[PATCH v2] fs/ntfs3: reject an oversized resident attribute on the inline iomap path

[Michael Bommarito <michael.bommarito@gmail.com>]

attr_data_get_block_locked() maps a resident attribute as a one-page
IOMAP_INLINE extent: it allocates a single page, copies the resident
value into it with memcpy(), and hands that page to ntfs_iomap_begin()
as the inline data.  The copy length is the on-disk resident value
length (res.data_size), and mi_enum_attr() only bounds that length
against the MFT record size.  A corrupted volume with MFT records larger
than a page can therefore present a resident $DATA whose length exceeds
PAGE_SIZE, and the memcpy() then writes past the single destination page.

Impact: reading or writing a resident file on a mounted crafted NTFS
volume whose MFT records are larger than a page overflows the one-page
inline buffer in attr_data_get_block_locked() with attacker-controlled
bytes.

Reject such an attribute in attr_data_get_block_locked(), before the
page is allocated and the copy is made.

Fixes: 099ef9ab9203 ("fs/ntfs3: implement iomap-based file operations")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-8
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
---
Changes since v1:
- Reworded the code comment and changelog.  The oversized resident length
  overflows the single alloc_page() page that attr_data_get_block_locked()
  copies the resident value into; v1 described the older
  iomap_write_end_inline() BUG_ON(!iomap_inline_data_valid()), which was
  removed in 7.2-rc1.  Thanks to Mihai Brodschi for catching the stale
  reference.  The fix itself is unchanged.

The unbounded resident length dates to the iomap conversion; only the
manifestation changed when the inline buffer became a single alloc_page():
  - 7.2-rc: the memcpy() into the one-page buffer overflows the page
    (out-of-bounds write, below).
  - 7.0..7.1: the buffer was kmemdup()ed at the real size and the oversized
    length instead tripped BUG_ON(!iomap_inline_data_valid()) in
    iomap_write_end_inline() (a denial of service).
The same data_size > PAGE_SIZE check prevents both, so it is the correct fix
across the stable range.

Reproduced under QEMU + KASAN on 7.2-rc2.  A resident $DATA value length
above PAGE_SIZE, as a volume with MFT records larger than a page presents,
was supplied while the alloc_page() + memcpy() path ran unchanged; the stock
kernel faults with a KASAN out-of-bounds write in
attr_data_get_block_locked():

  BUG: KASAN: out-of-bounds in attr_data_get_block_locked
  Write of size <data_size> by task ...
   __asan_memcpy
   attr_data_get_block_locked
   attr_data_get_block
   ntfs_iomap_begin
   iomap_iter
   iomap_file_buffered_write

With the patch the same access returns -EINVAL and does not fault.  An
in-bounds resident file (control) and a non-resident file are unaffected on
both the stock and patched kernels.
 fs/ntfs3/attrib.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/fs/ntfs3/attrib.c b/fs/ntfs3/attrib.c
index c621a4c582f9e..2f73714a2db97 100644
--- a/fs/ntfs3/attrib.c
+++ b/fs/ntfs3/attrib.c
@@ -1034,6 +1034,21 @@ int attr_data_get_block_locked(struct ntfs_inode *ni, CLST vcn, CLST clen,
 
 	if (!attr_b->non_res) {
 		u32 data_size = le32_to_cpu(attr_b->res.data_size);
+
+		/*
+		 * A resident attribute is copied into a single page below
+		 * (alloc_page() + memcpy()) and mapped as a one-page
+		 * IOMAP_INLINE extent by ntfs_iomap_begin().  mi_enum_attr()
+		 * only bounds the resident value length against the MFT record
+		 * size, so a corrupted volume whose records are larger than a
+		 * page can report data_size > PAGE_SIZE; copying that many bytes
+		 * would overflow the single destination page.  Reject it.
+		 */
+		if (data_size > PAGE_SIZE) {
+			err = -EINVAL;
+			goto out;
+		}
+
 		*lcn = RESIDENT_LCN;
 		*len = data_size;
 		if (res && data_size) {
-- 
2.53.0