"mount -o remount" non-compliant?
Paul Fox
pgf at brightstareng.com
Fri Nov 11 13:13:06 UTC 2005
> Hmmm... Think think think... Ok, at first glance the semantics seem
> ambiguous, because when you do:
>
> mount -o remount,rw sub
>
> Right now, it's looking for sub in /etc/fstab because we only gave one
> non-option argument. _BUT_, we also said remount, and remount is a special
> case. "This sucker is already mounted." So in theory we don't care about
> fstab (because if we were going to get any info out of it, we already did)...
right.
> So, special case remount. Except what if you "-o remount,fred" and the sucker
> was previously read only? Should you have to specify ro again in the -o? Or
> should it look it up out of fstab? What if fstab says it's read only but you
> previously overrode to make it writeable, and then you remount "-o
> remount,fred" without specifying ro or rw either way?
>
> Sigh. I wonder if there's a way to query the current flags from a mount
> point? Yes there is, /proc/mounts fourth field. Ok, so in theory we grab
> that, append the new flags, and let the kernel trim out duplicates?
>
> Boy this is a pain to get right. But you're correct that we need to
> parse /etc/mtab. Ok...
is remount the only flag this applies to? i confess i haven't been
keeping track of the ever-growing number of mount options. in any case,
it really does seem like the code would change very little if you just
substitute the mtab file for /etc/fstab when starting that parse loop.
almost by definition, on a remount you want to use information
that's "current", not information that's "default".
paul
=---------------------
paul fox, pgf at brightstareng.com
More information about the busybox
mailing list