Merge tag 'drm-xe-fixes-2026-01-22' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-fixes

UAPI Changes:
- Disallow bind-queue sharing across multiple VMs (Matt Auld)

Core Changes:
- Fix xe userptr in the absence of CONFIG_DEVICE_PRIVATE (Thomas)

Driver Changes:
- Fix a missed page count update (Matt Brost)
- Fix a confused argument to alloc_workqueue() (Marco Crivellari)
- Kernel-doc fixes (Jani)
- Disable a workaround on VFs (Matt Brost)
- Fix a job lock assert (Matt Auld)
- Update wedged.mode only after successful reset policy change (Lukasz)
- Select CONFIG_DEVICE_PRIVATE when DRM_XE_GPUSVM is selected (Thomas)

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Thomas Hellstrom <thomas.hellstrom@linux.intel.com>
Link: https://patch.msgid.link/aXIdiXaY-RxoaviV@fedora

Author: airlied

drivers/gpu/drm/Kconfig

@@ -210,7 +210,7 @@ config DRM_GPUVM
 
 config DRM_GPUSVM
 	tristate
-	depends on DRM && DEVICE_PRIVATE
+	depends on DRM
 	select HMM_MIRROR
 	select MMU_NOTIFIER
 	help

drivers/gpu/drm/Makefile

@@ -108,8 +108,10 @@ obj-$(CONFIG_DRM_EXEC) += drm_exec.o
 obj-$(CONFIG_DRM_GPUVM) += drm_gpuvm.o
 
 drm_gpusvm_helper-y := \
-	drm_gpusvm.o\
+	drm_gpusvm.o
+drm_gpusvm_helper-$(CONFIG_ZONE_DEVICE) += \
 	drm_pagemap.o
+
 obj-$(CONFIG_DRM_GPUSVM) += drm_gpusvm_helper.o
 
 obj-$(CONFIG_DRM_BUDDY) += drm_buddy.o

drivers/gpu/drm/xe/Kconfig

@@ -39,7 +39,7 @@ config DRM_XE
 	select DRM_TTM
 	select DRM_TTM_HELPER
 	select DRM_EXEC
-	select DRM_GPUSVM if !UML && DEVICE_PRIVATE
+	select DRM_GPUSVM if !UML
 	select DRM_GPUVM
 	select DRM_SCHED
 	select MMU_NOTIFIER
@@ -80,8 +80,9 @@ config DRM_XE_GPUSVM
 	bool "Enable CPU to GPU address mirroring"
 	depends on DRM_XE
 	depends on !UML
-	depends on DEVICE_PRIVATE
+	depends on ZONE_DEVICE
 	default y
+	select DEVICE_PRIVATE
 	select DRM_GPUSVM
 	help
 	  Enable this option if you want support for CPU to GPU address

drivers/gpu/drm/xe/xe_bo.c

@@ -1055,6 +1055,7 @@ static long xe_bo_shrink_purge(struct ttm_operation_ctx *ctx,
 			       unsigned long *scanned)
 {
 	struct xe_device *xe = ttm_to_xe_device(bo->bdev);
+	struct ttm_tt *tt = bo->ttm;
 	long lret;
 
 	/* Fake move to system, without copying data. */
@@ -1079,8 +1080,10 @@ static long xe_bo_shrink_purge(struct ttm_operation_ctx *ctx,
 			      .writeback = false,
 			      .allow_move = false});
 
-	if (lret > 0)
+	if (lret > 0) {
 		xe_ttm_tt_account_subtract(xe, bo->ttm);
+		update_global_total_pages(bo->bdev, -(long)tt->num_pages);
+	}
 
 	return lret;
 }
@@ -1166,8 +1169,10 @@ long xe_bo_shrink(struct ttm_operation_ctx *ctx, struct ttm_buffer_object *bo,
 	if (needs_rpm)
 		xe_pm_runtime_put(xe);
 
-	if (lret > 0)
+	if (lret > 0) {
 		xe_ttm_tt_account_subtract(xe, tt);
+		update_global_total_pages(bo->bdev, -(long)tt->num_pages);
+	}
 
 out_unref:
 	xe_bo_put(xe_bo);

drivers/gpu/drm/xe/xe_debugfs.c

@@ -256,14 +256,64 @@ static ssize_t wedged_mode_show(struct file *f, char __user *ubuf,
 	return simple_read_from_buffer(ubuf, size, pos, buf, len);
 }
 
+static int __wedged_mode_set_reset_policy(struct xe_gt *gt, enum xe_wedged_mode mode)
+{
+	bool enable_engine_reset;
+	int ret;
+
+	enable_engine_reset = (mode != XE_WEDGED_MODE_UPON_ANY_HANG_NO_RESET);
+	ret = xe_guc_ads_scheduler_policy_toggle_reset(&gt->uc.guc.ads,
+						       enable_engine_reset);
+	if (ret)
+		xe_gt_err(gt, "Failed to update GuC ADS scheduler policy (%pe)\n", ERR_PTR(ret));
+
+	return ret;
+}
+
+static int wedged_mode_set_reset_policy(struct xe_device *xe, enum xe_wedged_mode mode)
+{
+	struct xe_gt *gt;
+	int ret;
+	u8 id;
+
+	guard(xe_pm_runtime)(xe);
+	for_each_gt(gt, xe, id) {
+		ret = __wedged_mode_set_reset_policy(gt, mode);
+		if (ret) {
+			if (id > 0) {
+				xe->wedged.inconsistent_reset = true;
+				drm_err(&xe->drm, "Inconsistent reset policy state between GTs\n");
+			}
+			return ret;
+		}
+	}
+
+	xe->wedged.inconsistent_reset = false;
+
+	return 0;
+}
+
+static bool wedged_mode_needs_policy_update(struct xe_device *xe, enum xe_wedged_mode mode)
+{
+	if (xe->wedged.inconsistent_reset)
+		return true;
+
+	if (xe->wedged.mode == mode)
+		return false;
+
+	if (xe->wedged.mode == XE_WEDGED_MODE_UPON_ANY_HANG_NO_RESET ||
+	    mode == XE_WEDGED_MODE_UPON_ANY_HANG_NO_RESET)
+		return true;
+
+	return false;
+}
+
 static ssize_t wedged_mode_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_gt *gt;
 	u32 wedged_mode;
 	ssize_t ret;
-	u8 id;
 
 	ret = kstrtouint_from_user(ubuf, size, 0, &wedged_mode);
 	if (ret)
@@ -272,22 +322,14 @@ static ssize_t wedged_mode_set(struct file *f, const char __user *ubuf,
 	if (wedged_mode > 2)
 		return -EINVAL;
 
-	if (xe->wedged.mode == wedged_mode)
-		return size;
+	if (wedged_mode_needs_policy_update(xe, wedged_mode)) {
+		ret = wedged_mode_set_reset_policy(xe, wedged_mode);
+		if (ret)
+			return ret;
+	}
 
 	xe->wedged.mode = wedged_mode;
 
-	xe_pm_runtime_get(xe);
-	for_each_gt(gt, xe, id) {
-		ret = xe_guc_ads_scheduler_policy_toggle_reset(&gt->uc.guc.ads);
-		if (ret) {
-			xe_gt_err(gt, "Failed to update GuC ADS scheduler policy. GuC may still cause engine reset even with wedged_mode=2\n");
-			xe_pm_runtime_put(xe);
-			return -EIO;
-		}
-	}
-	xe_pm_runtime_put(xe);
-
 	return size;
 }
 

drivers/gpu/drm/xe/xe_device_types.h

@@ -44,6 +44,22 @@ struct xe_pat_ops;
 struct xe_pxp;
 struct xe_vram_region;
 
+/**
+ * enum xe_wedged_mode - possible wedged modes
+ * @XE_WEDGED_MODE_NEVER: Device will never be declared wedged.
+ * @XE_WEDGED_MODE_UPON_CRITICAL_ERROR: Device will be declared wedged only
+ *	when critical error occurs like GT reset failure or firmware failure.
+ *	This is the default mode.
+ * @XE_WEDGED_MODE_UPON_ANY_HANG_NO_RESET: Device will be declared wedged on
+ *	any hang. In this mode, engine resets are disabled to avoid automatic
+ *	recovery attempts. This mode is primarily intended for debugging hangs.
+ */
+enum xe_wedged_mode {
+	XE_WEDGED_MODE_NEVER = 0,
+	XE_WEDGED_MODE_UPON_CRITICAL_ERROR = 1,
+	XE_WEDGED_MODE_UPON_ANY_HANG_NO_RESET = 2,
+};
+
 #define XE_BO_INVALID_OFFSET	LONG_MAX
 
 #define GRAPHICS_VER(xe) ((xe)->info.graphics_verx100 / 100)
@@ -587,6 +603,8 @@ struct xe_device {
 		int mode;
 		/** @wedged.method: Recovery method to be sent in the drm device wedged uevent */
 		unsigned long method;
+		/** @wedged.inconsistent_reset: Inconsistent reset policy state between GTs */
+		bool inconsistent_reset;
 	} wedged;
 
 	/** @bo_device: Struct to control async free of BOs */

drivers/gpu/drm/xe/xe_exec_queue.c

@@ -328,6 +328,7 @@ struct xe_exec_queue *xe_exec_queue_create_class(struct xe_device *xe, struct xe
  * @xe: Xe device.
  * @tile: tile which bind exec queue belongs to.
  * @flags: exec queue creation flags
+ * @user_vm: The user VM which this exec queue belongs to
  * @extensions: exec queue creation extensions
  *
  * Normalize bind exec queue creation. Bind exec queue is tied to migration VM
@@ -341,6 +342,7 @@ struct xe_exec_queue *xe_exec_queue_create_class(struct xe_device *xe, struct xe
  */
 struct xe_exec_queue *xe_exec_queue_create_bind(struct xe_device *xe,
 						struct xe_tile *tile,
+						struct xe_vm *user_vm,
 						u32 flags, u64 extensions)
 {
 	struct xe_gt *gt = tile->primary_gt;
@@ -377,6 +379,9 @@ struct xe_exec_queue *xe_exec_queue_create_bind(struct xe_device *xe,
 			xe_exec_queue_put(q);
 			return ERR_PTR(err);
 		}
+
+		if (user_vm)
+			q->user_vm = xe_vm_get(user_vm);
 	}
 
 	return q;
@@ -407,6 +412,11 @@ void xe_exec_queue_destroy(struct kref *ref)
 			xe_exec_queue_put(eq);
 	}
 
+	if (q->user_vm) {
+		xe_vm_put(q->user_vm);
+		q->user_vm = NULL;
+	}
+
 	q->ops->destroy(q);
 }
 
@@ -742,16 +752,34 @@ int xe_exec_queue_create_ioctl(struct drm_device *dev, void *data,
 		    XE_IOCTL_DBG(xe, eci[0].engine_instance != 0))
 			return -EINVAL;
 
+		vm = xe_vm_lookup(xef, args->vm_id);
+		if (XE_IOCTL_DBG(xe, !vm))
+			return -ENOENT;
+
+		err = down_read_interruptible(&vm->lock);
+		if (err) {
+			xe_vm_put(vm);
+			return err;
+		}
+
+		if (XE_IOCTL_DBG(xe, xe_vm_is_closed_or_banned(vm))) {
+			up_read(&vm->lock);
+			xe_vm_put(vm);
+			return -ENOENT;
+		}
+
 		for_each_tile(tile, xe, id) {
 			struct xe_exec_queue *new;
 
 			flags |= EXEC_QUEUE_FLAG_VM;
 			if (id)
 				flags |= EXEC_QUEUE_FLAG_BIND_ENGINE_CHILD;
 
-			new = xe_exec_queue_create_bind(xe, tile, flags,
+			new = xe_exec_queue_create_bind(xe, tile, vm, flags,
 							args->extensions);
 			if (IS_ERR(new)) {
+				up_read(&vm->lock);
+				xe_vm_put(vm);
 				err = PTR_ERR(new);
 				if (q)
 					goto put_exec_queue;
@@ -763,6 +791,8 @@ int xe_exec_queue_create_ioctl(struct drm_device *dev, void *data,
 				list_add_tail(&new->multi_gt_list,
 					      &q->multi_gt_link);
 		}
+		up_read(&vm->lock);
+		xe_vm_put(vm);
 	} else {
 		logical_mask = calc_validate_logical_mask(xe, eci,
 							  args->width,

drivers/gpu/drm/xe/xe_exec_queue.h

@@ -28,6 +28,7 @@ struct xe_exec_queue *xe_exec_queue_create_class(struct xe_device *xe, struct xe
 						 u32 flags, u64 extensions);
 struct xe_exec_queue *xe_exec_queue_create_bind(struct xe_device *xe,
 						struct xe_tile *tile,
+						struct xe_vm *user_vm,
 						u32 flags, u64 extensions);
 
 void xe_exec_queue_fini(struct xe_exec_queue *q);

drivers/gpu/drm/xe/xe_exec_queue_types.h

@@ -54,6 +54,12 @@ struct xe_exec_queue {
 	struct kref refcount;
 	/** @vm: VM (address space) for this exec queue */
 	struct xe_vm *vm;
+	/**
+	 * @user_vm: User VM (address space) for this exec queue (bind queues
+	 * only)
+	 */
+	struct xe_vm *user_vm;
+
 	/** @class: class of this exec queue */
 	enum xe_engine_class class;
 	/**

drivers/gpu/drm/xe/xe_ggtt.c

@@ -322,7 +322,7 @@ int xe_ggtt_init_early(struct xe_ggtt *ggtt)
 	else
 		ggtt->pt_ops = &xelp_pt_ops;
 
-	ggtt->wq = alloc_workqueue("xe-ggtt-wq", 0, WQ_MEM_RECLAIM);
+	ggtt->wq = alloc_workqueue("xe-ggtt-wq", WQ_MEM_RECLAIM, 0);
 	if (!ggtt->wq)
 		return -ENOMEM;