linux-kernel - Re: [PATCH v3] intel_idle: Add Comet Lake 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: <CAJZ5v0g6fAPZxX3MKGomjTWfWVMV-A+8yUC=ydOg5YwD-moDvQ@mail.gmail.com>
Date:   Tue, 10 Mar 2020 09:37:14 +0100
From:   "Rafael J. Wysocki" <rafael@...nel.org>
To:     "Pan, Harry" <harry.pan@...el.com>
Cc:     "Rafael J. Wysocki" <rafael@...nel.org>,
        LKML <linux-kernel@...r.kernel.org>,
        Harry Pan <gs0622@...il.com>,
        Jacob Pan <jacob.jun.pan@...ux.intel.com>,
        Len Brown <lenb@...nel.org>,
        Linux PM <linux-pm@...r.kernel.org>
Subject: Re: [PATCH v3] intel_idle: Add Comet Lake support

On Mon, Mar 9, 2020 at 10:07 AM Pan, Harry <harry.pan@...el.com> wrote:
>
> Hi Rafael,
>
> Thanks for the comments.
> I have some questions, I am wondering if you can share upstream thought w.r.t the future development of intel_idle.
>
>  - It looks to me since v5.6 intel_idle will prefer _CST of ACPI rather than the general table embedded in this driver.

Not exactly.

The rules are as follows:
 * If there is a built-in table for the given processor in the driver,
it will be used, but the C-states that are not exposed in _CST will be
disabled by default (the state parameters come from the built-in table
for all C-states).
 * Otherwise (i.e. the driver does not recognized the given
processor), C-state definitions will be based on the _CST data.

See https://www.kernel.org/doc/html/latest/admin-guide/pm/intel_idle.html
for details.

>  - Any pros and cons of using the tables of _CST in firmware and embedded one in the kernel?

Both can be used at the same time (see above).

The built-in table in the kernel must be suitable for all of the
platforms shipped with the given processor (SoC).

Obviously, the _CST table allows intel_idle to work if the processor
included in the platform is not recognized by it.

Generally, our planned strategy is to provide built-in C-state tables
for all new "mainstream" processors with the use_acpi flag set to
avoid enabling C-states that may have not been validated on a given
platform by default.

>  - Can the table in _CST archive more optimal idle states management?

Possibly, but not likely.

> If there is already any reference, documents I missed, kindly enlighten me then I would like to read it first before refining the questions.

See above. :-)

Thanks!


___________________________________
> 從: Rafael J. Wysocki [rafael@...nel.org]
> 寄件日期: 2020年3月5日 上午 05:14
> 至: Pan, Harry
> 副本: Rafael J. Wysocki; LKML; Harry Pan; Jacob Pan; Len Brown; Linux PM
> 主旨: Re: [PATCH v3] intel_idle: Add Comet Lake support
>
> On Wed, Mar 4, 2020 at 12:57 PM Pan, Harry <harry.pan@...el.com> wrote:
> >
> > Hi Rafael,
> >
> > Yes, I skipped it considering to align CML-U V0 and A0 stepping w/ the same table; I sent v4 for your review.
>
> Skipping that flag is risky, because it may cause some C-states to be
> enabled on systems where they have not been validated (e.g. systems
> shipping with other OSes which only use _CST C-states).  There were
> problems related to that in the past which is one of the reasons for
> adding _CST support to intel_idle.  use_acpi should be set for all new
> platforms going forward as a rule.
>
> > In the other hand, I am proposing using _CST as long term plan in CrOS dev teams.
>
> That I obviously agree with. :-)
>
>
> > > On Mar 4, 2020, at 17:53, Rafael J. Wysocki <rafael@...nel.org> wrote:
> > >
> > > On Tue, Mar 3, 2020 at 10:10 AM Harry Pan <harry.pan@...el.com> wrote:
> > >>
> > >> Add a general C-state table in order to support Comet Lake.
> > >>
> > >> Signed-off-by: Harry Pan <harry.pan@...el.com>
> > >>
> > >> ---
> > >>
> > >> drivers/idle/intel_idle.c | 7 +++++++
> > >> 1 file changed, 7 insertions(+)
> > >>
> > >> diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c
> > >> index d55606608ac8..05bce595fafe 100644
> > >> --- a/drivers/idle/intel_idle.c
> > >> +++ b/drivers/idle/intel_idle.c
> > >> @@ -1067,6 +1067,11 @@ static const struct idle_cpu idle_cpu_dnv = {
> > >>        .use_acpi = true,
> > >> };
> > >>
> > >> +static const struct idle_cpu idle_cpu_cml = {
> > >> +       .state_table = skl_cstates,
> > >> +       .disable_promotion_to_c1e = true,
> > >
> > > .use_acpi = true,
> > >
> > > missing?  Otherwise you can just use idle_cpu_skl as is, can't you?
> > >
> > >> +};
> > >> +
> > >> static const struct x86_cpu_id intel_idle_ids[] __initconst = {
> > >>        INTEL_CPU_FAM6(NEHALEM_EP,              idle_cpu_nhx),
> > >>        INTEL_CPU_FAM6(NEHALEM,                 idle_cpu_nehalem),
> > >> @@ -1105,6 +1110,8 @@ static const struct x86_cpu_id intel_idle_ids[] __initconst = {
> > >>        INTEL_CPU_FAM6(ATOM_GOLDMONT_PLUS,      idle_cpu_bxt),
> > >>        INTEL_CPU_FAM6(ATOM_GOLDMONT_D,         idle_cpu_dnv),
> > >>        INTEL_CPU_FAM6(ATOM_TREMONT_D,          idle_cpu_dnv),
> > >> +       INTEL_CPU_FAM6(COMETLAKE_L,             idle_cpu_cml),
> > >> +       INTEL_CPU_FAM6(COMETLAKE,               idle_cpu_cml),
> > >>        {}
> > >> };
> > >>
> > >> --
> > >> 2.24.1
> > >>