Re: How to get multistrap rootfs up and running?
On 2 February 2011 23:44, Wookey <wookey@wookware.org> wrote:
> +++ Bjørn Forsman [2011-02-02 22:13 +0100]:
>> On 2 February 2011 20:42, Marcus Osdoba <marcus.osdoba@googlemail.com> wrote:
>>
>> [snip]
>>
>> > The default kernels I found in Debian and Emdebian do not support rootfs
>> > over NFS. Which kernel do you use? I dont't think that init=/bin/sh is
>> > sufficient to get dpkg working. Maybe you should try it with standard init
>> > (/sbin/init ?) and/or reconfigure the rootfs before the first mount by the
>> > kernel (like I do in my example above - ok it's not baked but anyway).
>>
>> I built mainline Linux kernel for my board.
>>
>> Booting with /sbin/init is not working as the init package is not
>> configured. There is no /etc/inittab and init freaks out.
>
> That's right. So there are two approaches to the problem of postinst
> scripts needing to be run native.
> 1) Use Qemu-static to chroot into the rootfs and configure it. I'm
> planning to add this feature to both multistrap and schroot as it's
> convenient sometimes to have it done at chroot creation time and at
> chroot-entering time.
> 2) Use the script support to manually fettle the files needed to boot
> (inittab, fstab, /etc/securetty etc) enough for it to boot, then you
> can run dpkg --configure -a
> setupscript=/path/to/setup.sh
> configscript=/path/to/config.sh
> See machine::variant support in the man page
> This will help with the NFS mount method too.
That info helps a lot, thanks. (Sorry for not reading the manpage
before, don't know how I managed to miss it...)
BTW, having multistrap do configure with qemu-static would be really nice :-)
Best regards,
Bjørn Forsman
Reply to: