linux-kernel - Re: [PATCH v2] PM/hibernation: set the default image size for large memory

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 for Android: free password hash cracker in your pocket

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

Message-ID: <CAJZ5v0jdCx3PB1JOgEc8ZhDvvXNBzcfYYsSy5PB4s29PfoQ8Rg@mail.gmail.com>
Date:   Mon, 27 Mar 2023 18:01:24 +0200
From:   "Rafael J. Wysocki" <rafael@...nel.org>
To:     Xueqin Luo <luoxueqin@...inos.cn>
Cc:     rafael@...nel.org, len.brown@...el.com, pavel@....cz,
        linux-pm@...r.kernel.org, linux-kernel@...r.kernel.org,
        xiongxin@...inos.cn
Subject: Re: [PATCH v2] PM/hibernation: set the default image size for large memory

On Fri, Mar 3, 2023 at 6:51 AM Xueqin Luo <luoxueqin@...inos.cn> wrote:
>
> From: xueqin Luo <luoxueqin@...inos.cn>
>
> As computers have more and more memory, they can store larger and
> larger images. This poses a problem. Let's take a 16GB computer as
> an example. When the number of pre-allocated pages is greater than
> 900,000 and smaller than image_size, the duration of S4 increases
> with the number of saved pages. When the number of pre-allocated
> pages reaches 1.2 million, the duration of S4 increases by 8-10s.

Well, I'm not quite sure what the problem is from the description
above.  Can you please explain?

> We found that the image size is generally more than 900,000 pages,
> so we set the default size of image_size for 16GB and above
> computers to 1 million pages, which can reduce the running time
> of S4 under certain conditions.
>
> This is the test data for 5 hours after the computer is turned on:
>
> Original kernel begin:
>
> [2023-02-24 19:16:56] [   46.105423][ 2] [ T3075] PM: hibernation entry
> [2023-02-24 19:16:59] [   47.632000][ 1] [ T3075] done (allocated
> 959239 pages)
> [2023-02-24 19:18:05] [   56.987043][ 2] [ T3075] PM: hibernation exit
>
> Five hours later:
>
> [2023-02-25 00:22:48] [18069.651640][ 4] [ T7590] PM: hibernation entry
> [2023-02-25 01:23:59] [21671.194049][ 0] [ T8563] done (allocated
> 1228878 pages)
> [2023-02-25 00:24:06] [18080.639889][ 2] [ T7590] PM: hibernation exit
>
> After 5h, you can see that the S4 takes 8s more time.
>
> Five hours later, the modified kernel test data:
>
> [2023-02-24 15:52:08] [18190.171183][ 7] [T11151] PM: hibernation entry
> [2023-02-24 15:52:11] [18191.677021][ 7] [T11151] done (allocated
> 792597 pages)
> [2023-02-24 15:53:15] [18201.028488][ 2] [T11151] PM: hibernation exit
>
> You can see that after five hours, the time has changed little compared
> to the initial test data.
>
> Signed-off-by: xueqin Luo <luoxueqin@...inos.cn>
> ---
>
> v3: modify some comments
>
> v2: add test data
>
>  kernel/power/snapshot.c | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/power/snapshot.c b/kernel/power/snapshot.c
> index cd8b7b35f1e8..fa3950b19849 100644
> --- a/kernel/power/snapshot.c
> +++ b/kernel/power/snapshot.c
> @@ -136,7 +136,14 @@ unsigned long image_size;
>
>  void __init hibernate_image_size_init(void)
>  {
> -       image_size = ((totalram_pages() * 2) / 5) * PAGE_SIZE;
> +       /* The totalram pages() for a computer of 16 memory size is
> +        * equal to 4032990 pages. And according to our observation,
> +        * the average image size is less than 1 million pages.
> +        */
> +       if (totalram_pages() < 4032990)
> +               image_size = ((totalram_pages() * 2) / 5) * PAGE_SIZE;
> +       else
> +               image_size = 1000000 * PAGE_SIZE;
>  }
>
>  /*
> --
> 2.25.1
>