Re: root= kernel cmdline parameter

[Giorgio Dal Molin <giorgio.nicole@arcor.de>]

Hi,

> On August 9, 2018 at 9:45 AM Giorgio Dal Molin <giorgio.nicole@arcor.de> wrote:
> 
> 
> Hi,
> 
> > On August 8, 2018 at 9:52 AM Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > 
> > 
> > On Tue, Jul 31, 2018 at 01:20:12PM +0200, Giorgio Dal Molin wrote:
> > > Hi all,
> > > 
> > > I'm trying to find out the best method to define the root=
> > > parameter when booting the linux kernel from barebox.
> > > 
> > > My system boots from an sd card partitioned with a GPT.
> > > The sd card has two independent set of partitions with
> > > the following labels:
> > > 
> > > 1)   boot_1 + rootfs_1
> > > 2)   boot_2 + rootfs_2
> > > 
> > > for redondance in case one set gets damaged while updating.
> > > 
> > > 'boot_1' is formatted with ext4 and contains the kernel + dtb images
> > > for the userland in 'rootfs_1'; and the same for 'boot_2' and 'rootfs_2'.
> > > 
> > > Currently I hardcode the 'root' parameter to '/dev/mmcblk0p3' or
> > > '/dev/mmcblk0p4' corresponding to the kernel in 'boot_1' or
> > > 'boot_2' and it actually works but is there a better way to do
> > > this ?
> > 
> > You could use bootloader spec. That would at least allow you to store a
> > suitable root= parameter in the config files in the boot_1 and boot_2
> > partitions.
> > 
> > Not sure if that's what you are looking for though.
> > 
> > Sascha
> > 
> 
> as I already wrote to Uwe Kleine-Koenig I wanted to keep the kernel
> image and the dtb in a separate partition because these are the files
> the bootloader *must* be able to read to boot the system. Having the
> userland in a separate partition lets me choose an image format for it
> independently from what barebox supports; as an example I could pack
> my userland in a btrfs image. Moreover I can be more selective when
> updating the firmware: I can update only the kernel+dtb or only the
> userland image.
> 
> What I would like to have is an automatic way to set the root= kernel
> command line parameter similar to what barebox does with the 'bootm.appendroot'
> but a bit more flexible.
> Looking at the kernel function name_to_dev_t() in the source init/do_mounts.c,
> it accepts also the format 'PARTUUID=00112233-4455-6677-8899-AABBCCDDEEFF':
> my idea was to somehow tell barebox 'use the partition with label "rootfs_1"
> for the "root" parameter' and barebox should be able to match my label with
> the partition UUID and automatically generate the root= value.
> 
> giorgio

grepping a bit through the barebox sources I've found that what I need is
actually already implemented!

I know that at least one of 'rootfs_1' and 'rootfs_2' must contain a valid
userland image (a squashfs image); now, in my 'init' I try to 'mount -t squashfs'
rootfs_1: if it works I found that barebox defines a global variable '$squashfs0.linux.bootargs'
with exactly the content I need (root=PARTUUID=...). I just have to use this var.
to define my global linux.bootarg:

global linux.bootargs.root=$squashfs0.linux.bootargs

giorgio

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox