Re: [PATCH v1 5/5] mm, hwpoison: enable memory error handling on 1GB hugepage

["HORIGUCHI NAOYA(堀口 直也)" <naoya.horiguchi@nec.com>]

On Tue, Jun 07, 2022 at 10:11:24PM +0800, Miaohe Lin wrote:
> On 2022/6/2 13:06, Naoya Horiguchi wrote:
> > From: Naoya Horiguchi <naoya.horiguchi@nec.com>
> > 
> > Now error handling code is prepared, so remove the blocking code and
> > enable memory error handling on 1GB hugepage.
> > 
> 
> I'm nervous about this change. It seems there are many code paths not awared of pud swap entry.
> I browsed some of them:
> apply_to_pud_range called from apply_to_page_range:
> 
> apply_to_pud_range:
> 	next = pud_addr_end(addr, end);
> 	if (pud_none(*pud) && !create)
> 		continue;
> 	if (WARN_ON_ONCE(pud_leaf(*pud)))
> 		return -EINVAL;
> 	if (!pud_none(*pud) && WARN_ON_ONCE(pud_bad(*pud))) {
> 		if (!create)
> 			continue;
> 		pud_clear_bad(pud);
> 	}
> 	err = apply_to_pmd_range(mm, pud, addr, next,
> 				 fn, data, create, mask);
> 
> For !pud_present case, it will mostly reach apply_to_pmd_range and call pmd_offset on it. And invalid
> pointer will be de-referenced.

apply_to_pmd_range() has BUG_ON(pud_huge(*pud)) and apply_to_pte_range() has
BUG_ON(pmd_huge(*pmd)), so this page table walking code seems to not expect
to handle pmd/pud level mapping.

> 
> Another example might be copy_pud_range and so on. So I think it might not be prepared to enable the
> 1GB hugepage or all of these places should be fixed?

I think that most of page table walker for user address space should first
check is_vm_hugetlb_page() and call hugetlb specific walking code for vma
with VM_HUGETLB.
copy_page_range() is a good example.  It calls copy_hugetlb_page_range()
for vma with VM_HUGETLB and the function should support hwpoison entry.
But I feel that I need testing for confirmation.

And I'm not sure that all other are prepared for non-present pud-mapping,
so I'll need somehow code inspection and testing for each.

Thanks,
Naoya Horiguchi