On Fri, May 17, 2024 at 02:36:29PM +0200, Hannes Reinecke wrote:
> > +#define ZERO_FSB_SIZE (65536)
> > +#define ZERO_FSB_ORDER (get_order(ZERO_FSB_SIZE))
> > +extern struct page *zero_fs_block;
> > +
> > /*
> > * char_dev.c
> > */
> But why?
> We already have a perfectly fine hugepage zero page in huge_memory.c.
> Shouldn't we rather export that one and use it?
> (Actually I have some patches for doing so...)
But we don't necessarily. We only have it if
do_huge_pmd_anonymous_page() satisfies:
if (!(vmf->flags & FAULT_FLAG_WRITE) &&
!mm_forbids_zeropage(vma->vm_mm) &&
transparent_hugepage_use_zero_page()) {
ie we've taken a page fault on a PMD hole in a VMA, that VMA permits
PMD mappings to exist, the page fault was for read, the
forbids-huge-zeropage isn't set for this vma, and using the hugetlb zero
page isn't forbidden.
I'd like to see stats for how much the PMD-zero-page is actually used,
because I suspect it's not really used very much.
