Merge tag 'for-linus-iommufd' of git://git.kernel.org/pub/scm/linux/kernel/git/jgg/iommufd

Pull iommufd fixes from Jason Gunthorpe:

 - Fix dirty tracking bitmap collection when using reporting bitmaps
   that are not neatly aligned to u64's or match the IO page table radix
   tree layout.

 - Add self tests to cover the cases that were found to be broken.

 - Add missing enforcement of invalidation type in the uapi.

 - Fix selftest config generation

* tag 'for-linus-iommufd' of git://git.kernel.org/pub/scm/linux/kernel/git/jgg/iommufd:
  selftests/iommu: fix the config fragment
  iommufd: Reject non-zero data_type if no data_len is provided
  iommufd/iova_bitmap: Consider page offset for the pages to be pinned
  iommufd/selftest: Add mock IO hugepages tests
  iommufd/selftest: Hugepage mock domain support
  iommufd/selftest: Refactor mock_domain_read_and_clear_dirty()
  iommufd/selftest: Refactor dirty bitmap tests
  iommufd/iova_bitmap: Handle recording beyond the mapped pages
  iommufd/selftest: Test u64 unaligned bitmaps
  iommufd/iova_bitmap: Switch iova_bitmap::bitmap to an u8 array
  iommufd/iova_bitmap: Bounds check mapped::pages access

7 files changed, 210 insertions, 63 deletions

diff --git a/drivers/iommu/iommufd/hw_pagetable.c b/drivers/iommu/iommufd/hw_pagetable.c
index 3f3f1fa1a0a9..33d142f8057d 100644
--- a/drivers/iommu/iommufd/hw_pagetable.c
+++ b/drivers/iommu/iommufd/hw_pagetable.c
@@ -263,7 +263,8 @@ int iommufd_hwpt_alloc(struct iommufd_ucmd *ucmd)
 
 	if (cmd->__reserved)
 		return -EOPNOTSUPP;
-	if (cmd->data_type == IOMMU_HWPT_DATA_NONE && cmd->data_len)
+	if ((cmd->data_type == IOMMU_HWPT_DATA_NONE && cmd->data_len) ||
+	    (cmd->data_type != IOMMU_HWPT_DATA_NONE && !cmd->data_len))
 		return -EINVAL;
 
 	idev = iommufd_get_device(ucmd, cmd->dev_id);
diff --git a/drivers/iommu/iommufd/iommufd_test.h b/drivers/iommu/iommufd/iommufd_test.h
index 482d4059f5db..e854d3f67205 100644
--- a/drivers/iommu/iommufd/iommufd_test.h
+++ b/drivers/iommu/iommufd/iommufd_test.h
@@ -45,6 +45,7 @@ enum {
 
 enum {
 	MOCK_FLAGS_DEVICE_NO_DIRTY = 1 << 0,
+	MOCK_FLAGS_DEVICE_HUGE_IOVA = 1 << 1,
 };
 
 enum {
diff --git a/drivers/iommu/iommufd/iova_bitmap.c b/drivers/iommu/iommufd/iova_bitmap.c
index 0a92c9eeaf7f..db8c46bee155 100644
--- a/drivers/iommu/iommufd/iova_bitmap.c
+++ b/drivers/iommu/iommufd/iova_bitmap.c
@@ -100,7 +100,7 @@ struct iova_bitmap {
 	struct iova_bitmap_map mapped;
 
 	/* userspace address of the bitmap */
-	u64 __user *bitmap;
+	u8 __user *bitmap;
 
 	/* u64 index that @mapped points to */
 	unsigned long mapped_base_index;
@@ -113,6 +113,9 @@ struct iova_bitmap {
 
 	/* length of the IOVA range for the whole bitmap */
 	size_t length;
+
+	/* length of the IOVA range set ahead the pinned pages */
+	unsigned long set_ahead_length;
 };
 
 /*
@@ -162,7 +165,7 @@ static int iova_bitmap_get(struct iova_bitmap *bitmap)
 {
 	struct iova_bitmap_map *mapped = &bitmap->mapped;
 	unsigned long npages;
-	u64 __user *addr;
+	u8 __user *addr;
 	long ret;
 
 	/*
@@ -176,17 +179,18 @@ static int iova_bitmap_get(struct iova_bitmap *bitmap)
 			       sizeof(*bitmap->bitmap), PAGE_SIZE);
 
 	/*
-	 * We always cap at max number of 'struct page' a base page can fit.
-	 * This is, for example, on x86 means 2M of bitmap data max.
-	 */
-	npages = min(npages,  PAGE_SIZE / sizeof(struct page *));
-
-	/*
 	 * Bitmap address to be pinned is calculated via pointer arithmetic
 	 * with bitmap u64 word index.
 	 */
 	addr = bitmap->bitmap + bitmap->mapped_base_index;
 
+	/*
+	 * We always cap at max number of 'struct page' a base page can fit.
+	 * This is, for example, on x86 means 2M of bitmap data max.
+	 */
+	npages = min(npages + !!offset_in_page(addr),
+		     PAGE_SIZE / sizeof(struct page *));
+
 	ret = pin_user_pages_fast((unsigned long)addr, npages,
 				  FOLL_WRITE, mapped->pages);
 	if (ret <= 0)
@@ -247,7 +251,7 @@ struct iova_bitmap *iova_bitmap_alloc(unsigned long iova, size_t length,
 
 	mapped = &bitmap->mapped;
 	mapped->pgshift = __ffs(page_size);
-	bitmap->bitmap = data;
+	bitmap->bitmap = (u8 __user *)data;
 	bitmap->mapped_total_index =
 		iova_bitmap_offset_to_index(bitmap, length - 1) + 1;
 	bitmap->iova = iova;
@@ -304,7 +308,7 @@ static unsigned long iova_bitmap_mapped_remaining(struct iova_bitmap *bitmap)
 
 	remaining = bitmap->mapped_total_index - bitmap->mapped_base_index;
 	remaining = min_t(unsigned long, remaining,
-			  bytes / sizeof(*bitmap->bitmap));
+			  DIV_ROUND_UP(bytes, sizeof(*bitmap->bitmap)));
 
 	return remaining;
 }
@@ -341,6 +345,32 @@ static bool iova_bitmap_done(struct iova_bitmap *bitmap)
 	return bitmap->mapped_base_index >= bitmap->mapped_total_index;
 }
 
+static int iova_bitmap_set_ahead(struct iova_bitmap *bitmap,
+				 size_t set_ahead_length)
+{
+	int ret = 0;
+
+	while (set_ahead_length > 0 && !iova_bitmap_done(bitmap)) {
+		unsigned long length = iova_bitmap_mapped_length(bitmap);
+		unsigned long iova = iova_bitmap_mapped_iova(bitmap);
+
+		ret = iova_bitmap_get(bitmap);
+		if (ret)
+			break;
+
+		length = min(length, set_ahead_length);
+		iova_bitmap_set(bitmap, iova, length);
+
+		set_ahead_length -= length;
+		bitmap->mapped_base_index +=
+			iova_bitmap_offset_to_index(bitmap, length - 1) + 1;
+		iova_bitmap_put(bitmap);
+	}
+
+	bitmap->set_ahead_length = 0;
+	return ret;
+}
+
 /*
  * Advances to the next range, releases the current pinned
  * pages and pins the next set of bitmap pages.
@@ -357,6 +387,15 @@ static int iova_bitmap_advance(struct iova_bitmap *bitmap)
 	if (iova_bitmap_done(bitmap))
 		return 0;
 
+	/* Iterate, set and skip any bits requested for next iteration */
+	if (bitmap->set_ahead_length) {
+		int ret;
+
+		ret = iova_bitmap_set_ahead(bitmap, bitmap->set_ahead_length);
+		if (ret)
+			return ret;
+	}
+
 	/* When advancing the index we pin the next set of bitmap pages */
 	return iova_bitmap_get(bitmap);
 }
@@ -409,6 +448,7 @@ void iova_bitmap_set(struct iova_bitmap *bitmap,
 			mapped->pgshift) + mapped->pgoff * BITS_PER_BYTE;
 	unsigned long last_bit = (((iova + length - 1) - mapped->iova) >>
 			mapped->pgshift) + mapped->pgoff * BITS_PER_BYTE;
+	unsigned long last_page_idx = mapped->npages - 1;
 
 	do {
 		unsigned int page_idx = cur_bit / BITS_PER_PAGE;
@@ -417,10 +457,18 @@ void iova_bitmap_set(struct iova_bitmap *bitmap,
 					 last_bit - cur_bit + 1);
 		void *kaddr;
 
+		if (unlikely(page_idx > last_page_idx))
+			break;
+
 		kaddr = kmap_local_page(mapped->pages[page_idx]);
 		bitmap_set(kaddr, offset, nbits);
 		kunmap_local(kaddr);
 		cur_bit += nbits;
 	} while (cur_bit <= last_bit);
+
+	if (unlikely(cur_bit <= last_bit)) {
+		bitmap->set_ahead_length =
+			((last_bit - cur_bit + 1) << bitmap->mapped.pgshift);
+	}
 }
 EXPORT_SYMBOL_NS_GPL(iova_bitmap_set, IOMMUFD);
diff --git a/drivers/iommu/iommufd/selftest.c b/drivers/iommu/iommufd/selftest.c
index d9e9920c7eba..8abf9747773e 100644
--- a/drivers/iommu/iommufd/selftest.c
+++ b/drivers/iommu/iommufd/selftest.c
@@ -41,6 +41,7 @@ static atomic_t mock_dev_num;
 enum {
 	MOCK_DIRTY_TRACK = 1,
 	MOCK_IO_PAGE_SIZE = PAGE_SIZE / 2,
+	MOCK_HUGE_PAGE_SIZE = 512 * MOCK_IO_PAGE_SIZE,
 
 	/*
 	 * Like a real page table alignment requires the low bits of the address
@@ -53,6 +54,7 @@ enum {
 	MOCK_PFN_START_IOVA = _MOCK_PFN_START,
 	MOCK_PFN_LAST_IOVA = _MOCK_PFN_START,
 	MOCK_PFN_DIRTY_IOVA = _MOCK_PFN_START << 1,
+	MOCK_PFN_HUGE_IOVA = _MOCK_PFN_START << 2,
 };
 
 /*
@@ -191,6 +193,34 @@ static int mock_domain_set_dirty_tracking(struct iommu_domain *domain,
 	return 0;
 }
 
+static bool mock_test_and_clear_dirty(struct mock_iommu_domain *mock,
+				      unsigned long iova, size_t page_size,
+				      unsigned long flags)
+{
+	unsigned long cur, end = iova + page_size - 1;
+	bool dirty = false;
+	void *ent, *old;
+
+	for (cur = iova; cur < end; cur += MOCK_IO_PAGE_SIZE) {
+		ent = xa_load(&mock->pfns, cur / MOCK_IO_PAGE_SIZE);
+		if (!ent || !(xa_to_value(ent) & MOCK_PFN_DIRTY_IOVA))
+			continue;
+
+		dirty = true;
+		/* Clear dirty */
+		if (!(flags & IOMMU_DIRTY_NO_CLEAR)) {
+			unsigned long val;
+
+			val = xa_to_value(ent) & ~MOCK_PFN_DIRTY_IOVA;
+			old = xa_store(&mock->pfns, cur / MOCK_IO_PAGE_SIZE,
+				       xa_mk_value(val), GFP_KERNEL);
+			WARN_ON_ONCE(ent != old);
+		}
+	}
+
+	return dirty;
+}
+
 static int mock_domain_read_and_clear_dirty(struct iommu_domain *domain,
 					    unsigned long iova, size_t size,
 					    unsigned long flags,
@@ -198,31 +228,31 @@ static int mock_domain_read_and_clear_dirty(struct iommu_domain *domain,
 {
 	struct mock_iommu_domain *mock =
 		container_of(domain, struct mock_iommu_domain, domain);
-	unsigned long i, max = size / MOCK_IO_PAGE_SIZE;
-	void *ent, *old;
+	unsigned long end = iova + size;
+	void *ent;
 
 	if (!(mock->flags & MOCK_DIRTY_TRACK) && dirty->bitmap)
 		return -EINVAL;
 
-	for (i = 0; i < max; i++) {
-		unsigned long cur = iova + i * MOCK_IO_PAGE_SIZE;
+	do {
+		unsigned long pgsize = MOCK_IO_PAGE_SIZE;
+		unsigned long head;
 
-		ent = xa_load(&mock->pfns, cur / MOCK_IO_PAGE_SIZE);
-		if (ent && (xa_to_value(ent) & MOCK_PFN_DIRTY_IOVA)) {
-			/* Clear dirty */
-			if (!(flags & IOMMU_DIRTY_NO_CLEAR)) {
-				unsigned long val;
-
-				val = xa_to_value(ent) & ~MOCK_PFN_DIRTY_IOVA;
-				old = xa_store(&mock->pfns,
-					       cur / MOCK_IO_PAGE_SIZE,
-					       xa_mk_value(val), GFP_KERNEL);
-				WARN_ON_ONCE(ent != old);
-			}
-			iommu_dirty_bitmap_record(dirty, cur,
-						  MOCK_IO_PAGE_SIZE);
+		ent = xa_load(&mock->pfns, iova / MOCK_IO_PAGE_SIZE);
+		if (!ent) {
+			iova += pgsize;
+			continue;
 		}
-	}
+
+		if (xa_to_value(ent) & MOCK_PFN_HUGE_IOVA)
+			pgsize = MOCK_HUGE_PAGE_SIZE;
+		head = iova & ~(pgsize - 1);
+
+		/* Clear dirty */
+		if (mock_test_and_clear_dirty(mock, head, pgsize, flags))
+			iommu_dirty_bitmap_record(dirty, head, pgsize);
+		iova = head + pgsize;
+	} while (iova < end);
 
 	return 0;
 }
@@ -234,6 +264,7 @@ const struct iommu_dirty_ops dirty_ops = {
 
 static struct iommu_domain *mock_domain_alloc_paging(struct device *dev)
 {
+	struct mock_dev *mdev = container_of(dev, struct mock_dev, dev);
 	struct mock_iommu_domain *mock;
 
 	mock = kzalloc(sizeof(*mock), GFP_KERNEL);
@@ -242,6 +273,8 @@ static struct iommu_domain *mock_domain_alloc_paging(struct device *dev)
 	mock->domain.geometry.aperture_start = MOCK_APERTURE_START;
 	mock->domain.geometry.aperture_end = MOCK_APERTURE_LAST;
 	mock->domain.pgsize_bitmap = MOCK_IO_PAGE_SIZE;
+	if (dev && mdev->flags & MOCK_FLAGS_DEVICE_HUGE_IOVA)
+		mock->domain.pgsize_bitmap |= MOCK_HUGE_PAGE_SIZE;
 	mock->domain.ops = mock_ops.default_domain_ops;
 	mock->domain.type = IOMMU_DOMAIN_UNMANAGED;
 	xa_init(&mock->pfns);
@@ -287,7 +320,7 @@ mock_domain_alloc_user(struct device *dev, u32 flags,
 			return ERR_PTR(-EOPNOTSUPP);
 		if (user_data || (has_dirty_flag && no_dirty_ops))
 			return ERR_PTR(-EOPNOTSUPP);
-		domain = mock_domain_alloc_paging(NULL);
+		domain = mock_domain_alloc_paging(dev);
 		if (!domain)
 			return ERR_PTR(-ENOMEM);
 		if (has_dirty_flag)
@@ -350,6 +383,9 @@ static int mock_domain_map_pages(struct iommu_domain *domain,
 
 			if (pgcount == 1 && cur + MOCK_IO_PAGE_SIZE == pgsize)
 				flags = MOCK_PFN_LAST_IOVA;
+			if (pgsize != MOCK_IO_PAGE_SIZE) {
+				flags |= MOCK_PFN_HUGE_IOVA;
+			}
 			old = xa_store(&mock->pfns, iova / MOCK_IO_PAGE_SIZE,
 				       xa_mk_value((paddr / MOCK_IO_PAGE_SIZE) |
 						   flags),
@@ -604,7 +640,8 @@ static struct mock_dev *mock_dev_create(unsigned long dev_flags)
 	struct mock_dev *mdev;
 	int rc;
 
-	if (dev_flags & ~(MOCK_FLAGS_DEVICE_NO_DIRTY))
+	if (dev_flags &
+	    ~(MOCK_FLAGS_DEVICE_NO_DIRTY | MOCK_FLAGS_DEVICE_HUGE_IOVA))
 		return ERR_PTR(-EINVAL);
 
 	mdev = kzalloc(sizeof(*mdev), GFP_KERNEL);
diff --git a/tools/testing/selftests/iommu/config b/tools/testing/selftests/iommu/config
index 6c4f901d6fed..110d73917615 100644
--- a/tools/testing/selftests/iommu/config
+++ b/tools/testing/selftests/iommu/config
@@ -1,2 +1,3 @@
-CONFIG_IOMMUFD
-CONFIG_IOMMUFD_TEST
+CONFIG_IOMMUFD=y
+CONFIG_FAULT_INJECTION=y
+CONFIG_IOMMUFD_TEST=y
diff --git a/tools/testing/selftests/iommu/iommufd.c b/tools/testing/selftests/iommu/iommufd.c
index 1a881e7a21d1..edf1c99c9936 100644
--- a/tools/testing/selftests/iommu/iommufd.c
+++ b/tools/testing/selftests/iommu/iommufd.c
@@ -12,6 +12,7 @@
 static unsigned long HUGEPAGE_SIZE;
 
 #define MOCK_PAGE_SIZE (PAGE_SIZE / 2)
+#define MOCK_HUGE_PAGE_SIZE (512 * MOCK_PAGE_SIZE)
 
 static unsigned long get_huge_page_size(void)
 {
@@ -1716,10 +1717,12 @@ FIXTURE(iommufd_dirty_tracking)
 FIXTURE_VARIANT(iommufd_dirty_tracking)
 {
 	unsigned long buffer_size;
+	bool hugepages;
 };
 
 FIXTURE_SETUP(iommufd_dirty_tracking)
 {
+	int mmap_flags;
 	void *vrc;
 	int rc;
 
@@ -1732,25 +1735,41 @@ FIXTURE_SETUP(iommufd_dirty_tracking)
 			   variant->buffer_size, rc);
 	}
 
+	mmap_flags = MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED;
+	if (variant->hugepages) {
+		/*
+		 * MAP_POPULATE will cause the kernel to fail mmap if THPs are
+		 * not available.
+		 */
+		mmap_flags |= MAP_HUGETLB | MAP_POPULATE;
+	}
 	assert((uintptr_t)self->buffer % HUGEPAGE_SIZE == 0);
 	vrc = mmap(self->buffer, variant->buffer_size, PROT_READ | PROT_WRITE,
-		   MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
+		   mmap_flags, -1, 0);
 	assert(vrc == self->buffer);
 
 	self->page_size = MOCK_PAGE_SIZE;
 	self->bitmap_size =
 		variant->buffer_size / self->page_size / BITS_PER_BYTE;
 
-	/* Provision with an extra (MOCK_PAGE_SIZE) for the unaligned case */
+	/* Provision with an extra (PAGE_SIZE) for the unaligned case */
 	rc = posix_memalign(&self->bitmap, PAGE_SIZE,
-			    self->bitmap_size + MOCK_PAGE_SIZE);
+			    self->bitmap_size + PAGE_SIZE);
 	assert(!rc);
 	assert(self->bitmap);
 	assert((uintptr_t)self->bitmap % PAGE_SIZE == 0);
 
 	test_ioctl_ioas_alloc(&self->ioas_id);
-	test_cmd_mock_domain(self->ioas_id, &self->stdev_id, &self->hwpt_id,
-			     &self->idev_id);
+	/* Enable 1M mock IOMMU hugepages */
+	if (variant->hugepages) {
+		test_cmd_mock_domain_flags(self->ioas_id,
+					   MOCK_FLAGS_DEVICE_HUGE_IOVA,
+					   &self->stdev_id, &self->hwpt_id,
+					   &self->idev_id);
+	} else {
+		test_cmd_mock_domain(self->ioas_id, &self->stdev_id,
+				     &self->hwpt_id, &self->idev_id);
+	}
 }
 
 FIXTURE_TEARDOWN(iommufd_dirty_tracking)
@@ -1784,12 +1803,26 @@ FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty128M)
 	.buffer_size = 128UL * 1024UL * 1024UL,
 };
 
+FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty128M_huge)
+{
+	/* 4K bitmap (128M IOVA range) */
+	.buffer_size = 128UL * 1024UL * 1024UL,
+	.hugepages = true,
+};
+
 FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty256M)
 {
 	/* 8K bitmap (256M IOVA range) */
 	.buffer_size = 256UL * 1024UL * 1024UL,
 };
 
+FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty256M_huge)
+{
+	/* 8K bitmap (256M IOVA range) */
+	.buffer_size = 256UL * 1024UL * 1024UL,
+	.hugepages = true,
+};
+
 TEST_F(iommufd_dirty_tracking, enforce_dirty)
 {
 	uint32_t ioas_id, stddev_id, idev_id;
@@ -1849,65 +1882,80 @@ TEST_F(iommufd_dirty_tracking, device_dirty_capability)
 
 TEST_F(iommufd_dirty_tracking, get_dirty_bitmap)
 {
-	uint32_t stddev_id;
+	uint32_t page_size = MOCK_PAGE_SIZE;
 	uint32_t hwpt_id;
 	uint32_t ioas_id;
 
+	if (variant->hugepages)
+		page_size = MOCK_HUGE_PAGE_SIZE;
+
 	test_ioctl_ioas_alloc(&ioas_id);
 	test_ioctl_ioas_map_fixed_id(ioas_id, self->buffer,
 				     variant->buffer_size, MOCK_APERTURE_START);
 
 	test_cmd_hwpt_alloc(self->idev_id, ioas_id,
 			    IOMMU_HWPT_ALLOC_DIRTY_TRACKING, &hwpt_id);
-	test_cmd_mock_domain(hwpt_id, &stddev_id, NULL, NULL);
 
 	test_cmd_set_dirty_tracking(hwpt_id, true);
 
 	test_mock_dirty_bitmaps(hwpt_id, variant->buffer_size,
-				MOCK_APERTURE_START, self->page_size,
+				MOCK_APERTURE_START, self->page_size, page_size,
 				self->bitmap, self->bitmap_size, 0, _metadata);
 
 	/* PAGE_SIZE unaligned bitmap */
 	test_mock_dirty_bitmaps(hwpt_id, variant->buffer_size,
-				MOCK_APERTURE_START, self->page_size,
+				MOCK_APERTURE_START, self->page_size, page_size,
 				self->bitmap + MOCK_PAGE_SIZE,
 				self->bitmap_size, 0, _metadata);
 
-	test_ioctl_destroy(stddev_id);
+	/* u64 unaligned bitmap */
+	test_mock_dirty_bitmaps(hwpt_id, variant->buffer_size,
+				MOCK_APERTURE_START, self->page_size, page_size,
+				self->bitmap + 0xff1, self->bitmap_size, 0,
+				_metadata);
+
 	test_ioctl_destroy(hwpt_id);
 }
 
 TEST_F(iommufd_dirty_tracking, get_dirty_bitmap_no_clear)
 {
-	uint32_t stddev_id;
+	uint32_t page_size = MOCK_PAGE_SIZE;
 	uint32_t hwpt_id;
 	uint32_t ioas_id;
 
+	if (variant->hugepages)
+		page_size = MOCK_HUGE_PAGE_SIZE;
+
 	test_ioctl_ioas_alloc(&ioas_id);
 	test_ioctl_ioas_map_fixed_id(ioas_id, self->buffer,
 				     variant->buffer_size, MOCK_APERTURE_START);
 
 	test_cmd_hwpt_alloc(self->idev_id, ioas_id,
 			    IOMMU_HWPT_ALLOC_DIRTY_TRACKING, &hwpt_id);
-	test_cmd_mock_domain(hwpt_id, &stddev_id, NULL, NULL);
 
 	test_cmd_set_dirty_tracking(hwpt_id, true);
 
 	test_mock_dirty_bitmaps(hwpt_id, variant->buffer_size,
-				MOCK_APERTURE_START, self->page_size,
+				MOCK_APERTURE_START, self->page_size, page_size,
 				self->bitmap, self->bitmap_size,
 				IOMMU_HWPT_GET_DIRTY_BITMAP_NO_CLEAR,
 				_metadata);
 
 	/* Unaligned bitmap */
 	test_mock_dirty_bitmaps(hwpt_id, variant->buffer_size,
-				MOCK_APERTURE_START, self->page_size,
+				MOCK_APERTURE_START, self->page_size, page_size,
 				self->bitmap + MOCK_PAGE_SIZE,
 				self->bitmap_size,
 				IOMMU_HWPT_GET_DIRTY_BITMAP_NO_CLEAR,
 				_metadata);
 
-	test_ioctl_destroy(stddev_id);
+	/* u64 unaligned bitmap */
+	test_mock_dirty_bitmaps(hwpt_id, variant->buffer_size,
+				MOCK_APERTURE_START, self->page_size, page_size,
+				self->bitmap + 0xff1, self->bitmap_size,
+				IOMMU_HWPT_GET_DIRTY_BITMAP_NO_CLEAR,
+				_metadata);
+
 	test_ioctl_destroy(hwpt_id);
 }
 
diff --git a/tools/testing/selftests/iommu/iommufd_utils.h b/tools/testing/selftests/iommu/iommufd_utils.h
index c646264aa41f..8d2b46b2114d 100644
--- a/tools/testing/selftests/iommu/iommufd_utils.h
+++ b/tools/testing/selftests/iommu/iommufd_utils.h
@@ -344,16 +344,19 @@ static int _test_cmd_mock_domain_set_dirty(int fd, __u32 hwpt_id, size_t length,
 						  page_size, bitmap, nr))
 
 static int _test_mock_dirty_bitmaps(int fd, __u32 hwpt_id, size_t length,
-				    __u64 iova, size_t page_size, __u64 *bitmap,
+				    __u64 iova, size_t page_size,
+				    size_t pte_page_size, __u64 *bitmap,
 				    __u64 bitmap_size, __u32 flags,
 				    struct __test_metadata *_metadata)
 {
-	unsigned long i, nbits = bitmap_size * BITS_PER_BYTE;
-	unsigned long nr = nbits / 2;
+	unsigned long npte = pte_page_size / page_size, pteset = 2 * npte;
+	unsigned long nbits = bitmap_size * BITS_PER_BYTE;
+	unsigned long j, i, nr = nbits / pteset ?: 1;
 	__u64 out_dirty = 0;
 
 	/* Mark all even bits as dirty in the mock domain */
-	for (i = 0; i < nbits; i += 2)
+	memset(bitmap, 0, bitmap_size);
+	for (i = 0; i < nbits; i += pteset)
 		set_bit(i, (unsigned long *)bitmap);
 
 	test_cmd_mock_domain_set_dirty(fd, hwpt_id, length, iova, page_size,
@@ -365,8 +368,12 @@ static int _test_mock_dirty_bitmaps(int fd, __u32 hwpt_id, size_t length,
 	test_cmd_get_dirty_bitmap(fd, hwpt_id, length, iova, page_size, bitmap,
 				  flags);
 	/* Beware ASSERT_EQ() is two statements -- braces are not redundant! */
-	for (i = 0; i < nbits; i++) {
-		ASSERT_EQ(!(i % 2), test_bit(i, (unsigned long *)bitmap));
+	for (i = 0; i < nbits; i += pteset) {
+		for (j = 0; j < pteset; j++) {
+			ASSERT_EQ(j < npte,
+				  test_bit(i + j, (unsigned long *)bitmap));
+		}
+		ASSERT_EQ(!(i % pteset), test_bit(i, (unsigned long *)bitmap));
 	}
 
 	memset(bitmap, 0, bitmap_size);
@@ -374,19 +381,23 @@ static int _test_mock_dirty_bitmaps(int fd, __u32 hwpt_id, size_t length,
 				  flags);
 
 	/* It as read already -- expect all zeroes */
-	for (i = 0; i < nbits; i++) {
-		ASSERT_EQ(!(i % 2) && (flags &
-				       IOMMU_HWPT_GET_DIRTY_BITMAP_NO_CLEAR),
-			  test_bit(i, (unsigned long *)bitmap));
+	for (i = 0; i < nbits; i += pteset) {
+		for (j = 0; j < pteset; j++) {
+			ASSERT_EQ(
+				(j < npte) &&
+					(flags &
+					 IOMMU_HWPT_GET_DIRTY_BITMAP_NO_CLEAR),
+				test_bit(i + j, (unsigned long *)bitmap));
+		}
 	}
 
 	return 0;
 }
-#define test_mock_dirty_bitmaps(hwpt_id, length, iova, page_size, bitmap,      \
-				bitmap_size, flags, _metadata)                 \
+#define test_mock_dirty_bitmaps(hwpt_id, length, iova, page_size, pte_size,\
+				bitmap, bitmap_size, flags, _metadata)     \
 	ASSERT_EQ(0, _test_mock_dirty_bitmaps(self->fd, hwpt_id, length, iova, \
-					      page_size, bitmap, bitmap_size,  \
-					      flags, _metadata))
+					      page_size, pte_size, bitmap,     \
+					      bitmap_size, flags, _metadata))
 
 static int _test_cmd_create_access(int fd, unsigned int ioas_id,
 				   __u32 *access_id, unsigned int flags)