Re: [PATCH 1/4] rmap: introduce rmap_walk_locked()

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

On Thu, 4 Feb 2016 01:45:07 +0300 "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> wrote:

> On Wed, Feb 03, 2016 at 02:40:19PM -0800, Andrew Morton wrote:
> > On Wed,  3 Feb 2016 18:14:16 +0300 "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> wrote:
> > 
> > > rmap_walk_locked() is the same as rmap_walk(), but caller takes care
> > > about relevant rmap lock. It only supports anonymous pages for now.
> > > 
> > > It's preparation to switch THP splitting from custom rmap walk in
> > > freeze_page()/unfreeze_page() to generic one.
> > > 
> > > ...
> > >
> > > +/* Like rmap_walk, but caller holds relevant rmap lock */
> > > +int rmap_walk_locked(struct page *page, struct rmap_walk_control *rwc)
> > > +{
> > > +	/* only for anon pages for now */
> > > +	VM_BUG_ON_PAGE(!PageAnon(page) || PageKsm(page), page);
> > > +	return rmap_walk_anon(page, rwc, true);
> > > +}
> > 
> > Should be rmap_walk_anon_locked()?
> 
> I leave interface open for further extension for file mappings, once it
> will be needed. Interface is mirroring plain rmap_walk()

hm, yes, I see.

> If you prefer to rename the function, I can do it too.

Well, what does "unlocked" mean in the context of rmap_walk_ksm() and
rmap_walk_file()?  That the caller holds totally different locks.  I
expect that sitting down and writing out the interface definition for
such an rmap_walk_locked() would reveal that we shouldn't have created
it.

I mean, if the caller is to call such an rmap_walk_locked(), he first
needs to work out if it's a ksm page or an anon page or a file page,
then take the appropriate lock and then call rmap_walk_locked(). 
That's silly - at this point he should directly call
rmap_walk_ksm_locked()?

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>