Message-ID: <CAAYoRsXK79PspEUh9pqgj2OGQnxQONkEeK-7af3=5frBzAqULQ@mail.gmail.com>
Date:   Tue, 7 Sep 2021 08:45:37 -0700
From:   Doug Smythies <dsmythies@...us.net>
To:     Giovanni Gherdovich <ggherdovich@...e.cz>,
        "Rafael J . Wysocki" <rjw@...ysocki.net>,
        Srinivas Pandruvada <srinivas.pandruvada@...ux.intel.com>
Cc:     Viresh Kumar <viresh.kumar@...aro.org>,
        Len Brown <lenb@...nel.org>,
        Linux PM list <linux-pm@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2] cpufreq: intel_pstate: Add Icelake servers support in
 no-HWP mode

Hi all,

Recent ASUS BIOS updates have changed the default system
response for this old thread, rendering "intel_pstate=no_hwp" useless.

It also raises a question: If BIOS has forced HWP, then how do we
prevent the acpi-cpufreq driver from being used? Read on.

On Fri, May 14, 2021 at 3:12 PM Doug Smythies <dsmythies@...us.net> wrote:
>
> On Fri, May 14, 2021 at 1:33 PM Giovanni Gherdovich <ggherdovich@...e.cz> wrote:
> > On Fri, 2021-05-14 at 08:31 -0700, Doug Smythies wrote:
...
> >
> > when COMETLAKE is not in that list, can you confirm that if you go into the
> > BIOS config at boot, and disable HWP from there, then intel_pstate does *not* load?
>
> Yes, already tested before my original reply.
>
> > Does it say "intel_pstate: CPU model not supported" in the dmesg log?
>
> That I did not check, but if I boot now with an unmodified kernel
> 5.13-rc1 (i.e. without this patch):
>
> [    0.369323] intel_pstate: CPU model not supported
>
> > The control may be somewhere around "power mangement" in the BIOS config, and
> > may be called "Enable/disable Intel Speed Shift".
>
> Yes.
>
> > I'm asking because I've just checked on two Dell laptops, one Skylake and the
> > other Kabylake, and the menu is there in the BIOS config to disable HWP,
> > but if I disable it... nothing happens. "lscpu" shows all the hwp flags as usual:
>
> Motherboard here is ASUS PRIME Z490-A.
> CPU: Intel(R) Core(TM) i5-10600K CPU @ 4.10GHz
>
> >     # lscpu | grep Flags | tr ' ' '\n' | grep hwp
> >     hwp
> >     hwp_notify
> >     hwp_act_window
> >     hwp_epp
>
> Here, for some reason I have to do it this way (sudo) or your command
> doesn't work properly. Results herein confirmed by looking at the
> "Flags" output manually without filtering:
>
> intel_speed_shift = Disabled in BIOS:
>
> doug@s19:~$ sudo lscpu | tr ' ' '\n' | grep hwp
> doug@s19:~$
>
> intel_speed_shift = Auto in BIOS
>
> $ sudo lscpu | tr ' ' '\n' | grep hwp
> hwp
> hwp_notify
> hwp_act_window
> hwp_epp
>
> > and turbostat gives me:
> >
> >     # turbostat -Summary -i 1 : 2>&1 | grep MSR_PM_ENABLE
> >     cpu0: MSR_PM_ENABLE: 0x00000001 (HWP)
>
> Here:
>
> intel_speed_shift = Disabled in BIOS:
>
> root@s19:/home/doug#
> /home/doug/temp-k-git/linux/tools/power/x86/turbostat/turbostat
> -Summary -i 1 : 2>&1 | grep MSR_PM_ENABLE
> root@s19:/home/doug#
>
> intel_speed_shift = Auto in BIOS (the default setting)
>
> root@s19:/home/doug#
> /home/doug/temp-k-git/linux/tools/power/x86/turbostat/turbostat
> -Summary -i 1 : 2>&1 | grep MSR_PM_ENABLE
> cpu0: MSR_PM_ENABLE: 0x00000001 (HWP)
>
> or with "intel_pstate=no_hwp"
>
> root@s19:/home/doug#
> /home/doug/temp-k-git/linux/tools/power/x86/turbostat/turbostat
> -Summary -i 1 : 2>&1 | grep MSR_PM_ENABLE
> cpu0: MSR_PM_ENABLE: 0x00000000 (No-HWP)
>
> > Which is to say, on the Intel client machines I have, the firmware doesn't
> > seem to be able to hide HWP from the OS. Buggy BIOS? Maybe, the fact of the
> > matter is, I wouldn't need to add, say, KABYLAKE to that list, based on my
> > experience.
>
> My experience (hardware) differs from yours with respect to this.
>
> > The other side of the issue is that, from my understanding, the
> > preferred/supported way to disable HWP is to boot with intel_pstate=no_hwp,

Previous correspondence was with BIOS version 1003. There have been 3 BIOS
releases since then (at least that I know of), 2103, 2201, 2301, and all of them
have changed the behaviour of the "Auto" setting for Intel Speed Shift
Technology BIOS setting, forcing it on upon transfer of control to the OS.

Where with "intel_pstate=no_hwp" one used to get 0 for MSR_PM_ENABLE
(0x770) they now get 1.

That was for my ASUS Z490-P Motherboard.
For my ASUS Z390-A Motherboard (actually a windows 10 computer now),
with an older BIOS, things still work properly for "intel_pstate=no_hwp", and
while there is a newer BIOS, I will not install it as it is also not possible to
rollback.

I have had an escalation in progress with ASUS about this for about a month,
and while they have been responsive we have communications/language
issues and have yet to even agree that there is an issue.

I have been working with this incomplete patch:

diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c
index bb4549959b11..7dcc51ee56ea 100644
--- a/drivers/cpufreq/intel_pstate.c
+++ b/drivers/cpufreq/intel_pstate.c
@@ -3347,17 +3347,27 @@ device_initcall(intel_pstate_init);

 static int __init intel_pstate_setup(char *str)
 {
+       /* The following needs to be conditional on CPUID with EAX 6 */
+       /* CPUID.06H:EAX[bit 7], which I do not know how to do. */
+       /* Avoid doing checking and printing multiple times, */
+       /* which I do not know why it does. */
+       if(!force_load){
+               if(intel_pstate_hwp_is_enabled()){
+                       pr_info("HWP enabled by BIOS\n");
+                       force_load = 1;
+               }
+       }
        if (!str)
                return -EINVAL;

-       if (!strcmp(str, "disable"))
+       if (!strcmp(str, "disable") && !force_load)
                no_load = 1;
-       else if (!strcmp(str, "active"))
+       if (!strcmp(str, "active") && !no_load)
                default_driver = &intel_pstate;
        else if (!strcmp(str, "passive"))
                default_driver = &intel_cpufreq;

-       if (!strcmp(str, "no_hwp")) {
+       if (!strcmp(str, "no_hwp") && !force_load) {
                pr_info("HWP disabled\n");
                no_hwp = 1;
        }

Which has troubles when HWP has been disabled by BIOS, but otherwise works well.

HWP Disabled by BIOS:

doug@s19:~$ sudo dmesg | grep intel_pstate
[sudo] password for doug:
[    0.000000] Command line: BOOT_IMAGE=/boot/vmlinuz-5.14.0-ipstate4
root=UUID=0ac356c1-caa9-4c2e-8229-4408bd998dbd ro ipv6.disable=1
consoleblank=314 intel_pstate=active intel_pstate=no_hwp
msr.allow_writes=on cpuidle.governor=teo
[    0.000000]  intel_pstate_setup+0x24/0x151
[    0.000000]  intel_pstate_setup+0x24/0x151
[    0.000000] intel_pstate: HWP disabled
[    0.051278] Kernel command line:
BOOT_IMAGE=/boot/vmlinuz-5.14.0-ipstate4
root=UUID=0ac356c1-caa9-4c2e-8229-4408bd998dbd ro ipv6.disable=1
consoleblank=314 intel_pstate=active intel_pstate=no_hwp
msr.allow_writes=on cpuidle.governor=teo
[    0.393236] intel_pstate: Intel P-state driver initializing

HWP force enabled by BIOS, which is the only other option with the newer BIOS:

doug@s19:~$ sudo dmesg | grep intel_pstate
[sudo] password for doug:
[    0.000000] Command line: BOOT_IMAGE=/boot/vmlinuz-5.14.0-ipstate4
root=UUID=0ac356c1-caa9-4c2e-8229-4408bd998dbd ro ipv6.disable=1
consoleblank=314 intel_pstate=active intel_pstate=no_hwp
msr.allow_writes=on cpuidle.governor=teo
[    0.000000] intel_pstate: HWP enabled by BIOS
[    0.049205] Kernel command line:
BOOT_IMAGE=/boot/vmlinuz-5.14.0-ipstate4
root=UUID=0ac356c1-caa9-4c2e-8229-4408bd998dbd ro ipv6.disable=1
consoleblank=314 intel_pstate=active intel_pstate=no_hwp
msr.allow_writes=on cpuidle.governor=teo
[    0.370662] intel_pstate: Intel P-state driver initializing
[    0.371590] intel_pstate: HWP enabled

HWP forced by BIOS, disable intel_pstate:

doug@s19:~$ sudo dmesg | grep intel_pstate
[    0.000000] Command line: BOOT_IMAGE=/boot/vmlinuz-5.14.0-ipstate4
root=UUID=0ac356c1-caa9-4c2e-8229-4408bd998dbd ro ipv6.disable=1
consoleblank=300 intel_pstate=disable msr.allow_writes=on
cpuidle.governor=teo
[    0.000000] intel_pstate: HWP enabled by BIOS
[    0.049133] Kernel command line:
BOOT_IMAGE=/boot/vmlinuz-5.14.0-ipstate4
root=UUID=0ac356c1-caa9-4c2e-8229-4408bd998dbd ro ipv6.disable=1
consoleblank=300 intel_pstate=disable msr.allow_writes=on
cpuidle.governor=teo
[    0.370519] intel_pstate: Intel P-state driver initializing
[    0.371451] intel_pstate: HWP enabled
doug@s19:~$ cat /sys/devices/system/cpu/cpu5/cpufreq/scaling_driver
intel_pstate

HWP forced by BIOS, disable intel_pstate:, unpatched kernel 5.14:

doug@s19:~$ sudo dmesg | grep intel_pstate
[    0.000000] Command line: BOOT_IMAGE=/boot/vmlinuz-5.14.0-stock
root=UUID=0ac356c1-caa9-4c2e-8229-4408bd998dbd ro ipv6.disable=1
consoleblank=300 intel_pstate=disable msr.allow_writes=on
cpuidle.governor=teo
[    0.049108] Kernel command line:
BOOT_IMAGE=/boot/vmlinuz-5.14.0-stock
root=UUID=0ac356c1-caa9-4c2e-8229-4408bd998dbd ro ipv6.disable=1
consoleblank=300 intel_pstate=disable msr.allow_writes=on
cpuidle.governor=teo
doug@s19:~$ cat /sys/devices/system/cpu/cpu5/cpufreq/scaling_driver
acpi-cpufreq
doug@s19:~$ sudo rdmsr 0x770
1

Observe the driver ended up as acpi-cpufreq, but with HWP enabled.

... Doug

Powered by blists - more mailing lists