Thread (5 messages) 5 messages, 2 authors, 2017-02-06

Re: [PATCH] ext4: don't allow encrypted operations without keys

From: Eric Biggers <hidden>
Date: 2017-01-05 19:26:22
Also in: linux-fsdevel

Hi Ted,

On Wed, Dec 28, 2016 at 12:22:52AM -0500, Theodore Ts'o wrote:
quoted hunk ↗ jump to hunk
While we allow deletes without the key, the following should not be
permitted:

# cd /vdc/encrypted-dir-without-key
# ls -l
total 4
-rw-r--r-- 1 root root   0 Dec 27 22:35 6,LKNRJsp209FbXoSvJWzB
-rw-r--r-- 1 root root 286 Dec 27 22:35 uRJ5vJh9gE7vcomYMqTAyD
# mv uRJ5vJh9gE7vcomYMqTAyD  6,LKNRJsp209FbXoSvJWzB

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
---
 fs/ext4/namei.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index eadba919f26b..45a5ba558074 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -3525,6 +3525,12 @@ static int ext4_rename(struct inode *old_dir, struct dentry *old_dentry,
 			EXT4_I(old_dentry->d_inode)->i_projid)))
 		return -EXDEV;
 
+	if ((ext4_encrypted_inode(old_dir) &&
+	     !fscrypt_has_encryption_key(old_dir)) ||
+	    (ext4_encrypted_inode(new_dir) &&
+	     !fscrypt_has_encryption_key(new_dir)))
+		return -ENOKEY;
+
 	retval = dquot_initialize(old.dir);
 	if (retval)
 		return retval;
@@ -3725,6 +3731,12 @@ static int ext4_cross_rename(struct inode *old_dir, struct dentry *old_dentry,
 	int retval;
 	struct timespec ctime;
 
+	if ((ext4_encrypted_inode(old_dir) &&
+	     !fscrypt_has_encryption_key(old_dir)) ||
+	    (ext4_encrypted_inode(new_dir) &&
+	     !fscrypt_has_encryption_key(new_dir)))
+		return -ENOKEY;
+
 	if ((ext4_encrypted_inode(old_dir) ||
 	     ext4_encrypted_inode(new_dir)) &&
 	    (old_dir != new_dir) &&
I'm fine with this, with the understanding that it relies on ext4_lookup()
calling fscrypt_get_encryption_info() (via fscrypt_has_permitted_context()) when
looking up the directory.  I also suggest moving the fscrypt_permitted_context()
check in ext4_rename() up to be next to the new check, so that the fscrypt hooks
are grouped together and are consistent with ext4_cross_rename().

I can also write/update an xfstest to test this.

Something I'm thinking about is making things easier for filesystems by having
functions like "fscrypt_rename_hook()" which would handle all these needed
checks.  It would be easy to do with out-of-line functions in fs/crypto/, but we
don't want to be making ->is_encrypted() calls through the fscrypt_operations
all the time, when an inlined call to ext4_encrypted_inode() (or f2fs or
ubifs_encrypted_inode()) is much faster.  I think it could be implemented as
efficiently as now if the hooks were defined in a header and called a macro like
"fs_encrypted_inode()" which filesystems would have to #define first.  It would
be a little ugly, but at least it would be less error-prone than having multiple
filesystems replicate these increasingly complex checks.

Eric
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help