Re: [patch] arch/i386/setup.S bug

[Tigran Aivazian]

On Tue, 4 Jan 2000, Chris Noe wrote:
> Yep, definitely my bad. The problem that I saw was that although movsd's
> are generally always a win over movsw's, throughout the bootcode we're in
> .code16 which means we get an addr32 prefix byte on each one (with gas;
> I'm assuming as86 did as well).

I realized that. In fact, as soon as I saw those movsw back in there I
quickly changed them back to movsd and send a patch to Linus. That was my
mistake because with .code16 it wouldn't even compile (and I always shout
that "test your patches before sending to Linus!" :). However, all-patient
and all-kind Linus didn't chastise me for that but instead took the time
to explain why we should use movsw instead of movsd in bootsect.S. Thanks
to Linus - at the end it always turns out that he was right and we (or I)
was just too impatient to figure everything out on my own.

Having said that, however, in protected mode, (in kernel/head.S) we should
really use movsd and not movsb. I sent a patch to Linus (and then
also to Alan) but so far it is refused. Can you please have a look at this
and see what is wrong with it:

   http://www.ocston.org/~tigran/patches/head-2.3.36-p6.patch

it is just copying command line and as it is huge (2048 bytes) I suppose
the faster the better.

Thanks,
------
Tigran A. Aivazian           | http://www.sco.com
Escalations Research Group   | tel: [phone]
Santa Cruz Operation Ltd     | http://www.ocston.org/~tigran