[PATCH: NEW ARCHITECTURE FOR 2.5.3] support for NCR voyager

[PATCH: NEW ARCHITECTURE FOR 2.5.3] support for NCR voyager

[James Bottomley]

This is essentially the same as the 2.5.2 patch except that it adds support 
for the new migration cross processor interrupt.

As far as the new scheduler goes, the process affinity properties are much 
better for the voyager, which is configured to have a fairly huge L3 cache 
shared by several CPUs.  (My current voyager has 2 CPU cards with 4 pentium 
CPUs each and 32Mb of L3 cache on each card).

The patch (157k) is available here

http://www.hansenpartnership.com/voyager/files/voyager-2.5.3.diff

If there's any interest from the other architecture groups, I can also put 
together the incremental diff between 2.5.2 and 2.5.3 which shows what changes 
needed to be made in the arch specific boot sequence and smp code for the new 
scheduler.

James Bottomley

Re: [PATCH: NEW ARCHITECTURE FOR 2.5.3] support for NCR voyager

[Pavel Machek]

Hi!

> This is essentially the same as the 2.5.2 patch except that it adds support 
> for the new migration cross processor interrupt.
> 
> As far as the new scheduler goes, the process affinity properties are much 
> better for the voyager, which is configured to have a fairly huge L3 cache 
> shared by several CPUs.  (My current voyager has 2 CPU cards with 4 pentium 
> CPUs each and 32Mb of L3 cache on each card).

Maybe you should stop calling it new architecture? Its mostly
i386-compatible, right?

BTW are those "current" machines, or is their production already
stopped?
									Pavel

-- 
(about SSSCA) "I don't say this lightly.  However, I really think that the U.S.
no longer is classifiable as a democracy, but rather as a plutocracy." --hpa

Re: [PATCH: NEW ARCHITECTURE FOR 2.5.3] support for NCR voyager

[James Bottomley]

pavel@suse.cz said:
> Maybe you should stop calling it new architecture? Its mostly
> i386-compatible, right? 

It's x86 based, and like all good x86 machines, it has a mode where it will 
boot on a single CPU and look like a microchannel PC.  However, the voyagers 
were designed to be large multi-CPU SMP machines.  For SMP, it is completely 
incompatible with the APIC/IO-APIC architecture which linux supports, so the 
voyager patches provide a completely different SMP HAL, which had to be 
written from scratch; it's not merely a modification of the existing one.

> BTW are those "current" machines, or is their production already
> stopped? 

I believe the last data centre machines (the 32 CPU 51xx series) rolled off 
NCR's production lines in 2001, so it is pretty much an obsolete machine now.

James

Re: [PATCH: NEW ARCHITECTURE FOR 2.5.3] support for NCR voyager

[Andrey Panin]

[-- Attachment #1: Type: text/plain, Size: 564 bytes --]


Hi,

can you please explain, what does this config.in fragment mean:

if [ "$CONFIG_VISWS" != "y" ]; then
   bool 'MCA support' CONFIG_MCA
   if [ "$CONFIG_MCA" = "y" ]; then
	bool '   Support for the NCR Voyager Architecture' CONFIG_VOYAGER
	define_bool CONFIG_X86_TSC n
   fi
else
   define_bool CONFIG_MCA n
fi

How MCA and NCR Voyager support related to SGI Visual Workstations support
(CONFIG_VISWS) ?

Best regards.

-- 
Andrey Panin            | Embedded systems software engineer
pazke@orbita1.ru        | PGP key: wwwkeys.eu.pgp.net

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: [PATCH: NEW ARCHITECTURE FOR 2.5.3] support for NCR voyager

[Dave Jones]

On Mon, Feb 11, 2002 at 12:00:21PM +0300, Andrey Panin wrote:

 > if [ "$CONFIG_VISWS" != "y" ]; then
 >    bool 'MCA support' CONFIG_MCA
 >    if [ "$CONFIG_MCA" = "y" ]; then
 > 	bool '   Support for the NCR Voyager Architecture' CONFIG_VOYAGER
 > 
 > How MCA and NCR Voyager support related to SGI Visual Workstations support
 > (CONFIG_VISWS) ?

 It is kinda horrible. What I parsed it as is VISWS and Voyager
 are mutually exclusive, both have MCA available, but Voyager
 is a 'must have' case.

 icky.

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: [PATCH: NEW ARCHITECTURE FOR 2.5.3] support for NCR voyager

[Alan Cox]

> can you please explain, what does this config.in fragment mean:
> 
> if [ "$CONFIG_VISWS" !=3D "y" ]; then
>    bool 'MCA support' CONFIG_MCA
>    if [ "$CONFIG_MCA" =3D "y" ]; then
> 	bool '   Support for the NCR Voyager Architecture' CONFIG_VOYAGER
> 	define_bool CONFIG_X86_TSC n
>    fi
> else
>    define_bool CONFIG_MCA n
> fi
> 
> How MCA and NCR Voyager support related to SGI Visual Workstations support
> (CONFIG_VISWS) ?

Exactly as it says. If its a VISWS machine it can't be a voyager
If it hasn't got MCA bus it can't be a voyager
If it is a VISWS it doesnt have MCA
 

Re: [PATCH: NEW ARCHITECTURE FOR 2.5.3] support for NCR voyager

[James Bottomley]

It's historical creep.

Initially the question about being Voyager was asked immediately below 
CONFIG_MCA, which was then at the top level.  Then that got wrapped by if != 
VISW, so it ends up looking like this.

How about:

if [ "$CONFIG_VISWS" != "y" ]; then
    bool 'MCA support' CONFIG_MCA
else
    define_bool CONFIG_MCA n
fi

if [ "$CONFIG_MCA" = "y" ]; then
    bool '   Support for the NCR Voyager Architecture' CONFIG_VOYAGER
    define_bool CONFIG_X86_TSC n
fi

Actually, this also exposes a bug, the last statement should be:

if [ "$CONFIG_MCA" = "y" ]; then
    bool '   Support for the NCR Voyager Architecture' CONFIG_VOYAGER
    if [ "$CONFIG_VOYAGER" = "y"]; then
        define_bool CONFIG_X86_TSC n
    fi
fi

Since MCA machines may use the pentium TSC but voyager may not.

James