Re: [PATCH] mm: use PageAnon() and PageKsm() helpers in page_anon_vma()

["Kirill A. Shutemov" <kirill@shutemov.name>]

On Wed, Apr 01, 2015 at 03:06:53PM -0700, Andrew Morton wrote:
> On Thu, 2 Apr 2015 01:02:46 +0300 "Kirill A. Shutemov" <kirill@shutemov.name> wrote:
> 
> > On Wed, Apr 01, 2015 at 12:57:45PM -0700, Andrew Morton wrote:
> > > On Wed, 1 Apr 2015 14:50:54 +0300 "Kirill A. Shutemov" <kirill@shutemov.name> wrote:
> > > 
> > > > >From adc384977898173d65c2567fc5eb421da9b272e0 Mon Sep 17 00:00:00 2001
> > > > From: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
> > > > Date: Wed, 1 Apr 2015 14:33:56 +0300
> > > > Subject: [PATCH] mm: uninline and cleanup page-mapping related helpers
> > > > 
> > > > Most-used page->mapping helper -- page_mapping() -- has already
> > > > uninlined. Let's uninline also page_rmapping() and page_anon_vma().
> > > > It saves us depending on configuration around 400 bytes in text:
> > > > 
> > > >    text	   data	    bss	    dec	    hex	filename
> > > >  660318	  99254	 410000	1169572	 11d8a4	mm/built-in.o-before
> > > >  659854	  99254	 410000	1169108	 11d6d4	mm/built-in.o
> > > 
> > > Well, code size isn't the only thing to care about.  Some functions
> > > really should be inlined for performance reasons even if that makes the
> > > overall code larger.  But the changes you're proposing here look OK to
> > > me.
> > > 
> > > > As side effect page_anon_vma() now works properly on tail pages.
> > > 
> > > Let's fix the bug in a separate patch, please.  One which can be
> > > backported to earlier kernels if that should be needed.  ie: it should
> > > precede any uninlining.
> > 
> > The bug is not triggerable in current upsteam. AFAIK, we don't call
> > page_anon_vma() on tail pages of THP, since we don't map THP with PTEs
> > yet. For rest of cases we will get NULL, which is valid answer.
> 
> argh.  It rather helps if you can tell me when this happens (and which
> patch it fixes).  I sometimes spend quite a bit of time runnnig around
> in circles wondering what the heck tree/patch just got fixed.

Sorry for the mess.

> > Do you still want "page = compound_head(page);" line in separate patch?
> 
> I think that would be best.  That way the offending patch gets fixed
> and doesn't get bundled up with an unrelated change.

I've sent updated "mm: sanitize page->mapping for tail pages" patch that
care about this part. And this is uninlining patch on top of updated
patch.