Re: [PATCH v8 02/15] introduce khugepaged_collapse_single_pmd to unify khugepaged and madvise_collapse

[Baolin Wang]

On 2025/7/4 12:20, Nico Pache wrote:
> On Thu, Jul 3, 2025 at 9:51 PM Baolin Wang
> <baolin.wang@xxxxxxxxxxxxxxxxx> wrote:
> > On 2025/7/2 13:57, Nico Pache wrote:
> > > The khugepaged daemon and madvise_collapse have two different
> > > implementations that do almost the same thing.
> > >
> > > Create khugepaged_collapse_single_pmd to increase code reuse and create an
> > > entry point these two users.
> > >
> > > Refactor madvise_collapse and khugepaged_scan_mm_slot to use the new
> > > khugepaged_collapse_single_pmd function. This introduces a minor
> > > behavioral change that is most likely an undiscovered bug. The current
> > > implementation of khugepaged tests khugepaged_test_exit_or_disable
> > > before calling khugepaged_pte_mapped_thp, but we weren't doing it in the
> > > madvise_collapse case. By unifying these two callers madvise_collapse
> > > now also performs this check.
> > >
> > > Reviewed-by: Baolin Wang <baolin.wang@xxxxxxxxxxxxxxxxx>
> > > Signed-off-by: Nico Pache <npache@xxxxxxxxxx>
> > > ---
> > >    mm/khugepaged.c | 95 +++++++++++++++++++++++++------------------------
> > >    1 file changed, 49 insertions(+), 46 deletions(-)
> > >
> > > diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> > > index f27fe7ca9b86..bf69e81a3d82 100644
> > > --- a/mm/khugepaged.c
> > > +++ b/mm/khugepaged.c
> > > @@ -2354,6 +2354,50 @@ static int khugepaged_scan_file(struct mm_struct *mm, unsigned long addr,
> > >        return result;
> > >    }
> > >
> > > +/*
> > > + * Try to collapse a single PMD starting at a PMD aligned addr, and return
> > > + * the results.
> > > + */
> > > +static int khugepaged_collapse_single_pmd(unsigned long addr,
> > > +                                struct vm_area_struct *vma, bool *mmap_locked,
> > > +                                struct collapse_control *cc)
> > > +{
> > > +     int result = SCAN_FAIL;
> > > +     struct mm_struct *mm = vma->vm_mm;
> > > +
> > > +     if (IS_ENABLED(CONFIG_SHMEM) && !vma_is_anonymous(vma)) {
> >
> > Seems you still forgot to drop 'IS_ENABLED(CONFIG_SHMEM)' :)
> Hi Baolin,
>
> You had me questioning my sanity for a moment because I SWEAR I did do
> this! but now I see the issue, I accidentally merged the fixup into
> patch 6 (not entirely sure how or why I would have done that given my
> notes say it belongs in this patch...)

Ah, yes.

> I'm sorry about that :< I'll see if I can work with Andrew to get some
> fixups into mm-new (this was merged a few hours ago) for this patch
> and patch 6.

OK. Thanks.