Bug#589118: `rdev` setting ignored

[Elliott Mitchell <ehem@m5p.com>]

reopen 589118
quit

>From: Ben Hutchings <ben@decadent.org.uk>
> On Wed, 2010-07-14 at 18:11 -0700, Elliott Mitchell wrote:
> > Package: initramfs-tools
> > Version: 0.92o
> > 
> > Subject tells the story. Appears the images generated by initramfs-tools
> > completely ignore the `rdev` setting that the kernel was given to the
> > kernel. While 99% of users may be explicitly passing the root device via
> > passing "root=/dev/foo" through the bootloader, if that is absent one
> > would think the value from `rdev` would be honored.
> > 
> > (yeah, it's an ancient method, but not officially deprecated)
> 
> If the bootloader passes an initramfs to the kernel, that overrides any
> rdev parameter.  This is nothing to do with the contents of the
> initramfs.

Bzzzt! While the "initrd=" kernel command-line option and `rdev` kernel
settings are not completely orthogonal, they are mostly unrelated. The
"initrd=" option overrides the `rdev` setting in the same fashion the
"initrd=" option overrides the "root=" and all other kernel command-line
options. Mainly, the initramfs can ignore any and all options and use
ones built in, or it can implement all those options.  It is the "root="
option that is directly related to `rdev`.

Unlike the kernel command-line, I don't know how the `rdev` (and
accompanying) setting is passed along to initial ram disks, but I do know
it is (or was). I'm unsure whether Debian 4.0/Etch honored the `rdev`
setting, but I am pretty certain initial ram disks generated with Debian
3.1/Sage did honor the `rdev` setting unless overridden by the "root="
option.


-- 
(\___(\___(\______          --=> 8-) EHM <=--          ______/)___/)___/)
 \BS (    |         EHeM@gremlin.m5p.com PGP F6B23DE0         |    )   /
  \_CS\   |  _____  -O #include <stddisclaimer.h> O-   _____  |   /  _/
2477\___\_|_/DC21 03A0 5D61 985B <-PGP-> F2BE 6526 ABD2 F6B2\_|_/___/3DE0