Weird Mount Messages

Rob Landley rob at landley.net
Tue Feb 14 12:29:19 UTC 2006


On Tuesday 14 February 2006 11:53 am, Jason Schoon wrote:
> Rob, all,
>
> I saw the following message from mount while working on my development
> system today:
>
> ~ # mount /dev/sdb1
> mount: Can't find /dev/sdb1 in /etc/fstab
>
> : Success
>
> mount: Mounting /dev/sdb1 on (null) failed: Success
>
>
> This was caused by me accidentally leaving off the mountpoint on the
> command line.
>
> Attached is a patch which fixes the problem, although I find it a bit
> ugly.  Definately a Monday patch.

On my system mount's been undergoing rather extensive dental work for a couple 
weeks now (with a couple resets back to baseline as I go "aagh, wrong 
approach!").

One of the things it's growing is a test suite.  Dunno if I'll check it in 
before 1.1.1 because it requires a User Mode Linux wrapper to run it (it has 
to be run as root, in a chroot environment, with a known set of existing 
mount points) and I'm not quite sure how I want to plug that into the test 
harness yet.

There are rather a lot of things that require root access to test properly and 
the User Mode Linux approach is a nice way of handling that.  More a 1.1.2 
issue than 1.1.1, although I plan to upgrade the test suite so it works 
without the rest of busybox before 1.1.1.  Mostly there.

Banging on the bug list at the moment to close out old low hanging fruit...

> One other question.  There are 2 bb_perror_msg_and_die() calls that it
> seems to me will bypass some necessary cleanup if FEATURE_CLEANUP is
> defined.

They generally do.

> Would it make sense to change these to a bb_perror_msg() and 
> jump to the appropriate cleanup?

It would make things bigger.

Every bb_error_msg_and_die() call in the project bypasses FEATURE_CLEANUP.  At 
some point, if we really care, we have to register atexit() stuff to do some 
of this cleanup reliably, or perhaps do something funky with longjmp.

The thing is, how much do we care?  What are the advantages of 
FEATURE_CLEANUP?  (Apart from making valgrind happy and potentially working 
on machines where fork() and exit() are emulated via longjmp?)

How much effort is it worth?

Rob
-- 
Never bet against the cheap plastic solution.



More information about the busybox mailing list