Bug#382788: [Yaird-devel] Bug#382788: yaird does not uses the kernel command like to find out the root partition

[Sven Luther]

On Mon, Aug 14, 2006 at 04:03:31PM +0200, Jonas Smedegaard wrote:
> On Mon, 14 Aug 2006 14:57:21 +0200 Sven Luther wrote:
> > Jonas, i agree on this with maks and pierre, being able to override
> > root= is a very essential feature. Furthermore, it is not all that
> > difficult to implement (just check for an existing root= command line
> > before providing your own copy), probably 3 lines of perl or so,
> > since the comand line is available as /proc/cmdline. I am no perl
> > expert though.
> 
> There's a difference between being able to read the root option
> provided on commandline and actually supporting it.

Well, once you can read the root options, you can easily enough make an
initial effort to support it, even with the below caveats.

You just need to use this value instead of the detected at build time one, i
have not looked, but i can't imagine it is anything but trivial.

> The fundamental design of yaird causes it to not be able to support
> changing root at boot time. It may happen to work for some specific
> cases, but not in general.
> 
>  * Only kernel modules specifically needed to reach the default root
>    is included with the image.

Sure, but even with the same modules, it is still good to be allowed to
override root, be it only to be able to boot into the system and rerun yaird
or something. Imagine you moved the disk around on the same controler, or
inserted a new disk and the ordering changed. Imagine too you used a tool like
partimage to move the whole system to another partition of the same disk or of
a similar machine.

>  * Only device files specifically needed to reach the default root
>    is created while in the pre-root boot stage.

Indeed, but it would be trivial to create the command-line passed root instead
of the build-time detected one.

> I consider it a pretty damn good feature to have. But not essential.

Ok.

> Please provide pointers to official statements supporting the claim
> that this is a release-critical issue.

I will let Maxs reply to this.

Friendly,

Sven Luther