_reliable_ way to get the dev for a mount point?

_reliable_ way to get the dev for a mount point?

[Pozsar Balazs]

Hi all!

I found that in a chrooted environment the /proc/mounts file is
messed up badly. For example, if I do a
 chroot /path/to/somewhere
then in the chrooted environment's proc/mounts file will truncate
the /path/to/somewhere string from those mount point it can, and leaves
the rest as they are. (something like s|/path/to/somewhere|| )
Consequently you will have at least two devices showns as mounted to /.
But it can get worse (i know its extreme), if the directory of the
chrooted environment is overmounted.

For example, here's an output from inside the chroot environment:
# cat /proc/mounts
rootfs / rootfs rw 0 0
/dev/root / ext3 rw 0 0
none /dev devfs rw 0 0
none /proc proc rw 0 0
none /proc/bus/usb usbdevfs rw 0 0
none /root/chroot1/proc proc rw 0 0
/dev/cdrom / iso9660 ro 0 0
none /proc proc rw 0 0
/dev/hda2 / ext3 rw 0 0

As you can see, there are 3 devices shown as mounted to / (in reality,
only 2 in the chrooted env), and you cannot know from this which of these
3 is your real rootdir.


So, my question is there a way to get back the device for a directory,
_reliably_. (I want to know which devices holds the files my process sees
under an arbitrary /path/to/somewhere).

Thanks,
-- 
Balazs Pozsar

Re: _reliable_ way to get the dev for a mount point?

[Erik Andersen]

On Mon May 06, 2002 at 11:55:38AM +0200, Pozsar Balazs wrote:
> So, my question is there a way to get back the device for a directory,
> _reliably_. (I want to know which devices holds the files my process sees
> under an arbitrary /path/to/somewhere).

stat(mnt_point, &statbuf); then walk through /dev and stat each
device, check that it is a block device and that st_rdev matches
the statbuf.st_rdev.  When you get a match, you know a device
name for the directory.

 -Erik

--
Erik B. Andersen             http://codepoet-consulting.com/
--This message was written using 73% post-consumer electrons--

Re: _reliable_ way to get the dev for a mount point?

[Xavier Bestel]

Le lun 06/05/2002 à 19:38, Erik Andersen a écrit :
> On Mon May 06, 2002 at 11:55:38AM +0200, Pozsar Balazs wrote:
> > So, my question is there a way to get back the device for a directory,
> > _reliably_. (I want to know which devices holds the files my process sees
> > under an arbitrary /path/to/somewhere).
> 
> stat(mnt_point, &statbuf); then walk through /dev and stat each
> device, check that it is a block device and that st_rdev matches
> the statbuf.st_rdev.  When you get a match, you know a device
> name for the directory.

is devfs walkable ? I mean, no loops or infinite dynamically generated
directory ?

Re: _reliable_ way to get the dev for a mount point?

[Denis Vlasenko]

> > stat(mnt_point, &statbuf); then walk through /dev and stat each
> > device, check that it is a block device and that st_rdev matches
> > the statbuf.st_rdev.  When you get a match, you know a device
> > name for the directory.
>
> is devfs walkable ? I mean, no loops or infinite dynamically generated
> directory ?

If not, you still can do that but walk method must be clever
(do 1-level deep search, then 2-level, etc)
--
vda