On Thu, Jul 21, 2011 at 08:09:08AM -0400, Genes MailLists wrote:
On 07/21/2011 07:09 AM, Steve Clark wrote:
> Well what benefit(s) does the new 'df' provide, is it worth all the pain
> it brings?
>
I concur - the current df behavior is well .. goofy :-) - however this
may be tricky to fix in the new world - but should be fixed.
If this behavior is somehow desirable it would be preferable to make
it an option (like df --full or whatever) and make the default something
more sensible.
That said, it may be tricky in the new world;
where can you retrieve the info about a mount being a bind mount ?
How can you push the chrooted bind mounts into being less obtrusive (or
even optional, --show-chrooted-mounts)
/proc/mounts does not seem to distinguish bind mounts - so this may
have to be a kernel change and perhaps adding /proc/mounts/bind and
moving bind mounts 1 level down - this is not an area I know a lot about
however, so I'll leave this to the real experts.
I've already talked about it in this list... "bind" is operation, not
state of any mountpoint. Something like /proc/mounts/bind does not
make sense from kernel's point of view.
On Linux arbitrary filesystem could be mounted to more than one place
in VFS -- our userspace utils have to accept this fact...
Karel
--
Karel Zak <kzak(a)redhat.com>
http://karelzak.blogspot.com