[PATCH] mount: support -i; pass through -f and -n to helpers as necessary

[Rob Landley]

On Friday 19 June 2009 02:03:12 Gilles Espinasse wrote:
> ----- Original Message -----
> From: "Rob Landley" <rob at landley.net>
> To: <busybox at busybox.net>
> Cc: "Colin Watson" <cjwatson at ubuntu.com>
> Sent: Friday, June 19, 2009 8:54 AM
> Subject: Re: [PATCH] mount: support -i;pass through -f and -n to helpers as
> necessary
>
> > On Thursday 18 June 2009 19:09:26 Colin Watson wrote:
> > > When FEATURE_MOUNT_HELPERS is enabled, mount should support the -i
> > > option, which prevents calling the helper function. This is used by
> > > FUSE: it installs fusermount as a helper, and if it's configured to
> > > update /etc/mtab then it does so by calling 'mount -i -f' (i.e. don't
> > > call helper, fake the system call, but do update mtab).
> >
> > There's something that still uses mtab?  Wow.
>
> Yes mke2fs check if filesystem is mounted and fail the check without mtab

No, I mean things that don't work if you "ln -s /proc/self/mounts /etc/mtab".

The kernel exports an mtab-format file in /proc, listing all the mounted 
filesystems.  One of the reasons the kernel does this is mounts can vary from 
process to process (google for "linux shared subtree"), in which case no 
matter what a normal file version of /etc/mtab said it would be wrong for some 
of the processes.  (Plus you can screw up mtab by mounting and unmounting stuff 
in a chroot and so on, and the kernel never gets confused but the file won't 
always keep track)...

>     Gilles

Rob
-- 
Latency is more important than throughput. It's that simple. - Linus Torvalds