May 14, 2024 at 8:40 PM, "Johannes Weiner" <hannes@xxxxxxxxxxx> wrote:
>
> On Tue, May 14, 2024 at 02:59:33PM +0800, Yajun Deng wrote:
>
> >
> > The first parameter passed to set_pageblock_migratetype should be the
> >
> > buddy, not the page. Let's change it back correctly.
> >
>
> No it shouldn't be.
>
> >
> > Fixes: fd919a85cd55 ("mm: page_isolation: prepare for hygienic freelists")
> >
> > Signed-off-by: Yajun Deng <yajun.deng@xxxxxxxxx>
> >
> > ---
> >
> > mm/page_alloc.c | 2 +-
> >
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> >
> >
> > diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> >
> > index cd584aace6bf..5422f6f8975d 100644
> >
> > --- a/mm/page_alloc.c
> >
> > +++ b/mm/page_alloc.c
> >
> > @@ -1728,7 +1728,7 @@ bool move_freepages_block_isolate(struct zone *zone, struct page *page,
> >
> >
> >
> > del_page_from_free_list(buddy, zone, order,
> >
> > get_pfnblock_migratetype(buddy, pfn));
> >
> > - set_pageblock_migratetype(page, migratetype);
> >
> > + set_pageblock_migratetype(buddy, migratetype);
> >
> > split_large_buddy(zone, buddy, pfn, order);
> >
>
> [ pageblock | pageblock ]
>
> ^ ^
>
> buddy page
>
> We're trying to isolate the second pageblock, but it's part of a
>
> bigger buddy. Remove buddy from the free list, update the migratetype
>
> of page, then split and free the chunks back. This puts the first
>
> pageblock back on the original freelist and the second block onto the
>
> isolate list.
>
> In the else branch, it's the first block we're interested in:
>
> [ pageblock | pageblock ]
>
> ^
>
> page
>
> So again, take the whole thing off, but now mark the first block
>
> isolate, then split and free.
>
Got it, thank you for your detailed explanation.
