installer: fix the "umount: XXX not mounted" error message during postinst

installer: fix the "umount: XXX not mounted" error message during postinst

When cleaning loop devices, it may be not always mounted.
This CL makes it allowing an extra parameter to specify any additional
steps to be performed before releasing a loop device.

BUG=chromium-os:9278
TEST=Build an USB image with this change, and then /usr/sbin/chromeos-install
 Before: Seeing "umount: /dev/loop2: not mounted" and
                "umount: /dev/loop3: not mounted" in output messages
 After: No error message. Also verified "sudo losetup -a" does not leave
        any resources mounted.

Change-Id: I10dfa08b037bb11bd821754f32a7a39c155fe445

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=8bb0de1

[Will Drewry, 2010-11-18 09:20:15]

Looks ok to me with the given question.

However, I wonder if it'd make sense to just check if the device is mounted in
/proc/mounts.  On one hand I like the explicit indication of what needs to be
done.

Do you forsee a future use of "cleanup_command" separate from umount?  If so,
then this makes sense, but otherwise, I wonder if it'd make more sense to make
it check if "$2" = "umount"  or even "--umount" and if so, call umount instead
of calling it with the passed in arg.

Either way, thanks for the clean up!

http://codereview.chromium.org/5122002/diff/1/chromeos-setimage
File chromeos-setimage (right):

http://codereview.chromium.org/5122002/diff/1/chromeos-setimage#newcode132
chromeos-setimage:132: cleanup_loop "${KERN_LOOP}" ""
Do you need to specify empty $2s here?

[Hung-Te, 2010-11-18 09:49:02]

Looking at /proc/mount seems not a really good idea because we don't know if the
device file is already used by other scripts. (i.e., mounted before and passed
into this script as a device file)

At first I think we may have something like resigning, or invoking
enable/disable_rw_mount, so an arbitrary command looks better. However I think
that's not going to happen very soon (or maybe never), so I've changed to 2nd
param to a $FLAGS_TRUE / $FLAGS_FALSE.

Please take a look again, thanks!

http://codereview.chromium.org/5122002/diff/1/chromeos-setimage
File chromeos-setimage (right):

http://codereview.chromium.org/5122002/diff/1/chromeos-setimage#newcode132
chromeos-setimage:132: cleanup_loop "${KERN_LOOP}" ""
Although it may be not necessary, I think an explicit "" may help letting
developer to know "you have to pass something here"

[Will Drewry, 2010-11-23 01:36:48]

LGTM with the comparison nit.  Thanks!

http://codereview.chromium.org/5122002/diff/5001/chromeos-setimage
File chromeos-setimage (right):

http://codereview.chromium.org/5122002/diff/5001/chromeos-setimage#newcode123
chromeos-setimage:123: if [ "$need_umount" = ${FLAGS_TRUE} ]; then
nit: FLAGS_TRUE/FALSE is normally checked with -eq and not using a string
equality, though it should be equivalent.

[Hung-Te, 2010-11-23 02:14:52]

http://codereview.chromium.org/5122002/diff/5001/chromeos-setimage
File chromeos-setimage (right):

http://codereview.chromium.org/5122002/diff/5001/chromeos-setimage#newcode123
chromeos-setimage:123: if [ "$need_umount" = ${FLAGS_TRUE} ]; then
On 2010/11/23 01:36:50, Will Drewry wrote:
> nit: FLAGS_TRUE/FALSE is normally checked with -eq and not using a string
> equality, though it should be equivalent.

Done.