Re: [PATCH 3/9] x86/mwait-idle: latch struct idle_cpu contents

["Roger Pau Monné" <roger.pau@citrix.com>]

On Thu, Mar 12, 2026 at 05:55:22PM +0100, Jan Beulich wrote:
> Rather than storing a pointer (and needing to keep all struct instances in
> memory post-init), and rather than (like the Linux counterpart has it)
> keeping individual variables, simply copy the respective structure
> instance. By implication, subsequent updates now need doing to the copy.

Shouldn't the aim here be to move the myriad of per-arch cpuidle_state
arrays to the init section, so we can get rid of them after boot?

Overall I wonder whether we would rather attempt to sya in-sync with
what Linux does, simply because it's then easier to pick updates fro
Linux, and less likely to introduce bugs as part of the modification
needs to adapt the Linux code into our fork.

Not saying the change is bad, just wondering whether in general we
would better try to keep the code bases as similar as feasible.

Thanks, Roger.