Re: [PATCH 1/2] mm: Enable khugepaged to operate on non-writable VMAs

[Dev Jain <dev.jain@arm.com>]

On 03/09/25 2:45 pm, Dev Jain wrote:
>
> On 03/09/25 1:38 pm, Wei Yang wrote:
>> On Wed, Sep 03, 2025 at 11:16:34AM +0530, Dev Jain wrote:
>>> Currently khugepaged does not collapse a region which does not have a
>>> single writable page. This is wasteful since non-writable VMAs 
>>> mapped by
>>> the application won't benefit from THP collapse. Therefore, remove this
>>> restriction and allow khugepaged to collapse a VMA with arbitrary
>>> protections.
>>>
>>> Along with this, currently MADV_COLLAPSE does not perform a collapse 
>>> on a
>>> non-writable VMA, and this restriction is nowhere to be found on the
>>> manpage - the restriction itself sounds wrong to me since the user 
>>> knows
>>> the protection of the memory it has mapped, so collapsing read-only
>>> memory via madvise() should be a choice of the user which shouldn't
>>> be overriden by the kernel.
>>>
>>> On an arm64 machine, an average of 5% improvement is seen on some 
>>> mmtests
>>> benchmarks, particularly hackbench, with a maximum improvement of 12%.
>>>
>>> Signed-off-by: Dev Jain <dev.jain@arm.com>
>>> ---
>> [...]
>>> mm/khugepaged.c | 9 ++-------
>>> 1 file changed, 2 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
>>> index 4ec324a4c1fe..a0f1df2a7ae6 100644
>>> --- a/mm/khugepaged.c
>>> +++ b/mm/khugepaged.c
>>> @@ -676,9 +676,7 @@ static int __collapse_huge_page_isolate(struct 
>>> vm_area_struct *vma,
>>>             writable = true;
>>>     }
>>>
>>> -    if (unlikely(!writable)) {
>>> -        result = SCAN_PAGE_RO;
>>> -    } else if (unlikely(cc->is_khugepaged && !referenced)) {
>> Would this cause more memory usage in system?
>>
>> For example, one application would fork itself many times. It 
>> executable area
>> is read only, so all of them share one copy in memory.
>>
>> Now we may collapse the range and create one copy for each process.
>>
>> Ok, we have max_ptes_shared, while if some ptes are none, could it 
>> still do
>> collapse?
>>
>> Maybe this is not realistic, just curious.
>
> Misunderstood your concern - you mean to say that a parent forks and 
> the children
> VMAs are read-only pointing to the pages which were mapped by parent. 
> Hmm.

I meant to say, writable VMAs with wrprotected ptes. Maybe after this 
patch, people

can finally make some real use of the max_ptes_shared tunable :)


>
>>
>>> +    if (unlikely(cc->is_khugepaged && !referenced)) {
>>>         result = SCAN_LACK_REFERENCED_PAGE;
>>>     } else {
>>>         result = SCAN_SUCCEED;
>>> @@ -1421,9 +1419,7 @@ static int hpage_collapse_scan_pmd(struct 
>>> mm_struct *mm,
>>>              mmu_notifier_test_young(vma->vm_mm, _address)))
>>>             referenced++;
>>>     }
>>> -    if (!writable) {
>>> -        result = SCAN_PAGE_RO;
>>> -    } else if (cc->is_khugepaged &&
>>> +    if (cc->is_khugepaged &&
>>>            (!referenced ||
>>>             (unmapped && referenced < HPAGE_PMD_NR / 2))) {
>>>         result = SCAN_LACK_REFERENCED_PAGE;
>>> @@ -2830,7 +2826,6 @@ int madvise_collapse(struct vm_area_struct 
>>> *vma, unsigned long start,
>>>         case SCAN_PMD_NULL:
>>>         case SCAN_PTE_NON_PRESENT:
>>>         case SCAN_PTE_UFFD_WP:
>>> -        case SCAN_PAGE_RO:
>>>         case SCAN_LACK_REFERENCED_PAGE:
>>>         case SCAN_PAGE_NULL:
>>>         case SCAN_PAGE_COUNT:
>>> -- 
>>> 2.30.2
>>>
>