Skip to contentSkip to conversation

ntfs: add check for mft record size in superblock

PATCH In review Rustam Kovhaev 2 participants 3 messages Aug 23, 2020 – Aug 24, 2020 NTFS filesystem

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ntfs: add check for mft record size in superblock

number of bytes allocated for mft record should be equal to the mft record size stored in ntfs superblock
as reported by syzbot, userspace might trigger out-of-bounds read by dereferencing ctx->attr in ntfs_attr_find()

---
 fs/ntfs/inode.c | 6 ++++++
 1 file changed, 6 insertions(+)
fs/ntfs/inode.c +6 -0
index 9bb9f0952b18..6407af7c2e4f 100644
@@ -1810,6 +1810,12 @@ int ntfs_read_inode_mount(struct inode *vi)
 		brelse(bh);
 	}
 
+	if (m->bytes_allocated != vol->mft_record_size) {
+		ntfs_error(sb, "Incorrect mft record size [%u] in superblock, should be [%u].",
+				m->bytes_allocated, vol->mft_record_size);
+		goto err_out;
+	}
+
 	/* Apply the mst fixups. */
 	if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
 		/* FIXME: Try to use the $MFTMirr now. */
Signature · 1 line
2.28.0
Acked-by on v1 1/1
#1 lore raw

Hi Rustam,

Thank you for the patch but it introduces an endianness bug - you have to us le32_to_cpu(m->bytes_allocated) both when doing the comparison and then printing the message.

Also, please drop the square brackets. Wherever the driver prints such things it never uses brackets around the numbers and it would be better to have this consistent throughout.

Can you please resend with the above issues addressed? You can then also add to the commit message:

Thanks!

Best regards,

Anton

Quoted 34 lines · Rustam Kovhaev

number of bytes allocated for mft record should be equal to the mft record size stored in ntfs superblock
as reported by syzbot, userspace might trigger out-of-bounds read by dereferencing ctx->attr in ntfs_attr_find()

---
fs/ntfs/inode.c | 6 ++++++

1 file changed, 6 insertions(+)
fs/ntfs/inode.c +6 -0
index 9bb9f0952b18..6407af7c2e4f 100644
@@ -1810,6 +1810,12 @@ int ntfs_read_inode_mount(struct inode *vi)
 		brelse(bh);
 	}
 
+	if (m->bytes_allocated != vol->mft_record_size) {
+		ntfs_error(sb, "Incorrect mft record size [%u] in superblock, should be [%u].",
+				m->bytes_allocated, vol->mft_record_size);
+		goto err_out;
+	}
+
 	/* Apply the mst fixups. */
 	if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
 		/* FIXME: Try to use the $MFTMirr now. */
Signature · 1 line
2.28.0
Signature · 3 lines
Anton Altaparmakov <anton at tuxera.com> (replace at with @)
Lead in File System Development, Tuxera Inc., http://www.tuxera.com/
Linux NTFS maintainer
#2 lore raw
Quoted 8 lines · Anton Altaparmakov

Hi Rustam,

Thank you for the patch but it introduces an endianness bug - you have to us le32_to_cpu(m->bytes_allocated) both when doing the comparison and then printing the message.

Also, please drop the square brackets. Wherever the driver prints such things it never uses brackets around the numbers and it would be better to have this consistent throughout.

Can you please resend with the above issues addressed? You can then also add to the commit message:

Quoted 8 lines

Thanks!

Best regards,

Anton

hi Anton,
thank you for the review, my bad, i'll get it fixed and i'll resend the patch