linux-kernel - Re: [PATCH] mm/page_reporting: the "page" must not be the list head

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CAKgT0UdiRw4WMe8RxOZ3A6fcQJ9Tp7BAWBbuYkje7YWiT0cVOA@mail.gmail.com>
Date:   Tue, 18 Aug 2020 07:58:14 -0700
From:   Alexander Duyck <alexander.duyck@...il.com>
To:     Wei Yang <richard.weiyang@...ux.alibaba.com>
Cc:     Alexander Duyck <alexander.h.duyck@...ux.intel.com>,
        David Hildenbrand <david@...hat.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        linux-mm <linux-mm@...ck.org>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] mm/page_reporting: the "page" must not be the list head

On Mon, Aug 17, 2020 at 8:22 PM Wei Yang
<richard.weiyang@...ux.alibaba.com> wrote:
>
> On Mon, Aug 17, 2020 at 09:05:32AM -0700, Alexander Duyck wrote:
> >
> >
> >On 8/17/2020 2:35 AM, David Hildenbrand wrote:
> >> On 17.08.20 10:48, Wei Yang wrote:
> >> > If "page" is the list head, list_for_each_entry_safe() would stop
> >> > iteration.
> >> >
> >> > Signed-off-by: Wei Yang <richard.weiyang@...ux.alibaba.com>
> >> > ---
> >> >   mm/page_reporting.c | 2 +-
> >> >   1 file changed, 1 insertion(+), 1 deletion(-)
> >> >
> >> > diff --git a/mm/page_reporting.c b/mm/page_reporting.c
> >> > index 3bbd471cfc81..aaaa3605123d 100644
> >> > --- a/mm/page_reporting.c
> >> > +++ b/mm/page_reporting.c
> >> > @@ -178,7 +178,7 @@ page_reporting_cycle(struct page_reporting_dev_info *prdev, struct zone *zone,
> >> >             * the new head of the free list before we release the
> >> >             * zone lock.
> >> >             */
> >> > -          if (&page->lru != list && !list_is_first(&page->lru, list))
> >> > +          if (!list_is_first(&page->lru, list))
> >> >                    list_rotate_to_front(&page->lru, list);
> >> >            /* release lock before waiting on report processing */
> >> >
> >>
> >> Is this a fix or a cleanup? If it's a fix, can this be reproduced easily
> >> and what ere the effects?
> >>
> >
> >This should be a clean-up. Since the &page->lru != list will always be true.
> >
> >If I recall at some point the that was a check for &next->lru != list but I
> >think I pulled out an additional conditional check somewhere so that we just
> >go through the start of the loop again and iterate over reported pages until
> >we are guaranteed to have a non-reported page to rotate to the top of the
> >list with the general idea being that we wanted the allocator to pull
> >non-reported pages before reported pages.
>
> Hi, Alexander,
>
> I see you mentioned in the changelog, this change "mm/page_reporting: rotate
> reported pages to the tail of the list" brings some performance gain.
>
> Would you mind sharing more test detail? I would like to have a try at my
> side.
>
> Thanks :-)

I seem to recall my default test for most of this was the page_fault1
test from the will-it-scale suite of tests. Basically I was running
that while leaving page reporting enabled. However I don't know how
much visibility you would have into the performance impact as I seem
to recall I had to modify the frequency of scheduling for the
reporting polling task in order to see much of an impact.

Thanks.

- Alex