Re: + mm-page_alloc-free-pages-to-correct-buddy-list-after-pcp-lock-contention.patch added to mm-hotfixes-unstable branch

[Johannes Weiner <hannes@cmpxchg.org>]

On Tue, Sep 12, 2023 at 09:09:38AM -0700, Andrew Morton wrote:
> On Tue, 12 Sep 2023 09:50:29 -0400 Johannes Weiner <hannes@cmpxchg.org> wrote:
> 
> > > Cc: Johannes Weiner <hannes@cmpxchg.org>
> > > Cc: <stable@vger.kernel.org>
> > > Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> > 
> > This patch is superseded by the following patch you picked up:
> > mm-page_alloc-fix-cma-and-highatomic-landing-on-the-wrong-buddy-list.patch
> 
> OK.
> 
> > If you drop this patch here, you can also drop the fixlet to
> > free_unref_page(). The branch in there should look like this:
> > 
> > 	if (pcp)
> > 		free_unref_page_commit(..., pcpmigratetype, ...);
> > 	else
> > 		free_one_page(..., migratetype, ...);
> 
> Well kinda.  It's actually
> 
> 	if (pcp) {
> 		free_unref_page_commit(zone, pcp, page, migratetype, order);

							^^ pcpmigratetype

is what I was trying to highlight.

But yes, the pcp_spin_unlock() is needed too!

> 		pcp_spin_unlock(pcp);
> 	} else {
> 		free_one_page(zone, page, pfn, order, migratetype, FPI_NONE);
> 	}
>