Re: [PATCH v4 09/12] x86/mtrr: construct a memory map with cache modes

[Borislav Petkov]

On Wed, Mar 29, 2023 at 03:39:35PM +0200, Juergen Gross wrote:
> No. :-)

Because?

> The final form of the code is the result of an iterative process. :-)

I have a similar iterative process: until it hasn't been reviewed and
explained properly, this is not going anywhere.

So however you wanna do it, fine by me.

> I've reused the wording from cleanup.c (just above amd_special_default_mtrr()).

That got added with K8. K8 is ancient history so nothing magic about
that anymore. It is basically a bit in the SYSCFG MSR which says that

	[4G ... TOP_MEM2]

is WB.

> > Why not in mtrr_bp_init()? That is the first CPU.
> 
> Yeah, but generic_set_mtrr() can be called after boot, too.

That function sets a single MTRR register so you'd have to merge the
ranges, AFAICT. Not rebuild the whole map...

> Umm, not really. I want to do the copy even in the Xen PV case.

How about some comments? Or you're expecting me to be able to read your
mind?!

;-\

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette