linux-kernel - Re: [PATCH] intel_idle: Add ICL support

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

Message-ID: <CAJZ5v0jDY62HYWF-QKE8kH4kFx9Ympjwk1Dbhdm_VCFaP28RZg@mail.gmail.com>
Date:   Wed, 26 Aug 2020 18:41:44 +0200
From:   "Rafael J. Wysocki" <rafael@...nel.org>
To:     Artem Bityutskiy <dedekind1@...il.com>
Cc:     Guilhem Lettron <guilhem@...pilot.io>,
        Zhang Rui <rui.zhang@...el.com>,
        "Rafael J. Wysocki" <rafael@...nel.org>,
        Jacob Pan <jacob.jun.pan@...ux.intel.com>,
        Len Brown <lenb@...nel.org>,
        Linux PM <linux-pm@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] intel_idle: Add ICL support

On Wed, Aug 26, 2020 at 6:19 PM Artem Bityutskiy <dedekind1@...il.com> wrote:
>
> Indeed, when I compare them:
>
> acpi_idle (without the patch):

Does this come from the Guilhem's data?  It's intel_idle in both
cases, but in the "without the patch" case it uses ACPI.

> CPU%c1  CPU%c6  CPU%c7  CoreTmp PkgTmp  GFX%rc6 Pkg%pc2 Pkg%pc3 Pkg%pc6 Pkg%pc7 Pkg%pc8 Pkg%pc9 Pk%pc10 PkgWatt
> 29.48   0.00    60.71   58      58      97.96   16.96   0.00    0.00    0.00    0.00    0.00    0.00    6.08

and I get the same data here, but

> intel_idle (with the patch):
>
> CPU%c1  CPU%c6  CPU%c7  CoreTmp PkgTmp  GFX%rc6 Pkg%pc2 Pkg%pc3 Pkg%pc6 Pkg%pc7 Pkg%pc8 Pkg%pc9 Pk%pc10 PkgWatt
> 56      56      96.64   300     68.29   48.58   0.00    0.00    0.00    0.00    0.00    0.00    7.38    0.00

you seem to have columns wrong here.

I get something like this

CPU%c1     CPU%c6     CPU%c7     CoreTmp PkgTmp     GFX%rc6 GFXMHz
Totl%C0 Any%C0     GFX%C0     CPUGFX% Pkg%pc2    Pkg%pc3 Pkg%pc6
Pkg%pc7 Pkg%pc8 Pkg%pc9 Pk%pc10
16.07     26.24     49.10     56     56     96.64     300     68.29
 48.58     3.08     2.10     30.36     0.04     0.00     0.00     0.00
    0.00     0.00

so still no PC10 residency (and it would be rather strange to get PC10
residency without any PC6 or higher residency).  The 7.38 is the
PkgWatt number AFAICS.

> With intel_idle we reach PC10, without it we only go as deep as PC2 - huge difference.

Not really.  We don't get any PC10 residency in both cases.

> I really wonder why the BIOS does not expose deeper C-states...

It does expose C10.

> And if it does not, is this for a reason? And how windows works then?

It can only expose 3 C-states and it chose to expose C1, C7s and C10.

> May be there is a BIOS update that fixes this problem? May be Windows
> user get it quickly because stuff like this is often well-integrated in
> Windows? Would you please check if there is newer BIOS?

I doubt it.

Cheers!