Here's the current state of the unionmount patches.
Significant changes made include:
(1) Brought up to date with the current changes in the VFS (eg. RCU pathwalk,
mount/vfsmount struct split).
(2) Moved the copy-up for namei metadata syscalls that modify an inode
(eg. link(), utimes(), setxattr()) into pathwalk. These syscalls now
merely need to provide LOOKUP_COPY_UP to pathwalk and the copy up is done
there.
One thing I'm not certain of is that this (and the original patches)
would do the copy up, even if the caller didn't have permission to alter
the inode and the system call failed the permission check after the copy
up had taken place.
Truncate and open-write don't suffer that issue.
(3) Merged the in-development ext2 patches in and completed them - at least I
think I did, one of them didn't say what it did.
(4) Added some code to override the credentials around upper inode creation
to make sure the inode gets the right UID/GID. This doesn't help if the
lower inode has some sort of foreign user identifier.
Also, I'm not sure whether the LSM xattrs should be blindly copied up.
Should the LSM policies applicable to the lower fs's apply to the upper
fs too?
(5) Added a marker flag for mounts on the lower fs. There is the possibility
of having a file mounted over a file on the upperfs. I suspect some of
the logic will malfunction in such a case as the previous component will
be seen to be on the upperfs and the current component not on a unioned
fs - and may trigger a copyup attempt.
(6) Added a patch to pass the mount flags to sget() and thence to the compare
routine. sget() installs them in the superblock before returning it.
(7) Added a patch to combine the multiple chown syscalls.
(8) Moved the xattr copyup code earlier so that it can be used for directories
too (it is called there, but isn't introduced till a later patch).
(9) Added a patch to have a second lock class for use by unionmounts for
i_mutex and i_dir_mutex to stop complaints when a union is made of two
filesystems of the same type.
Unionmount over a unionmount will need special handling, and possibly
rejecting. It should just work, however if the two upper filesystems are
the same type, lockdep will incorrectly moan a lot.
Some issues:
(1) Need to handle automount points and managed directories. Probably simply
ignoring them is best.
(2) Need to better handle mountpoints. Currently it just calls
follow_mount(), which is probably wrong.
(3) do_lookup() needs to come up with the correct inode after
needs_lookup_union() is called. I think I have this right, but it could
do with checking.
(4) Should d_revalidate() be called on the lower fs objects under some
circumstances. I assume not, since we don't want to see the lower fs
changing.
David
---
David Howells (21):
fallthru: ext2 support for lookup of d_type/d_ino in fallthrus
ext2: Add whiteout and opaque directory support
ext2: Remove target inode pointer from ext2_add_entry()
union-mount: Implement union-aware truncate()
union-mount: Implement union-aware rename()
union-mount: Make various syscalls aware (link, chmod, chown, utimes & setxattr)
unionmount: Override creds when copying up a file to correctly set ownership
unionmount: Add LOOKUP_COPY_UP
union-mount: In-kernel file copyup routines
union-mount: Add wrapper for lookup_union_locked() and RCU hook
union-mount: Implement union mount
union-mount: Duplicate the i_{,dir_}mutex lock classes and use for upper layer
unionmount: Mark lower layers in union
union-mount: Add union_create_topmost_dir()
whiteout: Add vfs_whiteout() and whiteout inode operation
VFS: Split inode_permission()
VFS: Pass mount flags to sget()
VFS: Make lookup_hash() return a struct path
VFS: Comment mount following code
VFS: Make clone_mnt()/copy_tree()/collect_mounts() return errors
VFS: Make chown() and lchown() call fchownat()
Felix Fietkau (2):
jffs2: Add fallthru support
jffs2: Add whiteout support
Jan Blunck (6):
union-mount: Create IS_MNT_UNION()
union-mount: Free union stack on removal of topmost dentry from dcache
union-mount: Introduce MNT_UNION and MS_UNION flags
tmpfs: Add whiteout support
whiteout: Allow removal of a directory with whiteouts
whiteout/NFSD: Don't return information about whiteouts to userspace
Valerie Aurora (44):
fallthru: jffs2 support for lookup of d_type/d_ino in fallthrus
ext2: Add fallthru support
ext2: Split ext2_add_entry() from ext2_add_link()
ext2: Add ext2_dirent_in_use()
union-mount: Implement union-aware writable open()
union-mount: Implement union-aware access()/faccessat()
VFS: Create user_path_nd() to lookup both parent and target
fallthru: tmpfs support for lookup of d_type/d_ino in fallthrus
union-mount: Add generic_readdir_fallthru() helper
union-mount: Copy up directory entries on first readdir()
union-mount: Set opaque flag on new directories in unioned file systems
union-mount: Create whiteout on rmdir()
union-mount: Create whiteout on unlink()
union-mount: Call union lookup functions in lookup path
union-mount: Add lookup_union_locked()
union-mount: Follow mount in __lookup_union()
union-mount: Build union stack in __lookup_union()
union-mount: Return files found in lower layers in __lookup_union()
union-mount: Process negative dentries in __lookup_union()
union-mount: Basic infrastructure of __lookup_union()
union-mount: Temporarily disable some syscalls
union-mount: Prevent bind mounts of union mounts
union-mount: Prevent topmost file system from being mounted elsewhere
union-mount: Prevent improper union-related remounts
union-mount: Create prepare_mnt_union() and cleanup_mnt_union()
union-mount: Create build_root_union()
union-mount: Add clone_union_tree() and put_union_sb()
union-mount: Create check_topmost_union_mnt()
union-mount: Create needs_lookup_union()
union-mount: Create union_add_dir()
union-mount: Create d_free_unions()
union-mount: Add union_find_dir()
union-mount: Add union_alloc()
union-mount: Add two superblock fields for union mounts
union-mount: Create union_stack structure
union-mount: Add CONFIG_UNION_MOUNT option
union-mount: Union mounts documentation
tmpfs: Add fallthru support
VFS: Basic fallthru definitions
whiteout: Define flags and operations for opaque inodes
VFS: Add CL_MAKE_HARD_READONLY flag to clone_mnt()/copy_tree()
VFS: Add CL_NO_SLAVE flag to clone_mnt()/copy_tree()
VFS: Add CL_NO_SHARED flag to clone_mnt()/copy_tree()
VFS: Add hard read-only users count to superblock
Documentation/filesystems/union-mounts.txt | 712 ++++++++++++++++++++++
Documentation/filesystems/vfs.txt | 16
drivers/mtd/mtdsuper.c | 4
fs/9p/vfs_super.c | 4
fs/Kconfig | 12
fs/Makefile | 1
fs/afs/super.c | 3
fs/btrfs/super.c | 4
fs/ceph/super.c | 2
fs/cifs/cifsfs.c | 9
fs/compat.c | 9
fs/dcache.c | 28 +
fs/devpts/inode.c | 6
fs/ecryptfs/main.c | 3
fs/ext2/dir.c | 181 +++++-
fs/ext2/ext2.h | 3
fs/ext2/inode.c | 11
fs/ext2/namei.c | 73 ++
fs/ext2/super.c | 6
fs/gfs2/ops_fstype.c | 5
fs/inode.c | 48 +
fs/internal.h | 5
fs/jffs2/dir.c | 117 +++-
fs/jffs2/fs.c | 4
fs/jffs2/super.c | 2
fs/libfs.c | 25 +
fs/logfs/super.c | 3
fs/namei.c | 906 +++++++++++++++++++++++++---
fs/namespace.c | 391 ++++++++++--
fs/nfs/super.c | 10
fs/nfsd/nfs3xdr.c | 5
fs/nfsd/nfs4xdr.c | 5
fs/nfsd/nfsxdr.c | 4
fs/nilfs2/super.c | 4
fs/open.c | 131 +++-
fs/pnode.c | 5
fs/pnode.h | 4
fs/proc/root.c | 3
fs/proc_namespace.c | 1
fs/readdir.c | 18 +
fs/reiserfs/procfs.c | 2
fs/super.c | 42 +
fs/sysfs/mount.c | 3
fs/ubifs/super.c | 3
fs/union.c | 721 ++++++++++++++++++++++
fs/union.h | 189 ++++++
fs/utimes.c | 2
fs/xattr.c | 10
include/linux/dcache.h | 40 +
include/linux/ext2_fs.h | 8
include/linux/fs.h | 47 +
include/linux/jffs2.h | 8
include/linux/mount.h | 5
include/linux/namei.h | 3
kernel/audit_tree.c | 10
kernel/cgroup.c | 2
mm/shmem.c | 192 ++++++
57 files changed, 3750 insertions(+), 320 deletions(-)
create mode 100644 Documentation/filesystems/union-mounts.txt
create mode 100644 fs/union.c
create mode 100644 fs/union.h
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
