Skip to contentSkip to conversation

ufs: harden the mount path against malformed images

PATCH Applied Ali Ahmet Memis 3 participants 7 messages Aug 1 – Aug 25

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover ufs: harden the mount path against malformed images

Two robustness fixes for the ufs mount path, both reachable by mounting a crafted UFS image.

Patch 1 reorders ufs_fill_super() so the root dentry is created after the cylinder group metadata is loaded. Today a failure while reading the cylinder groups drops UFS_SB(sb) but leaves s_root installed, so the generic teardown oopses on a NULL pointer in ufs_sync_fs() and the put_super operation.

Patch 2 validates the cylinder group index and the rotor positions in ufs_read_cylinder() before they are cached. Unchecked, c_cgx indexes the cylinder summary array (a 32 bit write outside s_csp), and the rotors become bitmap scan offsets that can wrap an unsigned length and walk past the cylinder group buffers during an ordinary allocation.

Mounting a ufs image needs CAP_SYS_ADMIN and ufs is not unprivileged-mountable, so this is image-parsing hardening rather than a privilege boundary, handled in the open like the recent sibling work:

ufs: reject malformed cylinder summary geometry

    https://lore.kernel.org/all/[email protected]/
  ufs: reject oversized cylinder group metadata
    https://lore.kernel.org/all/[email protected]/

Those validate fs_cssize and fs_cgsize in ufs_fill_super(); the checks here live in ufs_read_cylinder() and do not overlap with them.

Both were reproduced under KASAN on an unpatched kernel by mounting a crafted UFS2 image.

Patch 1, an image whose first cylinder group fails the magic check, so ufs_read_cylinder_structures() fails after the root dentry is installed:

  BUG: KASAN: null-ptr-deref in mutex_lock+0x76/0xe0
   ufs_sync_fs+0x5b/0x6d0
  BUG: kernel NULL pointer dereference, address: 0000000000000100

Patch 2, an image whose on-disk cg_cgx is out of range, followed by a write that allocates a block:

  BUG: KASAN: slab-out-of-bounds in adjust_free_blocks+0x1bf0/0x2510
   adjust_free_blocks

With the patches applied, patch 1 loads the cylinder groups before publishing the root dentry and patch 2 rejects the mismatched cg_cgx in ufs_read_cylinder(), so neither faulting access is reached.

Ali Ahmet Memis (2):
ufs: create the root dentry after loading cylinder metadata
ufs: validate cylinder group metadata before caching it

 fs/ufs/cylinder.c | 10 ++++++++++
 fs/ufs/super.c    | 17 +++++++++--------
 2 files changed, 19 insertions(+), 8 deletions(-)
Signature · 1 line
2.54.0
#1 lore raw

PATCH 1/2 ufs: create the root dentry after loading cylinder metadata

ufs_fill_super() installed sb->s_root before it loaded the cylinder group structures for a writable mount:

	sb->s_root = d_make_root(inode);
	...
	if (!sb_rdonly(sb))
		if (!ufs_read_cylinder_structures(sb))
			goto failed;

When ufs_read_cylinder_structures() failed, the error path freed the in-core superblock information and set sb->s_fs_info to NULL while sb->s_root stayed installed. get_tree_bdev() then reached deactivate_locked_super(), and because s_root was present, generic_shutdown_super() called sync_filesystem() and the put_super operation. Both dereference UFS_SB(sb), which is now NULL, so a mount that fails only while reading the cylinder groups oopses during teardown. A crafted image whose first cylinder group cannot be read reaches this path.

Load the cylinder group metadata first and create the root dentry last, so the superblock is published to the VFS only once it is fully set up. ufs_setup_cstotal() and ufs_read_cylinder_structures() take only the super_block and do not use the root inode, so the reordering is safe.

Fixes 1da177e4c3f4 ("Linux-2.6.12-rc2")
---
 fs/ufs/super.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)
fs/ufs/super.c +9 -8
index c4831a8b9..7deecb395 100644
@@ -1199,6 +1199,15 @@ static int ufs_fill_super(struct super_block *sb, struct fs_context *fc)
 	sb->s_maxbytes = ufs_max_bytes(sb);
 	sb->s_max_links = UFS_LINK_MAX;
 
+	ufs_setup_cstotal(sb);
+	/*
+	 * Read cylinder group structures
+	 */
+	if (!sb_rdonly(sb))
+		if (!ufs_read_cylinder_structures(sb))
+			goto failed;
+
+	/* create the root dentry last, once UFS_SB(sb) is fully set up */
 	inode = ufs_iget(sb, UFS_ROOTINO);
 	if (IS_ERR(inode)) {
 		ret = PTR_ERR(inode);
@@ -1210,14 +1219,6 @@ static int ufs_fill_super(struct super_block *sb, struct fs_context *fc)
 		goto failed;
 	}
 
-	ufs_setup_cstotal(sb);
-	/*
-	 * Read cylinder group structures
-	 */
-	if (!sb_rdonly(sb))
-		if (!ufs_read_cylinder_structures(sb))
-			goto failed;
-
 	UFSD("EXIT\n");
 	return 0;
 
Signature · 1 line
2.54.0
reviewer Reviewed-by on v1 1/2
#2 lore raw

Re: [PATCH 1/2] ufs: create the root dentry after loading cylinder metadata

Quoted 19 lines · Ali Ahmet Memis

ufs_fill_super() installed sb->s_root before it loaded the cylinder group structures for a writable mount:

	sb->s_root = d_make_root(inode);
	...
	if (!sb_rdonly(sb))
		if (!ufs_read_cylinder_structures(sb))
			goto failed;

When ufs_read_cylinder_structures() failed, the error path freed the in-core superblock information and set sb->s_fs_info to NULL while sb->s_root stayed installed. get_tree_bdev() then reached deactivate_locked_super(), and because s_root was present, generic_shutdown_super() called sync_filesystem() and the put_super operation. Both dereference UFS_SB(sb), which is now NULL, so a mount that fails only while reading the cylinder groups oopses during teardown. A crafted image whose first cylinder group cannot be read reaches this path.

Quoted 8 lines

Load the cylinder group metadata first and create the root dentry last, so the superblock is published to the VFS only once it is fully set up. ufs_setup_cstotal() and ufs_read_cylinder_structures() take only the super_block and do not use the root inode, so the reordering is safe.

Fixes 1da177e4c3f4 ("Linux-2.6.12-rc2")

Looks good. Feel free to add:

Honza

Quoted 42 lines
---
 fs/ufs/super.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)
fs/ufs/super.c +9 -8
index c4831a8b9..7deecb395 100644
@@ -1199,6 +1199,15 @@ static int ufs_fill_super(struct super_block *sb, struct fs_context *fc)
 	sb->s_maxbytes = ufs_max_bytes(sb);
 	sb->s_max_links = UFS_LINK_MAX;
 
+	ufs_setup_cstotal(sb);
+	/*
+	 * Read cylinder group structures
+	 */
+	if (!sb_rdonly(sb))
+		if (!ufs_read_cylinder_structures(sb))
+			goto failed;
+
+	/* create the root dentry last, once UFS_SB(sb) is fully set up */
 	inode = ufs_iget(sb, UFS_ROOTINO);
 	if (IS_ERR(inode)) {
 		ret = PTR_ERR(inode);
@@ -1210,14 +1219,6 @@ static int ufs_fill_super(struct super_block *sb, struct fs_context *fc)
 		goto failed;
 	}
 
-	ufs_setup_cstotal(sb);
-	/*
-	 * Read cylinder group structures
-	 */
-	if (!sb_rdonly(sb))
-		if (!ufs_read_cylinder_structures(sb))
-			goto failed;
-
 	UFSD("EXIT\n");
 	return 0;
 
Signature · 1 line
2.54.0
Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
#3 lore raw

PATCH 2/2 ufs: validate cylinder group metadata before caching it

In reply to Ali Ahmet Memis

ufs_read_cylinder() copies the cylinder group index and the rotor positions straight from the on-disk group and caches them without any check:

	ucpi->c_cgx    = fs32_to_cpu(sb, ucg->cg_cgx);
	ucpi->c_rotor  = fs32_to_cpu(sb, ucg->cg_rotor);
	ucpi->c_frotor = fs32_to_cpu(sb, ucg->cg_frotor);
	ucpi->c_irotor = fs32_to_cpu(sb, ucg->cg_irotor);

They are then used as indices during allocation and free:

- c_cgx indexes the cylinder summary array as UFS_SB(sb)->fs_cs(ucpi->c_cgx), so a value past s_ncg writes a 32 bit count outside the s_csp allocation.

- c_frotor becomes a bitmap scan start, start = c_frotor >> 3, and then length = ((s_fpg + 7) >> 3) - start. A start beyond the block bitmap wraps the unsigned length to a huge value, so ubh_scanc() walks far past the cylinder group buffers. c_irotor drives the inode bitmap the same way.

A crafted image can set any of these freely, turning an ordinary allocation into an out of bounds access.

Reject a cylinder group whose recorded index does not match the group being read, or whose rotors fall outside the group, before the metadata is cached. Valid filesystems keep cg_cgx equal to the group number and the rotors within the group, so only malformed images are rejected.

Fixes 1da177e4c3f4 ("Linux-2.6.12-rc2")
---
 fs/ufs/cylinder.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
fs/ufs/cylinder.c +10 -0
index a2813270c..b930ee1cf 100644
@@ -68,6 +68,16 @@ static bool ufs_read_cylinder(struct super_block *sb,
 	ucpi->c_clustersumoff = fs32_to_cpu(sb, ucg->cg_u.cg_44.cg_clustersumoff);
 	ucpi->c_clusteroff = fs32_to_cpu(sb, ucg->cg_u.cg_44.cg_clusteroff);
 	ucpi->c_nclusterblks = fs32_to_cpu(sb, ucg->cg_u.cg_44.cg_nclusterblks);
+
+	/* these on-disk values become array and bitmap indices */
+	if (ucpi->c_cgx != cgno ||
+	    ucpi->c_rotor >= uspi->s_fpg ||
+	    ucpi->c_frotor >= uspi->s_fpg ||
+	    ucpi->c_irotor >= uspi->s_ipg) {
+		ufs_error(sb, __func__,
+			  "inconsistent metadata in cylinder group %u\n", cgno);
+		goto failed;
+	}
 	UFSD("EXIT\n");
 	return true;
 	
Signature · 1 line
2.54.0
reviewer Reviewed-by on v1 2/2
#4 lore raw

Re: [PATCH 2/2] ufs: validate cylinder group metadata before caching it

Quoted 24 lines · Ali Ahmet Memis

ufs_read_cylinder() copies the cylinder group index and the rotor positions straight from the on-disk group and caches them without any check:

	ucpi->c_cgx    = fs32_to_cpu(sb, ucg->cg_cgx);
	ucpi->c_rotor  = fs32_to_cpu(sb, ucg->cg_rotor);
	ucpi->c_frotor = fs32_to_cpu(sb, ucg->cg_frotor);
	ucpi->c_irotor = fs32_to_cpu(sb, ucg->cg_irotor);

They are then used as indices during allocation and free:

- c_cgx indexes the cylinder summary array as UFS_SB(sb)->fs_cs(ucpi->c_cgx), so a value past s_ncg writes a 32 bit count outside the s_csp allocation.

- c_frotor becomes a bitmap scan start, start = c_frotor >> 3, and then length = ((s_fpg + 7) >> 3) - start. A start beyond the block bitmap wraps the unsigned length to a huge value, so ubh_scanc() walks far past the cylinder group buffers. c_irotor drives the inode bitmap the same way.

A crafted image can set any of these freely, turning an ordinary allocation into an out of bounds access.

Quoted 8 lines

Reject a cylinder group whose recorded index does not match the group being read, or whose rotors fall outside the group, before the metadata is cached. Valid filesystems keep cg_cgx equal to the group number and the rotors within the group, so only malformed images are rejected.

Fixes 1da177e4c3f4 ("Linux-2.6.12-rc2")

Looks good. Feel free to add:

Honza

Quoted 28 lines
---
 fs/ufs/cylinder.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
fs/ufs/cylinder.c +10 -0
index a2813270c..b930ee1cf 100644
@@ -68,6 +68,16 @@ static bool ufs_read_cylinder(struct super_block *sb,
 	ucpi->c_clustersumoff = fs32_to_cpu(sb, ucg->cg_u.cg_44.cg_clustersumoff);
 	ucpi->c_clusteroff = fs32_to_cpu(sb, ucg->cg_u.cg_44.cg_clusteroff);
 	ucpi->c_nclusterblks = fs32_to_cpu(sb, ucg->cg_u.cg_44.cg_nclusterblks);
+
+	/* these on-disk values become array and bitmap indices */
+	if (ucpi->c_cgx != cgno ||
+	    ucpi->c_rotor >= uspi->s_fpg ||
+	    ucpi->c_frotor >= uspi->s_fpg ||
+	    ucpi->c_irotor >= uspi->s_ipg) {
+		ufs_error(sb, __func__,
+			  "inconsistent metadata in cylinder group %u\n", cgno);
+		goto failed;
+	}
 	UFSD("EXIT\n");
 	return true;
 
Signature · 1 line
2.54.0
Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
maintainer Applied
#5 lore raw
In reply to Ali Ahmet Memis
Quoted 10 lines · Ali Ahmet Memis

Two robustness fixes for the ufs mount path, both reachable by mounting a crafted UFS image.

Patch 1 reorders ufs_fill_super() so the root dentry is created after the cylinder group metadata is loaded. Today a failure while reading the cylinder groups drops UFS_SB(sb) but leaves s_root installed, so the generic teardown oopses on a NULL pointer in ufs_sync_fs() and the put_super operation.

[...]

In the future, please cut down and rewrite LLM generated cover letters and commit messages.

---

Applied to the vfs.fixes branch of the vfs/vfs.git tree. Patches in the vfs.fixes branch should appear in linux-next soon.

Please report any outstanding bugs that were missed during review in a new review to the original patch series allowing us to drop it.

It's encouraged to provide Acked-bys and Reviewed-bys even though the patch has now been applied. If possible patch trailers will be updated.

Note that commit hashes shown below are subject to change due to rebase, trailer updates or similar. If in doubt, please check the listed branch.

tree:   https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git
branch: vfs.fixes

[1/2] ufs: create the root dentry after loading cylinder metadata

      https://git.kernel.org/vfs/vfs/c/55a4c98abb96
[2/2] ufs: validate cylinder group metadata before caching it
      https://git.kernel.org/vfs/vfs/c/c9d263be2680
#6 lore raw
Quoted 2 lines · Christian Brauner

In the future, please cut down and rewrite LLM generated cover letters and commit messages.

Thanks for applying these. I wrote them myself (English isn't my first language, so I guess that can make them sound a bit artificial) but I'll keep the commit messages and cover letters shorter next time.