Template Version: @(#)sac_nextcase 1.66 04/17/08 SMI
This information is Copyright 2008 Sun Microsystems
1. Introduction
1.1. Project/Component Working Name:
lofi mount
1.2. Name of Document Author/Supplier:
Author: John Levon
1.3 Date of This Document:
29 April, 2008
4. Technical Description
1. Introduction
A long-standing RFE has been for direct support of loopback file
mounts in the mount(1m) utility. This case defines such a feature.
A patch binding is requested.
2. Discussion
On Solaris today, mounting a file system image is a two stage
process:
lofiadm -a /path/to/my.iso
mount -F hsfs /dev/lofi/1 /mnt/myiso
The same applies for the unmount process. Instead, it should be a
single operation:
mount -F hsfs /path/to/my.iso /mnt/myiso
Internally, this will create a lofi minor node for the file and
mount the filesystem from that node. Unmount will remove the lofi
node again.
When such a system is mounted, the user will see it listed in mount as
usual:
$ mount -v
...
/lomount/1 on /lomount/ufs.good
read/write/nosetuid/nodevices/intr/largefiles/logging/xattr/onerror=panic/dev=2400001
on Fri Apr 25 07:08:17 2008
Such entries may be placed in /etc/vfstab.
In terms of permissions, PRIV_SYS_MOUNT is required, plus the
ability to access the underlying file RO or RW, as needed. No
explicit permission to use lofi is needed.
Such mounts are always forced as 'nosuid' (no setuid and no
devices).
Unlike lofiadm, a lofi mount of an existing block or character
device is not supported.
On some operating systems such as Linux, this feature requires an
option "loop". This will not be necessary on Solaris, but in the
interests of approachability, it will be silently ignored. The
ability to explicitly specify the backing loop device is neither
present nor needed.
Zones do not support lofi(7d). This case doesn't change that.
There is an interaction with lofi(7d) crypto support. Currently,
this case does not propose to extend mount(1m) options to include
the ones needed for crypto support. This may happen in the future.
Compressed lofi(7d) support just works.
3. VFS interface
This feature requires explicit filesystem support in the relevant
VFS_MOUNT() routine.
In particular, a filesystem can call the following helper:
int vfs_get_lofi(vfs_t *, vnode_t **);
This returns the vnode for a lofi minor node corresponding to the
mounted file. A filesystem would then use this vnode as the device
to actually mount.
For this to work, such a filesystem needs to skip the block device
security check for the lofi node, as it could fail. However, the
VOP_ACCESS() check on the mount source (which is a file for the lofi
case) is still needed to ensure that the mount is secure.
For example, a process with PRIV_SYS_MOUNT and
PRIV_FILE_DAC_READ/WRITE would fail an explicit check to open the
lofi node, but succeed in the mount source check (unless the file is
root-owned, preserving standard semantics for least privilege).
Currently, filesystem support exists for ufs, pcfs, hsfs, and udfs.
ZFS support is unlikely to happen as it doesn't support traditional
mounting in this manner.
No change in behaviour should occur in filesystems that don't
support this feature.
4. lofi changes
When unmounting such a mount, we need to take down the created lofi
node. However, it may be held open. Rather than forcing removal of
the node via ->li_force, we add a new option, ->li_cleanup. This
only removes the lofi node when the last user closes it.
5. Interface table
lofi_ioctl Consolidation private, used and extended
by this project
New mount(1m) syntax. Committed.
6. References
PSARC 1999/463 lofi - fast-track
PSARC 2007/001 lofi(7d) crypto support
6384817 Need persistant lofi based mounts and direct mount(1m) support
for lofi
7. Manual page differences
--- mount.man Mon Apr 28 06:57:24 2008
+++ mount.new Mon Apr 28 06:55:53 2008
@@ -20,7 +20,6 @@
mount [-F FSType] [generic_options] [-o specific_options]
[-O] special mount_point
-
mount -a [-F FSType] [-V] [current_options]
[-o specific_options] [mount_point]...
@@ -83,10 +82,16 @@
/etc/dfs/fstypes file. After filling in missing arguments,
mount will invoke the FSType-specific mount module.
+ For file system types which support it, a file can be mounted
+ directly as a file system by specifying the full path to the
+ file as the special argument. If specific file-system support
+ for such loopback file mounts is not present, lofiadm(1M) can
+ still be used by hand to mount a file-system image.
- Only a super-user can mount or unmount file systems using
- mount and umount. However, any user can use mount to list
- mounted file systems and resources.
+ Only a user with sufficient privilege (at least PRIV_SYS_MOUNT)
+ can mount or unmount file systems using mount and umount.
+ However, any user can use mount to list mounted file systems
+ and resources.
OPTIONS
-F FSType
@@ -224,7 +229,10 @@
for files within the file system. The default
is exec.
+ loop
+ Ignored for compatibility.
+
nbmand | nonbmand
Allow or disallow non-blocking mandatory locking
@@ -411,9 +419,10 @@
SEE ALSO
mount_cachefs(1M), mount_hsfs(1M), mount_nfs(1M),
mount_pcfs(1M), mount_tmpfs(1M), mount_ufs(1M),
+ mount_udfs(1M),
mountall(1M), umountall(1M), fcntl(2), mmap(2), mnttab(4),
vfstab(4), attributes( 5), largefile(5), lofs(7FS),
- pcfs(7FS)
+ pcfs(7FS), privileges(5), lofiadm(1M)
NOTES
If the directory on which a file system is to be mounted is
--- mount_hsfs.man Mon Apr 28 06:57:45 2008
+++ mount_hsfs.new Mon Apr 28 06:55:09 2008
@@ -32,6 +32,7 @@
the missing arguments, including the FSType-
specific_options; see mount(1M) for more details.
+ hsfs supports loopback file mounts: see mount(1M).
A file system conforming to ISO 9660 can contain extensions
that allow it to overcome limitations of the original ISO
--- mount_pcfs.man Mon Apr 28 06:57:49 2008
+++ mount_pcfs.new Mon Apr 28 06:56:16 2008
@@ -24,6 +24,7 @@
mount operation, these are hidden until the file system is
unmounted.
+ pcfs supports loopback file mounts: see mount(1M).
If mount is invoked with special or mount_point as the only
arguments, mount will search /etc/vfstab to fill in the
--- mount_udfs.man Mon Apr 28 06:57:52 2008
+++ mount_udfs.new Mon Apr 28 06:56:32 2008
@@ -22,6 +22,7 @@
mount operation, these are hidden until the file system is
unmounted.
+ udfs supports loopback file mounts: see mount(1M).
If mount is invoked with either special or mount_point as
the only arguments, mount searches /etc/vfstab to fill in
--- mount_ufs.man Mon Apr 28 06:57:57 2008
+++ mount_ufs.new Mon Apr 28 06:56:42 2008
@@ -22,6 +22,7 @@
mount operation, these are hidden until the file system is
unmounted.
+ ufs supports loopback file mounts: see mount(1M).
If mount is invoked with special or mount_point as the only
arguments, mount will search /etc/vfstab to fill in the
6. Resources and Schedule
6.4. Steering Committee requested information
6.4.1. Consolidation C-team Name:
ON
6.5. ARC review type: FastTrack
6.6. ARC Exposure: open
