On Fri, Mar 20, 2026 at 12:03:30PM -0700, Dave Hansen wrote:
> This is old cruft, but it appears that having two copies of these
> MSR functions is enabling warnings to creep in[1].
>
> I know there's also been some work to pare down the XXL code, but
> it's obviously not merged yet and this is a good baby step.
>
> Create helpers that both paravirt and native can use in common code
> and remove the paravirt implementations of the helpers. This reduces
> the amount of logic that is duplicated in the paravirt code.
>
> The other thing I really like about this is that it puts the
> raw=>{native,paravirt} switch in one compact place in the code.
>
> Conceptually:
> - native: The bare-metal implementation. Might not be usable under
> paravirt XXL.
> - raw: The lowest-level function that is always usable. Might
> be native or paravirt under the hood.
I went through the patchset twice and I kinda get what you're trying to do but
the "raw" thing is confusing as hell.
To me "raw" means, the lowest level of the functionality - something like
__<function_name> with the two underscores. Or three, depending on the
indirection levels.
And those do *only* *raw* instructions - no more indirections.
But then how can "raw" be the lowest level and then still have something else
underneath - native_ and paravirt_?
I *think* this is only a naming issue and with "raw_" you probably wanna say
"native_or_paravirt_" depending on the ifdeffery... but shorter...
If so, I wouldn't call it "raw". I'd say
xx_read_msr()
xx_write_msr()
to denote that the "xx" resolves to either of the two types. But a better
name. I can't think of a good one now but I know that "raw" isn't it...
Hmmm.
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette
