+{
+	struct xe_sa_manager *bb_pool;
+	enum xe_sriov_vf_ccs_rw_ctxs ctx_id;
+
+	if (!IS_VF_CCS_READY(xe))
+		return;
+
+	xe_pm_runtime_get(xe);
+
+	for_each_ccs_rw_ctx(ctx_id) {
+		bb_pool = xe->sriov.vf.ccs.contexts[ctx_id].mem.ccs_bb_pool;
+		if (!bb_pool)
+			break;
+
+		drm_printf(p, "ccs %s bb suballoc info\n", ctx_id ? "write" : "read");
+		drm_printf(p, "-------------------------\n");
+		drm_suballoc_dump_debug_info(&bb_pool->base, p, xe_sa_manager_gpu_addr(bb_pool));
+		drm_puts(p, "\n");
+	}
+
+	xe_pm_runtime_put(xe);
 }

drivers/gpu/drm/xe/xe_sriov_vf_ccs.h

···
 #ifndef _XE_SRIOV_VF_CCS_H_
 #define _XE_SRIOV_VF_CCS_H_
 
+#include "xe_device_types.h"
+#include "xe_sriov.h"
+#include "xe_sriov_vf_ccs_types.h"
+
+struct drm_printer;
 struct xe_device;
 struct xe_bo;
 
···
 int xe_sriov_vf_ccs_attach_bo(struct xe_bo *bo);
 int xe_sriov_vf_ccs_detach_bo(struct xe_bo *bo);
 int xe_sriov_vf_ccs_register_context(struct xe_device *xe);
+void xe_sriov_vf_ccs_print(struct xe_device *xe, struct drm_printer *p);
+
+static inline bool xe_sriov_vf_ccs_ready(struct xe_device *xe)
+{
+	xe_assert(xe, IS_SRIOV_VF(xe));
+	return xe->sriov.vf.ccs.initialized;
+}
+
+#define IS_VF_CCS_READY(xe) ({ \
+	struct xe_device *xe__ = (xe); \
+	IS_SRIOV_VF(xe__) && xe_sriov_vf_ccs_ready(xe__); \
+	})
 
 #endif

drivers/gpu/drm/xe/xe_sriov_vf_ccs_types.h

···
 #ifndef _XE_SRIOV_VF_CCS_TYPES_H_
 #define _XE_SRIOV_VF_CCS_TYPES_H_
 
+#include <linux/types.h>
+
 #define for_each_ccs_rw_ctx(id__) \
 	for ((id__) = 0; (id__) < XE_SRIOV_VF_CCS_CTX_COUNT; (id__)++)
-
-#define IS_VF_CCS_READY(xe) ({ \
-		struct xe_device *___xe = (xe); \
-		xe_assert(___xe, IS_SRIOV_VF(___xe)); \
-		___xe->sriov.vf.ccs.initialized; \
-		})
-
-#define IS_VF_CCS_INIT_NEEDED(xe) ({\
-		struct xe_device *___xe = (xe); \
-		IS_SRIOV_VF(___xe) && !IS_DGFX(___xe) && \
-		xe_device_has_flat_ccs(___xe) && GRAPHICS_VER(___xe) >= 20; \
-		})
 
 enum xe_sriov_vf_ccs_rw_ctxs {
 	XE_SRIOV_VF_CCS_READ_CTX,
···
 	XE_SRIOV_VF_CCS_CTX_COUNT
 };
 
-#define IS_VF_CCS_BB_VALID(xe, bo) ({ \
-		struct xe_device *___xe = (xe); \
-		struct xe_bo *___bo = (bo); \
-		IS_SRIOV_VF(___xe) && \
-		___bo->bb_ccs[XE_SRIOV_VF_CCS_READ_CTX] && \
-		___bo->bb_ccs[XE_SRIOV_VF_CCS_WRITE_CTX]; \
-		})
-
 struct xe_migrate;
 struct xe_sa_manager;
 
-struct xe_tile_vf_ccs {
-	/** @id: Id to which context it belongs to */
+/**
+ * struct xe_sriov_vf_ccs_ctx - VF CCS migration context data.
+ */
+struct xe_sriov_vf_ccs_ctx {
+	/** @ctx_id: Id to which context it belongs to */
 	enum xe_sriov_vf_ccs_rw_ctxs ctx_id;
+
 	/** @mig_q: exec queues used for migration */
 	struct xe_exec_queue *mig_q;
 
+	/** @mem: memory data */
 	struct {
-		/** @ccs_bb_pool: Pool from which batch buffers are allocated. */
+		/** @mem.ccs_bb_pool: Pool from which batch buffers are allocated. */
 		struct xe_sa_manager *ccs_bb_pool;
 	} mem;
+};
+
+/**
+ * struct xe_sriov_vf_ccs - The VF CCS migration support data.
+ */
+struct xe_sriov_vf_ccs {
+	/** @contexts: CCS read and write contexts for VF. */
+	struct xe_sriov_vf_ccs_ctx contexts[XE_SRIOV_VF_CCS_CTX_COUNT];
+
+	/** @initialized: Initialization of VF CCS is completed or not. */
+	bool initialized;
 };
 
 #endif

drivers/gpu/drm/xe/xe_sriov_vf_types.h

···
 #include <linux/types.h>
 #include <linux/workqueue_types.h>
 
+#include "xe_sriov_vf_ccs_types.h"
+
 /**
  * struct xe_sriov_vf_relay_version - PF ABI version details.
  */
···
 		struct work_struct worker;
 		/** @migration.gt_flags: Per-GT request flags for VF migration recovery */
 		unsigned long gt_flags;
+		/**
+		 * @migration.enabled: flag indicating if migration support
+		 * was enabled or not due to missing prerequisites
+		 */
+		bool enabled;
 	} migration;
 
 	/** @ccs: VF CCS state data */
-	struct {
-		/** @ccs.initialized: Initilalization of VF CCS is completed or not */
-		bool initialized;
-	} ccs;
+	struct xe_sriov_vf_ccs ccs;
 };
 
 #endif

drivers/gpu/drm/xe/xe_survivability_mode.c

···
 	u32 data;
 	bool survivability_mode;
 
-	if (!IS_DGFX(xe) || IS_SRIOV_VF(xe))
+	if (!IS_DGFX(xe) || IS_SRIOV_VF(xe) || xe->info.platform < XE_BATTLEMAGE)
 		return false;
 
 	survivability_mode = xe_configfs_get_survivability_mode(pdev);
-
-	if (xe->info.platform < XE_BATTLEMAGE) {
-		if (survivability_mode) {
-			dev_err(&pdev->dev, "Survivability Mode is not supported on this card\n");
-			xe_configfs_clear_survivability_mode(pdev);
-		}
-		return false;
-	}
-
 	/* Enable survivability mode if set via configfs */
 	if (survivability_mode)
 		return true;

drivers/gpu/drm/xe/xe_svm.c

···
 #include <drm/drm_drv.h>
 
 #include "xe_bo.h"
+#include "xe_exec_queue_types.h"
 #include "xe_gt_stats.h"
 #include "xe_migrate.h"
 #include "xe_module.h"
···
 	 * memory.
 	 */
 
-	struct drm_gpusvm_range_flags flags = {
+	struct drm_gpusvm_pages_flags flags = {
 		/* Pairs with WRITE_ONCE in drm_gpusvm.c */
-		.__flags = READ_ONCE(range->base.flags.__flags),
+		.__flags = READ_ONCE(range->base.pages.flags.__flags),
 	};
 
 	return flags.has_devmem_pages;
···
 	return gpusvm_to_vm(r->gpusvm);
 }
 
-#define range_debug(r__, operaton__)					\
+#define range_debug(r__, operation__)					\
 	vm_dbg(&range_to_vm(&(r__)->base)->xe->drm,			\
 	       "%s: asid=%u, gpusvm=%p, vram=%d,%d, seqno=%lu, " \
 	       "start=0x%014lx, end=0x%014lx, size=%lu",		\
-	       (operaton__), range_to_vm(&(r__)->base)->usm.asid,	\
+	       (operation__), range_to_vm(&(r__)->base)->usm.asid,	\
 	       (r__)->base.gpusvm,					\
 	       xe_svm_range_in_vram((r__)) ? 1 : 0,			\
 	       xe_svm_range_has_vram_binding((r__)) ? 1 : 0,		\
-	       (r__)->base.notifier_seq,				\
+	       (r__)->base.pages.notifier_seq,				\
 	       xe_svm_range_start((r__)), xe_svm_range_end((r__)),	\
 	       xe_svm_range_size((r__)))
 
···
 		   &vm->svm.garbage_collector.work);
 }
 
+static void xe_svm_tlb_inval_count_stats_incr(struct xe_gt *gt)
+{
+	xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_TLB_INVAL_COUNT, 1);
+}
+
 static u8
 xe_svm_range_notifier_event_begin(struct xe_vm *vm, struct drm_gpusvm_range *r,
 				  const struct mmu_notifier_range *mmu_range,
···
 	range_debug(range, "NOTIFIER");
 
 	/* Skip if already unmapped or if no binding exist */
-	if (range->base.flags.unmapped || !range->tile_present)
+	if (range->base.pages.flags.unmapped || !range->tile_present)
 		return 0;
 
 	range_debug(range, "NOTIFIER - EXECUTE");
···
 	 */
 	for_each_tile(tile, xe, id)
 		if (xe_pt_zap_ptes_range(tile, vm, range)) {
-			tile_mask |= BIT(id);
 			/*
 			 * WRITE_ONCE pairs with READ_ONCE in
 			 * xe_vm_has_valid_gpu_mapping()
 			 */
 			WRITE_ONCE(range->tile_invalidated,
 				   range->tile_invalidated | BIT(id));
+
+			if (!(tile_mask & BIT(id))) {
+				xe_svm_tlb_inval_count_stats_incr(tile->primary_gt);
+				if (tile->media_gt)
+					xe_svm_tlb_inval_count_stats_incr(tile->media_gt);
+				tile_mask |= BIT(id);
+			}
 		}
 
 	return tile_mask;
···
 						   mmu_range);
 }
 
+static s64 xe_svm_stats_ktime_us_delta(ktime_t start)
+{
+	return IS_ENABLED(CONFIG_DEBUG_FS) ?
+		ktime_us_delta(ktime_get(), start) : 0;
+}
+
+static void xe_svm_tlb_inval_us_stats_incr(struct xe_gt *gt, ktime_t start)
+{
+	s64 us_delta = xe_svm_stats_ktime_us_delta(start);
+
+	xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_TLB_INVAL_US, us_delta);
+}
+
+static ktime_t xe_svm_stats_ktime_get(void)
+{
+	return IS_ENABLED(CONFIG_DEBUG_FS) ? ktime_get() : 0;
+}
+
 static void xe_svm_invalidate(struct drm_gpusvm *gpusvm,
 			      struct drm_gpusvm_notifier *notifier,
 			      const struct mmu_notifier_range *mmu_range)
···
 	struct xe_vm *vm = gpusvm_to_vm(gpusvm);
 	struct xe_device *xe = vm->xe;
 	struct drm_gpusvm_range *r, *first;
+	struct xe_tile *tile;
+	ktime_t start = xe_svm_stats_ktime_get();
 	u64 adj_start = mmu_range->start, adj_end = mmu_range->end;
-	u8 tile_mask = 0;
+	u8 tile_mask = 0, id;
 	long err;
 
 	xe_svm_assert_in_notifier(vm);
···
 	r = first;
 	drm_gpusvm_for_each_range(r, notifier, adj_start, adj_end)
 		xe_svm_range_notifier_event_end(vm, r, mmu_range);
+	for_each_tile(tile, xe, id) {
+		if (tile_mask & BIT(id)) {
+			xe_svm_tlb_inval_us_stats_incr(tile->primary_gt, start);
+			if (tile->media_gt)
+				xe_svm_tlb_inval_us_stats_incr(tile->media_gt, start);
+		}
+	}
 }
 
 static int __xe_svm_garbage_collector(struct xe_vm *vm,
···
 	if (xe_vm_is_closed_or_banned(vm))
 		return -ENOENT;
 
-	spin_lock(&vm->svm.garbage_collector.lock);
 	for (;;) {
+		spin_lock(&vm->svm.garbage_collector.lock);
 		range = list_first_entry_or_null(&vm->svm.garbage_collector.range_list,
 						 typeof(*range),
 						 garbage_collector_link);
···
 			else
 				return err;
 		}
-
-		spin_lock(&vm->svm.garbage_collector.lock);
 	}
 	spin_unlock(&vm->svm.garbage_collector.lock);
 
···
 	XE_SVM_COPY_TO_SRAM,
 };
 
+static void xe_svm_copy_kb_stats_incr(struct xe_gt *gt,
+				      const enum xe_svm_copy_dir dir,
+				      int kb)
+{
+	if (dir == XE_SVM_COPY_TO_VRAM)
+		xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_DEVICE_COPY_KB, kb);
+	else
+		xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_CPU_COPY_KB, kb);
+}
+
+static void xe_svm_copy_us_stats_incr(struct xe_gt *gt,
+				      const enum xe_svm_copy_dir dir,
+				      unsigned long npages,
+				      ktime_t start)
+{
+	s64 us_delta = xe_svm_stats_ktime_us_delta(start);
+
+	if (dir == XE_SVM_COPY_TO_VRAM) {
+		switch (npages) {
+		case 1:
+			xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_4K_DEVICE_COPY_US,
+					 us_delta);
+			break;
+		case 16:
+			xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_64K_DEVICE_COPY_US,
+					 us_delta);
+			break;
+		case 512:
+			xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_2M_DEVICE_COPY_US,
+					 us_delta);
+			break;
+		}
+		xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_DEVICE_COPY_US,
+				 us_delta);
+	} else {
+		switch (npages) {
+		case 1:
+			xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_4K_CPU_COPY_US,
+					 us_delta);
+			break;
+		case 16:
+			xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_64K_CPU_COPY_US,
+					 us_delta);
+			break;
+		case 512:
+			xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_2M_CPU_COPY_US,
+					 us_delta);
+			break;
+		}
+		xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_CPU_COPY_US,
+				 us_delta);
+	}
+}
+
 static int xe_svm_copy(struct page **pages,
 		       struct drm_pagemap_addr *pagemap_addr,
 		       unsigned long npages, const enum xe_svm_copy_dir dir)
 {
 	struct xe_vram_region *vr = NULL;
+	struct xe_gt *gt = NULL;
 	struct xe_device *xe;
 	struct dma_fence *fence = NULL;
 	unsigned long i;
···
 	u64 vram_addr = XE_VRAM_ADDR_INVALID;
 	int err = 0, pos = 0;
 	bool sram = dir == XE_SVM_COPY_TO_SRAM;
+	ktime_t start = xe_svm_stats_ktime_get();
 
 	/*
 	 * This flow is complex: it locates physically contiguous device pages,
···
 
 		if (!vr && spage) {
 			vr = page_to_vr(spage);
+			gt = xe_migrate_exec_queue(vr->migrate)->gt;
 			xe = vr->xe;
 		}
 		XE_WARN_ON(spage && page_to_vr(spage) != vr);
···
 			int incr = (match && last) ? 1 : 0;
 
 			if (vram_addr != XE_VRAM_ADDR_INVALID) {
+				xe_svm_copy_kb_stats_incr(gt, dir,
+							  (i - pos + incr) *
+							  (PAGE_SIZE / SZ_1K));
 				if (sram) {
 					vm_dbg(&xe->drm,
 					       "COPY TO SRAM - 0x%016llx -> 0x%016llx, NPAGES=%ld",
···
 
 			/* Extra mismatched device page, copy it */
 			if (!match && last && vram_addr != XE_VRAM_ADDR_INVALID) {
+				xe_svm_copy_kb_stats_incr(gt, dir,
+							  (PAGE_SIZE / SZ_1K));
 				if (sram) {
 					vm_dbg(&xe->drm,
 					       "COPY TO SRAM - 0x%016llx -> 0x%016llx, NPAGES=%d",
···
 		dma_fence_wait(fence, false);
 		dma_fence_put(fence);
 	}
+
+	/*
+	 * XXX: We can't derive the GT here (or anywhere in this functions, but
+	 * compute always uses the primary GT so accumlate stats on the likely
+	 * GT of the fault.
+	 */
+	if (gt)
+		xe_svm_copy_us_stats_incr(gt, dir, npages, start);
 
 	return err;
 #undef XE_MIGRATE_CHUNK_SIZE
···
 {
 	int err;
 
-	spin_lock_init(&vm->svm.garbage_collector.lock);
-	INIT_LIST_HEAD(&vm->svm.garbage_collector.range_list);
-	INIT_WORK(&vm->svm.garbage_collector.work,
-		  xe_svm_garbage_collector_work_func);
+	if (vm->flags & XE_VM_FLAG_FAULT_MODE) {
+		spin_lock_init(&vm->svm.garbage_collector.lock);
+		INIT_LIST_HEAD(&vm->svm.garbage_collector.range_list);
+		INIT_WORK(&vm->svm.garbage_collector.work,
+			  xe_svm_garbage_collector_work_func);
 
-	err = drm_gpusvm_init(&vm->svm.gpusvm, "Xe SVM", &vm->xe->drm,
-			      current->mm, xe_svm_devm_owner(vm->xe), 0,
-			      vm->size, xe_modparam.svm_notifier_size * SZ_1M,
-			      &gpusvm_ops, fault_chunk_sizes,
-			      ARRAY_SIZE(fault_chunk_sizes));
-	if (err)
-		return err;
+		err = drm_gpusvm_init(&vm->svm.gpusvm, "Xe SVM", &vm->xe->drm,
+				      current->mm, xe_svm_devm_owner(vm->xe), 0,
+				      vm->size,
+				      xe_modparam.svm_notifier_size * SZ_1M,
+				      &gpusvm_ops, fault_chunk_sizes,
+				      ARRAY_SIZE(fault_chunk_sizes));
+		drm_gpusvm_driver_set_lock(&vm->svm.gpusvm, &vm->lock);
+	} else {
+		err = drm_gpusvm_init(&vm->svm.gpusvm, "Xe SVM (simple)",
+				      &vm->xe->drm, NULL, NULL, 0, 0, 0, NULL,
+				      NULL, 0);
+	}
 
-	drm_gpusvm_driver_set_lock(&vm->svm.gpusvm, &vm->lock);
-
-	return 0;
+	return err;
 }
 
 /**
···
 	xe_svm_notifier_lock(vm);
 
 	ret = (range->tile_present & ~range->tile_invalidated & tile_mask) == tile_mask &&
-	       (devmem_preferred == range->base.flags.has_devmem_pages);
+	       (devmem_preferred == range->base.pages.flags.has_devmem_pages);
 
 	xe_svm_notifier_unlock(vm);
 
···
 	struct xe_device *xe = vr->xe;
 	struct device *dev = xe->drm.dev;
 	struct drm_buddy_block *block;
+	struct xe_validation_ctx vctx;
 	struct list_head *blocks;
+	struct drm_exec exec;
 	struct xe_bo *bo;
-	ktime_t time_end = 0;
-	int err, idx;
+	int err = 0, idx;
 
 	if (!drm_dev_enter(&xe->drm, &idx))
 		return -ENODEV;
 
 	xe_pm_runtime_get(xe);
 
- retry:
-	bo = xe_bo_create_locked(vr->xe, NULL, NULL, end - start,
-				 ttm_bo_type_device,
-				 (IS_DGFX(xe) ? XE_BO_FLAG_VRAM(vr) : XE_BO_FLAG_SYSTEM) |
-				 XE_BO_FLAG_CPU_ADDR_MIRROR);
-	if (IS_ERR(bo)) {
-		err = PTR_ERR(bo);
-		if (xe_vm_validate_should_retry(NULL, err, &time_end))
-			goto retry;
-		goto out_pm_put;
+	xe_validation_guard(&vctx, &xe->val, &exec, (struct xe_val_flags) {}, err) {
+		bo = xe_bo_create_locked(xe, NULL, NULL, end - start,
+					 ttm_bo_type_device,
+					 (IS_DGFX(xe) ? XE_BO_FLAG_VRAM(vr) : XE_BO_FLAG_SYSTEM) |
+					 XE_BO_FLAG_CPU_ADDR_MIRROR, &exec);
+		drm_exec_retry_on_contention(&exec);
+		if (IS_ERR(bo)) {
+			err = PTR_ERR(bo);
+			xe_validation_retry_on_oom(&vctx, &err);
+			break;
+		}
+
+		drm_pagemap_devmem_init(&bo->devmem_allocation, dev, mm,
+					&dpagemap_devmem_ops, dpagemap, end - start);
+
+		blocks = &to_xe_ttm_vram_mgr_resource(bo->ttm.resource)->blocks;
+		list_for_each_entry(block, blocks, link)
+			block->private = vr;
+
+		xe_bo_get(bo);
+
+		/* Ensure the device has a pm ref while there are device pages active. */
+		xe_pm_runtime_get_noresume(xe);
+		err = drm_pagemap_migrate_to_devmem(&bo->devmem_allocation, mm,
+						    start, end, timeslice_ms,
+						    xe_svm_devm_owner(xe));
+		if (err)
+			xe_svm_devmem_release(&bo->devmem_allocation);
+		xe_bo_unlock(bo);
+		xe_bo_put(bo);
 	}
-
-	drm_pagemap_devmem_init(&bo->devmem_allocation, dev, mm,
-				&dpagemap_devmem_ops, dpagemap, end - start);
-
-	blocks = &to_xe_ttm_vram_mgr_resource(bo->ttm.resource)->blocks;
-	list_for_each_entry(block, blocks, link)
-		block->private = vr;
-
-	xe_bo_get(bo);
-
-	/* Ensure the device has a pm ref while there are device pages active. */
-	xe_pm_runtime_get_noresume(xe);
-	err = drm_pagemap_migrate_to_devmem(&bo->devmem_allocation, mm,
-					    start, end, timeslice_ms,
-					    xe_svm_devm_owner(xe));
-	if (err)
-		xe_svm_devmem_release(&bo->devmem_allocation);
-
-	xe_bo_unlock(bo);
-	xe_bo_put(bo);
-
-out_pm_put:
 	xe_pm_runtime_put(xe);
 	drm_dev_exit(idx);
 
···
 	struct xe_vm *vm = range_to_vm(&range->base);
 	u64 range_size = xe_svm_range_size(range);
 
-	if (!range->base.flags.migrate_devmem || !preferred_region_is_vram)
+	if (!range->base.pages.flags.migrate_devmem || !preferred_region_is_vram)
 		return false;
 
 	xe_assert(vm->xe, IS_DGFX(vm->xe));
 
-	if (preferred_region_is_vram && xe_svm_range_in_vram(range)) {
+	if (xe_svm_range_in_vram(range)) {
 		drm_info(&vm->xe->drm, "Range is already in VRAM\n");
 		return false;
 	}
 
-	if (preferred_region_is_vram && range_size < SZ_64K && !supports_4K_migration(vm->xe)) {
+	if (range_size < SZ_64K && !supports_4K_migration(vm->xe)) {
 		drm_dbg(&vm->xe->drm, "Platform doesn't support SZ_4K range migration\n");
 		return false;
 	}
···
 	return true;
 }
 
+#define DECL_SVM_RANGE_COUNT_STATS(elem, stat) \
+static void xe_svm_range_##elem##_count_stats_incr(struct xe_gt *gt, \
+						   struct xe_svm_range *range) \
+{ \
+	switch (xe_svm_range_size(range)) { \
+	case SZ_4K: \
+		xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_4K_##stat##_COUNT, 1); \
+		break; \
+	case SZ_64K: \
+		xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_64K_##stat##_COUNT, 1); \
+		break; \
+	case SZ_2M: \
+		xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_2M_##stat##_COUNT, 1); \
+		break; \
+	} \
+} \
+
+DECL_SVM_RANGE_COUNT_STATS(fault, PAGEFAULT)
+DECL_SVM_RANGE_COUNT_STATS(valid_fault, VALID_PAGEFAULT)
+DECL_SVM_RANGE_COUNT_STATS(migrate, MIGRATE)
+
+#define DECL_SVM_RANGE_US_STATS(elem, stat) \
+static void xe_svm_range_##elem##_us_stats_incr(struct xe_gt *gt, \
+						struct xe_svm_range *range, \
+						ktime_t start) \
+{ \
+	s64 us_delta = xe_svm_stats_ktime_us_delta(start); \
+\
+	switch (xe_svm_range_size(range)) { \
+	case SZ_4K: \
+		xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_4K_##stat##_US, \
+				 us_delta); \
+		break; \
+	case SZ_64K: \
+		xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_64K_##stat##_US, \
+				 us_delta); \
+		break; \
+	case SZ_2M: \
+		xe_gt_stats_incr(gt, XE_GT_STATS_ID_SVM_2M_##stat##_US, \
+				 us_delta); \
+		break; \
+	} \
+} \
+
+DECL_SVM_RANGE_US_STATS(migrate, MIGRATE)
+DECL_SVM_RANGE_US_STATS(get_pages, GET_PAGES)
+DECL_SVM_RANGE_US_STATS(bind, BIND)
+DECL_SVM_RANGE_US_STATS(fault, PAGEFAULT)
+
 static int __xe_svm_handle_pagefault(struct xe_vm *vm, struct xe_vma *vma,
 				     struct xe_gt *gt, u64 fault_addr,
 				     bool need_vram)
 {
+	int devmem_possible = IS_DGFX(vm->xe) &&
+		IS_ENABLED(CONFIG_DRM_XE_PAGEMAP);
 	struct drm_gpusvm_ctx ctx = {
 		.read_only = xe_vma_read_only(vma),
-		.devmem_possible = IS_DGFX(vm->xe) &&
-			IS_ENABLED(CONFIG_DRM_XE_PAGEMAP),
-		.check_pages_threshold = IS_DGFX(vm->xe) &&
-			IS_ENABLED(CONFIG_DRM_XE_PAGEMAP) ? SZ_64K : 0,
-		.devmem_only = need_vram && IS_ENABLED(CONFIG_DRM_XE_PAGEMAP),
-		.timeslice_ms = need_vram && IS_DGFX(vm->xe) &&
-			IS_ENABLED(CONFIG_DRM_XE_PAGEMAP) ?
+		.devmem_possible = devmem_possible,
+		.check_pages_threshold = devmem_possible ? SZ_64K : 0,
+		.devmem_only = need_vram && devmem_possible,
+		.timeslice_ms = need_vram && devmem_possible ?
 			vm->xe->atomic_svm_timeslice_ms : 0,
 	};
+	struct xe_validation_ctx vctx;
+	struct drm_exec exec;
 	struct xe_svm_range *range;
 	struct dma_fence *fence;
 	struct drm_pagemap *dpagemap;
 	struct xe_tile *tile = gt_to_tile(gt);
 	int migrate_try_count = ctx.devmem_only ? 3 : 1;
-	ktime_t end = 0;
+	ktime_t start = xe_svm_stats_ktime_get(), bind_start, get_pages_start;
 	int err;
 
 	lockdep_assert_held_write(&vm->lock);
···
 	if (IS_ERR(range))
 		return PTR_ERR(range);
 
-	if (ctx.devmem_only && !range->base.flags.migrate_devmem)
-		return -EACCES;
+	xe_svm_range_fault_count_stats_incr(gt, range);
 
-	if (xe_svm_range_is_valid(range, tile, ctx.devmem_only))
-		return 0;
+	if (ctx.devmem_only && !range->base.pages.flags.migrate_devmem) {
+		err = -EACCES;
+		goto out;
+	}
+
+	if (xe_svm_range_is_valid(range, tile, ctx.devmem_only)) {
+		xe_svm_range_valid_fault_count_stats_incr(gt, range);
+		range_debug(range, "PAGE FAULT - VALID");
+		goto out;
+	}
 
 	range_debug(range, "PAGE FAULT");
 
 	dpagemap = xe_vma_resolve_pagemap(vma, tile);
 	if (--migrate_try_count >= 0 &&
 	    xe_svm_range_needs_migrate_to_vram(range, vma, !!dpagemap || ctx.devmem_only)) {
+		ktime_t migrate_start = xe_svm_stats_ktime_get();
+
 		/* TODO : For multi-device dpagemap will be used to find the
 		 * remote tile and remote device. Will need to modify
 		 * xe_svm_alloc_vram to use dpagemap for future multi-device
 		 * support.
 		 */
+		xe_svm_range_migrate_count_stats_incr(gt, range);
 		err = xe_svm_alloc_vram(tile, range, &ctx);
+		xe_svm_range_migrate_us_stats_incr(gt, range, migrate_start);
 		ctx.timeslice_ms <<= 1;	/* Double timeslice if we have to retry */
 		if (err) {
 			if (migrate_try_count || !ctx.devmem_only) {
···
 			}
 		}
 	}
+
+	get_pages_start = xe_svm_stats_ktime_get();
 
 	range_debug(range, "GET PAGES");
 	err = xe_svm_range_get_pages(vm, range, &ctx);
···
 	}
 	if (err) {
 		range_debug(range, "PAGE FAULT - FAIL PAGE COLLECT");
-		goto err_out;
+		goto out;
 	}
 
+	xe_svm_range_get_pages_us_stats_incr(gt, range, get_pages_start);
 	range_debug(range, "PAGE FAULT - BIND");
 
-retry_bind:
-	xe_vm_lock(vm, false);
-	fence = xe_vm_range_rebind(vm, vma, range, BIT(tile->id));
-	if (IS_ERR(fence)) {
-		xe_vm_unlock(vm);
-		err = PTR_ERR(fence);
-		if (err == -EAGAIN) {
-			ctx.timeslice_ms <<= 1;	/* Double timeslice if we have to retry */
-			range_debug(range, "PAGE FAULT - RETRY BIND");
-			goto retry;
+	bind_start = xe_svm_stats_ktime_get();
+	xe_validation_guard(&vctx, &vm->xe->val, &exec, (struct xe_val_flags) {}, err) {
+		err = xe_vm_drm_exec_lock(vm, &exec);
+		drm_exec_retry_on_contention(&exec);
+
+		xe_vm_set_validation_exec(vm, &exec);
+		fence = xe_vm_range_rebind(vm, vma, range, BIT(tile->id));
+		xe_vm_set_validation_exec(vm, NULL);
+		if (IS_ERR(fence)) {
+			drm_exec_retry_on_contention(&exec);
+			err = PTR_ERR(fence);
+			xe_validation_retry_on_oom(&vctx, &err);
+			xe_svm_range_bind_us_stats_incr(gt, range, bind_start);
+			break;
 		}
-		if (xe_vm_validate_should_retry(NULL, err, &end))
-			goto retry_bind;
-		goto err_out;
 	}
-	xe_vm_unlock(vm);
+	if (err)
+		goto err_out;
 
 	dma_fence_wait(fence, false);
 	dma_fence_put(fence);
+	xe_svm_range_bind_us_stats_incr(gt, range, bind_start);
+
+out:
+	xe_svm_range_fault_us_stats_incr(gt, range, start);
+	return 0;
 
 err_out:
+	if (err == -EAGAIN) {
+		ctx.timeslice_ms <<= 1;	/* Double timeslice if we have to retry */
+		range_debug(range, "PAGE FAULT - RETRY BIND");
+		goto retry;
+	}
 
 	return err;
 }
···
 	r = drm_gpusvm_range_find_or_insert(&vm->svm.gpusvm, max(addr, xe_vma_start(vma)),
 					    xe_vma_start(vma), xe_vma_end(vma), ctx);
 	if (IS_ERR(r))
-		return ERR_PTR(PTR_ERR(r));
+		return ERR_CAST(r);
 
 	return to_xe_range(r);
 }
···
 {
 	struct drm_pagemap *dpagemap;
 
-	xe_assert(tile_to_xe(tile), range->base.flags.migrate_devmem);
+	xe_assert(tile_to_xe(tile), range->base.pages.flags.migrate_devmem);
 	range_debug(range, "ALLOCATE VRAM");
 
 	dpagemap = tile_local_pagemap(tile);

drivers/gpu/drm/xe/xe_svm.h

···
 static inline bool xe_svm_range_has_dma_mapping(struct xe_svm_range *range)
 {
 	lockdep_assert_held(&range->base.gpusvm->notifier_lock);
-	return range->base.flags.has_dma_mapping;
+	return range->base.pages.flags.has_dma_mapping;
 }
 
 /**
···
 	return drm_gpusvm_range_size(&range->base);
 }
 
-#define xe_svm_assert_in_notifier(vm__) \
-	lockdep_assert_held_write(&(vm__)->svm.gpusvm.notifier_lock)
-
-#define xe_svm_notifier_lock(vm__)	\
-	drm_gpusvm_notifier_lock(&(vm__)->svm.gpusvm)
-
-#define xe_svm_notifier_unlock(vm__)	\
-	drm_gpusvm_notifier_unlock(&(vm__)->svm.gpusvm)
-
 void xe_svm_flush(struct xe_vm *vm);
 
 #else
 #include <linux/interval_tree.h>
+#include "xe_vm.h"
 
 struct drm_pagemap_addr;
 struct drm_gpusvm_ctx;
···
 struct xe_svm_range {
 	struct {
 		struct interval_tree_node itree;
-		const struct drm_pagemap_addr *dma_addr;
+		struct {
+			const struct drm_pagemap_addr *dma_addr;
+		} pages;
 	} base;
 	u32 tile_present;
 	u32 tile_invalidated;
···
 static inline
 int xe_svm_init(struct xe_vm *vm)
 {
+#if IS_ENABLED(CONFIG_DRM_GPUSVM)
+	return drm_gpusvm_init(&vm->svm.gpusvm, "Xe SVM (simple)", &vm->xe->drm,
+			       NULL, NULL, 0, 0, 0, NULL, NULL, 0);
+#else
 	return 0;
+#endif
 }
 
 static inline
 void xe_svm_fini(struct xe_vm *vm)
 {
+#if IS_ENABLED(CONFIG_DRM_GPUSVM)
+	xe_assert(vm->xe, xe_vm_is_closed(vm));
+	drm_gpusvm_fini(&vm->svm.gpusvm);
+#endif
 }
 
 static inline
···
 	return NULL;
 }
 
-#define xe_svm_assert_in_notifier(...) do {} while (0)
+static inline void xe_svm_flush(struct xe_vm *vm)
+{
+}
 #define xe_svm_range_has_dma_mapping(...) false
+#endif /* CONFIG_DRM_XE_GPUSVM */
+
+#if IS_ENABLED(CONFIG_DRM_GPUSVM) /* Need to support userptr without XE_GPUSVM */
+#define xe_svm_assert_in_notifier(vm__) \
+	lockdep_assert_held_write(&(vm__)->svm.gpusvm.notifier_lock)
+
+#define xe_svm_assert_held_read(vm__) \
+	lockdep_assert_held_read(&(vm__)->svm.gpusvm.notifier_lock)
+
+#define xe_svm_notifier_lock(vm__)	\
+	drm_gpusvm_notifier_lock(&(vm__)->svm.gpusvm)
+
+#define xe_svm_notifier_lock_interruptible(vm__)	\
+	down_read_interruptible(&(vm__)->svm.gpusvm.notifier_lock)
+
+#define xe_svm_notifier_unlock(vm__)	\
+	drm_gpusvm_notifier_unlock(&(vm__)->svm.gpusvm)
+
+#else
+#define xe_svm_assert_in_notifier(...) do {} while (0)
+
+static inline void xe_svm_assert_held_read(struct xe_vm *vm)
+{
+}
 
 static inline void xe_svm_notifier_lock(struct xe_vm *vm)
 {
 }
 
+static inline int xe_svm_notifier_lock_interruptible(struct xe_vm *vm)
+{
+	return 0;
+}
+
 static inline void xe_svm_notifier_unlock(struct xe_vm *vm)
 {
 }
+#endif /* CONFIG_DRM_GPUSVM */
 
-static inline void xe_svm_flush(struct xe_vm *vm)
-{
-}
-#endif
 #endif

drivers/gpu/drm/xe/xe_tile_debugfs.c

···
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#include <linux/debugfs.h>
+#include <drm/drm_debugfs.h>
+
+#include "xe_pm.h"
+#include "xe_sa.h"
+#include "xe_tile_debugfs.h"
+
+static struct xe_tile *node_to_tile(struct drm_info_node *node)
+{
+	return node->dent->d_parent->d_inode->i_private;
+}
+
+/**
+ * tile_debugfs_simple_show - A show callback for struct drm_info_list
+ * @m: the &seq_file
+ * @data: data used by the drm debugfs helpers
+ *
+ * This callback can be used in struct drm_info_list to describe debugfs
+ * files that are &xe_tile specific.
+ *
+ * It is assumed that those debugfs files will be created on directory entry
+ * which struct dentry d_inode->i_private points to &xe_tile.
+ *
+ *      /sys/kernel/debug/dri/0/
+ *      ├── tile0/		# tile = dentry->d_inode->i_private
+ *      │   │   ├── id		# tile = dentry->d_parent->d_inode->i_private
+ *
+ * This function assumes that &m->private will be set to the &struct
+ * drm_info_node corresponding to the instance of the info on a given &struct
+ * drm_minor (see struct drm_info_list.show for details).
+ *
+ * This function also assumes that struct drm_info_list.data will point to the
+ * function code that will actually print a file content::
+ *
+ *   int (*print)(struct xe_tile *, struct drm_printer *)
+ *
+ * Example::
+ *
+ *    int tile_id(struct xe_tile *tile, struct drm_printer *p)
+ *    {
+ *        drm_printf(p, "%u\n", tile->id);
+ *        return 0;
+ *    }
+ *
+ *    static const struct drm_info_list info[] = {
+ *        { name = "id", .show = tile_debugfs_simple_show, .data = tile_id },
+ *    };
+ *
+ *    dir = debugfs_create_dir("tile0", parent);
+ *    dir->d_inode->i_private = tile;
+ *    drm_debugfs_create_files(info, ARRAY_SIZE(info), dir, minor);
+ *
+ * Return: 0 on success or a negative error code on failure.
+ */
+static int tile_debugfs_simple_show(struct seq_file *m, void *data)
+{
+	struct drm_printer p = drm_seq_file_printer(m);
+	struct drm_info_node *node = m->private;
+	struct xe_tile *tile = node_to_tile(node);
+	int (*print)(struct xe_tile *, struct drm_printer *) = node->info_ent->data;
+
+	return print(tile, &p);
+}
+
+/**
+ * tile_debugfs_show_with_rpm - A show callback for struct drm_info_list
+ * @m: the &seq_file
+ * @data: data used by the drm debugfs helpers
+ *
+ * Similar to tile_debugfs_simple_show() but implicitly takes a RPM ref.
+ *
+ * Return: 0 on success or a negative error code on failure.
+ */
+static int tile_debugfs_show_with_rpm(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = m->private;
+	struct xe_tile *tile = node_to_tile(node);
+	struct xe_device *xe = tile_to_xe(tile);
+	int ret;
+
+	xe_pm_runtime_get(xe);
+	ret = tile_debugfs_simple_show(m, data);
+	xe_pm_runtime_put(xe);
+
+	return ret;
+}
+
+static int sa_info(struct xe_tile *tile, struct drm_printer *p)
+{
+	drm_suballoc_dump_debug_info(&tile->mem.kernel_bb_pool->base, p,
+				     xe_sa_manager_gpu_addr(tile->mem.kernel_bb_pool));
+
+	return 0;
+}
+
+/* only for debugfs files which can be safely used on the VF */
+static const struct drm_info_list vf_safe_debugfs_list[] = {
+	{ "sa_info", .show = tile_debugfs_show_with_rpm, .data = sa_info },
+};
+
+/**
+ * xe_tile_debugfs_register - Register tile's debugfs attributes
+ * @tile: the &xe_tile to register
+ *
+ * Create debugfs sub-directory with a name that includes a tile ID and
+ * then creates set of debugfs files (attributes) specific to this tile.
+ */
+void xe_tile_debugfs_register(struct xe_tile *tile)
+{
+	struct xe_device *xe = tile_to_xe(tile);
+	struct drm_minor *minor = xe->drm.primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[8];
+
+	snprintf(name, sizeof(name), "tile%u", tile->id);
+	tile->debugfs = debugfs_create_dir(name, root);
+	if (IS_ERR(tile->debugfs))
+		return;
+
+	/*
+	 * Store the xe_tile pointer as private data of the tile/ directory
+	 * node so other tile specific attributes under that directory may
+	 * refer to it by looking at its parent node private data.
+	 */
+	tile->debugfs->d_inode->i_private = tile;
+
+	drm_debugfs_create_files(vf_safe_debugfs_list,
+				 ARRAY_SIZE(vf_safe_debugfs_list),
+				 tile->debugfs, minor);
+}

drivers/gpu/drm/xe/xe_tile_debugfs.h

···
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#ifndef _XE_TILE_DEBUGFS_H_
+#define _XE_TILE_DEBUGFS_H_
+
+struct xe_tile;
+
+void xe_tile_debugfs_register(struct xe_tile *tile);
+
+#endif

drivers/gpu/drm/xe/xe_tile_printk.h

···
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#ifndef _xe_tile_printk_H_
+#define _xe_tile_printk_H_
+
+#include "xe_printk.h"
+
+#define __XE_TILE_PRINTK_FMT(_tile, _fmt, _args...)	"Tile%u: " _fmt, (_tile)->id, ##_args
+
+#define xe_tile_printk(_tile, _level, _fmt, ...) \
+	xe_printk((_tile)->xe, _level, __XE_TILE_PRINTK_FMT((_tile), _fmt, ##__VA_ARGS__))
+
+#define xe_tile_err(_tile, _fmt, ...) \
+	xe_tile_printk((_tile), err, _fmt, ##__VA_ARGS__)
+
+#define xe_tile_err_once(_tile, _fmt, ...) \
+	xe_tile_printk((_tile), err_once, _fmt, ##__VA_ARGS__)
+
+#define xe_tile_err_ratelimited(_tile, _fmt, ...) \
+	xe_tile_printk((_tile), err_ratelimited, _fmt, ##__VA_ARGS__)
+
+#define xe_tile_warn(_tile, _fmt, ...) \
+	xe_tile_printk((_tile), warn, _fmt, ##__VA_ARGS__)
+
+#define xe_tile_notice(_tile, _fmt, ...) \
+	xe_tile_printk((_tile), notice, _fmt, ##__VA_ARGS__)
+
+#define xe_tile_info(_tile, _fmt, ...) \
+	xe_tile_printk((_tile), info, _fmt, ##__VA_ARGS__)
+
+#define xe_tile_dbg(_tile, _fmt, ...) \
+	xe_tile_printk((_tile), dbg, _fmt, ##__VA_ARGS__)
+
+#define xe_tile_WARN_type(_tile, _type, _condition, _fmt, ...) \
+	xe_WARN##_type((_tile)->xe, _condition, _fmt, ## __VA_ARGS__)
+
+#define xe_tile_WARN(_tile, _condition, _fmt, ...) \
+	xe_tile_WARN_type((_tile),, _condition, __XE_TILE_PRINTK_FMT((_tile), _fmt, ##__VA_ARGS__))
+
+#define xe_tile_WARN_ONCE(_tile, _condition, _fmt, ...) \
+	xe_tile_WARN_type((_tile), _ONCE, _condition, __XE_TILE_PRINTK_FMT((_tile), _fmt, ##__VA_ARGS__))
+
+#define xe_tile_WARN_ON(_tile, _condition) \
+	xe_tile_WARN((_tile), _condition, "%s(%s)", "WARN_ON", __stringify(_condition))
+
+#define xe_tile_WARN_ON_ONCE(_tile, _condition) \
+	xe_tile_WARN_ONCE((_tile), _condition, "%s(%s)", "WARN_ON_ONCE", __stringify(_condition))
+
+static inline void __xe_tile_printfn_err(struct drm_printer *p, struct va_format *vaf)
+{
+	struct xe_tile *tile = p->arg;
+
+	xe_tile_err(tile, "%pV", vaf);
+}
+
+static inline void __xe_tile_printfn_info(struct drm_printer *p, struct va_format *vaf)
+{
+	struct xe_tile *tile = p->arg;
+
+	xe_tile_info(tile, "%pV", vaf);
+}
+
+static inline void __xe_tile_printfn_dbg(struct drm_printer *p, struct va_format *vaf)
+{
+	struct xe_tile *tile = p->arg;
+	struct drm_printer dbg;
+
+	/*
+	 * The original xe_tile_dbg() callsite annotations are useless here,
+	 * redirect to the tweaked xe_dbg_printer() instead.
+	 */
+	dbg = xe_dbg_printer(tile->xe);
+	dbg.origin = p->origin;
+
+	drm_printf(&dbg, __XE_TILE_PRINTK_FMT(tile, "%pV", vaf));
+}
+
+/**
+ * xe_tile_err_printer - Construct a &drm_printer that outputs to xe_tile_err()
+ * @tile: the &xe_tile pointer to use in xe_tile_err()
+ *
+ * Return: The &drm_printer object.
+ */
+static inline struct drm_printer xe_tile_err_printer(struct xe_tile *tile)
+{
+	struct drm_printer p = {
+		.printfn = __xe_tile_printfn_err,
+		.arg = tile,
+	};
+	return p;
+}
+
+/**
+ * xe_tile_info_printer - Construct a &drm_printer that outputs to xe_tile_info()
+ * @tile: the &xe_tile pointer to use in xe_tile_info()
+ *
+ * Return: The &drm_printer object.
+ */
+static inline struct drm_printer xe_tile_info_printer(struct xe_tile *tile)
+{
+	struct drm_printer p = {
+		.printfn = __xe_tile_printfn_info,
+		.arg = tile,
+	};
+	return p;
+}
+
+/**
+ * xe_tile_dbg_printer - Construct a &drm_printer that outputs like xe_tile_dbg()
+ * @tile: the &xe_tile pointer to use in xe_tile_dbg()
+ *
+ * Return: The &drm_printer object.
+ */
+static inline struct drm_printer xe_tile_dbg_printer(struct xe_tile *tile)
+{
+	struct drm_printer p = {
+		.printfn = __xe_tile_printfn_dbg,
+		.arg = tile,
+		.origin = (const void *)_THIS_IP_,
+	};
+	return p;
+}
+
+#endif

drivers/gpu/drm/xe/xe_tile_sysfs.c

···
 	kt->tile = tile;
 
 	err = kobject_add(&kt->base, &dev->kobj, "tile%d", tile->id);
-	if (err) {
-		kobject_put(&kt->base);
-		return err;
-	}
+	if (err)
+		goto err_object;
 
 	tile->sysfs = &kt->base;
 
 	err = xe_vram_freq_sysfs_init(tile);
 	if (err)
-		return err;
+		goto err_object;
 
 	return devm_add_action_or_reset(xe->drm.dev, tile_sysfs_fini, tile);
+
+err_object:
+	kobject_put(&kt->base);
+	return err;
 }

drivers/gpu/drm/xe/xe_tlb_inval.c

···
 #include "xe_force_wake.h"
 #include "xe_gt.h"
 #include "xe_gt_printk.h"
+#include "xe_gt_stats.h"
 #include "xe_guc.h"
 #include "xe_guc_ct.h"
 #include "xe_guc_tlb_inval.h"
-#include "xe_gt_stats.h"
-#include "xe_tlb_inval.h"
 #include "xe_mmio.h"
 #include "xe_pm.h"
 #include "xe_tlb_inval.h"

drivers/gpu/drm/xe/xe_uc_fw.c

···
 #define XE_GT_TYPE_ANY XE_GT_TYPE_UNINITIALIZED
 
 #define XE_GUC_FIRMWARE_DEFS(fw_def, mmp_ver, major_ver)					\
-	fw_def(PANTHERLAKE,	GT_TYPE_ANY,	major_ver(xe,	guc,	ptl,	70, 47, 0))	\
-	fw_def(BATTLEMAGE,	GT_TYPE_ANY,	major_ver(xe,	guc,	bmg,	70, 45, 2))	\
+	fw_def(PANTHERLAKE,	GT_TYPE_ANY,	major_ver(xe,	guc,	ptl,	70, 49, 4))	\
+	fw_def(BATTLEMAGE,	GT_TYPE_ANY,	major_ver(xe,	guc,	bmg,	70, 49, 4))	\
 	fw_def(LUNARLAKE,	GT_TYPE_ANY,	major_ver(xe,	guc,	lnl,	70, 45, 2))	\
 	fw_def(METEORLAKE,	GT_TYPE_ANY,	major_ver(i915,	guc,	mtl,	70, 44, 1))	\
 	fw_def(DG2,		GT_TYPE_ANY,	major_ver(i915,	guc,	dg2,	70, 45, 2))	\
···
 	xe_uc_fw_change_status(uc_fw, XE_UC_FIRMWARE_SELECTED);
 }
 
-static int guc_read_css_info(struct xe_uc_fw *uc_fw, struct uc_css_header *css)
+static int guc_read_css_info(struct xe_uc_fw *uc_fw, struct uc_css_guc_info *guc_info)
 {
 	struct xe_gt *gt = uc_fw_to_gt(uc_fw);
 	struct xe_uc_fw_version *release = &uc_fw->versions.found[XE_UC_FW_VER_RELEASE];
···
 		return -EINVAL;
 	}
 
-	compatibility->major = FIELD_GET(CSS_SW_VERSION_UC_MAJOR, css->submission_version);
-	compatibility->minor = FIELD_GET(CSS_SW_VERSION_UC_MINOR, css->submission_version);
-	compatibility->patch = FIELD_GET(CSS_SW_VERSION_UC_PATCH, css->submission_version);
+	compatibility->major = FIELD_GET(CSS_SW_VERSION_UC_MAJOR, guc_info->submission_version);
+	compatibility->minor = FIELD_GET(CSS_SW_VERSION_UC_MINOR, guc_info->submission_version);
+	compatibility->patch = FIELD_GET(CSS_SW_VERSION_UC_PATCH, guc_info->submission_version);
 
-	uc_fw->private_data_size = css->private_data_size;
+	uc_fw->build_type = FIELD_GET(CSS_UKERNEL_INFO_BUILDTYPE, guc_info->ukernel_info);
+	uc_fw->private_data_size = guc_info->private_data_size;
 
 	return 0;
 }
···
 	css = (struct uc_css_header *)fw_data;
 
 	/* Check integrity of size values inside CSS header */
-	size = (css->header_size_dw - css->key_size_dw - css->modulus_size_dw -
-		css->exponent_size_dw) * sizeof(u32);
+	size = (css->header_size_dw - css->rsa_info.key_size_dw - css->rsa_info.modulus_size_dw -
+		css->rsa_info.exponent_size_dw) * sizeof(u32);
 	if (unlikely(size != sizeof(struct uc_css_header))) {
 		drm_warn(&xe->drm,
 			 "%s firmware %s: unexpected header size: %zu != %zu\n",
···
 	uc_fw->ucode_size = (css->size_dw - css->header_size_dw) * sizeof(u32);
 
 	/* now RSA */
-	uc_fw->rsa_size = css->key_size_dw * sizeof(u32);
+	uc_fw->rsa_size = css->rsa_info.key_size_dw * sizeof(u32);
 
 	/* At least, it should have header, uCode and RSA. Size of all three. */
 	size = sizeof(struct uc_css_header) + uc_fw->ucode_size +
···
 	}
 
 	/* Get version numbers from the CSS header */
-	release->major = FIELD_GET(CSS_SW_VERSION_UC_MAJOR, css->sw_version);
-	release->minor = FIELD_GET(CSS_SW_VERSION_UC_MINOR, css->sw_version);
-	release->patch = FIELD_GET(CSS_SW_VERSION_UC_PATCH, css->sw_version);
+	release->major = FIELD_GET(CSS_SW_VERSION_UC_MAJOR, css->guc_info.sw_version);
+	release->minor = FIELD_GET(CSS_SW_VERSION_UC_MINOR, css->guc_info.sw_version);
+	release->patch = FIELD_GET(CSS_SW_VERSION_UC_PATCH, css->guc_info.sw_version);
 
 	if (uc_fw->type == XE_UC_FW_TYPE_GUC)
-		return guc_read_css_info(uc_fw, css);
+		return guc_read_css_info(uc_fw, &css->guc_info);
 
 	return 0;
 }

drivers/gpu/drm/xe/xe_uc_fw_abi.h

···
  *    in fw. So driver will load a truncated firmware in this case.
  */
 
+struct uc_css_rsa_info {
+	u32 key_size_dw;
+	u32 modulus_size_dw;
+	u32 exponent_size_dw;
+} __packed;
+
+struct uc_css_guc_info {
+	u32 time;
+#define CSS_TIME_HOUR				(0xFF << 0)
+#define CSS_TIME_MIN				(0xFF << 8)
+#define CSS_TIME_SEC				(0xFFFF << 16)
+	u32 reserved0[5];
+	u32 sw_version;
+#define CSS_SW_VERSION_UC_MAJOR			(0xFF << 16)
+#define CSS_SW_VERSION_UC_MINOR			(0xFF << 8)
+#define CSS_SW_VERSION_UC_PATCH			(0xFF << 0)
+	u32 submission_version;
+	u32 reserved1[11];
+	u32 header_info;
+#define CSS_HEADER_INFO_SVN			(0xFF)
+#define CSS_HEADER_INFO_COPY_VALID		(0x1 << 31)
+	u32 private_data_size;
+	u32 ukernel_info;
+#define CSS_UKERNEL_INFO_DEVICEID		(0xFFFF << 16)
+#define CSS_UKERNEL_INFO_PRODKEY		(0xFF << 8)
+#define CSS_UKERNEL_INFO_BUILDTYPE		(0x3 << 2)
+#define CSS_UKERNEL_INFO_BUILDTYPE_PROD		0
+#define CSS_UKERNEL_INFO_BUILDTYPE_PREPROD	1
+#define CSS_UKERNEL_INFO_BUILDTYPE_DEBUG	2
+#define CSS_UKERNEL_INFO_ENCSTATUS		(0x1 << 1)
+#define CSS_UKERNEL_INFO_COPY_VALID		(0x1 << 0)
+} __packed;
+
 struct uc_css_header {
 	u32 module_type;
 	/*
···
 	 */
 	u32 header_size_dw;
 	u32 header_version;
-	u32 module_id;
+	u32 reserved0;
 	u32 module_vendor;
 	u32 date;
-#define CSS_DATE_DAY			(0xFF << 0)
-#define CSS_DATE_MONTH			(0xFF << 8)
-#define CSS_DATE_YEAR			(0xFFFF << 16)
+#define CSS_DATE_DAY				(0xFF << 0)
+#define CSS_DATE_MONTH				(0xFF << 8)
+#define CSS_DATE_YEAR				(0xFFFF << 16)
 	u32 size_dw; /* uCode plus header_size_dw */
-	u32 key_size_dw;
-	u32 modulus_size_dw;
-	u32 exponent_size_dw;
-	u32 time;
-#define CSS_TIME_HOUR			(0xFF << 0)
-#define CSS_DATE_MIN			(0xFF << 8)
-#define CSS_DATE_SEC			(0xFFFF << 16)
-	char username[8];
-	char buildnumber[12];
-	u32 sw_version;
-#define CSS_SW_VERSION_UC_MAJOR		(0xFF << 16)
-#define CSS_SW_VERSION_UC_MINOR		(0xFF << 8)
-#define CSS_SW_VERSION_UC_PATCH		(0xFF << 0)
 	union {
-		u32 submission_version; /* only applies to GuC */
-		u32 reserved2;
+		u32 reserved1[3];
+		struct uc_css_rsa_info rsa_info;
 	};
-	u32 reserved0[12];
 	union {
-		u32 private_data_size; /* only applies to GuC */
-		u32 reserved1;
+		u32 reserved2[22];
+		struct uc_css_guc_info guc_info;
 	};
-	u32 header_info;
 } __packed;
 static_assert(sizeof(struct uc_css_header) == 128);
 
···
 	u8 reserved3[56];
 	u32 modulus_size; /* in dwords */
 	u32 exponent_size; /* in dwords */
+} __packed;
+
+/**
+ * DOC: Late binding Firmware Layout
+ *
+ * The Late binding binary starts with FPT header, which contains locations
+ * of various partitions of the binary. Here we're interested in finding out
+ * manifest version. To the manifest version, we need to locate CPD header
+ * one of the entry in CPD header points to manifest header. Manifest header
+ * contains the version.
+ *
+ *      +================================================+
+ *      |  FPT Header                                    |
+ *      +================================================+
+ *      |  FPT entries[]                                 |
+ *      |      entry1                                    |
+ *      |      ...                                       |
+ *      |      entryX                                    |
+ *      |          "LTES"                                |
+ *      |          ...                                   |
+ *      |          offset  >-----------------------------|------o
+ *      +================================================+      |
+ *                                                              |
+ *      +================================================+      |
+ *      |  CPD Header                                    |<-----o
+ *      +================================================+
+ *      |  CPD entries[]                                 |
+ *      |      entry1                                    |
+ *      |      ...                                       |
+ *      |      entryX                                    |
+ *      |          "LTES.man"                            |
+ *      |           ...                                  |
+ *      |           offset  >----------------------------|------o
+ *      +================================================+      |
+ *                                                              |
+ *      +================================================+      |
+ *      |  Manifest Header                               |<-----o
+ *      |      ...                                       |
+ *      |      FW version                                |
+ *      |      ...                                       |
+ *      +================================================+
+ */
+
+/* FPT Headers */
+struct csc_fpt_header {
+	u32 header_marker;
+#define CSC_FPT_HEADER_MARKER 0x54504624
+	u32 num_of_entries;
+	u8 header_version;
+	u8 entry_version;
+	u8 header_length; /* in bytes */
+	u8 flags;
+	u16 ticks_to_add;
+	u16 tokens_to_add;
+	u32 uma_size;
+	u32 crc32;
+	struct gsc_version fitc_version;
+} __packed;
+
+struct csc_fpt_entry {
+	u8 name[4]; /* partition name */
+	u32 reserved1;
+	u32 offset; /* offset from beginning of CSE region */
+	u32 length; /* partition length in bytes */
+	u32 reserved2[3];
+	u32 partition_flags;
 } __packed;
 
 #endif

drivers/gpu/drm/xe/xe_uc_fw_types.h

···
 
 	/** @private_data_size: size of private data found in uC css header */
 	u32 private_data_size;
+
+	/** @build_type: Firmware build type (see CSS_UKERNEL_INFO_BUILDTYPE for definitions) */
+	u32 build_type;
 };
 
 #endif

drivers/gpu/drm/xe/xe_userptr.c

···
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#include "xe_userptr.h"
+
+#include <linux/mm.h>
+
+#include "xe_trace_bo.h"
+
+/**
+ * xe_vma_userptr_check_repin() - Advisory check for repin needed
+ * @uvma: The userptr vma
+ *
+ * Check if the userptr vma has been invalidated since last successful
+ * repin. The check is advisory only and can the function can be called
+ * without the vm->svm.gpusvm.notifier_lock held. There is no guarantee that the
+ * vma userptr will remain valid after a lockless check, so typically
+ * the call needs to be followed by a proper check under the notifier_lock.
+ *
+ * Return: 0 if userptr vma is valid, -EAGAIN otherwise; repin recommended.
+ */
+int xe_vma_userptr_check_repin(struct xe_userptr_vma *uvma)
+{
+	return mmu_interval_check_retry(&uvma->userptr.notifier,
+					uvma->userptr.pages.notifier_seq) ?
+		-EAGAIN : 0;
+}
+
+/**
+ * __xe_vm_userptr_needs_repin() - Check whether the VM does have userptrs
+ * that need repinning.
+ * @vm: The VM.
+ *
+ * This function checks for whether the VM has userptrs that need repinning,
+ * and provides a release-type barrier on the svm.gpusvm.notifier_lock after
+ * checking.
+ *
+ * Return: 0 if there are no userptrs needing repinning, -EAGAIN if there are.
+ */
+int __xe_vm_userptr_needs_repin(struct xe_vm *vm)
+{
+	lockdep_assert_held_read(&vm->svm.gpusvm.notifier_lock);
+
+	return (list_empty(&vm->userptr.repin_list) &&
+		list_empty(&vm->userptr.invalidated)) ? 0 : -EAGAIN;
+}
+
+int xe_vma_userptr_pin_pages(struct xe_userptr_vma *uvma)
+{
+	struct xe_vma *vma = &uvma->vma;
+	struct xe_vm *vm = xe_vma_vm(vma);
+	struct xe_device *xe = vm->xe;
+	struct drm_gpusvm_ctx ctx = {
+		.read_only = xe_vma_read_only(vma),
+	};
+
+	lockdep_assert_held(&vm->lock);
+	xe_assert(xe, xe_vma_is_userptr(vma));
+
+	if (vma->gpuva.flags & XE_VMA_DESTROYED)
+		return 0;
+
+	return drm_gpusvm_get_pages(&vm->svm.gpusvm, &uvma->userptr.pages,
+				    uvma->userptr.notifier.mm,
+				    &uvma->userptr.notifier,
+				    xe_vma_userptr(vma),
+				    xe_vma_userptr(vma) + xe_vma_size(vma),
+				    &ctx);
+}
+
+static void __vma_userptr_invalidate(struct xe_vm *vm, struct xe_userptr_vma *uvma)
+{
+	struct xe_userptr *userptr = &uvma->userptr;
+	struct xe_vma *vma = &uvma->vma;
+	struct dma_resv_iter cursor;
+	struct dma_fence *fence;
+	struct drm_gpusvm_ctx ctx = {
+		.in_notifier = true,
+		.read_only = xe_vma_read_only(vma),
+	};
+	long err;
+
+	/*
+	 * Tell exec and rebind worker they need to repin and rebind this
+	 * userptr.
+	 */
+	if (!xe_vm_in_fault_mode(vm) &&
+	    !(vma->gpuva.flags & XE_VMA_DESTROYED)) {
+		spin_lock(&vm->userptr.invalidated_lock);
+		list_move_tail(&userptr->invalidate_link,
+			       &vm->userptr.invalidated);
+		spin_unlock(&vm->userptr.invalidated_lock);
+	}
+
+	/*
+	 * Preempt fences turn into schedule disables, pipeline these.
+	 * Note that even in fault mode, we need to wait for binds and
+	 * unbinds to complete, and those are attached as BOOKMARK fences
+	 * to the vm.
+	 */
+	dma_resv_iter_begin(&cursor, xe_vm_resv(vm),
+			    DMA_RESV_USAGE_BOOKKEEP);
+	dma_resv_for_each_fence_unlocked(&cursor, fence)
+		dma_fence_enable_sw_signaling(fence);
+	dma_resv_iter_end(&cursor);
+
+	err = dma_resv_wait_timeout(xe_vm_resv(vm),
+				    DMA_RESV_USAGE_BOOKKEEP,
+				    false, MAX_SCHEDULE_TIMEOUT);
+	XE_WARN_ON(err <= 0);
+
+	if (xe_vm_in_fault_mode(vm) && userptr->initial_bind) {
+		err = xe_vm_invalidate_vma(vma);
+		XE_WARN_ON(err);
+	}
+
+	drm_gpusvm_unmap_pages(&vm->svm.gpusvm, &uvma->userptr.pages,
+			       xe_vma_size(vma) >> PAGE_SHIFT, &ctx);
+}
+
+static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni,
+				   const struct mmu_notifier_range *range,
+				   unsigned long cur_seq)
+{
+	struct xe_userptr_vma *uvma = container_of(mni, typeof(*uvma), userptr.notifier);
+	struct xe_vma *vma = &uvma->vma;
+	struct xe_vm *vm = xe_vma_vm(vma);
+
+	xe_assert(vm->xe, xe_vma_is_userptr(vma));
+	trace_xe_vma_userptr_invalidate(vma);
+
+	if (!mmu_notifier_range_blockable(range))
+		return false;
+
+	vm_dbg(&xe_vma_vm(vma)->xe->drm,
+	       "NOTIFIER: addr=0x%016llx, range=0x%016llx",
+		xe_vma_start(vma), xe_vma_size(vma));
+
+	down_write(&vm->svm.gpusvm.notifier_lock);
+	mmu_interval_set_seq(mni, cur_seq);
+
+	__vma_userptr_invalidate(vm, uvma);
+	up_write(&vm->svm.gpusvm.notifier_lock);
+	trace_xe_vma_userptr_invalidate_complete(vma);
+
+	return true;
+}
+
+static const struct mmu_interval_notifier_ops vma_userptr_notifier_ops = {
+	.invalidate = vma_userptr_invalidate,
+};
+
+#if IS_ENABLED(CONFIG_DRM_XE_USERPTR_INVAL_INJECT)
+/**
+ * xe_vma_userptr_force_invalidate() - force invalidate a userptr
+ * @uvma: The userptr vma to invalidate
+ *
+ * Perform a forced userptr invalidation for testing purposes.
+ */
+void xe_vma_userptr_force_invalidate(struct xe_userptr_vma *uvma)
+{
+	struct xe_vm *vm = xe_vma_vm(&uvma->vma);
+
+	/* Protect against concurrent userptr pinning */
+	lockdep_assert_held(&vm->lock);
+	/* Protect against concurrent notifiers */
+	lockdep_assert_held(&vm->svm.gpusvm.notifier_lock);
+	/*
+	 * Protect against concurrent instances of this function and
+	 * the critical exec sections
+	 */
+	xe_vm_assert_held(vm);
+
+	if (!mmu_interval_read_retry(&uvma->userptr.notifier,
+				     uvma->userptr.pages.notifier_seq))
+		uvma->userptr.pages.notifier_seq -= 2;
+	__vma_userptr_invalidate(vm, uvma);
+}
+#endif
+
+int xe_vm_userptr_pin(struct xe_vm *vm)
+{
+	struct xe_userptr_vma *uvma, *next;
+	int err = 0;
+
+	xe_assert(vm->xe, !xe_vm_in_fault_mode(vm));
+	lockdep_assert_held_write(&vm->lock);
+
+	/* Collect invalidated userptrs */
+	spin_lock(&vm->userptr.invalidated_lock);
+	xe_assert(vm->xe, list_empty(&vm->userptr.repin_list));
+	list_for_each_entry_safe(uvma, next, &vm->userptr.invalidated,
+				 userptr.invalidate_link) {
+		list_del_init(&uvma->userptr.invalidate_link);
+		list_add_tail(&uvma->userptr.repin_link,
+			      &vm->userptr.repin_list);
+	}
+	spin_unlock(&vm->userptr.invalidated_lock);
+
+	/* Pin and move to bind list */
+	list_for_each_entry_safe(uvma, next, &vm->userptr.repin_list,
+				 userptr.repin_link) {
+		err = xe_vma_userptr_pin_pages(uvma);
+		if (err == -EFAULT) {
+			list_del_init(&uvma->userptr.repin_link);
+			/*
+			 * We might have already done the pin once already, but
+			 * then had to retry before the re-bind happened, due
+			 * some other condition in the caller, but in the
+			 * meantime the userptr got dinged by the notifier such
+			 * that we need to revalidate here, but this time we hit
+			 * the EFAULT. In such a case make sure we remove
+			 * ourselves from the rebind list to avoid going down in
+			 * flames.
+			 */
+			if (!list_empty(&uvma->vma.combined_links.rebind))
+				list_del_init(&uvma->vma.combined_links.rebind);
+
+			/* Wait for pending binds */
+			xe_vm_lock(vm, false);
+			dma_resv_wait_timeout(xe_vm_resv(vm),
+					      DMA_RESV_USAGE_BOOKKEEP,
+					      false, MAX_SCHEDULE_TIMEOUT);
+
+			down_read(&vm->svm.gpusvm.notifier_lock);
+			err = xe_vm_invalidate_vma(&uvma->vma);
+			up_read(&vm->svm.gpusvm.notifier_lock);
+			xe_vm_unlock(vm);
+			if (err)
+				break;
+		} else {
+			if (err)
+				break;
+
+			list_del_init(&uvma->userptr.repin_link);
+			list_move_tail(&uvma->vma.combined_links.rebind,
+				       &vm->rebind_list);
+		}
+	}
+
+	if (err) {
+		down_write(&vm->svm.gpusvm.notifier_lock);
+		spin_lock(&vm->userptr.invalidated_lock);
+		list_for_each_entry_safe(uvma, next, &vm->userptr.repin_list,
+					 userptr.repin_link) {
+			list_del_init(&uvma->userptr.repin_link);
+			list_move_tail(&uvma->userptr.invalidate_link,
+				       &vm->userptr.invalidated);
+		}
+		spin_unlock(&vm->userptr.invalidated_lock);
+		up_write(&vm->svm.gpusvm.notifier_lock);
+	}
+	return err;
+}
+
+/**
+ * xe_vm_userptr_check_repin() - Check whether the VM might have userptrs
+ * that need repinning.
+ * @vm: The VM.
+ *
+ * This function does an advisory check for whether the VM has userptrs that
+ * need repinning.
+ *
+ * Return: 0 if there are no indications of userptrs needing repinning,
+ * -EAGAIN if there are.
+ */
+int xe_vm_userptr_check_repin(struct xe_vm *vm)
+{
+	return (list_empty_careful(&vm->userptr.repin_list) &&
+		list_empty_careful(&vm->userptr.invalidated)) ? 0 : -EAGAIN;
+}
+
+int xe_userptr_setup(struct xe_userptr_vma *uvma, unsigned long start,
+		     unsigned long range)
+{
+	struct xe_userptr *userptr = &uvma->userptr;
+	int err;
+
+	INIT_LIST_HEAD(&userptr->invalidate_link);
+	INIT_LIST_HEAD(&userptr->repin_link);
+
+	err = mmu_interval_notifier_insert(&userptr->notifier, current->mm,
+					   start, range,
+					   &vma_userptr_notifier_ops);
+	if (err)
+		return err;
+
+	userptr->pages.notifier_seq = LONG_MAX;
+
+	return 0;
+}
+
+void xe_userptr_remove(struct xe_userptr_vma *uvma)
+{
+	struct xe_vm *vm = xe_vma_vm(&uvma->vma);
+	struct xe_userptr *userptr = &uvma->userptr;
+
+	drm_gpusvm_free_pages(&vm->svm.gpusvm, &uvma->userptr.pages,
+			      xe_vma_size(&uvma->vma) >> PAGE_SHIFT);
+
+	/*
+	 * Since userptr pages are not pinned, we can't remove
+	 * the notifier until we're sure the GPU is not accessing
+	 * them anymore
+	 */
+	mmu_interval_notifier_remove(&userptr->notifier);
+}
+
+void xe_userptr_destroy(struct xe_userptr_vma *uvma)
+{
+	struct xe_vm *vm = xe_vma_vm(&uvma->vma);
+
+	spin_lock(&vm->userptr.invalidated_lock);
+	xe_assert(vm->xe, list_empty(&uvma->userptr.repin_link));
+	list_del(&uvma->userptr.invalidate_link);
+	spin_unlock(&vm->userptr.invalidated_lock);
+}

drivers/gpu/drm/xe/xe_userptr.h

···
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#ifndef _XE_USERPTR_H_
+#define _XE_USERPTR_H_
+
+#include <linux/list.h>
+#include <linux/mutex.h>
+#include <linux/notifier.h>
+#include <linux/scatterlist.h>
+#include <linux/spinlock.h>
+
+#include <drm/drm_gpusvm.h>
+
+struct xe_vm;
+struct xe_vma;
+struct xe_userptr_vma;
+
+/** struct xe_userptr_vm - User pointer VM level state */
+struct xe_userptr_vm {
+	/**
+	 * @userptr.repin_list: list of VMAs which are user pointers,
+	 * and needs repinning. Protected by @lock.
+	 */
+	struct list_head repin_list;
+	/**
+	 * @userptr.invalidated_lock: Protects the
+	 * @userptr.invalidated list.
+	 */
+	spinlock_t invalidated_lock;
+	/**
+	 * @userptr.invalidated: List of invalidated userptrs, not yet
+	 * picked
+	 * up for revalidation. Protected from access with the
+	 * @invalidated_lock. Removing items from the list
+	 * additionally requires @lock in write mode, and adding
+	 * items to the list requires either the @svm.gpusvm.notifier_lock in
+	 * write mode, OR @lock in write mode.
+	 */
+	struct list_head invalidated;
+};
+
+/** struct xe_userptr - User pointer */
+struct xe_userptr {
+	/** @invalidate_link: Link for the vm::userptr.invalidated list */
+	struct list_head invalidate_link;
+	/** @userptr: link into VM repin list if userptr. */
+	struct list_head repin_link;
+	/**
+	 * @pages: gpusvm pages for this user pointer.
+	 */
+	struct drm_gpusvm_pages pages;
+	/**
+	 * @notifier: MMU notifier for user pointer (invalidation call back)
+	 */
+	struct mmu_interval_notifier notifier;
+
+	/**
+	 * @initial_bind: user pointer has been bound at least once.
+	 * write: vm->svm.gpusvm.notifier_lock in read mode and vm->resv held.
+	 * read: vm->svm.gpusvm.notifier_lock in write mode or vm->resv held.
+	 */
+	bool initial_bind;
+#if IS_ENABLED(CONFIG_DRM_XE_USERPTR_INVAL_INJECT)
+	u32 divisor;
+#endif
+};
+
+#if IS_ENABLED(CONFIG_DRM_GPUSVM)
+void xe_userptr_remove(struct xe_userptr_vma *uvma);
+int xe_userptr_setup(struct xe_userptr_vma *uvma, unsigned long start,
+		     unsigned long range);
+void xe_userptr_destroy(struct xe_userptr_vma *uvma);
+
+int xe_vm_userptr_pin(struct xe_vm *vm);
+int __xe_vm_userptr_needs_repin(struct xe_vm *vm);
+int xe_vm_userptr_check_repin(struct xe_vm *vm);
+int xe_vma_userptr_pin_pages(struct xe_userptr_vma *uvma);
+int xe_vma_userptr_check_repin(struct xe_userptr_vma *uvma);
+#else
+static inline void xe_userptr_remove(struct xe_userptr_vma *uvma) {}
+
+static inline int xe_userptr_setup(struct xe_userptr_vma *uvma,
+				   unsigned long start, unsigned long range)
+{
+	return -ENODEV;
+}
+
+static inline void xe_userptr_destroy(struct xe_userptr_vma *uvma) {}
+
+static inline int xe_vm_userptr_pin(struct xe_vm *vm) { return 0; }
+static inline int __xe_vm_userptr_needs_repin(struct xe_vm *vm) { return 0; }
+static inline int xe_vm_userptr_check_repin(struct xe_vm *vm) { return 0; }
+static inline int xe_vma_userptr_pin_pages(struct xe_userptr_vma *uvma) { return -ENODEV; }
+static inline int xe_vma_userptr_check_repin(struct xe_userptr_vma *uvma) { return -ENODEV; };
+#endif
+
+#if IS_ENABLED(CONFIG_DRM_XE_USERPTR_INVAL_INJECT)
+void xe_vma_userptr_force_invalidate(struct xe_userptr_vma *uvma);
+#else
+static inline void xe_vma_userptr_force_invalidate(struct xe_userptr_vma *uvma)
+{
+}
+#endif
+#endif

drivers/gpu/drm/xe/xe_validation.c

···
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2024 Intel Corporation
+ */
+#include "xe_bo.h"
+#include <drm/drm_exec.h>
+#include <drm/drm_gem.h>
+#include <drm/drm_gpuvm.h>
+
+#include "xe_assert.h"
+#include "xe_validation.h"
+
+#ifdef CONFIG_DRM_XE_DEBUG
+/**
+ * xe_validation_assert_exec() - Assert that the drm_exec pointer is suitable
+ * for validation.
+ * @xe: Pointer to the xe device.
+ * @exec: The drm_exec pointer to check.
+ * @obj: Pointer to the object subject to validation.
+ *
+ * NULL exec pointers are not allowed.
+ * For XE_VALIDATION_UNIMPLEMENTED, no checking.
+ * For XE_VLIDATION_OPT_OUT, check that the caller is a kunit test
+ * For XE_VALIDATION_UNSUPPORTED, check that the object subject to
+ * validation is a dma-buf, for which support for ww locking is
+ * not in place in the dma-buf layer.
+ */
+void xe_validation_assert_exec(const struct xe_device *xe,
+			       const struct drm_exec *exec,
+			       const struct drm_gem_object *obj)
+{
+	xe_assert(xe, exec);
+	if (IS_ERR(exec)) {
+		switch (PTR_ERR(exec)) {
+		case __XE_VAL_UNIMPLEMENTED:
+			break;
+		case __XE_VAL_UNSUPPORTED:
+			xe_assert(xe, !!obj->dma_buf);
+			break;
+#if IS_ENABLED(CONFIG_KUNIT)
+		case __XE_VAL_OPT_OUT:
+			xe_assert(xe, current->kunit_test);
+			break;
+#endif
+		default:
+			xe_assert(xe, false);
+		}
+	}
+}
+#endif
+
+static int xe_validation_lock(struct xe_validation_ctx *ctx)
+{
+	struct xe_validation_device *val = ctx->val;
+	int ret = 0;
+
+	if (ctx->val_flags.interruptible) {
+		if (ctx->request_exclusive)
+			ret = down_write_killable(&val->lock);
+		else
+			ret = down_read_interruptible(&val->lock);
+	} else {
+		if (ctx->request_exclusive)
+			down_write(&val->lock);
+		else
+			down_read(&val->lock);
+	}
+
+	if (!ret) {
+		ctx->lock_held = true;
+		ctx->lock_held_exclusive = ctx->request_exclusive;
+	}
+
+	return ret;
+}
+
+static int xe_validation_trylock(struct xe_validation_ctx *ctx)
+{
+	struct xe_validation_device *val = ctx->val;
+	bool locked;
+
+	if (ctx->request_exclusive)
+		locked = down_write_trylock(&val->lock);
+	else
+		locked = down_read_trylock(&val->lock);
+
+	if (locked) {
+		ctx->lock_held = true;
+		ctx->lock_held_exclusive = ctx->request_exclusive;
+	}
+
+	return locked ? 0 : -EWOULDBLOCK;
+}
+
+static void xe_validation_unlock(struct xe_validation_ctx *ctx)
+{
+	if (!ctx->lock_held)
+		return;
+
+	if (ctx->lock_held_exclusive)
+		up_write(&ctx->val->lock);
+	else
+		up_read(&ctx->val->lock);
+
+	ctx->lock_held = false;
+}
+
+/**
+ * xe_validation_ctx_init() - Initialize an xe_validation_ctx
+ * @ctx: The xe_validation_ctx to initialize.
+ * @val: The xe_validation_device representing the validation domain.
+ * @exec: The struct drm_exec to use for the transaction. May be NULL.
+ * @flags: The flags to use for initialization.
+ *
+ * Initialize and lock a an xe_validation transaction using the validation domain
+ * represented by @val. Also initialize the drm_exec object forwarding parts of
+ * @flags to the drm_exec initialization. The @flags.exclusive flag should
+ * typically be set to false to avoid locking out other validators from the
+ * domain until an OOM is hit. For testing- or final attempt purposes it can,
+ * however, be set to true.
+ *
+ * Return: %0 on success, %-EINTR if interruptible initial locking failed with a
+ * signal pending. If @flags.no_block is set to true, a failed trylock
+ * returns %-EWOULDBLOCK.
+ */
+int xe_validation_ctx_init(struct xe_validation_ctx *ctx, struct xe_validation_device *val,
+			   struct drm_exec *exec, const struct xe_val_flags flags)
+{
+	int ret;
+
+	ctx->exec = exec;
+	ctx->val = val;
+	ctx->lock_held = false;
+	ctx->lock_held_exclusive = false;
+	ctx->request_exclusive = flags.exclusive;
+	ctx->val_flags = flags;
+	ctx->exec_flags = 0;
+	ctx->nr = 0;
+
+	if (flags.no_block)
+		ret = xe_validation_trylock(ctx);
+	else
+		ret = xe_validation_lock(ctx);
+	if (ret)
+		return ret;
+
+	if (exec) {
+		if (flags.interruptible)
+			ctx->exec_flags |= DRM_EXEC_INTERRUPTIBLE_WAIT;
+		if (flags.exec_ignore_duplicates)
+			ctx->exec_flags |= DRM_EXEC_IGNORE_DUPLICATES;
+		drm_exec_init(exec, ctx->exec_flags, ctx->nr);
+	}
+
+	return 0;
+}
+
+#ifdef CONFIG_DEBUG_WW_MUTEX_SLOWPATH
+/*
+ * This abuses both drm_exec and ww_mutex internals and should be
+ * replaced by checking for -EDEADLK when we can make TTM
+ * stop converting -EDEADLK to -ENOMEM.
+ * An alternative is to not have exhaustive eviction with
+ * CONFIG_DEBUG_WW_MUTEX_SLOWPATH until that happens.
+ */
+static bool xe_validation_contention_injected(struct drm_exec *exec)
+{
+	return !!exec->ticket.contending_lock;
+}
+
+#else
+
+static bool xe_validation_contention_injected(struct drm_exec *exec)
+{
+	return false;
+}
+
+#endif
+
+static bool __xe_validation_should_retry(struct xe_validation_ctx *ctx, int ret)
+{
+	if (ret == -ENOMEM &&
+	    ((ctx->request_exclusive &&
+	      xe_validation_contention_injected(ctx->exec)) ||
+	     !ctx->request_exclusive)) {
+		ctx->request_exclusive = true;
+		return true;
+	}
+
+	return false;
+}
+
+/**
+ * xe_validation_exec_lock() - Perform drm_gpuvm_exec_lock within a validation
+ * transaction.
+ * @ctx: An uninitialized xe_validation_ctx.
+ * @vm_exec: An initialized struct vm_exec.
+ * @val: The validation domain.
+ *
+ * The drm_gpuvm_exec_lock() function internally initializes its drm_exec
+ * transaction and therefore doesn't lend itself very well to be using
+ * xe_validation_ctx_init(). Provide a helper that takes an uninitialized
+ * xe_validation_ctx and calls drm_gpuvm_exec_lock() with OOM retry.
+ *
+ * Return: %0 on success, negative error code on failure.
+ */
+int xe_validation_exec_lock(struct xe_validation_ctx *ctx,
+			    struct drm_gpuvm_exec *vm_exec,
+			    struct xe_validation_device *val)
+{
+	int ret;
+
+	memset(ctx, 0, sizeof(*ctx));
+	ctx->exec = &vm_exec->exec;
+	ctx->exec_flags = vm_exec->flags;
+	ctx->val = val;
+	if (ctx->exec_flags & DRM_EXEC_INTERRUPTIBLE_WAIT)
+		ctx->val_flags.interruptible = 1;
+	if (ctx->exec_flags & DRM_EXEC_IGNORE_DUPLICATES)
+		ctx->val_flags.exec_ignore_duplicates = 1;
+retry:
+	ret = xe_validation_lock(ctx);
+	if (ret)
+		return ret;
+
+	ret = drm_gpuvm_exec_lock(vm_exec);
+	if (ret) {
+		xe_validation_unlock(ctx);
+		if (__xe_validation_should_retry(ctx, ret))
+			goto retry;
+	}
+
+	return ret;
+}
+
+/**
+ * xe_validation_ctx_fini() - Finalize a validation transaction
+ * @ctx: The Validation transaction to finalize.
+ *
+ * Finalize a validation transaction and its related drm_exec transaction.
+ */
+void xe_validation_ctx_fini(struct xe_validation_ctx *ctx)
+{
+	if (ctx->exec)
+		drm_exec_fini(ctx->exec);
+	xe_validation_unlock(ctx);
+}
+
+/**
+ * xe_validation_should_retry() - Determine if a validation transaction should retry
+ * @ctx: The validation transaction.
+ * @ret: Pointer to a return value variable.
+ *
+ * Determines whether a validation transaction should retry based on the
+ * internal transaction state and the return value pointed to by @ret.
+ * If a validation should be retried, the transaction is prepared for that,
+ * and the validation locked might be re-locked in exclusive mode, and *@ret
+ * is set to %0. If the re-locking errors, typically due to interruptible
+ * locking with signal pending, *@ret is instead set to -EINTR and the
+ * function returns %false.
+ *
+ * Return: %true if validation should be retried, %false otherwise.
+ */
+bool xe_validation_should_retry(struct xe_validation_ctx *ctx, int *ret)
+{
+	if (__xe_validation_should_retry(ctx, *ret)) {
+		drm_exec_fini(ctx->exec);
+		*ret = 0;
+		if (ctx->request_exclusive != ctx->lock_held_exclusive) {
+			xe_validation_unlock(ctx);
+			*ret = xe_validation_lock(ctx);
+		}
+		drm_exec_init(ctx->exec, ctx->exec_flags, ctx->nr);
+		return !*ret;
+	}
+
+	return false;
+}

drivers/gpu/drm/xe/xe_validation.h

···
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2024 Intel Corporation
+ */
+#ifndef _XE_VALIDATION_H_
+#define _XE_VALIDATION_H_
+
+#include <linux/dma-resv.h>
+#include <linux/types.h>
+#include <linux/rwsem.h>
+
+struct drm_exec;
+struct drm_gem_object;
+struct drm_gpuvm_exec;
+struct xe_device;
+
+#ifdef CONFIG_PROVE_LOCKING
+/**
+ * xe_validation_lockdep() - Assert that a drm_exec locking transaction can
+ * be initialized at this point.
+ */
+static inline void xe_validation_lockdep(void)
+{
+	struct ww_acquire_ctx ticket;
+
+	ww_acquire_init(&ticket, &reservation_ww_class);
+	ww_acquire_fini(&ticket);
+}
+#else
+static inline void xe_validation_lockdep(void)
+{
+}
+#endif
+
+/*
+ * Various values of the drm_exec pointer where we've not (yet)
+ * implemented full ww locking.
+ *
+ * XE_VALIDATION_UNIMPLEMENTED means implementation is pending.
+ * A lockdep check is made to assure that a drm_exec locking
+ * transaction can actually take place where the macro is
+ * used. If this asserts, the exec pointer needs to be assigned
+ * higher up in the callchain and passed down.
+ *
+ * XE_VALIDATION_UNSUPPORTED is for dma-buf code only where
+ * the dma-buf layer doesn't support WW locking.
+ *
+ * XE_VALIDATION_OPT_OUT is for simplification of kunit tests where
+ * exhaustive eviction isn't necessary.
+ */
+#define __XE_VAL_UNIMPLEMENTED -EINVAL
+#define XE_VALIDATION_UNIMPLEMENTED (xe_validation_lockdep(),		\
+				     (struct drm_exec *)ERR_PTR(__XE_VAL_UNIMPLEMENTED))
+
+#define __XE_VAL_UNSUPPORTED -EOPNOTSUPP
+#define XE_VALIDATION_UNSUPPORTED ((struct drm_exec *)ERR_PTR(__XE_VAL_UNSUPPORTED))
+
+#define __XE_VAL_OPT_OUT -ENOMEM
+#define XE_VALIDATION_OPT_OUT (xe_validation_lockdep(), \
+			       (struct drm_exec *)ERR_PTR(__XE_VAL_OPT_OUT))
+#ifdef CONFIG_DRM_XE_DEBUG
+void xe_validation_assert_exec(const struct xe_device *xe, const struct drm_exec *exec,
+			       const struct drm_gem_object *obj);
+#else
+#define xe_validation_assert_exec(_xe, _exec, _obj)	\
+	do {						\
+		(void)_xe; (void)_exec; (void)_obj;	\
+	} while (0)
+#endif
+
+/**
+ * struct xe_validation_device - The domain for exhaustive eviction
+ * @lock: The lock used to exclude other processes from allocating graphics memory
+ *
+ * The struct xe_validation_device represents the domain for which we want to use
+ * exhaustive eviction. The @lock is typically grabbed in read mode for allocations
+ * but when graphics memory allocation fails, it is retried with the write mode held.
+ */
+struct xe_validation_device {
+	struct rw_semaphore lock;
+};
+
+/**
+ * struct xe_val_flags - Flags for xe_validation_ctx_init().
+ * @exclusive: Start the validation transaction by locking out all other validators.
+ * @no_block:  Don't block on initialization.
+ * @interruptible: Block interruptible if blocking. Implies initializing the drm_exec
+ * context with the DRM_EXEC_INTERRUPTIBLE_WAIT flag.
+ * @exec_ignore_duplicates: Initialize the drm_exec context with the
+ * DRM_EXEC_IGNORE_DUPLICATES flag.
+ */
+struct xe_val_flags {
+	u32 exclusive :1;
+	u32 no_block :1;
+	u32 interruptible :1;
+	u32 exec_ignore_duplicates :1;
+};
+
+/**
+ * struct xe_validation_ctx - A struct drm_exec subclass with support for
+ * exhaustive eviction
+ * @exec: The drm_exec object base class. Note that we use a pointer instead of
+ * embedding to avoid diamond inheritance.
+ * @val: The exhaustive eviction domain.
+ * @val_flags: Copy of the struct xe_val_flags passed to xe_validation_ctx_init.
+ * @lock_held: Whether The domain lock is currently held.
+ * @lock_held_exclusive: Whether the domain lock is held in exclusive mode.
+ * @request_exclusive: Whether to lock exclusively (write mode) the next time
+ * the domain lock is locked.
+ * @exec_flags: The drm_exec flags used for drm_exec (re-)initialization.
+ * @nr: The drm_exec nr parameter used for drm_exec (re-)initializaiton.
+ */
+struct xe_validation_ctx {
+	struct drm_exec *exec;
+	struct xe_validation_device *val;
+	struct xe_val_flags val_flags;
+	bool lock_held;
+	bool lock_held_exclusive;
+	bool request_exclusive;
+	u32 exec_flags;
+	unsigned int nr;
+};
+
+int xe_validation_ctx_init(struct xe_validation_ctx *ctx, struct xe_validation_device *val,
+			   struct drm_exec *exec, const struct xe_val_flags flags);
+
+int xe_validation_exec_lock(struct xe_validation_ctx *ctx, struct drm_gpuvm_exec *vm_exec,
+			    struct xe_validation_device *val);
+
+void xe_validation_ctx_fini(struct xe_validation_ctx *ctx);
+
+bool xe_validation_should_retry(struct xe_validation_ctx *ctx, int *ret);
+
+/**
+ * xe_validation_retry_on_oom() - Retry on oom in an xe_validaton transaction
+ * @_ctx: Pointer to the xe_validation_ctx
+ * @_ret: The current error value possibly holding -ENOMEM
+ *
+ * Use this in way similar to drm_exec_retry_on_contention().
+ * If @_ret contains -ENOMEM the tranaction is restarted once in a way that
+ * blocks other transactions and allows exhastive eviction. If the transaction
+ * was already restarted once, Just return the -ENOMEM. May also set
+ * _ret to -EINTR if not retrying and waits are interruptible.
+ * May only be used within a drm_exec_until_all_locked() loop.
+ */
+#define xe_validation_retry_on_oom(_ctx, _ret)				\
+	do {								\
+		if (xe_validation_should_retry(_ctx, _ret))		\
+			goto *__drm_exec_retry_ptr;			\
+	} while (0)
+
+/**
+ * xe_validation_device_init - Initialize a struct xe_validation_device
+ * @val: The xe_validation_device to init.
+ */
+static inline void
+xe_validation_device_init(struct xe_validation_device *val)
+{
+	init_rwsem(&val->lock);
+}
+
+/*
+ * Make guard() and scoped_guard() work with xe_validation_ctx
+ * so that we can exit transactions without caring about the
+ * cleanup.
+ */
+DEFINE_CLASS(xe_validation, struct xe_validation_ctx *,
+	     if (_T) xe_validation_ctx_fini(_T);,
+	     ({_ret = xe_validation_ctx_init(_ctx, _val, _exec, _flags);
+	       _ret ? NULL : _ctx; }),
+	     struct xe_validation_ctx *_ctx, struct xe_validation_device *_val,
+	     struct drm_exec *_exec, const struct xe_val_flags _flags, int _ret);
+static inline void *class_xe_validation_lock_ptr(class_xe_validation_t *_T)
+{return *_T; }
+#define class_xe_validation_is_conditional true
+
+/**
+ * xe_validation_guard() - An auto-cleanup xe_validation_ctx transaction
+ * @_ctx: The xe_validation_ctx.
+ * @_val: The xe_validation_device.
+ * @_exec: The struct drm_exec object
+ * @_flags: Flags for the xe_validation_ctx initialization.
+ * @_ret: Return in / out parameter. May be set by this macro. Typicall 0 when called.
+ *
+ * This macro is will initiate a drm_exec transaction with additional support for
+ * exhaustive eviction.
+ */
+#define xe_validation_guard(_ctx, _val, _exec, _flags, _ret)		\
+	scoped_guard(xe_validation, _ctx, _val, _exec, _flags, _ret) \
+	drm_exec_until_all_locked(_exec)
+
+#endif

drivers/gpu/drm/xe/xe_vm.c

···
 #include "xe_tlb_inval.h"
 #include "xe_trace_bo.h"
 #include "xe_wa.h"
-#include "xe_hmm.h"
 
 static struct drm_gem_object *xe_vm_obj(struct xe_vm *vm)
 {
···
 }
 
 /**
- * xe_vma_userptr_check_repin() - Advisory check for repin needed
- * @uvma: The userptr vma
+ * xe_vm_drm_exec_lock() - Lock the vm's resv with a drm_exec transaction
+ * @vm: The vm whose resv is to be locked.
+ * @exec: The drm_exec transaction.
  *
- * Check if the userptr vma has been invalidated since last successful
- * repin. The check is advisory only and can the function can be called
- * without the vm->userptr.notifier_lock held. There is no guarantee that the
- * vma userptr will remain valid after a lockless check, so typically
- * the call needs to be followed by a proper check under the notifier_lock.
+ * Helper to lock the vm's resv as part of a drm_exec transaction.
  *
- * Return: 0 if userptr vma is valid, -EAGAIN otherwise; repin recommended.
+ * Return: %0 on success. See drm_exec_lock_obj() for error codes.
  */
-int xe_vma_userptr_check_repin(struct xe_userptr_vma *uvma)
+int xe_vm_drm_exec_lock(struct xe_vm *vm, struct drm_exec *exec)
 {
-	return mmu_interval_check_retry(&uvma->userptr.notifier,
-					uvma->userptr.notifier_seq) ?
-		-EAGAIN : 0;
-}
-
-int xe_vma_userptr_pin_pages(struct xe_userptr_vma *uvma)
-{
-	struct xe_vma *vma = &uvma->vma;
-	struct xe_vm *vm = xe_vma_vm(vma);
-	struct xe_device *xe = vm->xe;
-
-	lockdep_assert_held(&vm->lock);
-	xe_assert(xe, xe_vma_is_userptr(vma));
-
-	return xe_hmm_userptr_populate_range(uvma, false);
+	return drm_exec_lock_obj(exec, xe_vm_obj(vm));
 }
 
 static bool preempt_fences_waiting(struct xe_vm *vm)
···
 		.num_fences = 1,
 	};
 	struct drm_exec *exec = &vm_exec.exec;
+	struct xe_validation_ctx ctx;
 	struct dma_fence *pfence;
 	int err;
 	bool wait;
···
 	xe_assert(vm->xe, xe_vm_in_preempt_fence_mode(vm));
 
 	down_write(&vm->lock);
-	err = drm_gpuvm_exec_lock(&vm_exec);
+	err = xe_validation_exec_lock(&ctx, &vm_exec, &vm->xe->val);
 	if (err)
 		goto out_up_write;
 
 	pfence = xe_preempt_fence_create(q, q->lr.context,
 					 ++q->lr.seqno);
-	if (!pfence) {
-		err = -ENOMEM;
+	if (IS_ERR(pfence)) {
+		err = PTR_ERR(pfence);
 		goto out_fini;
 	}
 
···
 	++vm->preempt.num_exec_queues;
 	q->lr.pfence = pfence;
 
-	down_read(&vm->userptr.notifier_lock);
+	xe_svm_notifier_lock(vm);
 
 	drm_gpuvm_resv_add_fence(&vm->gpuvm, exec, pfence,
 				 DMA_RESV_USAGE_BOOKKEEP, DMA_RESV_USAGE_BOOKKEEP);
···
 	if (wait)
 		dma_fence_enable_sw_signaling(pfence);
 
-	up_read(&vm->userptr.notifier_lock);
+	xe_svm_notifier_unlock(vm);
 
 out_fini:
-	drm_exec_fini(exec);
+	xe_validation_ctx_fini(&ctx);
 out_up_write:
 	up_write(&vm->lock);
 
···
 		q->lr.pfence = NULL;
 	}
 	up_write(&vm->lock);
-}
-
-/**
- * __xe_vm_userptr_needs_repin() - Check whether the VM does have userptrs
- * that need repinning.
- * @vm: The VM.
- *
- * This function checks for whether the VM has userptrs that need repinning,
- * and provides a release-type barrier on the userptr.notifier_lock after
- * checking.
- *
- * Return: 0 if there are no userptrs needing repinning, -EAGAIN if there are.
- */
-int __xe_vm_userptr_needs_repin(struct xe_vm *vm)
-{
-	lockdep_assert_held_read(&vm->userptr.notifier_lock);
-
-	return (list_empty(&vm->userptr.repin_list) &&
-		list_empty(&vm->userptr.invalidated)) ? 0 : -EAGAIN;
 }
 
 #define XE_VM_REBIND_RETRY_TIMEOUT_MS 1000
···
 	/* TODO: Inform user the VM is banned */
 }
 
-/**
- * xe_vm_validate_should_retry() - Whether to retry after a validate error.
- * @exec: The drm_exec object used for locking before validation.
- * @err: The error returned from ttm_bo_validate().
- * @end: A ktime_t cookie that should be set to 0 before first use and
- * that should be reused on subsequent calls.
- *
- * With multiple active VMs, under memory pressure, it is possible that
- * ttm_bo_validate() run into -EDEADLK and in such case returns -ENOMEM.
- * Until ttm properly handles locking in such scenarios, best thing the
- * driver can do is retry with a timeout. Check if that is necessary, and
- * if so unlock the drm_exec's objects while keeping the ticket to prepare
- * for a rerun.
- *
- * Return: true if a retry after drm_exec_init() is recommended;
- * false otherwise.
- */
-bool xe_vm_validate_should_retry(struct drm_exec *exec, int err, ktime_t *end)
-{
-	ktime_t cur;
-
-	if (err != -ENOMEM)
-		return false;
-
-	cur = ktime_get();
-	*end = *end ? : ktime_add_ms(cur, XE_VM_REBIND_RETRY_TIMEOUT_MS);
-	if (!ktime_before(cur, *end))
-		return false;
-
-	msleep(20);
-	return true;
-}
-
 static int xe_gpuvm_validate(struct drm_gpuvm_bo *vm_bo, struct drm_exec *exec)
 {
 	struct xe_vm *vm = gpuvm_to_vm(vm_bo->vm);
···
 	if (!try_wait_for_completion(&vm->xe->pm_block))
 		return -EAGAIN;
 
-	ret = xe_bo_validate(gem_to_xe_bo(vm_bo->obj), vm, false);
+	ret = xe_bo_validate(gem_to_xe_bo(vm_bo->obj), vm, false, exec);
 	if (ret)
 		return ret;
 
···
 static void preempt_rebind_work_func(struct work_struct *w)
 {
 	struct xe_vm *vm = container_of(w, struct xe_vm, preempt.rebind_work);
+	struct xe_validation_ctx ctx;
 	struct drm_exec exec;
 	unsigned int fence_count = 0;
 	LIST_HEAD(preempt_fences);
-	ktime_t end = 0;
 	int err = 0;
 	long wait;
 	int __maybe_unused tries = 0;
···
 			goto out_unlock_outer;
 	}
 
-	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0);
+	err = xe_validation_ctx_init(&ctx, &vm->xe->val, &exec,
+				     (struct xe_val_flags) {.interruptible = true});
+	if (err)
+		goto out_unlock_outer;
 
 	drm_exec_until_all_locked(&exec) {
 		bool done = false;
 
 		err = xe_preempt_work_begin(&exec, vm, &done);
 		drm_exec_retry_on_contention(&exec);
+		xe_validation_retry_on_oom(&ctx, &err);
 		if (err || done) {
-			drm_exec_fini(&exec);
-			if (err && xe_vm_validate_should_retry(&exec, err, &end))
-				err = -EAGAIN;
-
+			xe_validation_ctx_fini(&ctx);
 			goto out_unlock_outer;
 		}
 	}
···
 	if (err)
 		goto out_unlock;
 
+	xe_vm_set_validation_exec(vm, &exec);
 	err = xe_vm_rebind(vm, true);
+	xe_vm_set_validation_exec(vm, NULL);
 	if (err)
 		goto out_unlock;
 
···
 	(!(__tries)++ || __xe_vm_userptr_needs_repin(__vm)) : \
 	__xe_vm_userptr_needs_repin(__vm))
 
-	down_read(&vm->userptr.notifier_lock);
+	xe_svm_notifier_lock(vm);
 	if (retry_required(tries, vm)) {
-		up_read(&vm->userptr.notifier_lock);
+		xe_svm_notifier_unlock(vm);
 		err = -EAGAIN;
 		goto out_unlock;
 	}
···
 	/* Point of no return. */
 	arm_preempt_fences(vm, &preempt_fences);
 	resume_and_reinstall_preempt_fences(vm, &exec);
-	up_read(&vm->userptr.notifier_lock);
+	xe_svm_notifier_unlock(vm);
 
 out_unlock:
-	drm_exec_fini(&exec);
+	xe_validation_ctx_fini(&ctx);
 out_unlock_outer:
 	if (err == -EAGAIN) {
 		trace_xe_vm_rebind_worker_retry(vm);
···
 	free_preempt_fences(&preempt_fences);
 
 	trace_xe_vm_rebind_worker_exit(vm);
-}
-
-static void __vma_userptr_invalidate(struct xe_vm *vm, struct xe_userptr_vma *uvma)
-{
-	struct xe_userptr *userptr = &uvma->userptr;
-	struct xe_vma *vma = &uvma->vma;
-	struct dma_resv_iter cursor;
-	struct dma_fence *fence;
-	long err;
-
-	/*
-	 * Tell exec and rebind worker they need to repin and rebind this
-	 * userptr.
-	 */
-	if (!xe_vm_in_fault_mode(vm) &&
-	    !(vma->gpuva.flags & XE_VMA_DESTROYED)) {
-		spin_lock(&vm->userptr.invalidated_lock);
-		list_move_tail(&userptr->invalidate_link,
-			       &vm->userptr.invalidated);
-		spin_unlock(&vm->userptr.invalidated_lock);
-	}
-
-	/*
-	 * Preempt fences turn into schedule disables, pipeline these.
-	 * Note that even in fault mode, we need to wait for binds and
-	 * unbinds to complete, and those are attached as BOOKMARK fences
-	 * to the vm.
-	 */
-	dma_resv_iter_begin(&cursor, xe_vm_resv(vm),
-			    DMA_RESV_USAGE_BOOKKEEP);
-	dma_resv_for_each_fence_unlocked(&cursor, fence)
-		dma_fence_enable_sw_signaling(fence);
-	dma_resv_iter_end(&cursor);
-
-	err = dma_resv_wait_timeout(xe_vm_resv(vm),
-				    DMA_RESV_USAGE_BOOKKEEP,
-				    false, MAX_SCHEDULE_TIMEOUT);
-	XE_WARN_ON(err <= 0);
-
-	if (xe_vm_in_fault_mode(vm) && userptr->initial_bind) {
-		err = xe_vm_invalidate_vma(vma);
-		XE_WARN_ON(err);
-	}
-
-	xe_hmm_userptr_unmap(uvma);
-}
-
-static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni,
-				   const struct mmu_notifier_range *range,
-				   unsigned long cur_seq)
-{
-	struct xe_userptr_vma *uvma = container_of(mni, typeof(*uvma), userptr.notifier);
-	struct xe_vma *vma = &uvma->vma;
-	struct xe_vm *vm = xe_vma_vm(vma);
-
-	xe_assert(vm->xe, xe_vma_is_userptr(vma));
-	trace_xe_vma_userptr_invalidate(vma);
-
-	if (!mmu_notifier_range_blockable(range))
-		return false;
-
-	vm_dbg(&xe_vma_vm(vma)->xe->drm,
-	       "NOTIFIER: addr=0x%016llx, range=0x%016llx",
-		xe_vma_start(vma), xe_vma_size(vma));
-
-	down_write(&vm->userptr.notifier_lock);
-	mmu_interval_set_seq(mni, cur_seq);
-
-	__vma_userptr_invalidate(vm, uvma);
-	up_write(&vm->userptr.notifier_lock);
-	trace_xe_vma_userptr_invalidate_complete(vma);
-
-	return true;
-}
-
-static const struct mmu_interval_notifier_ops vma_userptr_notifier_ops = {
-	.invalidate = vma_userptr_invalidate,
-};
-
-#if IS_ENABLED(CONFIG_DRM_XE_USERPTR_INVAL_INJECT)
-/**
- * xe_vma_userptr_force_invalidate() - force invalidate a userptr
- * @uvma: The userptr vma to invalidate
- *
- * Perform a forced userptr invalidation for testing purposes.
- */
-void xe_vma_userptr_force_invalidate(struct xe_userptr_vma *uvma)
-{
-	struct xe_vm *vm = xe_vma_vm(&uvma->vma);
-
-	/* Protect against concurrent userptr pinning */
-	lockdep_assert_held(&vm->lock);
-	/* Protect against concurrent notifiers */
-	lockdep_assert_held(&vm->userptr.notifier_lock);
-	/*
-	 * Protect against concurrent instances of this function and
-	 * the critical exec sections
-	 */
-	xe_vm_assert_held(vm);
-
-	if (!mmu_interval_read_retry(&uvma->userptr.notifier,
-				     uvma->userptr.notifier_seq))
-		uvma->userptr.notifier_seq -= 2;
-	__vma_userptr_invalidate(vm, uvma);
-}
-#endif
-
-int xe_vm_userptr_pin(struct xe_vm *vm)
-{
-	struct xe_userptr_vma *uvma, *next;
-	int err = 0;
-
-	xe_assert(vm->xe, !xe_vm_in_fault_mode(vm));
-	lockdep_assert_held_write(&vm->lock);
-
-	/* Collect invalidated userptrs */
-	spin_lock(&vm->userptr.invalidated_lock);
-	xe_assert(vm->xe, list_empty(&vm->userptr.repin_list));
-	list_for_each_entry_safe(uvma, next, &vm->userptr.invalidated,
-				 userptr.invalidate_link) {
-		list_del_init(&uvma->userptr.invalidate_link);
-		list_add_tail(&uvma->userptr.repin_link,
-			      &vm->userptr.repin_list);
-	}
-	spin_unlock(&vm->userptr.invalidated_lock);
-
-	/* Pin and move to bind list */
-	list_for_each_entry_safe(uvma, next, &vm->userptr.repin_list,
-				 userptr.repin_link) {
-		err = xe_vma_userptr_pin_pages(uvma);
-		if (err == -EFAULT) {
-			list_del_init(&uvma->userptr.repin_link);
-			/*
-			 * We might have already done the pin once already, but
-			 * then had to retry before the re-bind happened, due
-			 * some other condition in the caller, but in the
-			 * meantime the userptr got dinged by the notifier such
-			 * that we need to revalidate here, but this time we hit
-			 * the EFAULT. In such a case make sure we remove
-			 * ourselves from the rebind list to avoid going down in
-			 * flames.
-			 */
-			if (!list_empty(&uvma->vma.combined_links.rebind))
-				list_del_init(&uvma->vma.combined_links.rebind);
-
-			/* Wait for pending binds */
-			xe_vm_lock(vm, false);
-			dma_resv_wait_timeout(xe_vm_resv(vm),
-					      DMA_RESV_USAGE_BOOKKEEP,
-					      false, MAX_SCHEDULE_TIMEOUT);
-
-			down_read(&vm->userptr.notifier_lock);
-			err = xe_vm_invalidate_vma(&uvma->vma);
-			up_read(&vm->userptr.notifier_lock);
-			xe_vm_unlock(vm);
-			if (err)
-				break;
-		} else {
-			if (err)
-				break;
-
-			list_del_init(&uvma->userptr.repin_link);
-			list_move_tail(&uvma->vma.combined_links.rebind,
-				       &vm->rebind_list);
-		}
-	}
-
-	if (err) {
-		down_write(&vm->userptr.notifier_lock);
-		spin_lock(&vm->userptr.invalidated_lock);
-		list_for_each_entry_safe(uvma, next, &vm->userptr.repin_list,
-					 userptr.repin_link) {
-			list_del_init(&uvma->userptr.repin_link);
-			list_move_tail(&uvma->userptr.invalidate_link,
-				       &vm->userptr.invalidated);
-		}
-		spin_unlock(&vm->userptr.invalidated_lock);
-		up_write(&vm->userptr.notifier_lock);
-	}
-	return err;
-}
-
-/**
- * xe_vm_userptr_check_repin() - Check whether the VM might have userptrs
- * that need repinning.
- * @vm: The VM.
- *
- * This function does an advisory check for whether the VM has userptrs that
- * need repinning.
- *
- * Return: 0 if there are no indications of userptrs needing repinning,
- * -EAGAIN if there are.
- */
-int xe_vm_userptr_check_repin(struct xe_vm *vm)
-{
-	return (list_empty_careful(&vm->userptr.repin_list) &&
-		list_empty_careful(&vm->userptr.invalidated)) ? 0 : -EAGAIN;
 }
 
 static int xe_vma_ops_alloc(struct xe_vma_ops *vops, bool array_of_binds)
···
 		drm_gpuvm_bo_put(vm_bo);
 	} else /* userptr or null */ {
 		if (!is_null && !is_cpu_addr_mirror) {
-			struct xe_userptr *userptr = &to_userptr_vma(vma)->userptr;
+			struct xe_userptr_vma *uvma = to_userptr_vma(vma);
 			u64 size = end - start + 1;
 			int err;
 
-			INIT_LIST_HEAD(&userptr->invalidate_link);
-			INIT_LIST_HEAD(&userptr->repin_link);
 			vma->gpuva.gem.offset = bo_offset_or_userptr;
-			mutex_init(&userptr->unmap_mutex);
 
-			err = mmu_interval_notifier_insert(&userptr->notifier,
-							   current->mm,
-							   xe_vma_userptr(vma), size,
-							   &vma_userptr_notifier_ops);
+			err = xe_userptr_setup(uvma, xe_vma_userptr(vma), size);
 			if (err) {
 				xe_vma_free(vma);
 				return ERR_PTR(err);
 			}
-
-			userptr->notifier_seq = LONG_MAX;
 		}
 
 		xe_vm_get(vm);
···
 
 	if (xe_vma_is_userptr(vma)) {
 		struct xe_userptr_vma *uvma = to_userptr_vma(vma);
-		struct xe_userptr *userptr = &uvma->userptr;
 
-		if (userptr->sg)
-			xe_hmm_userptr_free_sg(uvma);
-
-		/*
-		 * Since userptr pages are not pinned, we can't remove
-		 * the notifier until we're sure the GPU is not accessing
-		 * them anymore
-		 */
-		mmu_interval_notifier_remove(&userptr->notifier);
-		mutex_destroy(&userptr->unmap_mutex);
+		xe_userptr_remove(uvma);
 		xe_vm_put(vm);
 	} else if (xe_vma_is_null(vma) || xe_vma_is_cpu_addr_mirror(vma)) {
 		xe_vm_put(vm);
···
 
 	if (xe_vma_is_userptr(vma)) {
 		xe_assert(vm->xe, vma->gpuva.flags & XE_VMA_DESTROYED);
-
-		spin_lock(&vm->userptr.invalidated_lock);
-		xe_assert(vm->xe, list_empty(&to_userptr_vma(vma)->userptr.repin_link));
-		list_del(&to_userptr_vma(vma)->userptr.invalidate_link);
-		spin_unlock(&vm->userptr.invalidated_lock);
+		xe_userptr_destroy(to_userptr_vma(vma));
 	} else if (!xe_vma_is_null(vma) && !xe_vma_is_cpu_addr_mirror(vma)) {
 		xe_bo_assert_held(xe_vma_bo(vma));
 
···
 
 static void xe_vma_destroy_unlocked(struct xe_vma *vma)
 {
+	struct xe_device *xe = xe_vma_vm(vma)->xe;
+	struct xe_validation_ctx ctx;
 	struct drm_exec exec;
-	int err;
+	int err = 0;
 
-	drm_exec_init(&exec, 0, 0);
-	drm_exec_until_all_locked(&exec) {
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {}, err) {
 		err = xe_vm_lock_vma(&exec, vma);
 		drm_exec_retry_on_contention(&exec);
 		if (XE_WARN_ON(err))
 			break;
+		xe_vma_destroy(vma, NULL);
 	}
-
-	xe_vma_destroy(vma, NULL);
-
-	drm_exec_fini(&exec);
+	xe_assert(xe, !err);
 }
 
 struct xe_vma *
···
  * @xe: xe device.
  * @tile: tile to set up for.
  * @vm: vm to set up for.
+ * @exec: The struct drm_exec object used to lock the vm resv.
  *
  * Sets up a pagetable tree with one page-table per level and a single
  * leaf PTE. All pagetable entries point to the single page-table or,
···
  * Return: 0 on success, negative error code on error.
  */
 static int xe_vm_create_scratch(struct xe_device *xe, struct xe_tile *tile,
-				struct xe_vm *vm)
+				struct xe_vm *vm, struct drm_exec *exec)
 {
 	u8 id = tile->id;
 	int i;
 
 	for (i = MAX_HUGEPTE_LEVEL; i < vm->pt_root[id]->level; i++) {
-		vm->scratch_pt[id][i] = xe_pt_create(vm, tile, i);
+		vm->scratch_pt[id][i] = xe_pt_create(vm, tile, i, exec);
 		if (IS_ERR(vm->scratch_pt[id][i])) {
 			int err = PTR_ERR(vm->scratch_pt[id][i]);
 
 			vm->scratch_pt[id][i] = NULL;
 			return err;
 		}
-
 		xe_pt_populate_empty(tile, vm, vm->scratch_pt[id][i]);
 	}
 
···
 	}
 }
 
+static void xe_vm_pt_destroy(struct xe_vm *vm)
+{
+	struct xe_tile *tile;
+	u8 id;
+
+	xe_vm_assert_held(vm);
+
+	for_each_tile(tile, vm->xe, id) {
+		if (vm->pt_root[id]) {
+			xe_pt_destroy(vm->pt_root[id], vm->flags, NULL);
+			vm->pt_root[id] = NULL;
+		}
+	}
+}
+
 struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags, struct xe_file *xef)
 {
 	struct drm_gem_object *vm_resv_obj;
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
 	struct xe_vm *vm;
 	int err, number_tiles = 0;
 	struct xe_tile *tile;
···
 
 	INIT_LIST_HEAD(&vm->userptr.repin_list);
 	INIT_LIST_HEAD(&vm->userptr.invalidated);
-	init_rwsem(&vm->userptr.notifier_lock);
 	spin_lock_init(&vm->userptr.invalidated_lock);
 
 	ttm_lru_bulk_move_init(&vm->lru_bulk_move);
···
 		INIT_LIST_HEAD(&vm->preempt.pm_activate_link);
 	}
 
-	if (flags & XE_VM_FLAG_FAULT_MODE) {
-		err = xe_svm_init(vm);
-		if (err)
-			goto err_no_resv;
-	}
+	err = xe_svm_init(vm);
+	if (err)
+		goto err_no_resv;
 
 	vm_resv_obj = drm_gpuvm_resv_object_alloc(&xe->drm);
 	if (!vm_resv_obj) {
···
 
 	drm_gem_object_put(vm_resv_obj);
 
-	err = xe_vm_lock(vm, true);
-	if (err)
-		goto err_close;
+	err = 0;
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {.interruptible = true},
+			    err) {
+		err = xe_vm_drm_exec_lock(vm, &exec);
+		drm_exec_retry_on_contention(&exec);
 
-	if (IS_DGFX(xe) && xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K)
-		vm->flags |= XE_VM_FLAG_64K;
+		if (IS_DGFX(xe) && xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K)
+			vm->flags |= XE_VM_FLAG_64K;
 
-	for_each_tile(tile, xe, id) {
-		if (flags & XE_VM_FLAG_MIGRATION &&
-		    tile->id != XE_VM_FLAG_TILE_ID(flags))
-			continue;
+		for_each_tile(tile, xe, id) {
+			if (flags & XE_VM_FLAG_MIGRATION &&
+			    tile->id != XE_VM_FLAG_TILE_ID(flags))
+				continue;
 
-		vm->pt_root[id] = xe_pt_create(vm, tile, xe->info.vm_max_level);
-		if (IS_ERR(vm->pt_root[id])) {
-			err = PTR_ERR(vm->pt_root[id]);
-			vm->pt_root[id] = NULL;
-			goto err_unlock_close;
+			vm->pt_root[id] = xe_pt_create(vm, tile, xe->info.vm_max_level,
+						       &exec);
+			if (IS_ERR(vm->pt_root[id])) {
+				err = PTR_ERR(vm->pt_root[id]);
+				vm->pt_root[id] = NULL;
+				xe_vm_pt_destroy(vm);
+				drm_exec_retry_on_contention(&exec);
+				xe_validation_retry_on_oom(&ctx, &err);
+				break;
+			}
 		}
-	}
+		if (err)
+			break;
 
-	if (xe_vm_has_scratch(vm)) {
+		if (xe_vm_has_scratch(vm)) {
+			for_each_tile(tile, xe, id) {
+				if (!vm->pt_root[id])
+					continue;
+
+				err = xe_vm_create_scratch(xe, tile, vm, &exec);
+				if (err) {
+					xe_vm_free_scratch(vm);
+					xe_vm_pt_destroy(vm);
+					drm_exec_retry_on_contention(&exec);
+					xe_validation_retry_on_oom(&ctx, &err);
+					break;
+				}
+			}
+			if (err)
+				break;
+			vm->batch_invalidate_tlb = true;
+		}
+
+		if (vm->flags & XE_VM_FLAG_LR_MODE) {
+			INIT_WORK(&vm->preempt.rebind_work, preempt_rebind_work_func);
+			vm->batch_invalidate_tlb = false;
+		}
+
+		/* Fill pt_root after allocating scratch tables */
 		for_each_tile(tile, xe, id) {
 			if (!vm->pt_root[id])
 				continue;
 
-			err = xe_vm_create_scratch(xe, tile, vm);
-			if (err)
-				goto err_unlock_close;
+			xe_pt_populate_empty(tile, vm, vm->pt_root[id]);
 		}
-		vm->batch_invalidate_tlb = true;
 	}
-
-	if (vm->flags & XE_VM_FLAG_LR_MODE)
-		vm->batch_invalidate_tlb = false;
-
-	/* Fill pt_root after allocating scratch tables */
-	for_each_tile(tile, xe, id) {
-		if (!vm->pt_root[id])
-			continue;
-
-		xe_pt_populate_empty(tile, vm, vm->pt_root[id]);
-	}
-	xe_vm_unlock(vm);
+	if (err)
+		goto err_close;
 
 	/* Kernel migration VM shouldn't have a circular loop.. */
 	if (!(flags & XE_VM_FLAG_MIGRATION)) {
···
 				      &xe->usm.next_asid, GFP_KERNEL);
 		up_write(&xe->usm.lock);
 		if (err < 0)
-			goto err_unlock_close;
+			goto err_close;
 
 		vm->usm.asid = asid;
 	}
···
 
 	return vm;
 
-err_unlock_close:
-	xe_vm_unlock(vm);
 err_close:
 	xe_vm_close_and_put(vm);
 	return ERR_PTR(err);
···
 		vma = gpuva_to_vma(gpuva);
 
 		if (xe_vma_has_no_bo(vma)) {
-			down_read(&vm->userptr.notifier_lock);
+			xe_svm_notifier_lock(vm);
 			vma->gpuva.flags |= XE_VMA_DESTROYED;
-			up_read(&vm->userptr.notifier_lock);
+			xe_svm_notifier_unlock(vm);
 		}
 
 		xe_vm_remove_vma(vm, vma);
···
 	 * destroy the pagetables immediately.
 	 */
 	xe_vm_free_scratch(vm);
-
-	for_each_tile(tile, xe, id) {
-		if (vm->pt_root[id]) {
-			xe_pt_destroy(vm->pt_root[id], vm->flags, NULL);
-			vm->pt_root[id] = NULL;
-		}
-	}
+	xe_vm_pt_destroy(vm);
 	xe_vm_unlock(vm);
 
 	/*
···
 		xe_vma_destroy_unlocked(vma);
 	}
 
-	if (xe_vm_in_fault_mode(vm))
-		xe_svm_fini(vm);
+	xe_svm_fini(vm);
 
 	up_write(&vm->lock);
 
···
 
 	err = copy_to_user(attrs_user, mem_attrs,
 			   args->sizeof_mem_range_attr * args->num_mem_ranges);
+	if (err)
+		err = -EFAULT;
 
 free_mem_attrs:
 	kvfree(mem_attrs);
···
 static void prep_vma_destroy(struct xe_vm *vm, struct xe_vma *vma,
 			     bool post_commit)
 {
-	down_read(&vm->userptr.notifier_lock);
+	xe_svm_notifier_lock(vm);
 	vma->gpuva.flags |= XE_VMA_DESTROYED;
-	up_read(&vm->userptr.notifier_lock);
+	xe_svm_notifier_unlock(vm);
 	if (post_commit)
 		xe_vm_remove_vma(vm, vma);
 }
···
 			      struct xe_vma_mem_attr *attr, unsigned int flags)
 {
 	struct xe_bo *bo = op->gem.obj ? gem_to_xe_bo(op->gem.obj) : NULL;
+	struct xe_validation_ctx ctx;
 	struct drm_exec exec;
 	struct xe_vma *vma;
 	int err = 0;
···
 	lockdep_assert_held_write(&vm->lock);
 
 	if (bo) {
-		drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0);
-		drm_exec_until_all_locked(&exec) {
-			err = 0;
+		err = 0;
+		xe_validation_guard(&ctx, &vm->xe->val, &exec,
+				    (struct xe_val_flags) {.interruptible = true}, err) {
 			if (!bo->vm) {
 				err = drm_exec_lock_obj(&exec, xe_vm_obj(vm));
 				drm_exec_retry_on_contention(&exec);
···
 				err = drm_exec_lock_obj(&exec, &bo->ttm.base);
 				drm_exec_retry_on_contention(&exec);
 			}
-			if (err) {
-				drm_exec_fini(&exec);
+			if (err)
 				return ERR_PTR(err);
+
+			vma = xe_vma_create(vm, bo, op->gem.offset,
+					    op->va.addr, op->va.addr +
+					    op->va.range - 1, attr, flags);
+			if (IS_ERR(vma))
+				return vma;
+
+			if (!bo->vm) {
+				err = add_preempt_fences(vm, bo);
+				if (err) {
+					prep_vma_destroy(vm, vma, false);
+					xe_vma_destroy(vma, NULL);
+				}
 			}
 		}
+		if (err)
+			return ERR_PTR(err);
+	} else {
+		vma = xe_vma_create(vm, NULL, op->gem.offset,
+				    op->va.addr, op->va.addr +
+				    op->va.range - 1, attr, flags);
+		if (IS_ERR(vma))
+			return vma;
+
+		if (xe_vma_is_userptr(vma))
+			err = xe_vma_userptr_pin_pages(to_userptr_vma(vma));
 	}
-	vma = xe_vma_create(vm, bo, op->gem.offset,
-			    op->va.addr, op->va.addr +
-			    op->va.range - 1, attr, flags);
-	if (IS_ERR(vma))
-		goto err_unlock;
-
-	if (xe_vma_is_userptr(vma))
-		err = xe_vma_userptr_pin_pages(to_userptr_vma(vma));
-	else if (!xe_vma_has_no_bo(vma) && !bo->vm)
-		err = add_preempt_fences(vm, bo);
-
-err_unlock:
-	if (bo)
-		drm_exec_fini(&exec);
-
 	if (err) {
 		prep_vma_destroy(vm, vma, false);
 		xe_vma_destroy_unlocked(vma);
···
 		struct xe_vma *vma = gpuva_to_vma(op->base.unmap.va);
 
 		if (vma) {
-			down_read(&vm->userptr.notifier_lock);
+			xe_svm_notifier_lock(vm);
 			vma->gpuva.flags &= ~XE_VMA_DESTROYED;
-			up_read(&vm->userptr.notifier_lock);
+			xe_svm_notifier_unlock(vm);
 			if (post_commit)
 				xe_vm_insert_vma(vm, vma);
 		}
···
 			xe_vma_destroy_unlocked(op->remap.next);
 		}
 		if (vma) {
-			down_read(&vm->userptr.notifier_lock);
+			xe_svm_notifier_lock(vm);
 			vma->gpuva.flags &= ~XE_VMA_DESTROYED;
-			up_read(&vm->userptr.notifier_lock);
+			xe_svm_notifier_unlock(vm);
 			if (post_commit)
 				xe_vm_insert_vma(vm, vma);
 		}
···
 			err = drm_exec_lock_obj(exec, &bo->ttm.base);
 		if (!err && validate)
 			err = xe_bo_validate(bo, vm,
-					     !xe_vm_in_preempt_fence_mode(vm));
+					     !xe_vm_in_preempt_fence_mode(vm), exec);
 	}
 
 	return err;
···
 					    false);
 		if (!err && !xe_vma_has_no_bo(vma))
 			err = xe_bo_migrate(xe_vma_bo(vma),
-					    region_to_mem_type[region]);
+					    region_to_mem_type[region],
+					    NULL,
+					    exec);
 		break;
 	}
 	default:
···
 static struct dma_fence *vm_bind_ioctl_ops_execute(struct xe_vm *vm,
 						   struct xe_vma_ops *vops)
 {
+	struct xe_validation_ctx ctx;
 	struct drm_exec exec;
 	struct dma_fence *fence;
-	int err;
+	int err = 0;
 
 	lockdep_assert_held_write(&vm->lock);
 
-	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
-		      DRM_EXEC_IGNORE_DUPLICATES, 0);
-	drm_exec_until_all_locked(&exec) {
+	xe_validation_guard(&ctx, &vm->xe->val, &exec,
+			    ((struct xe_val_flags) {
+				    .interruptible = true,
+				    .exec_ignore_duplicates = true,
+			    }), err) {
 		err = vm_bind_ioctl_ops_lock_and_prep(&exec, vm, vops);
 		drm_exec_retry_on_contention(&exec);
-		if (err) {
-			fence = ERR_PTR(err);
-			goto unlock;
-		}
+		xe_validation_retry_on_oom(&ctx, &err);
+		if (err)
+			return ERR_PTR(err);
 
+		xe_vm_set_validation_exec(vm, &exec);
 		fence = ops_execute(vm, vops);
+		xe_vm_set_validation_exec(vm, NULL);
 		if (IS_ERR(fence)) {
 			if (PTR_ERR(fence) == -ENODATA)
 				vm_bind_ioctl_ops_fini(vm, vops, NULL);
-			goto unlock;
+			return fence;
 		}
 
 		vm_bind_ioctl_ops_fini(vm, vops, fence);
 	}
 
-unlock:
-	drm_exec_fini(&exec);
-	return fence;
+	return err ? ERR_PTR(err) : fence;
 }
 ALLOW_ERROR_INJECTION(vm_bind_ioctl_ops_execute, ERRNO);
 
···
 				 op == DRM_XE_VM_BIND_OP_MAP_USERPTR) ||
 		    XE_IOCTL_DBG(xe, coh_mode == XE_COH_NONE &&
 				 op == DRM_XE_VM_BIND_OP_MAP_USERPTR) ||
+		    XE_IOCTL_DBG(xe, op == DRM_XE_VM_BIND_OP_MAP_USERPTR &&
+				 !IS_ENABLED(CONFIG_DRM_GPUSVM)) ||
 		    XE_IOCTL_DBG(xe, obj &&
 				 op == DRM_XE_VM_BIND_OP_PREFETCH) ||
 		    XE_IOCTL_DBG(xe, prefetch_region &&
···
  */
 int xe_vm_lock(struct xe_vm *vm, bool intr)
 {
-	if (intr)
-		return dma_resv_lock_interruptible(xe_vm_resv(vm), NULL);
+	int ret;
 
-	return dma_resv_lock(xe_vm_resv(vm), NULL);
+	if (intr)
+		ret = dma_resv_lock_interruptible(xe_vm_resv(vm), NULL);
+	else
+		ret = dma_resv_lock(xe_vm_resv(vm), NULL);
+
+	return ret;
 }
 
 /**
···
 	 */
 	if (IS_ENABLED(CONFIG_PROVE_LOCKING)) {
 		if (xe_vma_is_userptr(vma)) {
-			lockdep_assert(lockdep_is_held_type(&vm->userptr.notifier_lock, 0) ||
-				       (lockdep_is_held_type(&vm->userptr.notifier_lock, 1) &&
+			lockdep_assert(lockdep_is_held_type(&vm->svm.gpusvm.notifier_lock, 0) ||
+				       (lockdep_is_held_type(&vm->svm.gpusvm.notifier_lock, 1) &&
 					lockdep_is_held(&xe_vm_resv(vm)->lock.base)));
 
 			WARN_ON_ONCE(!mmu_interval_check_retry
 				     (&to_userptr_vma(vma)->userptr.notifier,
-				      to_userptr_vma(vma)->userptr.notifier_seq));
+				      to_userptr_vma(vma)->userptr.pages.notifier_seq));
 			WARN_ON_ONCE(!dma_resv_test_signaled(xe_vm_resv(vm),
 							     DMA_RESV_USAGE_BOOKKEEP));
 

drivers/gpu/drm/xe/xe_vm.h

···
 int xe_vm_add_compute_exec_queue(struct xe_vm *vm, struct xe_exec_queue *q);
 void xe_vm_remove_compute_exec_queue(struct xe_vm *vm, struct xe_exec_queue *q);
 
-int xe_vm_userptr_pin(struct xe_vm *vm);
-
-int __xe_vm_userptr_needs_repin(struct xe_vm *vm);
-
-int xe_vm_userptr_check_repin(struct xe_vm *vm);
-
 int xe_vm_rebind(struct xe_vm *vm, bool rebind_worker);
 struct dma_fence *xe_vma_rebind(struct xe_vm *vm, struct xe_vma *vma,
 				u8 tile_mask);
···
 	}
 }
 
-int xe_vma_userptr_pin_pages(struct xe_userptr_vma *uvma);
-
-int xe_vma_userptr_check_repin(struct xe_userptr_vma *uvma);
-
-bool xe_vm_validate_should_retry(struct drm_exec *exec, int err, ktime_t *end);
-
 int xe_vm_lock_vma(struct drm_exec *exec, struct xe_vma *vma);
 
 int xe_vm_validate_rebind(struct xe_vm *vm, struct drm_exec *exec,
···
  */
 #define xe_vm_assert_held(vm) dma_resv_assert_held(xe_vm_resv(vm))
 
+int xe_vm_drm_exec_lock(struct xe_vm *vm, struct drm_exec *exec);
+
 #if IS_ENABLED(CONFIG_DRM_XE_DEBUG_VM)
 #define vm_dbg drm_dbg
 #else
···
 	if (vm && !allow_res_evict) {
 		xe_vm_assert_held(vm);
 		/* Pairs with READ_ONCE in xe_vm_is_validating() */
-		WRITE_ONCE(vm->validating, current);
+		WRITE_ONCE(vm->validation.validating, current);
 	}
 }
 
···
 {
 	if (vm && !allow_res_evict) {
 		/* Pairs with READ_ONCE in xe_vm_is_validating() */
-		WRITE_ONCE(vm->validating, NULL);
+		WRITE_ONCE(vm->validation.validating, NULL);
 	}
 }
 
···
 static inline bool xe_vm_is_validating(struct xe_vm *vm)
 {
 	/* Pairs with WRITE_ONCE in xe_vm_is_validating() */
-	if (READ_ONCE(vm->validating) == current) {
+	if (READ_ONCE(vm->validation.validating) == current) {
 		xe_vm_assert_held(vm);
 		return true;
 	}
 	return false;
+}
+
+/**
+ * xe_vm_set_validation_exec() - Accessor to set the drm_exec object
+ * @vm: The vm we want to register a drm_exec object with.
+ * @exec: The exec object we want to register.
+ *
+ * Set the drm_exec object used to lock the vm's resv.
+ */
+static inline void xe_vm_set_validation_exec(struct xe_vm *vm, struct drm_exec *exec)
+{
+	xe_vm_assert_held(vm);
+	xe_assert(vm->xe, !!exec ^ !!vm->validation._exec);
+	vm->validation._exec = exec;
+}
+
+/**
+ * xe_vm_set_validation_exec() - Accessor to read the drm_exec object
+ * @vm: The vm we want to register a drm_exec object with.
+ *
+ * Return: The drm_exec object used to lock the vm's resv. The value
+ * is a valid pointer, %NULL, or one of the special values defined in
+ * xe_validation.h.
+ */
+static inline struct drm_exec *xe_vm_validation_exec(struct xe_vm *vm)
+{
+	xe_vm_assert_held(vm);
+	return vm->validation._exec;
 }
 
 /**
···
 #define xe_vm_has_valid_gpu_mapping(tile, tile_present, tile_invalidated)	\
 	((READ_ONCE(tile_present) & ~READ_ONCE(tile_invalidated)) & BIT((tile)->id))
 
-#if IS_ENABLED(CONFIG_DRM_XE_USERPTR_INVAL_INJECT)
-void xe_vma_userptr_force_invalidate(struct xe_userptr_vma *uvma);
-#else
-static inline void xe_vma_userptr_force_invalidate(struct xe_userptr_vma *uvma)
-{
-}
-#endif
 #endif

drivers/gpu/drm/xe/xe_vm_madvise.c

···
 	u64 range;
 	struct xe_vma **vmas;
 	int num_vmas;
-	bool has_svm_vmas;
 	bool has_bo_vmas;
-	bool has_userptr_vmas;
+	bool has_svm_userptr_vmas;
 };
 
 static int get_vmas(struct xe_vm *vm, struct xe_vmas_in_madvise_range *madvise_range)
···
 
 		if (xe_vma_bo(vma))
 			madvise_range->has_bo_vmas = true;
-		else if (xe_vma_is_cpu_addr_mirror(vma))
-			madvise_range->has_svm_vmas = true;
-		else if (xe_vma_is_userptr(vma))
-			madvise_range->has_userptr_vmas = true;
+		else if (xe_vma_is_cpu_addr_mirror(vma) || xe_vma_is_userptr(vma))
+			madvise_range->has_svm_userptr_vmas = true;
 
 		if (madvise_range->num_vmas == max_vmas) {
 			max_vmas <<= 1;
···
 			vmas[i]->attr.atomic_access = op->atomic.val;
 		}
 
-		vmas[i]->attr.atomic_access = op->atomic.val;
-
 		bo = xe_vma_bo(vmas[i]);
 		if (!bo || bo->attr.atomic_access == op->atomic.val)
 			continue;
···
 				if (xe_pt_zap_ptes(tile, vma)) {
 					tile_mask |= BIT(id);
 
-				/*
-				 * WRITE_ONCE pairs with READ_ONCE
-				 * in xe_vm_has_valid_gpu_mapping()
-				 */
-				WRITE_ONCE(vma->tile_invalidated,
-					   vma->tile_invalidated | BIT(id));
+					/*
+					 * WRITE_ONCE pairs with READ_ONCE
+					 * in xe_vm_has_valid_gpu_mapping()
+					 */
+					WRITE_ONCE(vma->tile_invalidated,
+						   vma->tile_invalidated | BIT(id));
 				}
 			}
 		}
···
 		if (XE_IOCTL_DBG(xe, args->preferred_mem_loc.pad))
 			return false;
 
-		if (XE_IOCTL_DBG(xe, args->atomic.reserved))
+		if (XE_IOCTL_DBG(xe, args->preferred_mem_loc.reserved))
 			return false;
 		break;
 	}
···
 		}
 	}
 
-	if (madvise_range.has_userptr_vmas) {
-		err = down_read_interruptible(&vm->userptr.notifier_lock);
+	if (madvise_range.has_svm_userptr_vmas) {
+		err = xe_svm_notifier_lock_interruptible(vm);
 		if (err)
 			goto err_fini;
-	}
-
-	if (madvise_range.has_svm_vmas) {
-		err = down_read_interruptible(&vm->svm.gpusvm.notifier_lock);
-		if (err)
-			goto unlock_userptr;
 	}
 
 	attr_type = array_index_nospec(args->type, ARRAY_SIZE(madvise_funcs));
···
 
 	err = xe_vm_invalidate_madvise_range(vm, args->start, args->start + args->range);
 
-	if (madvise_range.has_svm_vmas)
+	if (madvise_range.has_svm_userptr_vmas)
 		xe_svm_notifier_unlock(vm);
 
-unlock_userptr:
-	if (madvise_range.has_userptr_vmas)
-		up_read(&vm->userptr.notifier_lock);
 err_fini:
 	if (madvise_range.has_bo_vmas)
 		drm_exec_fini(&exec);

drivers/gpu/drm/xe/xe_vm_types.h

···
 #include "xe_device_types.h"
 #include "xe_pt_types.h"
 #include "xe_range_fence.h"
+#include "xe_userptr.h"
 
 struct xe_bo;
 struct xe_svm_range;
···
 #define XE_VMA_PTE_COMPACT	(DRM_GPUVA_USERBITS << 7)
 #define XE_VMA_DUMPABLE		(DRM_GPUVA_USERBITS << 8)
 #define XE_VMA_SYSTEM_ALLOCATOR	(DRM_GPUVA_USERBITS << 9)
-
-/** struct xe_userptr - User pointer */
-struct xe_userptr {
-	/** @invalidate_link: Link for the vm::userptr.invalidated list */
-	struct list_head invalidate_link;
-	/** @userptr: link into VM repin list if userptr. */
-	struct list_head repin_link;
-	/**
-	 * @notifier: MMU notifier for user pointer (invalidation call back)
-	 */
-	struct mmu_interval_notifier notifier;
-	/** @sgt: storage for a scatter gather table */
-	struct sg_table sgt;
-	/** @sg: allocated scatter gather table */
-	struct sg_table *sg;
-	/** @notifier_seq: notifier sequence number */
-	unsigned long notifier_seq;
-	/** @unmap_mutex: Mutex protecting dma-unmapping */
-	struct mutex unmap_mutex;
-	/**
-	 * @initial_bind: user pointer has been bound at least once.
-	 * write: vm->userptr.notifier_lock in read mode and vm->resv held.
-	 * read: vm->userptr.notifier_lock in write mode or vm->resv held.
-	 */
-	bool initial_bind;
-	/** @mapped: Whether the @sgt sg-table is dma-mapped. Protected by @unmap_mutex. */
-	bool mapped;
-#if IS_ENABLED(CONFIG_DRM_XE_USERPTR_INVAL_INJECT)
-	u32 divisor;
-#endif
-};
 
 /**
  * struct xe_vma_mem_attr - memory attributes associated with vma
···
 
 	/**
 	 * @tile_invalidated: Tile mask of binding are invalidated for this VMA.
-	 * protected by BO's resv and for userptrs, vm->userptr.notifier_lock in
-	 * write mode for writing or vm->userptr.notifier_lock in read mode and
+	 * protected by BO's resv and for userptrs, vm->svm.gpusvm.notifier_lock in
+	 * write mode for writing or vm->svm.gpusvm.notifier_lock in read mode and
 	 * the vm->resv. For stable reading, BO's resv or userptr
-	 * vm->userptr.notifier_lock in read mode is required. Can be
+	 * vm->svm.gpusvm.notifier_lock in read mode is required. Can be
 	 * opportunistically read with READ_ONCE outside of locks.
 	 */
 	u8 tile_invalidated;
···
 	/**
 	 * @tile_present: Tile mask of binding are present for this VMA.
 	 * protected by vm->lock, vm->resv and for userptrs,
-	 * vm->userptr.notifier_lock for writing. Needs either for reading,
+	 * vm->svm.gpusvm.notifier_lock for writing. Needs either for reading,
 	 * but if reading is done under the vm->lock only, it needs to be held
 	 * in write mode.
 	 */
···
 	const struct xe_pt_ops *pt_ops;
 
 	/** @userptr: user pointer state */
-	struct {
-		/**
-		 * @userptr.repin_list: list of VMAs which are user pointers,
-		 * and needs repinning. Protected by @lock.
-		 */
-		struct list_head repin_list;
-		/**
-		 * @notifier_lock: protects notifier in write mode and
-		 * submission in read mode.
-		 */
-		struct rw_semaphore notifier_lock;
-		/**
-		 * @userptr.invalidated_lock: Protects the
-		 * @userptr.invalidated list.
-		 */
-		spinlock_t invalidated_lock;
-		/**
-		 * @userptr.invalidated: List of invalidated userptrs, not yet
-		 * picked
-		 * up for revalidation. Protected from access with the
-		 * @invalidated_lock. Removing items from the list
-		 * additionally requires @lock in write mode, and adding
-		 * items to the list requires either the @userptr.notifier_lock in
-		 * write mode, OR @lock in write mode.
-		 */
-		struct list_head invalidated;
-	} userptr;
+	struct xe_userptr_vm userptr;
 
 	/** @preempt: preempt state */
 	struct {
···
 	} error_capture;
 
 	/**
+	 * @validation: Validation data only valid with the vm resv held.
+	 * Note: This is really task state of the task holding the vm resv,
+	 * and moving forward we should
+	 * come up with a better way of passing this down the call-
+	 * chain.
+	 */
+	struct {
+		/**
+		 * @validation.validating: The task that is currently making bos resident.
+		 * for this vm.
+		 * Protected by the VM's resv for writing. Opportunistic reading can be done
+		 * using READ_ONCE. Note: This is a workaround for the
+		 * TTM eviction_valuable() callback not being passed a struct
+		 * ttm_operation_context(). Future work might want to address this.
+		 */
+		struct task_struct *validating;
+		/**
+		 *  @validation.exec The drm_exec context used when locking the vm resv.
+		 *  Protected by the vm's resv.
+		 */
+		struct drm_exec *_exec;
+	} validation;
+
+	/**
 	 * @tlb_flush_seqno: Required TLB flush seqno for the next exec.
 	 * protected by the vm resv.
 	 */
 	u64 tlb_flush_seqno;
-	/**
-	 * @validating: The task that is currently making bos resident for this vm.
-	 * Protected by the VM's resv for writing. Opportunistic reading can be done
-	 * using READ_ONCE. Note: This is a workaround for the
-	 * TTM eviction_valuable() callback not being passed a struct
-	 * ttm_operation_context(). Future work might want to address this.
-	 */
-	struct task_struct *validating;
 	/** @batch_invalidate_tlb: Always invalidate TLB before batch start */
 	bool batch_invalidate_tlb;
 	/** @xef: XE file handle for tracking this VM's drm client */

drivers/gpu/drm/xe/xe_wa.c

···
  *   Register Immediate commands) once when initializing the device and saved in
  *   the default context. That default context is then used on every context
  *   creation to have a "primed golden context", i.e. a context image that
- *   already contains the changes needed to all the registers.
+ *   already contains the changes needed to all the registers. See
+ *   drivers/gpu/drm/xe/xe_lrc.c for default context handling.
  *
  * - Engine workarounds: the list of these WAs is applied whenever the specific
  *   engine is reset. It's also possible that a set of engine classes share a
···
  *   them need to keeep the workaround programming: the approach taken in the
  *   driver is to tie those workarounds to the first compute/render engine that
  *   is registered.  When executing with GuC submission, engine resets are
- *   outside of kernel driver control, hence the list of registers involved in
+ *   outside of kernel driver control, hence the list of registers involved is
  *   written once, on engine initialization, and then passed to GuC, that
  *   saves/restores their values before/after the reset takes place. See
- *   ``drivers/gpu/drm/xe/xe_guc_ads.c`` for reference.
+ *   drivers/gpu/drm/xe/xe_guc_ads.c for reference.
  *
  * - GT workarounds: the list of these WAs is applied whenever these registers
  *   revert to their default values: on GPU reset, suspend/resume [1]_, etc.
···
  *   hardware on every HW context restore. These buffers are created and
  *   programmed in the default context so the hardware always go through those
  *   programming sequences when switching contexts. The support for workaround
- *   batchbuffers is enabled these hardware mechanisms:
+ *   batchbuffers is enabled via these hardware mechanisms:
  *
- *   #. INDIRECT_CTX: A batchbuffer and an offset are provided in the default
- *      context, pointing the hardware to jump to that location when that offset
- *      is reached in the context restore. Workaround batchbuffer in the driver
- *      currently uses this mechanism for all platforms.
+ *   #. INDIRECT_CTX (also known as **mid context restore bb**): A batchbuffer
+ *      and an offset are provided in the default context, pointing the hardware
+ *      to jump to that location when that offset is reached in the context
+ *      restore.  When a context is being restored, this is executed after the
+ *      ring context, in the middle (or beginning) of the engine context image.
  *
- *   #. BB_PER_CTX_PTR: A batchbuffer is provided in the default context,
- *      pointing the hardware to a buffer to continue executing after the
- *      engine registers are restored in a context restore sequence. This is
- *      currently not used in the driver.
+ *   #. BB_PER_CTX_PTR (also known as **post context restore bb**): A
+ *      batchbuffer is provided in the default context, pointing the hardware to
+ *      a buffer to continue executing after the engine registers are restored
+ *      in a context restore sequence.
+ *
+ *   Below is the timeline for a context restore sequence:
+ *
+ *   .. code::
+ *
+ *                        INDIRECT_CTX_OFFSET
+ *                   |----------->|
+ *      .------------.------------.-------------.------------.--------------.-----------.
+ *      |Ring        | Engine     | Mid-context | Engine     | Post-context | Ring      |
+ *      |Restore     | Restore (1)| BB Restore  | Restore (2)| BB Restore   | Execution |
+ *      `------------'------------'-------------'------------'--------------'-----------'
  *
  * - Other/OOB:  There are WAs that, due to their nature, cannot be applied from
  *   a central place. Those are peppered around the rest of the code, as needed.
- *   Workarounds related to the display IP are the main example.
+ *   There's a central place to control which workarounds are enabled:
+ *   drivers/gpu/drm/xe/xe_wa_oob.rules for GT workarounds and
+ *   drivers/gpu/drm/xe/xe_device_wa_oob.rules for device/SoC workarounds.
+ *   These files only record which workarounds are enabled: during early device
+ *   initialization those rules are evaluated and recorded by the driver. Then
+ *   later the driver checks with ``XE_GT_WA()`` and ``XE_DEVICE_WA()`` to
+ *   implement them.
  *
  * .. [1] Technically, some registers are powercontext saved & restored, so they
  *    survive a suspend/resume. In practice, writing them again is not too
···
 		       FUNC(xe_rtp_match_first_render_or_compute)),
 	  XE_RTP_ACTIONS(SET(TDL_TSL_CHICKEN, RES_CHK_SPR_DIS))
 	},
+	{ XE_RTP_NAME("18041344222"),
+	  XE_RTP_RULES(GRAPHICS_VERSION_RANGE(2001, 2002),
+		       FUNC(xe_rtp_match_first_render_or_compute),
+		       FUNC(xe_rtp_match_not_sriov_vf),
+		       FUNC(xe_rtp_match_gt_has_discontiguous_dss_groups)),
+	  XE_RTP_ACTIONS(SET(TDL_CHICKEN, EUSTALL_PERF_SAMPLING_DISABLE))
+	},
 
 	/* Xe2_LPM */
 
···
 	{ XE_RTP_NAME("14021402888"),
 	  XE_RTP_RULES(GRAPHICS_VERSION(3003), FUNC(xe_rtp_match_first_render_or_compute)),
 	  XE_RTP_ACTIONS(SET(HALF_SLICE_CHICKEN7, CLEAR_OPTIMIZATION_DISABLE))
+	},
+	{ XE_RTP_NAME("18041344222"),
+	  XE_RTP_RULES(GRAPHICS_VERSION_RANGE(3000, 3001),
+		       FUNC(xe_rtp_match_first_render_or_compute),
+		       FUNC(xe_rtp_match_not_sriov_vf),
+		       FUNC(xe_rtp_match_gt_has_discontiguous_dss_groups)),
+	  XE_RTP_ACTIONS(SET(TDL_CHICKEN, EUSTALL_PERF_SAMPLING_DISABLE))
 	},
 };
 
···
 			 SET(VFLSKPD,
 			     DIS_PARTIAL_AUTOSTRIP |
 			     DIS_AUTOSTRIP))
+	},
+	{ XE_RTP_NAME("22021007897"),
+	  XE_RTP_RULES(GRAPHICS_VERSION_RANGE(3000, 3003), ENGINE_CLASS(RENDER)),
+	  XE_RTP_ACTIONS(SET(COMMON_SLICE_CHICKEN4, SBE_PUSH_CONSTANT_BEHIND_FIX_ENABLE))
 	},
 };
 

drivers/gpu/drm/xe/xe_wa_oob.rules

···
 16023588340	GRAPHICS_VERSION(2001), FUNC(xe_rtp_match_not_sriov_vf)
 14019789679	GRAPHICS_VERSION(1255)
 		GRAPHICS_VERSION_RANGE(1270, 2004)
-no_media_l3	MEDIA_VERSION_RANGE(3000, 3002)
 14022866841	GRAPHICS_VERSION(3000), GRAPHICS_STEP(A0, B0)
 		MEDIA_VERSION(3000), MEDIA_STEP(A0, B0)
 16021333562	GRAPHICS_VERSION_RANGE(1200, 1274)

drivers/misc/mei/Kconfig

···
 	  This driver can also be built as a module. If so, the module
 	  will be called mei-vsc.
 
+config INTEL_MEI_LB
+	tristate "Intel Late Binding (LB) support on ME Interface"
+	depends on INTEL_MEI_ME
+	depends on DRM_XE
+	help
+	  Enable support for Intel Late Binding (LB) via the MEI interface.
+
+	  Late Binding is a method for applying firmware updates at runtime,
+	  allowing the Intel Xe driver to load firmware payloads such as
+	  fan controller or voltage regulator. These firmware updates are
+	  authenticated and versioned, and do not require firmware flashing
+	  or system reboot.
+
 source "drivers/misc/mei/hdcp/Kconfig"
 source "drivers/misc/mei/pxp/Kconfig"
 source "drivers/misc/mei/gsc_proxy/Kconfig"

drivers/misc/mei/Makefile

···
 obj-$(CONFIG_INTEL_MEI_HDCP) += hdcp/
 obj-$(CONFIG_INTEL_MEI_PXP) += pxp/
 obj-$(CONFIG_INTEL_MEI_GSC_PROXY) += gsc_proxy/
+obj-$(CONFIG_INTEL_MEI_LB) += mei_lb.o
 
 obj-$(CONFIG_INTEL_MEI_VSC_HW) += mei-vsc-hw.o
 mei-vsc-hw-y := vsc-tp.o

drivers/misc/mei/bus.c

···
 EXPORT_SYMBOL_GPL(mei_cldev_ver);
 
 /**
+ * mei_cldev_mtu - max message that client can send and receive
+ *
+ * @cldev: mei client device
+ *
+ * Return: mtu or 0 if client is not connected
+ */
+size_t mei_cldev_mtu(const struct mei_cl_device *cldev)
+{
+	return mei_cl_mtu(cldev->cl);
+}
+EXPORT_SYMBOL_GPL(mei_cldev_mtu);
+
+/**
  * mei_cldev_enabled - check whether the device is enabled
  *
  * @cldev: mei client device

drivers/misc/mei/mei_lb.c

···
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2025 Intel Corporation
+ */
+
+#include <linux/component.h>
+#include <linux/mei_cl_bus.h>
+#include <linux/module.h>
+#include <linux/overflow.h>
+#include <linux/pci.h>
+#include <linux/slab.h>
+#include <linux/uuid.h>
+
+#include <drm/intel/i915_component.h>
+#include <drm/intel/intel_lb_mei_interface.h>
+
+#include "mkhi.h"
+
+/**
+ * DOC: Late Binding Firmware Update/Upload
+ *
+ * Late Binding is a firmware update/upload mechanism that allows configuration
+ * payloads to be securely delivered and applied at runtime, rather than
+ * being embedded in the system firmware image (e.g., IFWI or SPI flash).
+ *
+ * This mechanism is used to update device-level configuration such as:
+ * - Fan controller
+ * - Voltage regulator (VR)
+ *
+ * Key Characteristics:
+ * ---------------------
+ * - Runtime Delivery:
+ *   Firmware blobs are loaded by the host driver (e.g., Xe KMD)
+ *   after the GPU or SoC has booted.
+ *
+ * - Secure and Authenticated:
+ *   All payloads are signed and verified by the authentication firmware.
+ *
+ * - No Firmware Flashing Required:
+ *   Updates are applied in volatile memory and do not require SPI flash
+ *   modification or system reboot.
+ *
+ * - Re-entrant:
+ *   Multiple updates of the same or different types can be applied
+ *   sequentially within a single boot session.
+ *
+ * - Version Controlled:
+ *   Each payload includes version and security version number (SVN)
+ *   metadata to support anti-rollback enforcement.
+ *
+ * Upload Flow:
+ * ------------
+ * 1. Host driver (KMD or user-space tool) loads the late binding firmware.
+ * 2. Firmware is passed to the MEI interface and forwarded to
+ *    authentication firmware.
+ * 3. Authentication firmware authenticates the payload and extracts
+ *    command and data arrays.
+ * 4. Authentication firmware delivers the configuration to PUnit/PCODE.
+ * 5. Status is returned back to the host via MEI.
+ */
+
+#define INTEL_LB_CMD	0x12
+#define INTEL_LB_RSP	(INTEL_LB_CMD | 0x80)
+
+#define INTEL_LB_SEND_TIMEOUT_MSEC 3000
+#define INTEL_LB_RECV_TIMEOUT_MSEC 3000
+
+/**
+ * struct mei_lb_req - Late Binding request structure
+ * @header: MKHI message header (see struct mkhi_msg_hdr)
+ * @type: Type of the Late Binding payload
+ * @flags: Flags to be passed to the authentication firmware (e.g. %INTEL_LB_FLAGS_IS_PERSISTENT)
+ * @reserved: Reserved for future use by authentication firmware, must be set to 0
+ * @payload_size: Size of the payload data in bytes
+ * @payload: Payload data to be sent to the authentication firmware
+ */
+struct mei_lb_req {
+	struct mkhi_msg_hdr header;
+	__le32 type;
+	__le32 flags;
+	__le32 reserved[2];
+	__le32 payload_size;
+	u8 payload[] __counted_by(payload_size);
+} __packed;
+
+/**
+ * struct mei_lb_rsp - Late Binding response structure
+ * @header: MKHI message header (see struct mkhi_msg_hdr)
+ * @type: Type of the Late Binding payload
+ * @reserved: Reserved for future use by authentication firmware, must be set to 0
+ * @status: Status returned by authentication firmware (see &enum intel_lb_status)
+ */
+struct mei_lb_rsp {
+	struct mkhi_msg_hdr header;
+	__le32 type;
+	__le32 reserved[2];
+	__le32 status;
+} __packed;
+
+static bool mei_lb_check_response(const struct device *dev, ssize_t bytes,
+				  struct mei_lb_rsp *rsp)
+{
+	/*
+	 * Received message size may be smaller than the full message size when
+	 * reply contains only MKHI header with result field set to the error code.
+	 * Check the header size and content first to output exact error, if needed,
+	 * and then process to the whole message.
+	 */
+	if (bytes < sizeof(rsp->header)) {
+		dev_err(dev, "Received less than header size from the firmware: %zd < %zu\n",
+			bytes, sizeof(rsp->header));
+		return false;
+	}
+	if (rsp->header.group_id != MKHI_GROUP_ID_GFX) {
+		dev_err(dev, "Mismatch group id: 0x%x instead of 0x%x\n",
+			rsp->header.group_id, MKHI_GROUP_ID_GFX);
+		return false;
+	}
+	if (rsp->header.command != INTEL_LB_RSP) {
+		dev_err(dev, "Mismatch command: 0x%x instead of 0x%x\n",
+			rsp->header.command, INTEL_LB_RSP);
+		return false;
+	}
+	if (rsp->header.result) {
+		dev_err(dev, "Error in result: 0x%x\n", rsp->header.result);
+		return false;
+	}
+	if (bytes < sizeof(*rsp)) {
+		dev_err(dev, "Received less than message size from the firmware: %zd < %zu\n",
+			bytes, sizeof(*rsp));
+		return false;
+	}
+
+	return true;
+}
+
+static int mei_lb_push_payload(struct device *dev,
+			       enum intel_lb_type type, u32 flags,
+			       const void *payload, size_t payload_size)
+{
+	struct mei_cl_device *cldev;
+	struct mei_lb_req *req = NULL;
+	struct mei_lb_rsp rsp;
+	size_t req_size;
+	ssize_t bytes;
+	int ret;
+
+	cldev = to_mei_cl_device(dev);
+
+	ret = mei_cldev_enable(cldev);
+	if (ret) {
+		dev_dbg(dev, "Failed to enable firmware client. %d\n", ret);
+		return ret;
+	}
+
+	req_size = struct_size(req, payload, payload_size);
+	if (req_size > mei_cldev_mtu(cldev)) {
+		dev_err(dev, "Payload is too big: %zu\n", payload_size);
+		ret = -EMSGSIZE;
+		goto end;
+	}
+
+	req = kmalloc(req_size, GFP_KERNEL);
+	if (!req) {
+		ret = -ENOMEM;
+		goto end;
+	}
+
+	req->header.group_id = MKHI_GROUP_ID_GFX;
+	req->header.command = INTEL_LB_CMD;
+	req->type = cpu_to_le32(type);
+	req->flags = cpu_to_le32(flags);
+	req->reserved[0] = 0;
+	req->reserved[1] = 0;
+	req->payload_size = cpu_to_le32(payload_size);
+	memcpy(req->payload, payload, payload_size);
+
+	bytes = mei_cldev_send_timeout(cldev, (u8 *)req, req_size,
+				       INTEL_LB_SEND_TIMEOUT_MSEC);
+	if (bytes < 0) {
+		dev_err(dev, "Failed to send late binding request to firmware. %zd\n", bytes);
+		ret = bytes;
+		goto end;
+	}
+
+	bytes = mei_cldev_recv_timeout(cldev, (u8 *)&rsp, sizeof(rsp),
+				       INTEL_LB_RECV_TIMEOUT_MSEC);
+	if (bytes < 0) {
+		dev_err(dev, "Failed to receive late binding reply from MEI firmware. %zd\n",
+			bytes);
+		ret = bytes;
+		goto end;
+	}
+	if (!mei_lb_check_response(dev, bytes, &rsp)) {
+		dev_err(dev, "Bad response from the firmware. header: %02x %02x %02x %02x\n",
+			rsp.header.group_id, rsp.header.command,
+			rsp.header.reserved, rsp.header.result);
+		ret = -EPROTO;
+		goto end;
+	}
+
+	dev_dbg(dev, "status = %u\n", le32_to_cpu(rsp.status));
+	ret = (int)le32_to_cpu(rsp.status);
+end:
+	mei_cldev_disable(cldev);
+	kfree(req);
+	return ret;
+}
+
+static const struct intel_lb_component_ops mei_lb_ops = {
+	.push_payload = mei_lb_push_payload,
+};
+
+static int mei_lb_component_master_bind(struct device *dev)
+{
+	return component_bind_all(dev, (void *)&mei_lb_ops);
+}
+
+static void mei_lb_component_master_unbind(struct device *dev)
+{
+	component_unbind_all(dev, (void *)&mei_lb_ops);
+}
+
+static const struct component_master_ops mei_lb_component_master_ops = {
+	.bind = mei_lb_component_master_bind,
+	.unbind = mei_lb_component_master_unbind,
+};
+
+static int mei_lb_component_match(struct device *dev, int subcomponent,
+				  void *data)
+{
+	/*
+	 * This function checks if requester is Intel %PCI_CLASS_DISPLAY_VGA or
+	 * %PCI_CLASS_DISPLAY_OTHER device, and checks if the requester is the
+	 * grand parent of mei_if i.e. late bind MEI device
+	 */
+	struct device *base = data;
+	struct pci_dev *pdev;
+
+	if (!dev)
+		return 0;
+
+	if (!dev_is_pci(dev))
+		return 0;
+
+	pdev = to_pci_dev(dev);
+
+	if (pdev->vendor != PCI_VENDOR_ID_INTEL)
+		return 0;
+
+	if (pdev->class != (PCI_CLASS_DISPLAY_VGA << 8) &&
+	    pdev->class != (PCI_CLASS_DISPLAY_OTHER << 8))
+		return 0;
+
+	if (subcomponent != INTEL_COMPONENT_LB)
+		return 0;
+
+	base = base->parent;
+	if (!base) /* mei device */
+		return 0;
+
+	base = base->parent; /* pci device */
+
+	return !!base && dev == base;
+}
+
+static int mei_lb_probe(struct mei_cl_device *cldev,
+			const struct mei_cl_device_id *id)
+{
+	struct component_match *master_match = NULL;
+	int ret;
+
+	component_match_add_typed(&cldev->dev, &master_match,
+				  mei_lb_component_match, &cldev->dev);
+	if (IS_ERR_OR_NULL(master_match))
+		return -ENOMEM;
+
+	ret = component_master_add_with_match(&cldev->dev,
+					      &mei_lb_component_master_ops,
+					      master_match);
+	if (ret < 0)
+		dev_err(&cldev->dev, "Failed to add late binding master component. %d\n", ret);
+
+	return ret;
+}
+
+static void mei_lb_remove(struct mei_cl_device *cldev)
+{
+	component_master_del(&cldev->dev, &mei_lb_component_master_ops);
+}
+
+#define MEI_GUID_MKHI UUID_LE(0xe2c2afa2, 0x3817, 0x4d19, \
+			      0x9d, 0x95, 0x6, 0xb1, 0x6b, 0x58, 0x8a, 0x5d)
+
+static const struct mei_cl_device_id mei_lb_tbl[] = {
+	{ .uuid = MEI_GUID_MKHI, .version = MEI_CL_VERSION_ANY },
+	{ }
+};
+MODULE_DEVICE_TABLE(mei, mei_lb_tbl);
+
+static struct mei_cl_driver mei_lb_driver = {
+	.id_table = mei_lb_tbl,
+	.name = "mei_lb",
+	.probe = mei_lb_probe,
+	.remove	= mei_lb_remove,
+};
+
+module_mei_cl_driver(mei_lb_driver);
+
+MODULE_AUTHOR("Intel Corporation");
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("MEI Late Binding Firmware Update/Upload");

include/drm/drm_gpusvm.h

···
 };
 
 /**
- * struct drm_gpusvm_range_flags - Structure representing a GPU SVM range flags
+ * struct drm_gpusvm_pages_flags - Structure representing a GPU SVM pages flags
  *
- * @migrate_devmem: Flag indicating whether the range can be migrated to device memory
- * @unmapped: Flag indicating if the range has been unmapped
- * @partial_unmap: Flag indicating if the range has been partially unmapped
- * @has_devmem_pages: Flag indicating if the range has devmem pages
- * @has_dma_mapping: Flag indicating if the range has a DMA mapping
- * @__flags: Flags for range in u16 form (used for READ_ONCE)
+ * @migrate_devmem: Flag indicating whether the pages can be migrated to device memory
+ * @unmapped: Flag indicating if the pages has been unmapped
+ * @partial_unmap: Flag indicating if the pages has been partially unmapped
+ * @has_devmem_pages: Flag indicating if the pages has devmem pages
+ * @has_dma_mapping: Flag indicating if the pages has a DMA mapping
+ * @__flags: Flags for pages in u16 form (used for READ_ONCE)
  */
-struct drm_gpusvm_range_flags {
+struct drm_gpusvm_pages_flags {
 	union {
 		struct {
 			/* All flags below must be set upon creation */
···
 };
 
 /**
+ * struct drm_gpusvm_pages - Structure representing a GPU SVM mapped pages
+ *
+ * @dma_addr: Device address array
+ * @dpagemap: The struct drm_pagemap of the device pages we're dma-mapping.
+ *            Note this is assuming only one drm_pagemap per range is allowed.
+ * @notifier_seq: Notifier sequence number of the range's pages
+ * @flags: Flags for range
+ * @flags.migrate_devmem: Flag indicating whether the range can be migrated to device memory
+ * @flags.unmapped: Flag indicating if the range has been unmapped
+ * @flags.partial_unmap: Flag indicating if the range has been partially unmapped
+ * @flags.has_devmem_pages: Flag indicating if the range has devmem pages
+ * @flags.has_dma_mapping: Flag indicating if the range has a DMA mapping
+ */
+struct drm_gpusvm_pages {
+	struct drm_pagemap_addr *dma_addr;
+	struct drm_pagemap *dpagemap;
+	unsigned long notifier_seq;
+	struct drm_gpusvm_pages_flags flags;
+};
+
+/**
  * struct drm_gpusvm_range - Structure representing a GPU SVM range
  *
  * @gpusvm: Pointer to the GPU SVM structure
···
  * @refcount: Reference count for the range
  * @itree: Interval tree node for the range (inserted in GPU SVM notifier)
  * @entry: List entry to fast interval tree traversal
- * @notifier_seq: Notifier sequence number of the range's pages
- * @dma_addr: Device address array
- * @dpagemap: The struct drm_pagemap of the device pages we're dma-mapping.
- *            Note this is assuming only one drm_pagemap per range is allowed.
- * @flags: Flags for range
+ * @pages: The pages for this range.
  *
  * This structure represents a GPU SVM range used for tracking memory ranges
  * mapped in a DRM device.
···
 	struct kref refcount;
 	struct interval_tree_node itree;
 	struct list_head entry;
-	unsigned long notifier_seq;
-	struct drm_pagemap_addr *dma_addr;
-	struct drm_pagemap *dpagemap;
-	struct drm_gpusvm_range_flags flags;
+	struct drm_gpusvm_pages pages;
 };
 
 /**
···
 
 void drm_gpusvm_range_set_unmapped(struct drm_gpusvm_range *range,
 				   const struct mmu_notifier_range *mmu_range);
+
+int drm_gpusvm_get_pages(struct drm_gpusvm *gpusvm,
+			 struct drm_gpusvm_pages *svm_pages,
+			 struct mm_struct *mm,
+			 struct mmu_interval_notifier *notifier,
+			 unsigned long pages_start, unsigned long pages_end,
+			 const struct drm_gpusvm_ctx *ctx);
+
+void drm_gpusvm_unmap_pages(struct drm_gpusvm *gpusvm,
+			    struct drm_gpusvm_pages *svm_pages,
+			    unsigned long npages,
+			    const struct drm_gpusvm_ctx *ctx);
+
+void drm_gpusvm_free_pages(struct drm_gpusvm *gpusvm,
+			   struct drm_gpusvm_pages *svm_pages,
+			   unsigned long npages);
 
 #ifdef CONFIG_LOCKDEP
 /**

include/drm/intel/i915_component.h

···
 	I915_COMPONENT_HDCP,
 	I915_COMPONENT_PXP,
 	I915_COMPONENT_GSC_PROXY,
+	INTEL_COMPONENT_LB,
 };
 
 /* MAX_PORT is the number of port

include/drm/intel/intel_lb_mei_interface.h

···
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (c) 2025 Intel Corporation
+ */
+
+#ifndef _INTEL_LB_MEI_INTERFACE_H_
+#define _INTEL_LB_MEI_INTERFACE_H_
+
+#include <linux/types.h>
+
+struct device;
+
+/**
+ * define INTEL_LB_FLAG_IS_PERSISTENT - Mark the payload as persistent
+ *
+ * This flag indicates that the late binding payload should be stored
+ * persistently in flash across warm resets.
+ */
+#define INTEL_LB_FLAG_IS_PERSISTENT	BIT(0)
+
+/**
+ * enum intel_lb_type - enum to determine late binding payload type
+ * @INTEL_LB_TYPE_FAN_CONTROL: Fan controller configuration
+ */
+enum intel_lb_type {
+	INTEL_LB_TYPE_FAN_CONTROL = 1,
+};
+
+/**
+ * enum intel_lb_status - Status codes returned on late binding transmissions
+ * @INTEL_LB_STATUS_SUCCESS: Operation completed successfully
+ * @INTEL_LB_STATUS_4ID_MISMATCH: Mismatch in the expected 4ID (firmware identity/token)
+ * @INTEL_LB_STATUS_ARB_FAILURE: Arbitration failure (e.g. conflicting access or state)
+ * @INTEL_LB_STATUS_GENERAL_ERROR: General firmware error not covered by other codes
+ * @INTEL_LB_STATUS_INVALID_PARAMS: One or more input parameters are invalid
+ * @INTEL_LB_STATUS_INVALID_SIGNATURE: Payload has an invalid or untrusted signature
+ * @INTEL_LB_STATUS_INVALID_PAYLOAD: Payload contents are not accepted by firmware
+ * @INTEL_LB_STATUS_TIMEOUT: Operation timed out before completion
+ */
+enum intel_lb_status {
+	INTEL_LB_STATUS_SUCCESS           = 0,
+	INTEL_LB_STATUS_4ID_MISMATCH      = 1,
+	INTEL_LB_STATUS_ARB_FAILURE       = 2,
+	INTEL_LB_STATUS_GENERAL_ERROR     = 3,
+	INTEL_LB_STATUS_INVALID_PARAMS    = 4,
+	INTEL_LB_STATUS_INVALID_SIGNATURE = 5,
+	INTEL_LB_STATUS_INVALID_PAYLOAD   = 6,
+	INTEL_LB_STATUS_TIMEOUT           = 7,
+};
+
+/**
+ * struct intel_lb_component_ops - Ops for late binding services
+ */
+struct intel_lb_component_ops {
+	/**
+	 * push_payload - Sends a payload to the authentication firmware
+	 * @dev: Device struct corresponding to the mei device
+	 * @type: Payload type (see &enum intel_lb_type)
+	 * @flags: Payload flags bitmap (e.g. %INTEL_LB_FLAGS_IS_PERSISTENT)
+	 * @payload: Pointer to payload buffer
+	 * @payload_size: Payload buffer size in bytes
+	 *
+	 * Return: 0 success, negative errno value on transport failure,
+	 *         positive status returned by firmware
+	 */
+	int (*push_payload)(struct device *dev, u32 type, u32 flags,
+			    const void *payload, size_t payload_size);
+};
+
+#endif /* _INTEL_LB_MEI_INTERFACE_H_ */

include/linux/mei_cl_bus.h

···
 				mei_cldev_cb_t notif_cb);
 
 u8 mei_cldev_ver(const struct mei_cl_device *cldev);
+size_t mei_cldev_mtu(const struct mei_cl_device *cldev);
 
 void *mei_cldev_get_drvdata(const struct mei_cl_device *cldev);
 void mei_cldev_set_drvdata(struct mei_cl_device *cldev, void *data);