On Fri, 2019-03-22 at 08:41 +0500, Mikhail Gavrilov wrote:
> On Fri, 22 Mar 2019 at 00:14, Qian Cai <cai@xxxxxx> wrote:
> >
> >
> > That is OK. The above debug patch may still be useful to figure out where
> > those
> > pages come from (or you could add those 3 pages address to the patch as
> > well).
> > They may be initialized in a similar fashion or uninitialized to begin with.
>
> Strange I modified patch for catch all 0xffffXXXXX07ce000 pages
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 03fcf73..8808e2a 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -1273,6 +1273,10 @@ static void free_one_page(struct zone *zone,
> static void __meminit __init_single_page(struct page *page, unsigned long
> pfn,
> unsigned long zone, int nid)
> {
> + if (0xffff00000fffffff & page == (void *)0xffff0000007ce000) {
> + printk("KK page = %px\n", page);
> + dump_stack();
> + }
> mm_zero_struct_page(page);
> set_page_links(page, zone, nid, pfn);
> init_page_count(page);
Those pages are not initialized at all which likely mean that memblock did not
even allocate them at the first place, so Mel's patch might work.
