Re: [tip:numa/core] mm: Count the number of pages affected in change_protection()

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 14 Nov 2012 06:25:56 -0800 tip-bot for Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> mm: Count the number of pages affected in change_protection()
> 
> This will be used by three kinds of purposes:
> 
>  - to optimize mprotect()
> 
>  - to speed up working set scanning for working set areas that
>    have not been touched
> 
>  - to more accurately scan per real working set
> 
> No change in functionality from this patch.

Seems simple, cheap and useful.

A bit lacking in interface comments though.  In particular, does the
return value (which is in units of PAGE_SIZE pages) represent the
number of pages which were inspected, or the number of pages which
actually had their permissions modified?

>From a quick look, it seems the answer is both.  Or neither.  If a
page's protection was written to but unaltered, that counts as a "1". 
But if a pmd couldn't be locked via __pmd_trans_huge_lock(), that
counts as a "0".

Anyway, please have a think about it, and describe exactly what these
return values mean?