Re: [RFT][PATCH 1/3] PM: QoS: Introduce frequency QoS

From: "Rafael J. Wysocki" <rafael@kernel.org>
To: Doug Smythies <dsmythies@telus.net>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>,
	Linux PM <linux-pm@vger.kernel.org>,
	Linux ACPI <linux-acpi@vger.kernel.org>,
	LKML <linux-kernel@vger.kernel.org>,
	Viresh Kumar <viresh.kumar@linaro.org>,
	Sudeep Holla <sudeep.holla@arm.com>,
	Dmitry Osipenko <digetx@gmail.com>
Subject: Re: [RFT][PATCH 1/3] PM: QoS: Introduce frequency QoS
Date: Tue, 19 Nov 2019 20:17:05 +0100	[thread overview]
Message-ID: <CAJZ5v0i1iAjpWju6FiCjP3RvspKDRfSwz4=b_3qgGhhfz8sSrw@mail.gmail.com> (raw)
In-Reply-To: <000401d59ee6$959e3da0$c0dab8e0$@net>

On Tue, Nov 19, 2019 at 3:35 PM Doug Smythies <dsmythies@telus.net> wrote:
>
> On 2019.11.17 08:13 Doug Smythies wrote:
> > On 2019.11.16 23:35 Doug Smythies wrote:
>
> >> Hi Rafael,
> >>
> >> Not sure, but I think it is this one that
> >> causes complaining when I try to set the
> >> intel_pstate driver to passive mode.
> >> I started from active mode, powersave governor,
> >> no HWP.
> >>
> >> Kernel: 5.4-rc7
> >>
> >> I did not go back and try previous 5.4 RCs.
>
> After looking at the git tags for this patch,
> I tried kernel 5.4-rc2, which was the closest
> Kernel I had to before the patch set was added.
> It worked fine, as expected.
>
> >> I did try kernel 5.3-rc8, because I already had
> >> it installed, and it worked fine.
> >>
> >> I use a script (for years), run as sudo:
> >>
> >> doug@s15:~/temp$ cat set_cpu_passive
> >> #! /bin/bash
> >> cat /sys/devices/system/cpu/intel_pstate/status
> >> echo passive > /sys/devices/system/cpu/intel_pstate/status
> >> cat /sys/devices/system/cpu/intel_pstate/status
> >>
> >> And I get this (very small excerpt):
> >>
> >> freq_qos_add_request() called for active request
> >> WARNING: CPU: 1 PID: 2758 at kernel/power/qos.c:763 freq_qos_add_request+0x4c/0xa0
> >> CPU: 1 PID: 2758 Comm: set_cpu_passive Not tainted 5.4.0-rc7-stock #727
> >> Failed to add freq constraint for CPU0 (-22)
> >>
> >> freq_qos_add_request() called for active request
> >> WARNING: CPU: 1 PID: 2758 at kernel/power/qos.c:763 freq_qos_add_request+0x4c/0xa0
> >> CPU: 1 PID: 2758 Comm: set_cpu_passive Tainted: G        W         5.4.0-rc7-stock #727
> >> Failed to add freq constraint for CPU1 (-22)
>
> Updated summary of previous emails:
> This patch or patch set breaks the after boot
> ability to change CPU frequency scaling drivers.
>
> Using a workaround of booting with
> "intel_pstate=passive" seems to prevent the errors.
>
> Changing between the intel_pstate and intel_cpufreq drivers
> (i.e. between active and passive modes)
> after boot, either way, causes the errors. i.e.
>
> Failed to add freq constraint for CPU7 (-22)
> (2 per CPU per attempt)

These messages come from acpi_processor_ppc_init() and
acpi_thermal_cpufreq_init(), AFAICS, which are invoked by
acpi_processor_notifier() and that is invoked by the
blocking_notifier_call_chain() in cpufreq_online() which tirggers for
new policies after adding the max freq QoS request to
policy->constraints.

The requests added by them should be removed by
acpi_processor_ppc_exit() and acpi_thermal_cpufreq_exit(),
respectively, invoked by the blocking_notifier_call_chain() in
cpufreq_policy_free(), but it looks like that doesn't happen.

However, I now also see that freq_qos_remove_request() doesn't clear
the qos field in req which is should do, so freq_qos_add_request()
will complain and fail if the object pointed to by req is passed to it
again.

I'll send a patch to test for this later today.