FS#32875 - usr hook doesn't mount ZFS filesystem

Attached to Project: Arch Linux
Opened by Con Tassios (ct) - Tuesday, 27 November 2012, 03:04 GMT
Last edited by Dave Reisner (falconindy) - Wednesday, 28 November 2012, 16:48 GMT
Task Type Bug Report
Category System
Status Closed
Assigned To Dave Reisner (falconindy)
Architecture All
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:

The usr hook, from mkinitcpio package, does not specify the filesystem type (-t option) when mounting the /usr filesystem. This is sufficient for most file system types as they are already recognised by mount/util-linux, but ZFS is not recognised.

# mount OS/usr /usr
mount: you must specify the filesystem type

Please consider some additions to the script, to determine the fstype of the /usr filesystem and mount it with -t option

Something like this -

$fstype=$(findmnt -snero fstype --tab-file="$realtab" /usr)
mount -t "$fstype" "$usr_source" /new_root/usr -o "$mountopts"


Additional info:
* package version(s)
* config and/or log files etc.


Steps to reproduce:
This task depends upon

Closed by  Dave Reisner (falconindy)
Wednesday, 28 November 2012, 16:48 GMT
Reason for closing:  Won't fix
Additional comments about closing:  See closing comments.
Comment by Dave Reisner (falconindy) - Tuesday, 27 November 2012, 03:19 GMT
The usr hook isn't what needs fixing here.

Does blkid detect the filesystem properly? Does 'zfs' appear in /proc/filesystems?
Comment by Con Tassios (ct) - Tuesday, 27 November 2012, 03:31 GMT
blkid shows TYPE="zfs_member" for disk partitions that are members of the ZFS pool.
zfs is listed in /proc/filesystems.
Comment by Dave Reisner (falconindy) - Tuesday, 27 November 2012, 03:35 GMT
Hmmm... so what happens when you try to mount it without the -t flag? Can i get the output of:

LIBMOUNT_DEBUG=0xffff mount /dev/whatever /new_root/usr
Comment by Con Tassios (ct) - Tuesday, 27 November 2012, 03:47 GMT
mount command says "you must specify the filesystem type" when mounting without -t option.

# LIBMOUNT_DEBUG=0xffff mount OS/usr1 /mnt
libmount: debug mask set to 0xffff.
16085: libmount: CXT: [0x1124060]: ----> allocate
16085: libmount: UTILS: mtab: /etc/mtab
16085: libmount: UTILS: /etc/mtab: irregular/non-writable
16085: libmount: UTILS: utab: /run/mount/utab
16085: libmount: CXT: [0x1124060]: mount: preparing
16085: libmount: CXT: [0x1124060]: use default optmode
16085: libmount: CXT: [0x1124060]: OPTSMODE: ignore=0, append=0, prepend=1, replace=0, force=0, fstab=1, mtab=1
16085: libmount: CXT: [0x1124060]: fstab not required -- skip
16085: libmount: CXT: [0x1124060]: merging mount flags
16085: libmount: CXT: [0x1124060]: final flags: VFS=00000000 user=00000000
16085: libmount: CXT: [0x1124060]: mount: evaluating permissions
16085: libmount: CXT: [0x1124060]: mount: fixing optstr
16085: libmount: CXT: [0x1124060]: mount: fixing vfs optstr
16085: libmount: CXT: applying 0x00000000 flags to '(null)'
16085: libmount: CXT: new optstr 'rw'
16085: libmount: CXT: [0x1124060]: mount: fixing user optstr
16085: libmount: CXT: applying 0x00000000 flags to '(null)'
16085: libmount: CXT: new optstr '(null)'
16085: libmount: CXT: [0x1124060]: fixed options [rc=0]: vfs: 'rw' fs: '(null)' user: '(null)', optstr: 'rw'
16085: libmount: CXT: [0x1124060]: preparing source path
16085: libmount: CXT: [0x1124060]: srcpath 'OS/usr1'
16085: libmount: CACHE: [0x1124290]: alloc
16085: libmount: CACHE: [0x1124290]: add entry [ 1] (path): OS/usr1: OS/usr1
16085: libmount: CXT: [0x1124060]: final srcpath 'OS/usr1'
16085: libmount: CXT: [0x1124060]: preparing target path
16085: libmount: CACHE: [0x1124290]: add entry [ 2] (path): /mnt: /mnt
16085: libmount: CXT: [0x1124060]: final target '/mnt'
16085: libmount: CXT: [0x1124060]: FS type: (null)
16085: libmount: CXT: [0x1124060]: prepare update
16085: libmount: UPDATE: [0x1124f10]: allocate
16085: libmount: UPDATE: [0x1124f10]: resetting FS [fs=0x0x1124150, target=(null), flags=0x00000000]
16085: libmount: UPDATE: [0x1124f10]: FS template:
16085: libmount: UPDATE: ------ fs: 0x1124150
source: OS/usr1
target: /mnt
fstype: (null)
optstr: rw
VFS-optstr: rw
16085: libmount: UPDATE: prepare utab entry
16085: libmount: UPDATE: utab entry unnecessary (no options)
16085: libmount: CXT: [0x1124060]: mount: do mount
16085: libmount: CXT: [0x1124060]: trying to mount by filesystems lists
mount: you must specify the filesystem type
16085: libmount: CXT: [0x1124060]: <---- reset [status=0] ---->
16085: libmount: CACHE: [0x1124290]: free
16085: libmount: UPDATE: [0x1124f10]: free
16085: libmount: CXT: [0x1124060]: <---- free
Comment by Dave Reisner (falconindy) - Tuesday, 27 November 2012, 03:51 GMT
Oh. So ZFS devices aren't legit devices in that they're magical things that get pulled out of thin air rather than the pool itself existing as a block device in /dev. Annoying. 9p does this as well.

Well shit. I guess I have to fix this in the usr hook.
Comment by Dave Reisner (falconindy) - Tuesday, 27 November 2012, 04:36 GMT
...But I will point out that you can work around this by adding 'zfs' to /etc/filesystems. Not sure if it's intentional that /proc/filesystems isn't read by mount when /etc/filesystems doesn't exist.

edit: and mount(8) confirms that /proc/filesystems should be read if /etc/filesystems doesn't exist. Wohoo.

edit edit: Hmmm, so mount from latest util-linux git will indeed read /proc/filesystems, but it only tries filesystems which aren't marked as nodev (which zfs and 9p probably are). So, the u-l fix doesn't even save me.
Comment by Dave Reisner (falconindy) - Wednesday, 28 November 2012, 16:48 GMT
This will be fixed in the zfs-utils AUR package. I'd rather lean on mount to figure out what the filesystem is, helping it along where necessary. Maybe in the future I'll add a usrfstype= flag to the kernel cmdline in case you really want to override the auto detection.

Loading...