On Fri, Mar 08, 2024 at 04:48:33PM +0800, Miaohe Lin wrote:
> On 2024/3/1 5:20, Matthew Wilcox (Oracle) wrote:
> > @@ -2277,8 +2277,8 @@ int memory_failure(unsigned long pfn, int flags)
> > }
> > }
> >
> > - hpage = compound_head(p);
> > - if (PageTransHuge(hpage)) {
> > + folio = page_folio(p);
> > + if (folio_test_large(folio)) {
> > /*
> > * The flag must be set after the refcount is bumped
> > * otherwise it may race with THP split.
[...]
> > @@ -2318,11 +2319,11 @@ int memory_failure(unsigned long pfn, int flags)
> > * race window. If this happens, we could try again to hopefully
> > * handle the page next round.
> > */
> > - if (PageCompound(p)) {
> > + if (folio_test_large(folio)) {
>
> folio_test_large() only checks whether PG_head is set but PageCompound() also checks PageTail().
> So folio_test_large() and PageCompound() are not equivalent?
Assuming we have a refcount on this page so it can't be simultaneously
split/freed/whatever, these three sequences are equivalent:
1 if (PageCompound(p))
2 struct page *head = compound_head(p);
2 if (PageHead(head))
3 struct folio *folio = page_folio(p);
3 if (folio_test_large(folio))
