Re: mount --bind behaviour

[Karel Zak <kzak@xxxxxxxxxx>]

On Tue, Sep 25, 2007 at 11:53:07AM +0200, Matthias Koenig wrote:
> As far as my understanding of bind mounts is, that they get the mount
> options of the source and cannot have different mount options.
> This is what is stated in mount(2): 
> "The filesystemtype, mountflags, and data arguments are ignored."
> 
> Lets say we have /test and bind mount it to /mnt
> # mount --bind /test /mnt
> # mount
> /test on /mnt type none (rw,bind)
> 
> Now some people have the idea to do the bind mount read-only, of course
> this is not possible since the mount syscall ignores additional options.

 but this has to be fixed. These patches are already in -mm. So ..
 2.6.24 (probably;-).

> But:
> # mount --bind -r /test /mnt
> # mount
> /test on /mnt type none (ro,bind)
> 
> The mount *program* thinks the mount is ro, which is opposite to
> what /proc/mounts states:
> # cat /proc/mounts
> /dev/sda3 /mnt ext3 rw,data=ordered 0 0
> 
> We cannot just ignore additional options in mount, because the source
> could have been mounted with different options (e.g. ro).
> I think we will have to clone the options of the source in case
> of bind mounts.

 Yes, I agree -- with exception for kernels with r/o bind mounts where
 we need to modify rw/ro option. I think we can implement the support
 for r/o bind mounts already now. Something like:

    if (linux_version_code() >= MAKE_VERSION(2,6,24)))
        ... modify rw/ro ...
    else
        ... blindly clone options ...


 Grrr, mtab is hell.

    Karel

-- 
 Karel Zak  <kzak@xxxxxxxxxx>
-
To unsubscribe from this list: send the line "unsubscribe util-linux-ng" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html