Re: [PATCH mm-new v3 1/1] mm/khugepaged: abort collapse scan on non-swap entries

[Zi Yan <ziy@nvidia.com>]

On 10 Oct 2025, at 11:34, Lance Yang wrote:

> On 2025/10/10 23:21, Zi Yan wrote:
>> On 7 Oct 2025, at 23:26, Lance Yang wrote:
>>
>>> From: Lance Yang <lance.yang@linux.dev>
>>>
>>> Currently, special non-swap entries (like PTE markers) are not caught
>>> early in hpage_collapse_scan_pmd(), leading to failures deep in the
>>> swap-in logic.
>>>
>>> A function that is called __collapse_huge_page_swapin() and documented
>>> to "Bring missing pages in from swap" will handle other types as well.
>>>
>>> As analyzed by David[1], we could have ended up with the following
>>> entry types right before do_swap_page():
>>>
>>>    (1) Migration entries. We would have waited.
>>>        -> Maybe worth it to wait, maybe not. We suspect we don't stumble
>>>           into that frequently such that we don't care. We could always
>>>           unlock this separately later.
>>>
>>>    (2) Device-exclusive entries. We would have converted to non-exclusive.
>>>        -> See make_device_exclusive(), we cannot tolerate PMD entries and
>>>           have to split them through FOLL_SPLIT_PMD. As popped up during
>>>           a recent discussion, collapsing here is actually
>>>           counter-productive, because the next conversion will PTE-map
>>>           it again.
>>>        -> Ok to not collapse.
>>>
>>>    (3) Device-private entries. We would have migrated to RAM.
>>>        -> Device-private still does not support THPs, so collapsing right
>>>           now just means that the next device access would split the
>>>           folio again.
>>>        -> Ok to not collapse.
>>>
>>>    (4) HWPoison entries
>>>        -> Cannot collapse
>>>
>>>    (5) Markers
>>>        -> Cannot collapse
>>>
>>> First, this patch adds an early check for these non-swap entries. If
>>> any one is found, the scan is aborted immediately with the
>>> SCAN_PTE_NON_PRESENT result, as Lorenzo suggested[2], avoiding wasted
>>> work. While at it, convert pte_swp_uffd_wp_any() to pte_swp_uffd_wp()
>>> since we are in the swap pte branch.
>>>
>>> Second, as Wei pointed out[3], we may have a chance to get a non-swap
>>> entry, since we will drop and re-acquire the mmap lock before
>>> __collapse_huge_page_swapin(). To handle this, we also add a
>>> non_swap_entry() check there.
>>>
>>> Note that we can unlock later what we really need, and not account it
>>> towards max_swap_ptes.
>>>
>>> [1] https://lore.kernel.org/linux-mm/09eaca7b-9988-41c7-8d6e-4802055b3f1e@redhat.com
>>> [2] https://lore.kernel.org/linux-mm/7df49fe7-c6b7-426a-8680-dcd55219c8bd@lucifer.local
>>> [3] https://lore.kernel.org/linux-mm/20251005010511.ysek2nqojebqngf3@master
>>>
>>> Acked-by: David Hildenbrand <david@redhat.com>
>>> Reviewed-by: Wei Yang <richard.weiyang@gmail.com>
>>> Reviewed-by: Dev Jain <dev.jain@arm.com>
>>> Suggested-by: David Hildenbrand <david@redhat.com>
>>> Suggested-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
>>> Signed-off-by: Lance Yang <lance.yang@linux.dev>
>>> ---
>>> v2 -> v3:
>>>   - Collect Acked-by from David - thanks!
>>>   - Collect Reviewed-by from Wei and Dev - thanks!
>>>   - Add a non_swap_entry() check in __collapse_huge_page_swapin() (per Wei
>>>     and David) - thanks!
>>>   - Rework the changelog to incorporate David's detailed analysis of
>>>     non-swap entry types - thanks!!!
>>>   - https://lore.kernel.org/linux-mm/20251001032251.85888-1-lance.yang@linux.dev/
>>>
>>> v1 -> v2:
>>>   - Skip all non-present entries except swap entries (per David) thanks!
>>>   - https://lore.kernel.org/linux-mm/20250924100207.28332-1-lance.yang@linux.dev/
>>>
>>>   mm/khugepaged.c | 37 +++++++++++++++++++++++--------------
>>>   1 file changed, 23 insertions(+), 14 deletions(-)
>>>
>>> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
>>> index abe54f0043c7..bec3e268dc76 100644
>>> --- a/mm/khugepaged.c
>>> +++ b/mm/khugepaged.c
>>> @@ -1020,6 +1020,11 @@ static int __collapse_huge_page_swapin(struct mm_struct *mm,
>>>   		if (!is_swap_pte(vmf.orig_pte))
>>>   			continue;
>>>
>>> +		if (non_swap_entry(pte_to_swp_entry(vmf.orig_pte))) {
>>> +			result = SCAN_PTE_NON_PRESENT;
>>> +			goto out;
>>> +		}
>>> +
>>>   		vmf.pte = pte;
>>>   		vmf.ptl = ptl;
>>>   		ret = do_swap_page(&vmf);
>>> @@ -1281,7 +1286,23 @@ static int hpage_collapse_scan_pmd(struct mm_struct *mm,
>>>   	for (addr = start_addr, _pte = pte; _pte < pte + HPAGE_PMD_NR;
>>>   	     _pte++, addr += PAGE_SIZE) {
>>>   		pte_t pteval = ptep_get(_pte);
>>> -		if (is_swap_pte(pteval)) {
>>> +		if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
>>> +			++none_or_zero;
>>> +			if (!userfaultfd_armed(vma) &&
>>> +			    (!cc->is_khugepaged ||
>>> +			     none_or_zero <= khugepaged_max_ptes_none)) {
>>> +				continue;
>>> +			} else {
>>> +				result = SCAN_EXCEED_NONE_PTE;
>>> +				count_vm_event(THP_SCAN_EXCEED_NONE_PTE);
>>> +				goto out_unmap;
>>> +			}
>>> +		} else if (!pte_present(pteval)) {
>>> +			if (non_swap_entry(pte_to_swp_entry(pteval))) { +				result = SCAN_PTE_NON_PRESENT;
>>> +				goto out_unmap;
>>> +			}
>>> +
>>>   			++unmapped;
>>>   			if (!cc->is_khugepaged ||
>>>   			    unmapped <= khugepaged_max_ptes_swap) {
>>> @@ -1290,7 +1311,7 @@ static int hpage_collapse_scan_pmd(struct mm_struct *mm,
>>>   				 * enabled swap entries.  Please see
>>>   				 * comment below for pte_uffd_wp().
>>>   				 */
>>> -				if (pte_swp_uffd_wp_any(pteval)) {
>>> +				if (pte_swp_uffd_wp(pteval)) {
>>
>> pte_swp_uffd_wp_any() returns true for both pte_swp_uffd_wp() and
>> pte_marker_uffd_wp(). Why is it OK to just check pte_swp_uffd_wp() here?
>
> +		} else if (!pte_present(pteval)) {
> +			if (non_swap_entry(pte_to_swp_entry(pteval))) { <--
> +				result = SCAN_PTE_NON_PRESENT;
> +				goto out_unmap;
> +			}
>
> IIUC, we have just handled all non-swap entries above (which would
> include pte_marker_uffd_wp()), right?

Got it. Thanks for the explanation.

Reviewed-by: Zi Yan <ziy@nvidia.com>

--
Best Regards,
Yan, Zi