Trying to avoid incompatability on a mount enhancement

Trying to avoid incompatability on a mount enhancement

[Isaac Dunham]

Hello,
I've been thinking about adding a feature to Busybox mount, but I'd
like to check on options that are already claimed in util-linux so I
don't cause a major incompatability.

The feature I'm thinking about is roughly analogous to "mkdir -p" or
"cpio -d":
it creates the mountpoints if needed.
This could be done by calling mkdir -p, or by calling mkdir() from mount.
The point is to allow one to mount /dev, /sys and all their subsystems
in one pass by using 
mount -T /etc/vfstab -a -d
or something like that where /etc/vfstab contains (for example):
===
# Kernel filesystems to mount for bringing up the system
proc	/proc	proc	noexec,nodev,nosuid	0 0
sysfs	/sys	sysfs	noexec,nodev,nosuid	0 0
devfs	/dev	devtmpfs defaults		0 0
devshm	/dev/shm tmpfs	nodev,nosuid		0 0
devpts	/dev/pts devpts	noexec,nodev,nosuid	0 0
===

I'm thinking that the option should be ignored with -o bind,
because that may take a file or a directory.

Do you have any comments about the concept?

As far as I can see, -d, -m, and -p are all available (at least in
the version of util-linux mount that ships in Jessie).
Are there any plans to use any of these options?

Also, is there any interest in seeing this in util-linux mount?
(I'd assume some of the corresponding code would be added to libmount,
but I haven't really looked at util-linux code before--just Busybox and
Toybox mount(8) implementations.)

Thanks,
Isaac Dunham

Re: Trying to avoid incompatability on a mount enhancement

[Karel Zak]

On Sat, Mar 14, 2015 at 10:02:50PM +0000, Isaac Dunham wrote:
> Hello,
> I've been thinking about adding a feature to Busybox mount, but I'd
> like to check on options that are already claimed in util-linux so I
> don't cause a major incompatability.
> 
> The feature I'm thinking about is roughly analogous to "mkdir -p" or
> "cpio -d":
> it creates the mountpoints if needed.

 man mount:

 x-mount.mkdir[=mode]

        Allow  to  make  a target directory (mountpoint).  The
        optional argument mode specifies the filesystem access mode
        used for mkdir(2) in octal notation.   The default  mode  is
        0755.  This functionality is supported only for root users.


 ... so:

        mount /dev/sda1 /mnt/foo -o x-mount.mkdir

   Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

Re: Trying to avoid incompatability on a mount enhancement

[Isaac Dunham]

On Mon, Mar 16, 2015 at 11:22:33AM +0100, Karel Zak wrote:
> On Sat, Mar 14, 2015 at 10:02:50PM +0000, Isaac Dunham wrote:
> > Hello,
> > I've been thinking about adding a feature to Busybox mount, but I'd
> > like to check on options that are already claimed in util-linux so I
> > don't cause a major incompatability.
> > 
> > The feature I'm thinking about is roughly analogous to "mkdir -p" or
> > "cpio -d":
> > it creates the mountpoints if needed.
> 
>  man mount:
> 
>  x-mount.mkdir[=mode]
> 
>         Allow  to  make  a target directory (mountpoint).  The
>         optional argument mode specifies the filesystem access mode
>         used for mkdir(2) in octal notation.   The default  mode  is
>         0755.  This functionality is supported only for root users.
> 
> 
>  ... so:
> 
>         mount /dev/sda1 /mnt/foo -o x-mount.mkdir
> 

Thank you for responding.
I read about that some time after I sent the email; now comes the question
of whether Busybox will accept it.

I've already gotten a comment to the effect that x-mount.mkdir is "ugly"
(no clarifications provided), and that it's lacking the ability to set
owner/group for the mountpoint.
(The individual who opposes this is arguing in favor of adding a similar
but more expanded capability to mdev, the Busybox hotplugger.)

Thanks,
Isaac Dunham

Re: Trying to avoid incompatability on a mount enhancement

[Karel Zak]

On Mon, Mar 16, 2015 at 05:19:59PM +0000, Isaac Dunham wrote:
> On Mon, Mar 16, 2015 at 11:22:33AM +0100, Karel Zak wrote:
> >  x-mount.mkdir[=mode]
> > 
> >         Allow  to  make  a target directory (mountpoint).  The
> >         optional argument mode specifies the filesystem access mode
> >         used for mkdir(2) in octal notation.   The default  mode  is
> >         0755.  This functionality is supported only for root users.
> > 
> > 
> >  ... so:
> > 
> >         mount /dev/sda1 /mnt/foo -o x-mount.mkdir
> > 
> 
> Thank you for responding.
> I read about that some time after I sent the email; now comes the question
> of whether Busybox will accept it.
> 
> I've already gotten a comment to the effect that x-mount.mkdir is "ugly"

 Well, x-* prefixes are generic solution for userspace mount options,
 we already use it for more things (for example systemd).

> (no clarifications provided), and that it's lacking the ability to set
> owner/group for the mountpoint.

 Good point, it would be possible to extend it for example to

   x-mount.mkdir[=mode:user:group]

 where :user and :group are optional and may be specified by UID/GID
 or by username/groupname. 

> (The individual who opposes this is arguing in favor of adding a similar
> but more expanded capability to mdev, the Busybox hotplugger.)

 I have no clue about busybox hotplugger, for standard udev we 
 usually don't want to do things like "mount" in udev context.

    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com