Message-ID: <CACZaFFPCZfxG+FK9rrefFi1UmtWxiOf2j6KXx+9JDJfcHBz8ww@mail.gmail.com>
Date: Fri, 23 Jan 2026 23:08:51 +0800
From: Vernon Yang <vernon2gm@...il.com>
To: Lance Yang <lance.yang@...ux.dev>
Cc: lorenzo.stoakes@...cle.com, ziy@...dia.com, dev.jain@....com, 
	baohua@...nel.org, linux-mm@...ck.org, linux-kernel@...r.kernel.org, 
	david@...nel.org, Vernon Yang <yanglincheng@...inos.cn>, akpm@...ux-foundation.org
Subject: Re: [PATCH mm-new v5 4/5] mm: khugepaged: skip lazy-free folios

On Fri, Jan 23, 2026 at 5:09 PM Lance Yang <lance.yang@...ux.dev> wrote:
>
> On 2026/1/23 16:22, Vernon Yang wrote:
> > From: Vernon Yang <yanglincheng@...inos.cn>
> >
> > For example, create three task: hot1 -> cold -> hot2. After all three
> > task are created, each allocate memory 128MB. the hot1/hot2 task
> > continuously access 128 MB memory, while the cold task only accesses
> > its memory briefly andthen call madvise(MADV_FREE). However, khugepaged
> > still prioritizes scanning the cold task and only scans the hot2 task
> > after completing the scan of the cold task.
> >
> > And if we collapse with a lazyfree page, that content will never be none
> > and the deferred shrinker cannot reclaim them.
> >
> > So if the user has explicitly informed us via MADV_FREE that this memory
> > will be freed, it is appropriate for khugepaged to skip it only, thereby
> > avoiding unnecessary scan and collapse operations to reducing CPU
> > wastage.
> >
> > Here are the performance test results:
> > (Throughput bigger is better, other smaller is better)
> >
> > Testing on x86_64 machine:
> >
> > | task hot2           | without patch | with patch    |  delta  |
> > |---------------------|---------------|---------------|---------|
> > | total accesses time |  3.14 sec     |  2.93 sec     | -6.69%  |
> > | cycles per access   |  4.96         |  2.21         | -55.44% |
> > | Throughput          |  104.38 M/sec |  111.89 M/sec | +7.19%  |
> > | dTLB-load-misses    |  284814532    |  69597236     | -75.56% |
> >
> > Testing on qemu-system-x86_64 -enable-kvm:
> >
> > | task hot2           | without patch | with patch    |  delta  |
> > |---------------------|---------------|---------------|---------|
> > | total accesses time |  3.35 sec     |  2.96 sec     | -11.64% |
> > | cycles per access   |  7.29         |  2.07         | -71.60% |
> > | Throughput          |  97.67 M/sec  |  110.77 M/sec | +13.41% |
> > | dTLB-load-misses    |  241600871    |  3216108      | -98.67% |
> >
> > Signed-off-by: Vernon Yang <yanglincheng@...inos.cn>
> > ---
> >   include/trace/events/huge_memory.h |  1 +
> >   mm/khugepaged.c                    | 11 +++++++++++
> >   2 files changed, 12 insertions(+)
> >
> > diff --git a/include/trace/events/huge_memory.h b/include/trace/events/huge_memory.h
> > index 384e29f6bef0..bcdc57eea270 100644
> > --- a/include/trace/events/huge_memory.h
> > +++ b/include/trace/events/huge_memory.h
> > @@ -25,6 +25,7 @@
> >       EM( SCAN_PAGE_LRU,              "page_not_in_lru")              \
> >       EM( SCAN_PAGE_LOCK,             "page_locked")                  \
> >       EM( SCAN_PAGE_ANON,             "page_not_anon")                \
> > +     EM( SCAN_PAGE_LAZYFREE,         "page_lazyfree")                \
> >       EM( SCAN_PAGE_COMPOUND,         "page_compound")                \
> >       EM( SCAN_ANY_PROCESS,           "no_process_for_page")          \
> >       EM( SCAN_VMA_NULL,              "vma_null")                     \
> > diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> > index de95029e3763..be1c09842ea2 100644
> > --- a/mm/khugepaged.c
> > +++ b/mm/khugepaged.c
> > @@ -46,6 +46,7 @@ enum scan_result {
> >       SCAN_PAGE_LRU,
> >       SCAN_PAGE_LOCK,
> >       SCAN_PAGE_ANON,
> > +     SCAN_PAGE_LAZYFREE,
> >       SCAN_PAGE_COMPOUND,
> >       SCAN_ANY_PROCESS,
> >       SCAN_VMA_NULL,
> > @@ -583,6 +584,11 @@ static enum scan_result __collapse_huge_page_isolate(struct vm_area_struct *vma,
> >               folio = page_folio(page);
> >               VM_BUG_ON_FOLIO(!folio_test_anon(folio), folio);
> >
> > +             if (!pte_dirty(pteval) && folio_test_lazyfree(folio)) {
>
> I'm wondering if we need "cc->is_khugepaged &&" as well here?
>
> We should allow users to enforce collapse via the madvise_collapse()
> path even if pages are marked lazyfree, IMHO.

$ man madvise
MADV_COLLAPSE
        Perform a best-effort synchronous collapse of the native pages
        mapped by the memory range into Transparent Huge Pages (THPs).

The semantics of MADV_COLLAPSE are best-effort and do not imply to enforce
collapsing, so we don't need "cc->is_khugepaged" here.

We can imagine that if a user simultaneously uses MADV_FREE and
MADV_COLLAPSE, it indicates a misunderstanding of their semantics.
As the kernel, we need to safeguard the baseline.

> > +                     result = SCAN_PAGE_LAZYFREE;
> > +                     goto out;
> > +             }
> > +
> >               /* See hpage_collapse_scan_pmd(). */
> >               if (folio_maybe_mapped_shared(folio)) {
> >                       ++shared;
> > @@ -1330,6 +1336,11 @@ static enum scan_result hpage_collapse_scan_pmd(struct mm_struct *mm,
> >               }
> >               folio = page_folio(page);
> >
> > +             if (!pte_dirty(pteval) && folio_test_lazyfree(folio)) {
>
> Ditto.
>
> > +                     result = SCAN_PAGE_LAZYFREE;
> > +                     goto out_unmap;
> > +             }
> > +
> >               if (!folio_test_anon(folio)) {
> >                       result = SCAN_PAGE_ANON;
> >                       goto out_unmap;
>

Powered by blists - more mailing lists