Re: [hatari-devel] Linux started to init in Hatari

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

I've pushed this.  IMHO I've provided pretty good / easy step-by-step
instructions now for building & testing Linux under Hatari.

I'm a bit surprised how buggy m68k Linux seems, see the "Bugs / TODO"
section at beginning of:
	doc/m68k-linux-for-hatari.txt


	- Eero

Next weeks I'll be discussing the kernel issues on Debian m68k list,
looking at the missing Hatari NVRAM regs support, disasm/memdump issues
with --lilo, trying kernels which have more features needed by Debian,
maybe try building my W window system for m68k Linux and see whether
it works under Hatari.

On 3/26/19 1:41 AM, Eero Tamminen wrote:
> I've been updating the documentation and added few helper scripts
> for building test root file systems, but I need to:
> * test them properly more before pushing them
> * test few issues before reporting them to m68k upstream
>
> I case somebody's interested to take a look, the latest patch
> series is attached.
>
>
>      - Eero
>
> On 3/23/19 2:27 AM, Eero Tamminen wrote:
> > Finally got Linux booting successfully to shell.
> >
> > I'm now using Geert's kernel, but I think the main issue was just
> > getting a working kernel configuration.  There was something causing
> > oopses in the old one.
> >
> > I also never got ramfs booting working correctly, but booting with
> > IDE rootfs i.e. what was working on real device, works now. [1]
> >
> > Attached are working config, and bootlog from where you can see
> > Hatari options.
> >
> > IDE content was just busybox in bin/, made into a FAT file system
> > with atari-hd-image script coming with Hatari, so it wasn't very
> > useful. :-)
> >
> > After I've tested few extra things and updated/cleaned the docs a bit,
> > I'll push the updated patches in, if there are no further comments /
> > objections about them.
> >
> >
> > [1] Trying to boot with that same FAT rootfs that works through IDE,
> > from a floppy, never finished. I just got warnings from m68k kernel
> > soft-watchdog every few minutes that things were (still) stuck
> > somewhere. -> I think kernel itself is buggy in some configs.
> >
> >
> >      - Eero
> >
> > On 3/21/19 1:35 AM, Eero Tamminen wrote:
> > > I got Linux starting enough that it showed logs also in Atari screen,
> > > not just debug console, and it tried running init, at which point it
> > > got NULL pointer Oops.  Log attached.
> > >
> > > I got this by building few days old Linux Git from upstream:
> > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
> > >
> > > Using the attached kernel config.
> > >
> > > (Maybe the Debian 3.6 & 4.9 binary kernels I tried earlier enabled
> > > some extra HW which probing disagrees with Hatari.)
> > >
> > > This new kernel build some weird extra objects for which there were
> > > no config options (e.g. some ARM omap stuff), but I didn't check
> > > whether linker discarded those from final binary or not.
> > >
> > > I'll probably try Geert's (m68k arch maintainer) kernel next:
> > > https://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k.git/log/?h=m68k-v5.0 
> > >
> > > And try to get newer m68k initrd content.