@@ -187,9 +187,21 @@ static bool btrfs_fscrypt_empty_dir(struct inode *inode)
return true;
}
+static void btrfs_fscrypt_get_iv(u8 *iv, int ivsize, struct inode *inode,
+ u64 lblk_num)
+{
+ /*
+ * For encryption that doesn't involve extent data, juse use the
+ * nonce already loaded into the iv buffer.
+ */
+ return;
+}
+
const struct fscrypt_operations btrfs_fscrypt_ops = {
+ .flags = FS_CFLG_ALLOW_PARTIAL,
.key_prefix = "btrfs:",
.get_context = btrfs_fscrypt_get_context,
.set_context = btrfs_fscrypt_set_context,
.empty_dir = btrfs_fscrypt_empty_dir,
+ .get_fs_defined_iv = btrfs_fscrypt_get_iv,
};
As btrfs cannot use the standard inode or logical block based encryption for data block encryption, it must provide a IV generation function and users must use the IV_FROM_FS policy. For filenames, we can just use the nonce that fscrypt stores per-inode, since these encrypted datum are not shared between inodes; later on, we will store an IV per file extent, and return it in this function for data encryption. Signed-off-by: Sweet Tea Dorminy <sweettea-kernel@dorminy.me> --- fs/btrfs/fscrypt.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)