Re: [PATCH][RFC] SGI VISWS support for 2.5

[Andrey Panin <pazke@orbita1.ru>]

[-- Attachment #1: Type: text/plain, Size: 1881 bytes --]

On Чтв, Июн 20, 2002 at 02:37:43 -0700, Jesse Barnes wrote:
> I gave it a go on a 320 that I have here and it gets into start_kernel
> somewhere, but I don't get any output on the serial console.  If you'd
> like to get it into 2.5, maybe we should clean it up a little first
> (as well as get it working)?

We definetely need to make it working and cleanup is a good thing too,
but i dont have a VISWS (right now) so i'm limited to trivial mergework.

> When I ported the last patch forward to 2.4.17, I noticed that there
> were quite a few parts that could use some work.  Looking through the
> patch, it seems like the head.S stuff could be done in a better way,

As far as i understand VISWS fragment in the head.S does 3 thing:
	1) fills page tables;
	2) twiddle some bits in cr3.
	3) loads gdt and idt;
My small brain can't understand why we must do 1 and 3 so early.
BTW looks like we are filling page tables many times (for every CPU).
And why we move stext and _stext for VISWS ?
These are great mysteries ...

> maybe a seperate head.S for machines that start in protected mode?
I'm doubtfull, does it worth the effort ?

> Then again, maybe the boot loader should just be changed? The
> interrupt handler code also needs to be fixed up.
> I haven't looked at the patch to seperate out i386 subarches yet, but
> maybe that would be a good first step to abstracting away some of the
> visws setup code?

2.5 with James's arch-split patch has most of VISWS code separated already.
The only big trouble that remains is VISWS hacks in parse_mem_cmdline() 
function.

Also when 2.5 will realy use platform driver to support reboot, halt
and power off, it will be possible to split VISWS code from process.c.

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

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