fuse: fix bad inode
authorMiklos Szeredi <mszeredi@redhat.com>
Thu, 10 Dec 2020 14:33:14 +0000 (15:33 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 27 Jan 2022 07:47:40 +0000 (08:47 +0100)
commit3a2f8823aa565cc67bdd00c4cd5e1d8ad81e8436
treed53404cfce282b422a024efcb8e6988dee3f17b1
parentab5edcdd0e37f03d2d95f0daef145c700c2a890d
fuse: fix bad inode

commit 5d069dbe8aaf2a197142558b6fb2978189ba3454 upstream.

Jan Kara's analysis of the syzbot report (edited):

  The reproducer opens a directory on FUSE filesystem, it then attaches
  dnotify mark to the open directory.  After that a fuse_do_getattr() call
  finds that attributes returned by the server are inconsistent, and calls
  make_bad_inode() which, among other things does:

          inode->i_mode = S_IFREG;

  This then confuses dnotify which doesn't tear down its structures
  properly and eventually crashes.

Avoid calling make_bad_inode() on a live inode: switch to a private flag on
the fuse inode.  Also add the test to ops which the bad_inode_ops would
have caught.

This bug goes back to the initial merge of fuse in 2.6.14...

Reported-by: syzbot+f427adf9324b92652ccc@syzkaller.appspotmail.com
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Tested-by: Jan Kara <jack@suse.cz>
Cc: <stable@vger.kernel.org>
[bwh: Backported to 4.9:
 - Drop changes in fuse_dir_fsync(), fuse_readahead(), fuse_evict_inode()
 - In fuse_get_link(), return ERR_PTR(-EIO) for bad inodes
 - Convert some additional calls to is_bad_inode()
 - Adjust filename, context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/fuse/acl.c
fs/fuse/dir.c
fs/fuse/file.c
fs/fuse/fuse_i.h
fs/fuse/inode.c
fs/fuse/xattr.c