Re: [Linux-visws-devel] Re: [PATCH] visws support for 2.5.59

[Andrey P. <pa...@or...>]

On Чтв, Янв 30, 2003 at 08:54:58 +0000, Christoph Hellwig wrote:
> Hi Andrey,
> 
> here some comments, I hope you'll submit visw support to Linus soon
Thank you for comments. I'll try to submit it some day.

> (and I'll find one somewhere.. :)
Please find a 540 if you can :))

> 	I don't think there's a reason to not allow UP kernels on visws.

Probably agreed. But I didn't even try to compile UP kernels for visws.
Added into todo list.

> 	IMHO the stext changes should be moved into a single patch in
> 	preparation of the visw support.
> 
> 	Dito for the i8259 and generic apic code changes (!?)
>  
> 	Again, GDT changes should be a separate preparation patch.

Agreed.
  
> @@ -11,8 +13,15 @@
>  ifdef	CONFIG_ACPI_PCI
>  obj-y		+= acpi.o
>  endif
> -obj-y		+= legacy.o
>  
> +ifndef	CONFIG_X86_VISWS
> +obj-y		+= legacy.o
> +endif
>  
>  endif		# CONFIG_X86_NUMAQ
> -obj-y		+= irq.o common.o
> +
> +ifndef	CONFIG_X86_VISWS
> +obj-y		+= irq.o
> +endif
> 
> What about grouping the two ifndefs?
 
I'l take a look at it.

>  
> +#ifdef CONFIG_X86_VISWS
> +#define USE_IO
> +#endif
> +
> 	Separate patch probably.  I think you should just submit this
> 	one to jgarzik ASAP, independant of the other bits.
> 	A small cleanup suggestion would be turning USE_IO into a config
> 	option that is implied by CONFIG_X86_VISWS...

I hope this kludge will go away. I need to figure out why memory mapped pci
io doesn't work on visws. Nobody knows is it fundamental visws hardware 
problem or the lithium asic needs more tweaking. And how to tweak asic
without documentation ? :-/

But people are reporting that pci video cards work under windows...
 
> diff -urN -X /usr/share/dontdiff linux-2.5.59.vanilla/drivers/video/sgivwfb.c linux-2.5.59/drivers/video/sgivwfb.c
> --- linux-2.5.59.vanilla/drivers/video/sgivwfb.c	Wed Jan 15 20:37:20 2003
> +++ linux-2.5.59/drivers/video/sgivwfb.c	Mon Jan 27 19:03:53 2003
> 
> 	The fb driver should be a separate patch again.  As it doesn't
> 	currently work I'd suggest submitting it to James Simmons ASAP.

The visws support is totally borken now, so why submit this ASAP ?

> 	I think this should be include/asm-i386/mach-visws/cobalt.h instead.
> 
> 	Dito (include/asm-i386/mach-visws/lithium.h)

Good point.
 
> diff -urN -X /usr/share/dontdiff linux-2.5.59.vanilla/include/asm-i386/sgi-piix.h linux-2.5.59/include/asm-i386/sgi-piix.h
> --- linux-2.5.59.vanilla/include/asm-i386/sgi-piix.h	Thu Jan  1 03:00:00 1970
> +++ linux-2.5.59/include/asm-i386/sgi-piix.h	Sun Jan 19 18:43:10 2003
> 
> 	What's really VisW-specific here (except that Linux doesn't poke the
> 	PIIX that much on normal PeeCees)?  IMHO it should be either
> 	include/asm-i386/piix.h or include/asm-i386/mach-visws/piix.h
> 	depending on that.

IIRC it contains some defines specific to ns super i/o chip and its 
connection to piix. So mach-visws/piix.h is better I think.

> +#ifdef CONFIG_X86_VISWS
> +#define x86_visws	1
> +#else
> +#define x86_visws	0
> +#endif
> 
> 	Do we really need more than CONFIG_X86_VISWS and the mach-$foo
> 	abstraction?

This varibable-like macro replaces two #ifndef's in smp.c and smpboot.c
I'm not sure are the really necessary and I can't test it because my visws
has one cpu only (and it's impossible to buy VRM and slot-1 P3 in Russia).

Best regards.

-- 
Andrey Panin		| Embedded systems software developer
pa...@or...	| PGP key: wwwkeys.pgp.net