drm/xe: Introduce a new DRM driver for Intel GPUs

Xe, is a new driver for Intel GPUs that supports both integrated and
discrete platforms starting with Tiger Lake (first Intel Xe Architecture).

The code is at a stage where it is already functional and has experimental
support for multiple platforms starting from Tiger Lake, with initial
support implemented in Mesa (for Iris and Anv, our OpenGL and Vulkan
drivers), as well as in NEO (for OpenCL and Level0).

The new Xe driver leverages a lot from i915.

As for display, the intent is to share the display code with the i915
driver so that there is maximum reuse there. But it is not added
in this patch.

This initial work is a collaboration of many people and unfortunately
the big squashed patch won't fully honor the proper credits. But let's
get some git quick stats so we can at least try to preserve some of the
credits:

Co-developed-by: Matthew Brost <matthew.brost@intel.com>
Co-developed-by: Matthew Auld <matthew.auld@intel.com>
Co-developed-by: Matt Roper <matthew.d.roper@intel.com>
Co-developed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Co-developed-by: Francois Dugast <francois.dugast@intel.com>
Co-developed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Co-developed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Co-developed-by: Philippe Lecluse <philippe.lecluse@intel.com>
Co-developed-by: Nirmoy Das <nirmoy.das@intel.com>
Co-developed-by: Jani Nikula <jani.nikula@intel.com>
Co-developed-by: José Roberto de Souza <jose.souza@intel.com>
Co-developed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Co-developed-by: Dave Airlie <airlied@redhat.com>
Co-developed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Co-developed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Co-developed-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>

8 files changed, 1081 insertions, 0 deletions

diff --git a/drivers/gpu/drm/xe/tests/Makefile b/drivers/gpu/drm/xe/tests/Makefile
new file mode 100644
index 000000000000..47056b6459e3
--- /dev/null
+++ b/drivers/gpu/drm/xe/tests/Makefile
@@ -0,0 +1,4 @@
+# SPDX-License-Identifier: GPL-2.0
+
+obj-$(CONFIG_DRM_XE_KUNIT_TEST) += xe_bo_test.o xe_dma_buf_test.o \
+	xe_migrate_test.o
diff --git a/drivers/gpu/drm/xe/tests/xe_bo.c b/drivers/gpu/drm/xe/tests/xe_bo.c
new file mode 100644
index 000000000000..87ac21cc8ca9
--- /dev/null
+++ b/drivers/gpu/drm/xe/tests/xe_bo.c
@@ -0,0 +1,303 @@
+// SPDX-License-Identifier: GPL-2.0 AND MIT
+/*
+ * Copyright © 2022 Intel Corporation
+ */
+
+#include <kunit/test.h>
+
+#include "xe_bo_evict.h"
+#include "xe_pci.h"
+
+static int ccs_test_migrate(struct xe_gt *gt, struct xe_bo *bo,
+			    bool clear, u64 get_val, u64 assign_val,
+			    struct kunit *test)
+{
+	struct dma_fence *fence;
+	struct ttm_tt *ttm;
+	struct page *page;
+	pgoff_t ccs_page;
+	long timeout;
+	u64 *cpu_map;
+	int ret;
+	u32 offset;
+
+	/* Move bo to VRAM if not already there. */
+	ret = xe_bo_validate(bo, NULL, false);
+	if (ret) {
+		KUNIT_FAIL(test, "Failed to validate bo.\n");
+		return ret;
+	}
+
+	/* Optionally clear bo *and* CCS data in VRAM. */
+	if (clear) {
+		fence = xe_migrate_clear(gt->migrate, bo, bo->ttm.resource, 0);
+		if (IS_ERR(fence)) {
+			KUNIT_FAIL(test, "Failed to submit bo clear.\n");
+			return PTR_ERR(fence);
+		}
+		dma_fence_put(fence);
+	}
+
+	/* Evict to system. CCS data should be copied. */
+	ret = xe_bo_evict(bo, true);
+	if (ret) {
+		KUNIT_FAIL(test, "Failed to evict bo.\n");
+		return ret;
+	}
+
+	/* Sync all migration blits */
+	timeout = dma_resv_wait_timeout(bo->ttm.base.resv,
+					DMA_RESV_USAGE_KERNEL,
+					true,
+					5 * HZ);
+	if (timeout <= 0) {
+		KUNIT_FAIL(test, "Failed to sync bo eviction.\n");
+		return -ETIME;
+	}
+
+	/*
+	 * Bo with CCS data is now in system memory. Verify backing store
+	 * and data integrity. Then assign for the next testing round while
+	 * we still have a CPU map.
+	 */
+	ttm = bo->ttm.ttm;
+	if (!ttm || !ttm_tt_is_populated(ttm)) {
+		KUNIT_FAIL(test, "Bo was not in expected placement.\n");
+		return -EINVAL;
+	}
+
+	ccs_page = xe_bo_ccs_pages_start(bo) >> PAGE_SHIFT;
+	if (ccs_page >= ttm->num_pages) {
+		KUNIT_FAIL(test, "No TTM CCS pages present.\n");
+		return -EINVAL;
+	}
+
+	page = ttm->pages[ccs_page];
+	cpu_map = kmap_local_page(page);
+
+	/* Check first CCS value */
+	if (cpu_map[0] != get_val) {
+		KUNIT_FAIL(test,
+			   "Expected CCS readout 0x%016llx, got 0x%016llx.\n",
+			   (unsigned long long)get_val,
+			   (unsigned long long)cpu_map[0]);
+		ret = -EINVAL;
+	}
+
+	/* Check last CCS value, or at least last value in page. */
+	offset = xe_device_ccs_bytes(gt->xe, bo->size);
+	offset = min_t(u32, offset, PAGE_SIZE) / sizeof(u64) - 1;
+	if (cpu_map[offset] != get_val) {
+		KUNIT_FAIL(test,
+			   "Expected CCS readout 0x%016llx, got 0x%016llx.\n",
+			   (unsigned long long)get_val,
+			   (unsigned long long)cpu_map[offset]);
+		ret = -EINVAL;
+	}
+
+	cpu_map[0] = assign_val;
+	cpu_map[offset] = assign_val;
+	kunmap_local(cpu_map);
+
+	return ret;
+}
+
+static void ccs_test_run_gt(struct xe_device *xe, struct xe_gt *gt,
+			    struct kunit *test)
+{
+	struct xe_bo *bo;
+	u32 vram_bit;
+	int ret;
+
+	/* TODO: Sanity check */
+	vram_bit = XE_BO_CREATE_VRAM0_BIT << gt->info.vram_id;
+	kunit_info(test, "Testing gt id %u vram id %u\n", gt->info.id,
+		   gt->info.vram_id);
+
+	bo = xe_bo_create_locked(xe, NULL, NULL, SZ_1M, ttm_bo_type_device,
+				 vram_bit);
+	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(gt, bo, false, 0ULL, 0xdeadbeefdeadbeefULL,
+			       test);
+	if (ret)
+		goto out_unlock;
+
+	kunit_info(test, "Verifying that CCS data survives migration.\n");
+	ret = ccs_test_migrate(gt, bo, false, 0xdeadbeefdeadbeefULL,
+			       0xdeadbeefdeadbeefULL, test);
+	if (ret)
+		goto out_unlock;
+
+	kunit_info(test, "Verifying that CCS data can be properly cleared.\n");
+	ret = ccs_test_migrate(gt, bo, true, 0ULL, 0ULL, test);
+
+out_unlock:
+	xe_bo_unlock_no_vm(bo);
+	xe_bo_put(bo);
+}
+
+static int ccs_test_run_device(struct xe_device *xe)
+{
+	struct kunit *test = xe_cur_kunit();
+	struct xe_gt *gt;
+	int id;
+
+	if (!xe_device_has_flat_ccs(xe)) {
+		kunit_info(test, "Skipping non-flat-ccs device.\n");
+		return 0;
+	}
+
+	for_each_gt(gt, xe, id)
+		ccs_test_run_gt(xe, gt, test);
+
+	return 0;
+}
+
+void xe_ccs_migrate_kunit(struct kunit *test)
+{
+	xe_call_for_each_device(ccs_test_run_device);
+}
+EXPORT_SYMBOL(xe_ccs_migrate_kunit);
+
+static int evict_test_run_gt(struct xe_device *xe, struct xe_gt *gt, struct kunit *test)
+{
+	struct xe_bo *bo, *external;
+	unsigned int bo_flags = XE_BO_CREATE_USER_BIT |
+		XE_BO_CREATE_VRAM_IF_DGFX(gt);
+	struct xe_vm *vm = xe_migrate_get_vm(xe->gt[0].migrate);
+	struct ww_acquire_ctx ww;
+	int err, i;
+
+	kunit_info(test, "Testing device %s gt id %u vram id %u\n",
+		   dev_name(xe->drm.dev), gt->info.id, gt->info.vram_id);
+
+	for (i = 0; i < 2; ++i) {
+		xe_vm_lock(vm, &ww, 0, false);
+		bo = xe_bo_create(xe, NULL, vm, 0x10000, ttm_bo_type_device,
+				  bo_flags);
+		xe_vm_unlock(vm, &ww);
+		if (IS_ERR(bo)) {
+			KUNIT_FAIL(test, "bo create err=%pe\n", bo);
+			break;
+		}
+
+		external = xe_bo_create(xe, NULL, NULL, 0x10000,
+					ttm_bo_type_device, bo_flags);
+		if (IS_ERR(external)) {
+			KUNIT_FAIL(test, "external bo create err=%pe\n", external);
+			goto cleanup_bo;
+		}
+
+		xe_bo_lock(external, &ww, 0, false);
+		err = xe_bo_pin_external(external);
+		xe_bo_unlock(external, &ww);
+		if (err) {
+			KUNIT_FAIL(test, "external bo pin err=%pe\n",
+				   ERR_PTR(err));
+			goto cleanup_external;
+		}
+
+		err = xe_bo_evict_all(xe);
+		if (err) {
+			KUNIT_FAIL(test, "evict err=%pe\n", ERR_PTR(err));
+			goto cleanup_all;
+		}
+
+		err = xe_bo_restore_kernel(xe);
+		if (err) {
+			KUNIT_FAIL(test, "restore kernel err=%pe\n",
+				   ERR_PTR(err));
+			goto cleanup_all;
+		}
+
+		err = xe_bo_restore_user(xe);
+		if (err) {
+			KUNIT_FAIL(test, "restore user err=%pe\n", ERR_PTR(err));
+			goto cleanup_all;
+		}
+
+		if (!xe_bo_is_vram(external)) {
+			KUNIT_FAIL(test, "external bo is not vram\n");
+			err = -EPROTO;
+			goto cleanup_all;
+		}
+
+		if (xe_bo_is_vram(bo)) {
+			KUNIT_FAIL(test, "bo is vram\n");
+			err = -EPROTO;
+			goto cleanup_all;
+		}
+
+		if (i) {
+			down_read(&vm->lock);
+			xe_vm_lock(vm, &ww, 0, false);
+			err = xe_bo_validate(bo, bo->vm, false);
+			xe_vm_unlock(vm, &ww);
+			up_read(&vm->lock);
+			if (err) {
+				KUNIT_FAIL(test, "bo valid err=%pe\n",
+					   ERR_PTR(err));
+				goto cleanup_all;
+			}
+			xe_bo_lock(external, &ww, 0, false);
+			err = xe_bo_validate(external, NULL, false);
+			xe_bo_unlock(external, &ww);
+			if (err) {
+				KUNIT_FAIL(test, "external bo valid err=%pe\n",
+					   ERR_PTR(err));
+				goto cleanup_all;
+			}
+		}
+
+		xe_bo_lock(external, &ww, 0, false);
+		xe_bo_unpin_external(external);
+		xe_bo_unlock(external, &ww);
+
+		xe_bo_put(external);
+		xe_bo_put(bo);
+		continue;
+
+cleanup_all:
+		xe_bo_lock(external, &ww, 0, false);
+		xe_bo_unpin_external(external);
+		xe_bo_unlock(external, &ww);
+cleanup_external:
+		xe_bo_put(external);
+cleanup_bo:
+		xe_bo_put(bo);
+		break;
+	}
+
+	xe_vm_put(vm);
+
+	return 0;
+}
+
+static int evict_test_run_device(struct xe_device *xe)
+{
+	struct kunit *test = xe_cur_kunit();
+	struct xe_gt *gt;
+	int id;
+
+	if (!IS_DGFX(xe)) {
+		kunit_info(test, "Skipping non-discrete device %s.\n",
+			   dev_name(xe->drm.dev));
+		return 0;
+	}
+
+	for_each_gt(gt, xe, id)
+		evict_test_run_gt(xe, gt, test);
+
+	return 0;
+}
+
+void xe_bo_evict_kunit(struct kunit *test)
+{
+	xe_call_for_each_device(evict_test_run_device);
+}
+EXPORT_SYMBOL(xe_bo_evict_kunit);
diff --git a/drivers/gpu/drm/xe/tests/xe_bo_test.c b/drivers/gpu/drm/xe/tests/xe_bo_test.c
new file mode 100644
index 000000000000..c8fa29b0b3b2
--- /dev/null
+++ b/drivers/gpu/drm/xe/tests/xe_bo_test.c
@@ -0,0 +1,25 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright © 2022 Intel Corporation
+ */
+
+#include <kunit/test.h>
+
+void xe_ccs_migrate_kunit(struct kunit *test);
+void xe_bo_evict_kunit(struct kunit *test);
+
+static struct kunit_case xe_bo_tests[] = {
+	KUNIT_CASE(xe_ccs_migrate_kunit),
+	KUNIT_CASE(xe_bo_evict_kunit),
+	{}
+};
+
+static struct kunit_suite xe_bo_test_suite = {
+	.name = "xe_bo",
+	.test_cases = xe_bo_tests,
+};
+
+kunit_test_suite(xe_bo_test_suite);
+
+MODULE_AUTHOR("Intel Corporation");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/xe/tests/xe_dma_buf.c b/drivers/gpu/drm/xe/tests/xe_dma_buf.c
new file mode 100644
index 000000000000..615d22e3f731
--- /dev/null
+++ b/drivers/gpu/drm/xe/tests/xe_dma_buf.c
@@ -0,0 +1,259 @@
+// SPDX-License-Identifier: GPL-2.0 AND MIT
+/*
+ * Copyright © 2022 Intel Corporation
+ */
+
+#include <kunit/test.h>
+
+#include "xe_pci.h"
+
+static bool p2p_enabled(struct dma_buf_test_params *params)
+{
+	return IS_ENABLED(CONFIG_PCI_P2PDMA) && params->attach_ops &&
+		params->attach_ops->allow_peer2peer;
+}
+
+static bool is_dynamic(struct dma_buf_test_params *params)
+{
+	return IS_ENABLED(CONFIG_DMABUF_MOVE_NOTIFY) && params->attach_ops &&
+		params->attach_ops->move_notify;
+}
+
+static void check_residency(struct kunit *test, struct xe_bo *exported,
+			    struct xe_bo *imported, struct dma_buf *dmabuf)
+{
+	struct dma_buf_test_params *params = to_dma_buf_test_params(test->priv);
+	u32 mem_type;
+	int ret;
+
+	xe_bo_assert_held(exported);
+	xe_bo_assert_held(imported);
+
+	mem_type = XE_PL_VRAM0;
+	if (!(params->mem_mask & XE_BO_CREATE_VRAM0_BIT))
+		/* No VRAM allowed */
+		mem_type = XE_PL_TT;
+	else if (params->force_different_devices && !p2p_enabled(params))
+		/* No P2P */
+		mem_type = XE_PL_TT;
+	else if (params->force_different_devices && !is_dynamic(params) &&
+		 (params->mem_mask & XE_BO_CREATE_SYSTEM_BIT))
+		/* Pin migrated to TT */
+		mem_type = XE_PL_TT;
+
+	if (!xe_bo_is_mem_type(exported, mem_type)) {
+		KUNIT_FAIL(test, "Exported bo was not in expected memory type.\n");
+		return;
+	}
+
+	if (xe_bo_is_pinned(exported))
+		return;
+
+	/*
+	 * Evict exporter. Note that the gem object dma_buf member isn't
+	 * set from xe_gem_prime_export(), and it's needed for the move_notify()
+	 * functionality, so hack that up here. Evicting the exported bo will
+	 * evict also the imported bo through the move_notify() functionality if
+	 * importer is on a different device. If they're on the same device,
+	 * the exporter and the importer should be the same bo.
+	 */
+	swap(exported->ttm.base.dma_buf, dmabuf);
+	ret = xe_bo_evict(exported, true);
+	swap(exported->ttm.base.dma_buf, dmabuf);
+	if (ret) {
+		if (ret != -EINTR && ret != -ERESTARTSYS)
+			KUNIT_FAIL(test, "Evicting exporter failed with err=%d.\n",
+				   ret);
+		return;
+	}
+
+	/* Verify that also importer has been evicted to SYSTEM */
+	if (!xe_bo_is_mem_type(imported, XE_PL_SYSTEM)) {
+		KUNIT_FAIL(test, "Importer wasn't properly evicted.\n");
+		return;
+	}
+
+	/* Re-validate the importer. This should move also exporter in. */
+	ret = xe_bo_validate(imported, NULL, false);
+	if (ret) {
+		if (ret != -EINTR && ret != -ERESTARTSYS)
+			KUNIT_FAIL(test, "Validating importer failed with err=%d.\n",
+				   ret);
+		return;
+	}
+
+	/*
+	 * If on different devices, the exporter is kept in system  if
+	 * possible, saving a migration step as the transfer is just
+	 * likely as fast from system memory.
+	 */
+	if (params->force_different_devices &&
+	    params->mem_mask & XE_BO_CREATE_SYSTEM_BIT)
+		KUNIT_EXPECT_TRUE(test, xe_bo_is_mem_type(exported, XE_PL_TT));
+	else
+		KUNIT_EXPECT_TRUE(test, xe_bo_is_mem_type(exported, mem_type));
+
+	if (params->force_different_devices)
+		KUNIT_EXPECT_TRUE(test, xe_bo_is_mem_type(imported, XE_PL_TT));
+	else
+		KUNIT_EXPECT_TRUE(test, exported == imported);
+}
+
+static void xe_test_dmabuf_import_same_driver(struct xe_device *xe)
+{
+	struct kunit *test = xe_cur_kunit();
+	struct dma_buf_test_params *params = to_dma_buf_test_params(test->priv);
+	struct drm_gem_object *import;
+	struct dma_buf *dmabuf;
+	struct xe_bo *bo;
+
+	/* No VRAM on this device? */
+	if (!ttm_manager_type(&xe->ttm, XE_PL_VRAM0) &&
+	    (params->mem_mask & XE_BO_CREATE_VRAM0_BIT))
+		return;
+
+	kunit_info(test, "running %s\n", __func__);
+	bo = xe_bo_create(xe, NULL, NULL, PAGE_SIZE, ttm_bo_type_device,
+			  XE_BO_CREATE_USER_BIT | params->mem_mask);
+	if (IS_ERR(bo)) {
+		KUNIT_FAIL(test, "xe_bo_create() failed with err=%ld\n",
+			   PTR_ERR(bo));
+		return;
+	}
+
+	dmabuf = xe_gem_prime_export(&bo->ttm.base, 0);
+	if (IS_ERR(dmabuf)) {
+		KUNIT_FAIL(test, "xe_gem_prime_export() failed with err=%ld\n",
+			   PTR_ERR(dmabuf));
+		goto out;
+	}
+
+	import = xe_gem_prime_import(&xe->drm, dmabuf);
+	if (!IS_ERR(import)) {
+		struct xe_bo *import_bo = gem_to_xe_bo(import);
+
+		/*
+		 * Did import succeed when it shouldn't due to lack of p2p support?
+		 */
+		if (params->force_different_devices &&
+		    !p2p_enabled(params) &&
+		    !(params->mem_mask & XE_BO_CREATE_SYSTEM_BIT)) {
+			KUNIT_FAIL(test,
+				   "xe_gem_prime_import() succeeded when it shouldn't have\n");
+		} else {
+			int err;
+
+			/* Is everything where we expect it to be? */
+			xe_bo_lock_no_vm(import_bo, NULL);
+			err = xe_bo_validate(import_bo, NULL, false);
+			if (err && err != -EINTR && err != -ERESTARTSYS)
+				KUNIT_FAIL(test,
+					   "xe_bo_validate() failed with err=%d\n", err);
+
+			check_residency(test, bo, import_bo, dmabuf);
+			xe_bo_unlock_no_vm(import_bo);
+		}
+		drm_gem_object_put(import);
+	} else if (PTR_ERR(import) != -EOPNOTSUPP) {
+		/* Unexpected error code. */
+		KUNIT_FAIL(test,
+			   "xe_gem_prime_import failed with the wrong err=%ld\n",
+			   PTR_ERR(import));
+	} else if (!params->force_different_devices ||
+		   p2p_enabled(params) ||
+		   (params->mem_mask & XE_BO_CREATE_SYSTEM_BIT)) {
+		/* Shouldn't fail if we can reuse same bo, use p2p or use system */
+		KUNIT_FAIL(test, "dynamic p2p attachment failed with err=%ld\n",
+			   PTR_ERR(import));
+	}
+	dma_buf_put(dmabuf);
+out:
+	drm_gem_object_put(&bo->ttm.base);
+}
+
+static const struct dma_buf_attach_ops nop2p_attach_ops = {
+	.allow_peer2peer = false,
+	.move_notify = xe_dma_buf_move_notify
+};
+
+/*
+ * We test the implementation with bos of different residency and with
+ * importers with different capabilities; some lacking p2p support and some
+ * lacking dynamic capabilities (attach_ops == NULL). We also fake
+ * different devices avoiding the import shortcut that just reuses the same
+ * gem object.
+ */
+static const struct dma_buf_test_params test_params[] = {
+	{.mem_mask = XE_BO_CREATE_VRAM0_BIT,
+	 .attach_ops = &xe_dma_buf_attach_ops},
+	{.mem_mask = XE_BO_CREATE_VRAM0_BIT,
+	 .attach_ops = &xe_dma_buf_attach_ops,
+	 .force_different_devices = true},
+
+	{.mem_mask = XE_BO_CREATE_VRAM0_BIT,
+	 .attach_ops = &nop2p_attach_ops},
+	{.mem_mask = XE_BO_CREATE_VRAM0_BIT,
+	 .attach_ops = &nop2p_attach_ops,
+	 .force_different_devices = true},
+
+	{.mem_mask = XE_BO_CREATE_VRAM0_BIT},
+	{.mem_mask = XE_BO_CREATE_VRAM0_BIT,
+	 .force_different_devices = true},
+
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT,
+	 .attach_ops = &xe_dma_buf_attach_ops},
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT,
+	 .attach_ops = &xe_dma_buf_attach_ops,
+	 .force_different_devices = true},
+
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT,
+	 .attach_ops = &nop2p_attach_ops},
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT,
+	 .attach_ops = &nop2p_attach_ops,
+	 .force_different_devices = true},
+
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT},
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT,
+	 .force_different_devices = true},
+
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT | XE_BO_CREATE_VRAM0_BIT,
+	 .attach_ops = &xe_dma_buf_attach_ops},
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT | XE_BO_CREATE_VRAM0_BIT,
+	 .attach_ops = &xe_dma_buf_attach_ops,
+	 .force_different_devices = true},
+
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT | XE_BO_CREATE_VRAM0_BIT,
+	 .attach_ops = &nop2p_attach_ops},
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT | XE_BO_CREATE_VRAM0_BIT,
+	 .attach_ops = &nop2p_attach_ops,
+	 .force_different_devices = true},
+
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT | XE_BO_CREATE_VRAM0_BIT},
+	{.mem_mask = XE_BO_CREATE_SYSTEM_BIT | XE_BO_CREATE_VRAM0_BIT,
+	 .force_different_devices = true},
+
+	{}
+};
+
+static int dma_buf_run_device(struct xe_device *xe)
+{
+	const struct dma_buf_test_params *params;
+	struct kunit *test = xe_cur_kunit();
+
+	for (params = test_params; params->mem_mask; ++params) {
+		struct dma_buf_test_params p = *params;
+
+		p.base.id = XE_TEST_LIVE_DMA_BUF;
+		test->priv = &p;
+		xe_test_dmabuf_import_same_driver(xe);
+	}
+
+	/* A non-zero return would halt iteration over driver devices */
+	return 0;
+}
+
+void xe_dma_buf_kunit(struct kunit *test)
+{
+	xe_call_for_each_device(dma_buf_run_device);
+}
+EXPORT_SYMBOL(xe_dma_buf_kunit);
diff --git a/drivers/gpu/drm/xe/tests/xe_dma_buf_test.c b/drivers/gpu/drm/xe/tests/xe_dma_buf_test.c
new file mode 100644
index 000000000000..7bb292da1193
--- /dev/null
+++ b/drivers/gpu/drm/xe/tests/xe_dma_buf_test.c
@@ -0,0 +1,23 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright © 2022 Intel Corporation
+ */
+
+#include <kunit/test.h>
+
+void xe_dma_buf_kunit(struct kunit *test);
+
+static struct kunit_case xe_dma_buf_tests[] = {
+	KUNIT_CASE(xe_dma_buf_kunit),
+	{}
+};
+
+static struct kunit_suite xe_dma_buf_test_suite = {
+	.name = "xe_dma_buf",
+	.test_cases = xe_dma_buf_tests,
+};
+
+kunit_test_suite(xe_dma_buf_test_suite);
+
+MODULE_AUTHOR("Intel Corporation");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/xe/tests/xe_migrate.c b/drivers/gpu/drm/xe/tests/xe_migrate.c
new file mode 100644
index 000000000000..0f3b819f0a34
--- /dev/null
+++ b/drivers/gpu/drm/xe/tests/xe_migrate.c
@@ -0,0 +1,378 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2020-2022 Intel Corporation
+ */
+
+#include <kunit/test.h>
+
+#include "xe_pci.h"
+
+static bool sanity_fence_failed(struct xe_device *xe, struct dma_fence *fence,
+				const char *str, struct kunit *test)
+{
+	long ret;
+
+	if (IS_ERR(fence)) {
+		KUNIT_FAIL(test, "Failed to create fence for %s: %li\n", str,
+			   PTR_ERR(fence));
+		return true;
+	}
+	if (!fence)
+		return true;
+
+	ret = dma_fence_wait_timeout(fence, false, 5 * HZ);
+	if (ret <= 0) {
+		KUNIT_FAIL(test, "Fence timed out for %s: %li\n", str, ret);
+		return true;
+	}
+
+	return false;
+}
+
+static int run_sanity_job(struct xe_migrate *m, struct xe_device *xe,
+			  struct xe_bb *bb, u32 second_idx, const char *str,
+			  struct kunit *test)
+{
+	struct xe_sched_job *job = xe_bb_create_migration_job(m->eng, bb,
+							      m->batch_base_ofs,
+							      second_idx);
+	struct dma_fence *fence;
+
+	if (IS_ERR(job)) {
+		KUNIT_FAIL(test, "Failed to allocate fake pt: %li\n",
+			   PTR_ERR(job));
+		return PTR_ERR(job);
+	}
+
+	xe_sched_job_arm(job);
+	fence = dma_fence_get(&job->drm.s_fence->finished);
+	xe_sched_job_push(job);
+
+	if (sanity_fence_failed(xe, fence, str, test))
+		return -ETIMEDOUT;
+
+	dma_fence_put(fence);
+	kunit_info(test, "%s: Job completed\n", str);
+	return 0;
+}
+
+static void
+sanity_populate_cb(struct xe_migrate_pt_update *pt_update,
+		   struct xe_gt *gt, struct iosys_map *map, void *dst,
+		   u32 qword_ofs, u32 num_qwords,
+		   const struct xe_vm_pgtable_update *update)
+{
+	int i;
+	u64 *ptr = dst;
+
+	for (i = 0; i < num_qwords; i++)
+		ptr[i] = (qword_ofs + i - update->ofs) * 0x1111111111111111ULL;
+}
+
+static const struct xe_migrate_pt_update_ops sanity_ops = {
+	.populate = sanity_populate_cb,
+};
+
+#define check(_retval, _expected, str, _test)				\
+	do { if ((_retval) != (_expected)) {				\
+			KUNIT_FAIL(_test, "Sanity check failed: " str	\
+				   " expected %llx, got %llx\n",	\
+				   (u64)(_expected), (u64)(_retval));	\
+		} } while (0)
+
+static void test_copy(struct xe_migrate *m, struct xe_bo *bo,
+		      struct kunit *test)
+{
+	struct xe_device *xe = gt_to_xe(m->gt);
+	u64 retval, expected = 0xc0c0c0c0c0c0c0c0ULL;
+	bool big = bo->size >= SZ_2M;
+	struct dma_fence *fence;
+	const char *str = big ? "Copying big bo" : "Copying small bo";
+	int err;
+
+	struct xe_bo *sysmem = xe_bo_create_locked(xe, m->gt, NULL,
+						   bo->size,
+						   ttm_bo_type_kernel,
+						   XE_BO_CREATE_SYSTEM_BIT);
+	if (IS_ERR(sysmem)) {
+		KUNIT_FAIL(test, "Failed to allocate sysmem bo for %s: %li\n",
+			   str, PTR_ERR(sysmem));
+		return;
+	}
+
+	err = xe_bo_validate(sysmem, NULL, false);
+	if (err) {
+		KUNIT_FAIL(test, "Failed to validate system bo for %s: %li\n",
+			   str, err);
+		goto out_unlock;
+	}
+
+	err = xe_bo_vmap(sysmem);
+	if (err) {
+		KUNIT_FAIL(test, "Failed to vmap system bo for %s: %li\n",
+			   str, err);
+		goto out_unlock;
+	}
+
+	xe_map_memset(xe, &sysmem->vmap, 0, 0xd0, sysmem->size);
+	fence = xe_migrate_clear(m, sysmem, sysmem->ttm.resource, 0xc0c0c0c0);
+	if (!sanity_fence_failed(xe, fence, big ? "Clearing sysmem big bo" :
+				 "Clearing sysmem small bo", test)) {
+		retval = xe_map_rd(xe, &sysmem->vmap, 0, u64);
+		check(retval, expected, "sysmem first offset should be cleared",
+		      test);
+		retval = xe_map_rd(xe, &sysmem->vmap, sysmem->size - 8, u64);
+		check(retval, expected, "sysmem last offset should be cleared",
+		      test);
+	}
+	dma_fence_put(fence);
+
+	/* Try to copy 0xc0 from sysmem to lmem with 2MB or 64KiB/4KiB pages */
+	xe_map_memset(xe, &sysmem->vmap, 0, 0xc0, sysmem->size);
+	xe_map_memset(xe, &bo->vmap, 0, 0xd0, bo->size);
+
+	fence = xe_migrate_copy(m, sysmem, sysmem->ttm.resource,
+				bo->ttm.resource);
+	if (!sanity_fence_failed(xe, fence, big ? "Copying big bo sysmem -> vram" :
+				 "Copying small bo sysmem -> vram", test)) {
+		retval = xe_map_rd(xe, &bo->vmap, 0, u64);
+		check(retval, expected,
+		      "sysmem -> vram bo first offset should be copied", test);
+		retval = xe_map_rd(xe, &bo->vmap, bo->size - 8, u64);
+		check(retval, expected,
+		      "sysmem -> vram bo offset should be copied", test);
+	}
+	dma_fence_put(fence);
+
+	/* And other way around.. slightly hacky.. */
+	xe_map_memset(xe, &sysmem->vmap, 0, 0xd0, sysmem->size);
+	xe_map_memset(xe, &bo->vmap, 0, 0xc0, bo->size);
+
+	fence = xe_migrate_copy(m, sysmem, bo->ttm.resource,
+				sysmem->ttm.resource);
+	if (!sanity_fence_failed(xe, fence, big ? "Copying big bo vram -> sysmem" :
+				 "Copying small bo vram -> sysmem", test)) {
+		retval = xe_map_rd(xe, &sysmem->vmap, 0, u64);
+		check(retval, expected,
+		      "vram -> sysmem bo first offset should be copied", test);
+		retval = xe_map_rd(xe, &sysmem->vmap, bo->size - 8, u64);
+		check(retval, expected,
+		      "vram -> sysmem bo last offset should be copied", test);
+	}
+	dma_fence_put(fence);
+
+	xe_bo_vunmap(sysmem);
+out_unlock:
+	xe_bo_unlock_no_vm(sysmem);
+	xe_bo_put(sysmem);
+}
+
+static void test_pt_update(struct xe_migrate *m, struct xe_bo *pt,
+			   struct kunit *test)
+{
+	struct xe_device *xe = gt_to_xe(m->gt);
+	struct dma_fence *fence;
+	u64 retval, expected;
+	int i;
+
+	struct xe_vm_pgtable_update update = {
+		.ofs = 1,
+		.qwords = 0x10,
+		.pt_bo = pt,
+	};
+	struct xe_migrate_pt_update pt_update = {
+		.ops = &sanity_ops,
+	};
+
+	/* Test xe_migrate_update_pgtables() updates the pagetable as expected */
+	expected = 0xf0f0f0f0f0f0f0f0ULL;
+	xe_map_memset(xe, &pt->vmap, 0, (u8)expected, pt->size);
+
+	fence = xe_migrate_update_pgtables(m, NULL, NULL, m->eng, &update, 1,
+					   NULL, 0, &pt_update);
+	if (sanity_fence_failed(xe, fence, "Migration pagetable update", test))
+		return;
+
+	dma_fence_put(fence);
+	retval = xe_map_rd(xe, &pt->vmap, 0, u64);
+	check(retval, expected, "PTE[0] must stay untouched", test);
+
+	for (i = 0; i < update.qwords; i++) {
+		retval = xe_map_rd(xe, &pt->vmap, (update.ofs + i) * 8, u64);
+		check(retval, i * 0x1111111111111111ULL, "PTE update", test);
+	}
+
+	retval = xe_map_rd(xe, &pt->vmap, 8 * (update.ofs + update.qwords),
+			   u64);
+	check(retval, expected, "PTE[0x11] must stay untouched", test);
+}
+
+static void xe_migrate_sanity_test(struct xe_migrate *m, struct kunit *test)
+{
+	struct xe_gt *gt = m->gt;
+	struct xe_device *xe = gt_to_xe(gt);
+	struct xe_bo *pt, *bo = m->pt_bo, *big, *tiny;
+	struct xe_res_cursor src_it;
+	struct dma_fence *fence;
+	u64 retval, expected;
+	struct xe_bb *bb;
+	int err;
+	u8 id = gt->info.id;
+
+	err = xe_bo_vmap(bo);
+	if (err) {
+		KUNIT_FAIL(test, "Failed to vmap our pagetables: %li\n",
+			   PTR_ERR(bo));
+		return;
+	}
+
+	big = xe_bo_create_pin_map(xe, m->gt, m->eng->vm, SZ_4M,
+				   ttm_bo_type_kernel,
+				   XE_BO_CREATE_VRAM_IF_DGFX(m->gt) |
+				   XE_BO_CREATE_PINNED_BIT);
+	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, m->gt, m->eng->vm, GEN8_PAGE_SIZE,
+				  ttm_bo_type_kernel,
+				  XE_BO_CREATE_VRAM_IF_DGFX(m->gt) |
+				  XE_BO_CREATE_PINNED_BIT);
+	if (IS_ERR(pt)) {
+		KUNIT_FAIL(test, "Failed to allocate fake pt: %li\n",
+			   PTR_ERR(pt));
+		goto free_big;
+	}
+
+	tiny = xe_bo_create_pin_map(xe, m->gt, m->eng->vm,
+				    2 * SZ_4K,
+				    ttm_bo_type_kernel,
+				    XE_BO_CREATE_VRAM_IF_DGFX(m->gt) |
+				    XE_BO_CREATE_PINNED_BIT);
+	if (IS_ERR(tiny)) {
+		KUNIT_FAIL(test, "Failed to allocate fake pt: %li\n",
+			   PTR_ERR(pt));
+		goto free_pt;
+	}
+
+	bb = xe_bb_new(m->gt, 32, xe->info.supports_usm);
+	if (IS_ERR(bb)) {
+		KUNIT_FAIL(test, "Failed to create batchbuffer: %li\n",
+			   PTR_ERR(bb));
+		goto free_tiny;
+	}
+
+	kunit_info(test, "Starting tests, top level PT addr: %llx, special pagetable base addr: %llx\n",
+		   xe_bo_main_addr(m->eng->vm->pt_root[id]->bo, GEN8_PAGE_SIZE),
+		   xe_bo_main_addr(m->pt_bo, GEN8_PAGE_SIZE));
+
+	/* First part of the test, are we updating our pagetable bo with a new entry? */
+	xe_map_wr(xe, &bo->vmap, GEN8_PAGE_SIZE * (NUM_KERNEL_PDE - 1), u64, 0xdeaddeadbeefbeef);
+	expected = gen8_pte_encode(NULL, pt, 0, XE_CACHE_WB, 0, 0);
+	if (m->eng->vm->flags & XE_VM_FLAGS_64K)
+		expected |= GEN12_PTE_PS64;
+	xe_res_first(pt->ttm.resource, 0, pt->size, &src_it);
+	emit_pte(m, bb, NUM_KERNEL_PDE - 1, xe_bo_is_vram(pt),
+		 &src_it, GEN8_PAGE_SIZE, pt);
+	run_sanity_job(m, xe, bb, bb->len, "Writing PTE for our fake PT", test);
+
+	retval = xe_map_rd(xe, &bo->vmap, GEN8_PAGE_SIZE * (NUM_KERNEL_PDE - 1),
+			   u64);
+	check(retval, expected, "PTE entry write", test);
+
+	/* Now try to write data to our newly mapped 'pagetable', see if it succeeds */
+	bb->len = 0;
+	bb->cs[bb->len++] = MI_BATCH_BUFFER_END;
+	xe_map_wr(xe, &pt->vmap, 0, u32, 0xdeaddead);
+	expected = 0x12345678U;
+
+	emit_clear(m->gt, bb, xe_migrate_vm_addr(NUM_KERNEL_PDE - 1, 0), 4, 4,
+		   expected, IS_DGFX(xe));
+	run_sanity_job(m, xe, bb, 1, "Writing to our newly mapped pagetable",
+		       test);
+
+	retval = xe_map_rd(xe, &pt->vmap, 0, u32);
+	check(retval, expected, "Write to PT after adding PTE", test);
+
+	/* Sanity checks passed, try the full ones! */
+
+	/* Clear a small bo */
+	kunit_info(test, "Clearing small buffer object\n");
+	xe_map_memset(xe, &tiny->vmap, 0, 0x22, tiny->size);
+	expected = 0x224488ff;
+	fence = xe_migrate_clear(m, tiny, tiny->ttm.resource, expected);
+	if (sanity_fence_failed(xe, fence, "Clearing small bo", test))
+		goto out;
+
+	dma_fence_put(fence);
+	retval = xe_map_rd(xe, &tiny->vmap, 0, u32);
+	check(retval, expected, "Command clear small first value", test);
+	retval = xe_map_rd(xe, &tiny->vmap, tiny->size - 4, u32);
+	check(retval, expected, "Command clear small last value", test);
+
+	if (IS_DGFX(xe)) {
+		kunit_info(test, "Copying small buffer object to system\n");
+		test_copy(m, tiny, test);
+	}
+
+	/* Clear a big bo with a fixed value */
+	kunit_info(test, "Clearing big buffer object\n");
+	xe_map_memset(xe, &big->vmap, 0, 0x11, big->size);
+	expected = 0x11223344U;
+	fence = xe_migrate_clear(m, big, big->ttm.resource, expected);
+	if (sanity_fence_failed(xe, fence, "Clearing big bo", test))
+		goto out;
+
+	dma_fence_put(fence);
+	retval = xe_map_rd(xe, &big->vmap, 0, u32);
+	check(retval, expected, "Command clear big first value", test);
+	retval = xe_map_rd(xe, &big->vmap, big->size - 4, u32);
+	check(retval, expected, "Command clear big last value", test);
+
+	if (IS_DGFX(xe)) {
+		kunit_info(test, "Copying big buffer object to system\n");
+		test_copy(m, big, test);
+	}
+
+	test_pt_update(m, pt, test);
+
+out:
+	xe_bb_free(bb, NULL);
+free_tiny:
+	xe_bo_unpin(tiny);
+	xe_bo_put(tiny);
+free_pt:
+	xe_bo_unpin(pt);
+	xe_bo_put(pt);
+free_big:
+	xe_bo_unpin(big);
+	xe_bo_put(