4.5-stable review patch. If anyone has any objections, please let me know.
------------------ From: Miklos Szeredi <[email protected]> commit 9dd78d8c9a7bd4bc341f5864db32d4331b8eae4c upstream. In f_op->open() lock on parent is not held, so there's no guarantee that parent dentry won't go away at any time. Even after this patch there's no guarantee that 'dir' will stay the parent of 'inode', but at least it won't be freed while being used. Fixes: ff978b09f973 ("ext4 crypto: move context consistency check to ext4_file_open()") Signed-off-by: Miklos Szeredi <[email protected]> Signed-off-by: Theodore Ts'o <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]> --- fs/ext4/file.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) --- a/fs/ext4/file.c +++ b/fs/ext4/file.c @@ -335,7 +335,7 @@ static int ext4_file_open(struct inode * struct super_block *sb = inode->i_sb; struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); struct vfsmount *mnt = filp->f_path.mnt; - struct inode *dir = filp->f_path.dentry->d_parent->d_inode; + struct dentry *dir; struct path path; char buf[64], *cp; int ret; @@ -379,14 +379,18 @@ static int ext4_file_open(struct inode * if (ext4_encryption_info(inode) == NULL) return -ENOKEY; } - if (ext4_encrypted_inode(dir) && - !ext4_is_child_context_consistent_with_parent(dir, inode)) { + + dir = dget_parent(filp->f_path.dentry); + if (ext4_encrypted_inode(d_inode(dir)) && + !ext4_is_child_context_consistent_with_parent(d_inode(dir), inode)) { ext4_warning(inode->i_sb, "Inconsistent encryption contexts: %lu/%lu\n", - (unsigned long) dir->i_ino, + (unsigned long) d_inode(dir)->i_ino, (unsigned long) inode->i_ino); + dput(dir); return -EPERM; } + dput(dir); /* * Set up the jbd2_inode if we are opening the inode for * writing and the journal is present
