On Fri, Oct 04, 2024 at 03:09:22PM +0200, Miguel Ojeda wrote: > Hi Matthew, > > I noticed riscv64 defconfig seems to reproducibly hit > `VM_BUG_ON_PAGE(PageAnonNotKsm(page))` in today's next-20241004 -- > please see below. > > I hope that helps! I can also reproduce this on arm64 with CONFIG_DEBUG_VM_PGFLAGS=y. Looks like an invert logic bug from: https://lore.kernel.org/linux-mm/20241002152533.1350629-5-willy@xxxxxxxxxxxxx/ I made the below changes and the problem goes away. Best regards, Nam diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h index fe9abdf26ab0..ccf3c78faefc 100644 --- a/include/linux/page-flags.h +++ b/include/linux/page-flags.h @@ -1131,14 +1131,14 @@ static __always_inline int PageAnonExclusive(const struct page *page) static __always_inline void SetPageAnonExclusive(struct page *page) { - VM_BUG_ON_PGFLAGS(PageAnonNotKsm(page), page); + VM_BUG_ON_PGFLAGS(!PageAnonNotKsm(page), page); VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); } static __always_inline void ClearPageAnonExclusive(struct page *page) { - VM_BUG_ON_PGFLAGS(PageAnonNotKsm(page), page); + VM_BUG_ON_PGFLAGS(!PageAnonNotKsm(page), page); VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); }
