Re: i386 initialization cleanup & freeing

Mark Hemment (markhe@nextd.demon.co.uk)
Sat, 12 Apr 1997 22:59:15 +0100 (BST)


Hi Martin/All,

On Sat, 12 Apr 1997, Martin Mares wrote:
> I've played a bit with the building procedure for i386:
>
> (1) Freeing of initialization code implemented in similar way to that
> currently present on Sparc. Added dummy headers for other archs =>
> adding of __initfunc to all generic code is now possible and welcome.

This is v. nice.

I'm having great fun with __initfunc (and __initdata) on the 2.1.33 kernel
(the patch applied cleanly to this - your original message didn't state a
kernel version).

> Other ideas to consider (comments welcome):

Only one (small) gripe. The boot message, which reports the memory used
by the kernel, now reports __initfunc functions as data. This is a bit
confusing.

> (b) Disabling writes to kernel code and R/O data -- maybe this could
> help catch a few bugs, but it's impossible to do with the Pentium
> 4MB pages.

Definitely _not_ worth it at the expensive of 4MB pages, unless it is made
a configurable option for when debugging support is needed (and in an
installation kernel).

> Please test this patch, especialy if you have a SMP machine... It's available
> on ftp://atrey.karlin.mff.cuni.cz/pub/local/mj/linux/patches/free.diff.

Sorry, I don't have an SMP box....

Thanks,

markhe

------------------------------------------------------------------
Mark Hemment, Unix/C Software Engineer (Contractor)
markhe@nextd.demon.co.uk http://www.nextd.demon.co.uk/
"Success has many fathers, failure is a B**TARD!" - anon
------------------------------------------------------------------