linux-apus-devel

MOL

[Michel <da...@re...>]

The commit notices are pending for unknown reasons, but I have applied the MOL
'patches' to the 2.2 tree. No branch, the changes are really minimal. So
minimal I can hardly believe it works (why is NO_RELOAD_HTAB undef'd in
mm/init.c while it remains defined in kernel/head.S?), but the ones who are so
eager about MOL should try.


Michel


-- 
Earthling Michel Dänzer (MrCooper)  \  CS student and free software enthusiast
Debian GNU/Linux (powerpc,i386) user \   member of XFree86 and the DRI project

MOL

[Roman Z. <zi...@fh...>]

Hi,

I got 2.3 booting with the MOL option enabled, it's the usual
memory-starts-0-assumption. There might be more in the MOL package itself.
A quick scan shows for example, that the ismolpatched tool will not work.
A committed the same fix to 2.2, but I didn't test it. Anyway, there's a
chance now that it works. :-)

bye, Roman

Re: MOL

[Alan B. <al...@ms...>]

hi,

> I got 2.3 booting with the MOL option enabled, it's the usual
> memory-starts-0-assumption. There might be more in the MOL package itself.
> A quick scan shows for example, that the ismolpatched tool will not work.
> A committed the same fix to 2.2, but I didn't test it. Anyway, there's a
> chance now that it works. :-)

I saw the committ for 2.2 mol-branch, i will be looking at it as soon as i
can

alan

MOL

[Glenn H. <gh...@c2...>]

Hi,

I have just tried MOL with the latest 2.4.0-test11 kernel.
When trying to load the mol.o module, it complains about two unresolved
symbols:

/usr/lib/mol/modules/mol.o: unresolved symbol find_path_device
/usr/lib/mol/modules/mol.o: unresolved symbol get_module_symbol

I have the latest mol rpm for the 2.4 kernel installed.
I get a warning saying that the mol.o module is build for the 2.4.0-test10
kernel. Could something have changed between test10 and test11 making the
module fail ?

Anyway, this was just to let you all know how it went.


I'd like to test with a 2.2 kernel too, but haven't downloaded the sources
yet.
What is the correct command to checkout the mol branch ?


-- 
--------------------------------------------------
Glenn Hisdal                homepage: http://home.c2i.net/ghisdal    
Trengereidfjord            ICQ UIN: 10801980
5658 Aarland               Amiga 4000 040/25, PPC/233, CyberVision64 
--------------------------------------------------

Re: MOL

[Alan B. <al...@ms...>]

hi,

> I'd like to test with a 2.2 kernel too, but haven't downloaded the sources
> yet.
> What is the correct command to checkout the mol branch ?

cvs co -r mol-branch 2.2

alan

Re: MOL

[Michel <da...@re...>]

Glenn Hisdal wrote:

> I have just tried MOL with the latest 2.4.0-test11 kernel.
> When trying to load the mol.o module, it complains about two unresolved
> symbols:
> 
> /usr/lib/mol/modules/mol.o: unresolved symbol find_path_device

I guess this for finding the Open Firmware (OF) path of a device. As we don't
have OF, I wonder if MOL works without it? Anyone would like to ask the MOL
developers? :)

> /usr/lib/mol/modules/mol.o: unresolved symbol get_module_symbol

Don't know about this one - a 'grep -l get_module_symbol' on the current tree
didn't show anything...

> I have the latest mol rpm for the 2.4 kernel installed.
> I get a warning saying that the mol.o module is build for the 2.4.0-test10
> kernel. Could something have changed between test10 and test11 making the
> module fail ?

I could imagine that something about get_module_symbol changed between test10
and test11, but even without that the issue with find_path_device remains...


> I'd like to test with a 2.2 kernel too, but haven't downloaded the sources
> yet.
> What is the correct command to checkout the mol branch ?

cvs -d[...] co -r mol-branch 2.2

or if you already have a 2.2 tree: (from the top level directory)

cvs upd -r mol-branch


Michel


-- 
Earthling Michel Dänzer (MrCooper)  \  CS student and free software enthusiast
Debian GNU/Linux (powerpc,i386) user \   member of XFree86 and the DRI project

Re: MOL

[Roman Z. <zi...@fh...>]

Hi,

On Mon, 27 Nov 2000, Glenn Hisdal wrote:

> /usr/lib/mol/modules/mol.o: unresolved symbol find_path_device

That's only defined if you have CONFIG_ALL_PPC selected, where I have no
idea whether that works... It might be needed to recompile the module for
apus and lets hope MOL can handle that there is no Open Firmware.

> /usr/lib/mol/modules/mol.o: unresolved symbol get_module_symbol

That function is gone with test11 and MOL needs to updated for that.

> I'd like to test with a 2.2 kernel too, but haven't downloaded the sources
> yet.
> What is the correct command to checkout the mol branch ?

If you already have the 2.2 sources, then 'cvs update -r mol-branch' is
enough (and the fastest way), you can get back to the  original with 'cvs
update -A'. If you need to check it out, append '-r mol-branch' to the
checkout command.

bye, Roman

Re: MOL

[Glenn H. <gh...@c2...>]

Hello 

On 27-Nov-2000, Roman wrote:


>> What is the correct command to checkout the mol branch ?

> If you already have the 2.2 sources, then 'cvs update -r mol-branch' is
> enough (and the fastest way), you can get back to the  original with 'cvs
> update -A'. If you need to check it out, append '-r mol-branch' to the
> checkout command.

ok. thanks :-)


- glenn

Re: MOL

[Roman Z. <zi...@fh...>]

Hi,

On Mon, 27 Nov 2000, Glenn Hisdal wrote:

> /usr/lib/mol/modules/mol.o: unresolved symbol find_path_device
> /usr/lib/mol/modules/mol.o: unresolved symbol get_module_symbol
> 
> I have the latest mol rpm for the 2.4 kernel installed.
> I get a warning saying that the mol.o module is build for the 2.4.0-test10
> kernel. Could something have changed between test10 and test11 making the
> module fail ?

BTW if someone wants to try it with test10, you can also go back to test10
with 'cvs up -r apus-2_4_0-test10' and then apply the change from test11
with 'cvs diff -r 1.1.1.1 -r 1.2 arch/ppc/lib/mol.h | patch -p0'.
But someone has to contact the MOL author, about find_path_device and
how/if it works without Open Firmware.

bye, Roman

MOL

[Samuel R. <sa...@ib...>]

I'm currently considering porting MOL to other platforms
than Linux/PPC (primarily *BSD and Darwin). Before doing that 
however, I though it might be nice to have it working on APUS.

The recent changes to MOL removes almost all assumptions
about the kernel. As far as I can tell, the only thing that
prevents MOL from working on APUS systems is the nonstandard
KERNELBASE stuff.

So, I have a few questions:

- Does the BK kernels at fsmlabs contain working APUS support?

- Any special quirks? In particular, why is the kernel
*not* located at physical address 0?

- Which cpu flavors are used? MOL currently lacks 603(e) support, 
but I intend to add support for this processor shortly.

To be able to add APUS support, I obviously need help with
testing and debugging since I don't have an APUS box.


Regards,

/Samuel


----------------------------------------------------------
 E-mail <sa...@ib...>  WWW: <http://www.ibrium.se>
  Phone/fax: (home) +46 8 4418431, (work) +46 8 7908470   
----------------------------------------------------------

Re: MOL

[Michel <mic...@ii...>]

Samuel Rydh wrote:
> 
> I'm currently considering porting MOL to other platforms
> than Linux/PPC (primarily *BSD and Darwin).

Time for a name change? ;)

> Before doing that however, I though it might be nice to have it working on
> APUS.

Thanks for thinking of us! :)

> The recent changes to MOL removes almost all assumptions
> about the kernel. As far as I can tell, the only thing that
> prevents MOL from working on APUS systems is the nonstandard
> KERNELBASE stuff.
> 
> So, I have a few questions:
> 
> - Does the BK kernels at fsmlabs contain working APUS support?

I doubt it. Only Jesper Skov used to have BK write permissions but he's been
out of APUS development for about a year. Roman Zippel would be the worthiest
successor, there's been some discussion about that but I don't know if
anything has come out of it yet.

The authoritative source for APUS is our CVS tree where Roman integrates BK
after each minor version bump. The current is 2.4.3.


> - Any special quirks? In particular, why is the kernel
> *not* located at physical address 0?

Because the memory starting at 0 is so-called 'chip memory' shared with the
custom chips and thus very slow. The kernel resides in 'fast memory'.


> - Which cpu flavors are used? MOL currently lacks 603(e) support,
> but I intend to add support for this processor shortly.

We have 603e and 604e so far.


> To be able to add APUS support, I obviously need help with
> testing and debugging since I don't have an APUS box.

We've been constantly hit by questions about when MOL was going to work on
APUS so there should be a lot of people who'd like to test it. :)


-- 
Earthling Michel Dänzer (MrCooper)    \   Debian GNU/Linux (powerpc) developer
CS student, Free Software enthusiast   \        XFree86 and DRI project member

Re: MOL

[Roman Z. <zi...@li...>]

Hi,

Samuel Rydh wrote:

> - Does the BK kernels at fsmlabs contain working APUS support?

No, although I started to send Cort patches, but it takes ages... :(

> - Any special quirks? In particular, why is the kernel
> *not* located at physical address 0?

The chipram is already there and it's to slow to be usable for normal
memory.
Patches are available to get the APUS kernel booting with MOL enabled.

> - Which cpu flavors are used? MOL currently lacks 603(e) support,
> but I intend to add support for this processor shortly.

603/604.

bye, Roman

Re: MOL

[Samuel R. <sa...@ib...>]

On Thu, Apr 19, 2001 at 10:58:04PM +0200, Roman Zippel wrote:
> Hi,
> 
> Samuel Rydh wrote:
> 
> > - Does the BK kernels at fsmlabs contain working APUS support?
> 
> No, although I started to send Cort patches, but it takes ages... :(
> 
> > - Any special quirks? In particular, why is the kernel
> > *not* located at physical address 0?

The exception vectors then, are they 

> 
> The chipram is already there and it's to slow to be usable for normal
> memory. Patches are available to get the APUS kernel booting 
> with MOL enabled.

Actually, the hooks which used to be present in the 2.4 kernel
are no longer used by MOL. The latest kernel only exports a few
extra symbols. Instead, MOL hooks the very first instruction 
of each exception vector (this makes the low-level code 
system independent).

So... where are the exception vectors on APUS? Are there
branches from 0xN00 to fastmem? I guess this would imply
that the kernel image is always in the range of an absolute
branch (which would simplify things quite a bit).

Cheers,

/Samuel


----------------------------------------------------------
 E-mail <sa...@ib...>  WWW: <http://www.ibrium.se>
  Phone/fax: (home) +46 8 4418431, (work) +46 8 7908470   
----------------------------------------------------------

Re: MOL

[Roman Z. <zi...@li...>]

Hi,

Samuel Rydh wrote:

> So... where are the exception vectors on APUS? Are there
> branches from 0xN00 to fastmem? I guess this would imply
> that the kernel image is always in the range of an absolute
> branch (which would simplify things quite a bit).

The IP bit is set in MSR for APUS, so you end up at 0xfff0xx00, from
where we jump back to the kernel. I'm not sure if it's in the 26bit
range, (checking...), nope, the kernel is at 0x8000000.

bye, Roman

Re: MOL

[Glenn H. <gh...@c2...>]

Hello 

On 19-Apr-01, Samuel wrote:

> I'm currently considering porting MOL to other platforms
> than Linux/PPC (primarily *BSD and Darwin). Before doing that 
> however, I though it might be nice to have it working on APUS.

That would be great :-)

> To be able to add APUS support, I obviously need help with
> testing and debugging since I don't have an APUS box.

I'll be happy to do some testing :-)


- glenn

Re: MOL

[Alan B. <al...@ms...>]

hi,

> I'm currently considering porting MOL to other platforms
> than Linux/PPC (primarily *BSD and Darwin). Before doing that 
> however, I though it might be nice to have it working on APUS.

good idea :-)
 
> - Does the BK kernels at fsmlabs contain working APUS support?

dont know.... BK changes get integrated into the APUS CVS (held at
sourceforge)...i dont think theres a reversal.
 
> - Any special quirks? In particular, why is the kernel
> *not* located at physical address 0?

LinuxAPUS-capable Amigas are of two varieties.... those with 2Mb of
ChipRAM and those with 1Mb of ChipRAM. This ChipRAM is what used to make
the Amiga so very good at its sound and gfx..its memory addressable
directly by the custom chips...and therefore if the CPU has to use it, it
falls into contention...having to wait 2 clockcycles for access.. it also
runs at a slow rate and is of slow speed.  A large chunk of that ChipRAM
is also used by the framebuffer for display (if the Amiga concerned does
not have a gfx card) and for sound (ditto - no sound card installed). 
from memory (no pun intended) its 1.2Mb taken up for custom chip
utilisation on an 'AGA' (2Mb installed) machine.

FastRAM is memory only accessible by the CPU..hence its name. there is no
contention with the cusotm chips, its local to the CPU on the PowerPC
cards and its usually the 60ns flavour. 

memory access speeds (measured under AmigaOS) are something like  ~8Mb/s
for ChipRAM and 40Mb/s upwards (depending on operation) for FastRAM
(604 PowerPC cards on Amigas get around 80Mb/s iirC)

 > - Which cpu flavors are used? MOL currently lacks 603(e) support,

no!!!! argh. when did this drop out? Whilst I was looking at MOL there was
603(e) support...but the kernel required a run-time patching...and hooks
were installed into the kernel to allow this.

 > but I intend to add support for this processor shortly. > 

great! this is what I have.

> To be able to add APUS support, I obviously need help with
> testing and debugging since I don't have an APUS box.

I can help you out...having the required machine AND MacOS CD's lying next
to it.
 
alan