ntfs: avoid calling post_write_mst_fixup() for invalid index_block

ntfs_icx_ib_sync_write() calls post_write_mst_fixup() when ntfs_ib_write()
returns an error, intending to restore the buffer after a failed write.

However, ntfs_ib_write() returns an error immediately if
pre_write_mst_fixup() validation fails. The caller,
ntfs_icx_ib_sync_write(), interprets any error as a write failure
requiring rollback. It does not differentiate between I/O errors and
validation failures, and calls post_write_mst_fixup() anyway.

Since post_write_mst_fixup() assumes that the index_block contents is
correct, it doesn't perform the boundary checks, which results in
out-of-bounds memory access.

An attacker can craft a malicious NTFS image with:
  - large index_block.usa_ofs offset, pointing outside the ntfs_record
  - index_block.usa_count = 0, causing integer underflow
  - or index_block.usa_count larger than actual number of sectors in the
    ntfs_record, causing out-of-bounds access

KASAN reports describing the memory corruption:
  ==================================================================
  BUG: KASAN: slab-out-of-bounds in post_write_mst_fixup+0x19c/0x1d0
  Read of size 2 at addr ffff8881586c9018 by task p/9428
  Call Trace:
   <TASK>
   dump_stack_lvl+0x100/0x190
   print_report+0x139/0x4ad
   ? post_write_mst_fixup+0x19c/0x1d0
   ? __virt_addr_valid+0x262/0x500
   ? post_write_mst_fixup+0x19c/0x1d0
   kasan_report+0xe4/0x1d0
   ? post_write_mst_fixup+0x19c/0x1d0
   post_write_mst_fixup+0x19c/0x1d0
   ntfs_icx_ib_sync_write+0x179/0x220
   ntfs_inode_sync_filename+0x83d/0x1080
   __ntfs_write_inode+0x1049/0x1480
   ntfs_file_fsync+0x131/0x9b0
  ==================================================================
  BUG: KASAN: slab-out-of-bounds in post_write_mst_fixup+0x1aa/0x1d0
  Write of size 2 at addr ffff8881586c91fe by task p/9428
  Call Trace:
   <TASK>
   dump_stack_lvl+0x100/0x190
   print_report+0x139/0x4ad
   ? post_write_mst_fixup+0x1aa/0x1d0
   ? __virt_addr_valid+0x262/0x500
   ? post_write_mst_fixup+0x1aa/0x1d0
   kasan_report+0xe4/0x1d0
   ? post_write_mst_fixup+0x1aa/0x1d0
   post_write_mst_fixup+0x1aa/0x1d0
   ntfs_icx_ib_sync_write+0x179/0x220
   ntfs_inode_sync_filename+0x83d/0x1080
   __ntfs_write_inode+0x1049/0x1480
   ntfs_file_fsync+0x131/0x9b0
  ==================================================================

Let's move the post_write_mst_fixup() call to ntfs_ib_write().
The ntfs_ib_write() function calls pre_write_mst_fixup() at the beginning.
If the index_block contents is invalid, pre_write_mst_fixup() fails and
ntfs_ib_write() returns early without calling post_write_mst_fixup() on
bad index_block.

Fixes: 0a8ac0c1fa0b ("ntfs: update directory operations")
Cc: stable@vger.kernel.org
Signed-off-by: Valeriy Yashnikov <yashnikov.valeriy@gmail.com>
Reviewed-by: Hyunchul Lee <hyc.lee@gmail.com>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>

diff --git a/fs/ntfs/index.c b/fs/ntfs/index.c
index c5f2cf75b750e998fcc8ec0b339cd2dffbe8dd57..faa7ee920a3ad073a8c672d5c8c2523f7fe3850b 100644 (file)
--- a/fs/ntfs/index.c
+++ b/fs/ntfs/index.c
@@ -110,6 +110,10 @@ static int ntfs_ib_write(struct ntfs_index_context *icx, struct index_block *ib)
        ret = ntfs_inode_attr_pwrite(VFS_I(icx->ia_ni),
                        ntfs_ib_vcn_to_pos(icx, vcn), icx->block_size,
                        (u8 *)ib, icx->sync_write);
+
+       /* Perform data restoration before returning */
+       post_write_mst_fixup((struct ntfs_record *)ib);
+
        if (ret != icx->block_size) {
                ntfs_debug("Failed to write index block %lld, inode %llu",
                                vcn, (unsigned long long)icx->idx_ni->mft_no);
@@ -147,7 +151,6 @@ int ntfs_icx_ib_sync_write(struct ntfs_index_context *icx)
                icx->ib = NULL;
                icx->ib_dirty = false;
        } else {
-               post_write_mst_fixup((struct ntfs_record *)icx->ib);
                icx->sync_write = false;
        }