Date:   Thu, 2 Mar 2023 15:21:50 -0800
From:   Zach O'Keefe <zokeefe@google.com>
To:     Yang Shi <shy828301@gmail.com>
Cc:     Peter Xu <peterx@redhat.com>, linux-kernel@vger.kernel.org,
        linux-mm@kvack.org, Andrew Morton <akpm@linux-foundation.org>,
        David Stevens <stevensd@chromium.org>,
        Johannes Weiner <hannes@cmpxchg.org>
Subject: Re: [PATCH v2] mm/khugepaged: alloc_charge_hpage() take care of mem
 charge errors
Message-ID: <20230302232150.vvmszlrdzqm5ndjq@google.com>
References: <20230222195247.791227-1-peterx@redhat.com>
 <CAHbLzkp3PwpXir=YUuWmxpW5VXEba81UKb3YOyq+snKTZGjqDA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAHbLzkp3PwpXir=YUuWmxpW5VXEba81UKb3YOyq+snKTZGjqDA@mail.gmail.com>
Precedence: bulk

On Feb 22 14:53, Yang Shi wrote:
> On Wed, Feb 22, 2023 at 11:52 AM Peter Xu <peterx@redhat.com> wrote:
> >
> > If memory charge failed, instead of returning the hpage but with an error,
> > allow the function to cleanup the folio properly, which is normally what a
> > function should do in this case - either return successfully, or return
> > with no side effect of partial runs with an indicated error.
> >
> > This will also avoid the caller calling mem_cgroup_uncharge() unnecessarily
> > with either anon or shmem path (even if it's safe to do so).
> 
> Thanks for the cleanup. Reviewed-by: Yang Shi <shy828301@gmail.com>
> 
> >
> > Cc: Yang Shi <shy828301@gmail.com>
> > Reviewed-by: David Stevens <stevensd@chromium.org>
> > Acked-by: Johannes Weiner <hannes@cmpxchg.org>
> > Signed-off-by: Peter Xu <peterx@redhat.com>
> > ---
> > v1->v2:
> > - Enhance commit message, drop "Fixes:" and "Cc: stable" tag, add R-bs.
> > ---
> >  mm/khugepaged.c | 9 ++++++++-
> >  1 file changed, 8 insertions(+), 1 deletion(-)
> >
> > diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> > index 8dbc39896811..941d1c7ea910 100644
> > --- a/mm/khugepaged.c
> > +++ b/mm/khugepaged.c
> > @@ -1063,12 +1063,19 @@ static int alloc_charge_hpage(struct page **hpage, struct mm_struct *mm,
> >         gfp_t gfp = (cc->is_khugepaged ? alloc_hugepage_khugepaged_gfpmask() :
> >                      GFP_TRANSHUGE);
> >         int node = hpage_collapse_find_target_node(cc);
> > +       struct folio *folio;
> >
> >         if (!hpage_collapse_alloc_page(hpage, gfp, node, &cc->alloc_nmask))
> >                 return SCAN_ALLOC_HUGE_PAGE_FAIL;
> > -       if (unlikely(mem_cgroup_charge(page_folio(*hpage), mm, gfp)))
> > +
> > +       folio = page_folio(*hpage);
> > +       if (unlikely(mem_cgroup_charge(folio, mm, gfp))) {
> > +               folio_put(folio);
> > +               *hpage = NULL;
> >                 return SCAN_CGROUP_CHARGE_FAIL;
> > +       }
> >         count_memcg_page_event(*hpage, THP_COLLAPSE_ALLOC);
> > +
> >         return SCAN_SUCCEED;
> >  }
> >
> > --
> > 2.39.1
> >
> 

Thanks, Peter.

Can we also get rid of the unnecessary mem_cgroup_uncharge() calls while we're
at it? Maybe this deserves a separate patch, but after Yang's cleanup of the
!NUMA case (where we would preallocate a hugepage) we can depend on put_page()
do take care of that for us.

Regardless, can have my

Reviewed-by: Zach O'Keefe <zokeefe@google.com>