Re: [PATCH v1] docs: filesystems: clarify KernelPageSize vs. MMUPageSize in smaps

["David Hildenbrand (Arm)" <david@kernel.org>]

On 3/4/26 21:22, Andi Kleen wrote:
> On Wed, Mar 04, 2026 at 04:56:36PM +0100, David Hildenbrand (Arm) wrote:
>> There was recently some confusion around THPs and the interaction with
>> KernelPageSize / MMUPageSize. Historically, these entries always
>> correspond to the smallest size we could encounter, not any current
>> usage of transparent huge pages or larger sizes used by the MMU.
> 
> It still seems like a bug to me, only documented now, but seems
> I'm in the minority on that.

I'd agree that something smarter and more future proof should have been
done if THPs wouldn't have been added (checks notes) 15 years ago and
how KernelPageSize / MMUPageSize behaves and what it means would already
be set in stone for a long time.

People decided to add things like "AnonHugePages" (which I also detest,
but here we are) instead.

I wish we would have never exposed MMUPageSize (in 2009) to handle some
ppc oddity.

While digging through some history, there were earlier conclusions
(around 10 years ago) that MMUPageSize [1] is not a good fit, and people
even tried exposing actual amounts (Ptes@4kB, Ptes@2MB). I don't think
that's the right approach either for various reasons raised already
(e.g., doesn't really tell you the full TLB/MMU story).

> 
> But anyways if you change this file you should probably remove
> the duplicated KernelPageSize/MMUPageSize entries in the example
> too. That triped me up the last time.

Good point. I also discovered that that PPC64 thing is still in place
(maybe it's dead code, but at least code-wise, it's not a thing from the
past ).

[1] https://lkml.org/lkml/2016/11/29/882

-- 
Cheers,

David