On Thu, Aug 22, 2024 at 05:04:14PM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <djwong@xxxxxxxxxx>
>
> Metadata inodes are private files and therefore cannot be exposed to
> userspace. This means no bulkstat, no open-by-handle, no linking them
> into the directory tree, and no feeding them to LSMs. As such, we mark
> them S_PRIVATE, which stops all that.
Can you merge this back up into the initial iget support code?
>
> While we're at it, put them in a separate lockdep class so that it won't
> get confused by "recursive" i_rwsem locking such as what happens when we
> write to a rt file and need to allocate from the rt bitmap file. The
> static function that we use to do this will be exported in the rtgroups
> patchset.
Stale commit message? There's nothing of the sort in this patch....
>
> Signed-off-by: Darrick J. Wong <djwong@xxxxxxxxxx>
> ---
> fs/xfs/scrub/tempfile.c | 8 ++++++++
> fs/xfs/xfs_iops.c | 15 ++++++++++++++-
> 2 files changed, 22 insertions(+), 1 deletion(-)
>
>
> diff --git a/fs/xfs/scrub/tempfile.c b/fs/xfs/scrub/tempfile.c
> index 177f922acfaf1..3c5a1d77fefae 100644
> --- a/fs/xfs/scrub/tempfile.c
> +++ b/fs/xfs/scrub/tempfile.c
> @@ -844,6 +844,14 @@ xrep_is_tempfile(
> const struct xfs_inode *ip)
> {
> const struct inode *inode = &ip->i_vnode;
> + struct xfs_mount *mp = ip->i_mount;
> +
> + /*
> + * Files in the metadata directory tree also have S_PRIVATE set and
> + * IOP_XATTR unset, so we must distinguish them separately.
> + */
> + if (xfs_has_metadir(mp) && (ip->i_diflags2 & XFS_DIFLAG2_METADATA))
> + return false;
Why do you need to check both xfs_has_metadir() and the inode flag
here? The latter should only be set if the former is set, yes?
If it's the other way around, then we have an on-disk corruption...
> if (IS_PRIVATE(inode) && !(inode->i_opflags & IOP_XATTR))
> return true;
> diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
> index 1cdc8034f54d9..c1686163299a0 100644
> --- a/fs/xfs/xfs_iops.c
> +++ b/fs/xfs/xfs_iops.c
> @@ -42,7 +42,9 @@
> * held. For regular files, the lock order is the other way around - the
> * mmap_lock is taken during the page fault, and then we lock the ilock to do
> * block mapping. Hence we need a different class for the directory ilock so
> - * that lockdep can tell them apart.
> + * that lockdep can tell them apart. Directories in the metadata directory
> + * tree get a separate class so that lockdep reports will warn us if someone
> + * ever tries to lock regular directories after locking metadata directories.
> */
> static struct lock_class_key xfs_nondir_ilock_class;
> static struct lock_class_key xfs_dir_ilock_class;
> @@ -1299,6 +1301,7 @@ xfs_setup_inode(
> {
> struct inode *inode = &ip->i_vnode;
> gfp_t gfp_mask;
> + bool is_meta = xfs_is_metadata_inode(ip);
>
> inode->i_ino = ip->i_ino;
> inode->i_state |= I_NEW;
> @@ -1310,6 +1313,16 @@ xfs_setup_inode(
> i_size_write(inode, ip->i_disk_size);
> xfs_diflags_to_iflags(ip, true);
>
> + /*
> + * Mark our metadata files as private so that LSMs and the ACL code
> + * don't try to add their own metadata or reason about these files,
> + * and users cannot ever obtain file handles to them.
> + */
> + if (is_meta) {
> + inode->i_flags |= S_PRIVATE;
> + inode->i_opflags &= ~IOP_XATTR;
> + }
No need for a temporary variable here.
-Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
