Re: [PATCH] dumpstack: x86: various small unification steps

["Alexander van Heukelum" <heukelum@fastmail.fm>]

On Sun, 5 Oct 2008 11:35:23 +0200, "Ingo Molnar" <mingo@elte.hu> said:
> [...]
> 
> regarding klibc, that's interesting: is that the in-kernel klibc from 
> hpa? Which tree are you using to pull that into tip/master? (unless i 
> misunderstood what you are doing)

Hi,

I just use hpa's git://git.kernel.org/pub/scm/libs/klibc/klibc.git,
and compile it 'stand-alone' (you need to create a symlink 'linux',
pointing to a configured kernel tree first.)

I put the klibc-library and all the binaries in the kernel-built-in
initramfs and add an 'init', which contains:

#! /bin/sh
mkdir proc
mount -t proc proc /proc
mkdir sys
mount -t sysfs sysfs /sys
mkdir debugfs
mount -t debugfs debugfs /debugfs
bin/sh -i
reboot

This way even the simplest test-kernel gets into userspace and
you can examine the contents of proc and sysfs.

As hpa is listening too... I have a feature request for klibc
to auto-generate a file that can be fed to the kernel's
INITRAMFS_SOURCE, containing all binaries and an init like
the one I gave above. Especially because the hash added to
the shared library name keeps changing :-/.

Adding a (staticly linked) binary for testing is then an
easy way to run a test in userspace. And usually I do that
using qemu.

Greetings,
    Alexander

> 	Ingo
-- 
  Alexander van Heukelum
  heukelum@fastmail.fm

-- 
http://www.fastmail.fm - One of many happy users:
  http://www.fastmail.fm/docs/quotes.html