Re: [PATCH v6 1/2] mm: migration: fix the FOLL_GET failure on following huge page

[Andrew Morton]

On Tue, 16 Aug 2022 10:21:00 +0800 Haiyue Wang <haiyue.wang@xxxxxxxxx> wrote:

> Not all huge page APIs support FOLL_GET option, so move_pages() syscall
> will fail to get the page node information for some huge pages.
> 
> Like x86 on linux 5.19 with 1GB huge page API follow_huge_pud(), it will
> return NULL page for FOLL_GET when calling move_pages() syscall with the
> NULL 'nodes' parameter, the 'status' parameter has '-2' error in array.
> 
> Note: follow_huge_pud() now supports FOLL_GET in linux 6.0.
>       Link: https://lore.kernel.org/all/20220714042420.1847125-3-naoya.horiguchi@xxxxxxxxx
> 
> But these huge page APIs don't support FOLL_GET:
>   1. follow_huge_pud() in arch/s390/mm/hugetlbpage.c

Let's tell the s390 maintainers.

>   2. follow_huge_addr() in arch/ia64/mm/hugetlbpage.c
>      It will cause WARN_ON_ONCE for FOLL_GET.

ia64 doesn't have maintainers :( Can we come up with something local to
arch/ia64 for this?

>   3. follow_huge_pgd() in mm/hugetlb.c

Hi, Mike.

> This is an temporary solution to mitigate the side effect of the race
> condition fix by calling follow_page() with FOLL_GET set for huge pages.
> 
> After supporting follow huge page by FOLL_GET is done, this fix can be
> reverted safely.
> 
> ...
>
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -1848,6 +1848,7 @@ static void do_pages_stat_array(struct mm_struct *mm, unsigned long nr_pages,
>  
>  	for (i = 0; i < nr_pages; i++) {
>  		unsigned long addr = (unsigned long)(*pages);
> +		unsigned int foll_flags = FOLL_DUMP;
>  		struct vm_area_struct *vma;
>  		struct page *page;
>  		int err = -EFAULT;
> @@ -1856,8 +1857,12 @@ static void do_pages_stat_array(struct mm_struct *mm, unsigned long nr_pages,
>  		if (!vma)
>  			goto set_status;
>  
> +		/* Not all huge page follow APIs support 'FOLL_GET' */
> +		if (!is_vm_hugetlb_page(vma))
> +			foll_flags |= FOLL_GET;
> +
>  		/* FOLL_DUMP to ignore special (like zero) pages */
> -		page = follow_page(vma, addr, FOLL_GET | FOLL_DUMP);
> +		page = follow_page(vma, addr, foll_flags);
>  
>  		err = PTR_ERR(page);
>  		if (IS_ERR(page))
> @@ -1865,7 +1870,8 @@ static void do_pages_stat_array(struct mm_struct *mm, unsigned long nr_pages,
>  
>  		if (page && !is_zone_device_page(page)) {
>  			err = page_to_nid(page);
> -			put_page(page);
> +			if (foll_flags & FOLL_GET)
> +				put_page(page);
>  		} else {
>  			err = -ENOENT;
>  		}

I would be better to fix this for real at those three client code sites?