vfs: rename: check backing inode being equal
authorMiklos Szeredi <mszeredi@redhat.com>
Tue, 10 May 2016 23:16:37 +0000 (01:16 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 19 May 2016 01:35:12 +0000 (18:35 -0700)
commit1c1c3e93daeec9989e50614e99697903d32942ed
tree8d320b0275f600a5dbb2dd636635953440b38354
parentad56dcb2447522f6a165cb9bccff379e96acca8d
vfs: rename: check backing inode being equal

commit 9409e22acdfc9153f88d9b1ed2bd2a5b34d2d3ca upstream.

If a file is renamed to a hardlink of itself POSIX specifies that rename(2)
should do nothing and return success.

This condition is checked in vfs_rename().  However it won't detect hard
links on overlayfs where these are given separate inodes on the overlayfs
layer.

Overlayfs itself detects this condition and returns success without doing
anything, but then vfs_rename() will proceed as if this was a successful
rename (detach_mounts(), d_move()).

The correct thing to do is to detect this condition before even calling
into overlayfs.  This patch does this by calling vfs_select_inode() to get
the underlying inodes.

Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/namei.c