From: "Rafael J. Wysocki" <rafael@kernel.org>
To: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Matthias Brugger <matthias.bgg@gmail.com>,
linux-pm@vger.kernel.org,
Vincent Guittot <vincent.guittot@linaro.org>,
regressions@leemhuis.info, daniel@makrotopia.org,
thomas.huehn@hs-nordhausen.de,
"v5 . 19+" <stable@vger.kernel.org>, Nick <vincent@systemli.org>,
linux-kernel@vger.kernel.org,
linux-arm-kernel@lists.infradead.org,
linux-mediatek@lists.infradead.org
Subject: Re: [PATCH] Revert "cpufreq: mediatek: Refine mtk_cpufreq_voltage_tracking()"
Date: Fri, 2 Dec 2022 20:46:25 +0100 [thread overview]
Message-ID: <CAJZ5v0hc8CsvqLKxi5iRq7iR0bkt25dRnLBd23mx-zdi2Sjgsw@mail.gmail.com> (raw)
In-Reply-To: <CAJZ5v0ixHocQbu6-zs3dMDsiw8vdPyv=8Re7N4kUckeGkLhUzg@mail.gmail.com>
On Fri, Dec 2, 2022 at 1:17 PM Rafael J. Wysocki <rafael@kernel.org> wrote:
>
> On Fri, Dec 2, 2022 at 6:26 AM Viresh Kumar <viresh.kumar@linaro.org> wrote:
> >
> > This reverts commit 6a17b3876bc8303612d7ad59ecf7cbc0db418bcd.
> >
> > This commit caused regression on Banana Pi R64 (MT7622), revert until
> > the problem is identified and fixed properly.
> >
> > Link: https://lore.kernel.org/all/930778a1-5e8b-6df6-3276-42dcdadaf682@systemli.org/
> > Cc: v5.19+ <stable@vger.kernel.org> # v5.19+
> > Reported-by: Nick <vincent@systemli.org>
> > Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
>
> Do you want me to push this revert for -rc8?
After all, I've decided to queue it up for 6.2, thanks!
> > ---
> > drivers/cpufreq/mediatek-cpufreq.c | 147 +++++++++++++++++++----------
> > 1 file changed, 96 insertions(+), 51 deletions(-)
> >
> > diff --git a/drivers/cpufreq/mediatek-cpufreq.c b/drivers/cpufreq/mediatek-cpufreq.c
> > index 7f2680bc9a0f..4466d0c91a6a 100644
> > --- a/drivers/cpufreq/mediatek-cpufreq.c
> > +++ b/drivers/cpufreq/mediatek-cpufreq.c
> > @@ -8,7 +8,6 @@
> > #include <linux/cpu.h>
> > #include <linux/cpufreq.h>
> > #include <linux/cpumask.h>
> > -#include <linux/minmax.h>
> > #include <linux/module.h>
> > #include <linux/of.h>
> > #include <linux/of_platform.h>
> > @@ -16,6 +15,8 @@
> > #include <linux/pm_opp.h>
> > #include <linux/regulator/consumer.h>
> >
> > +#define VOLT_TOL (10000)
> > +
> > struct mtk_cpufreq_platform_data {
> > int min_volt_shift;
> > int max_volt_shift;
> > @@ -55,7 +56,6 @@ struct mtk_cpu_dvfs_info {
> > unsigned int opp_cpu;
> > unsigned long current_freq;
> > const struct mtk_cpufreq_platform_data *soc_data;
> > - int vtrack_max;
> > bool ccifreq_bound;
> > };
> >
> > @@ -82,7 +82,6 @@ static int mtk_cpufreq_voltage_tracking(struct mtk_cpu_dvfs_info *info,
> > struct regulator *proc_reg = info->proc_reg;
> > struct regulator *sram_reg = info->sram_reg;
> > int pre_vproc, pre_vsram, new_vsram, vsram, vproc, ret;
> > - int retry = info->vtrack_max;
> >
> > pre_vproc = regulator_get_voltage(proc_reg);
> > if (pre_vproc < 0) {
> > @@ -90,44 +89,91 @@ static int mtk_cpufreq_voltage_tracking(struct mtk_cpu_dvfs_info *info,
> > "invalid Vproc value: %d\n", pre_vproc);
> > return pre_vproc;
> > }
> > + /* Vsram should not exceed the maximum allowed voltage of SoC. */
> > + new_vsram = min(new_vproc + soc_data->min_volt_shift,
> > + soc_data->sram_max_volt);
> > +
> > + if (pre_vproc < new_vproc) {
> > + /*
> > + * When scaling up voltages, Vsram and Vproc scale up step
> > + * by step. At each step, set Vsram to (Vproc + 200mV) first,
> > + * then set Vproc to (Vsram - 100mV).
> > + * Keep doing it until Vsram and Vproc hit target voltages.
> > + */
> > + do {
> > + pre_vsram = regulator_get_voltage(sram_reg);
> > + if (pre_vsram < 0) {
> > + dev_err(info->cpu_dev,
> > + "invalid Vsram value: %d\n", pre_vsram);
> > + return pre_vsram;
> > + }
> > + pre_vproc = regulator_get_voltage(proc_reg);
> > + if (pre_vproc < 0) {
> > + dev_err(info->cpu_dev,
> > + "invalid Vproc value: %d\n", pre_vproc);
> > + return pre_vproc;
> > + }
> >
> > - pre_vsram = regulator_get_voltage(sram_reg);
> > - if (pre_vsram < 0) {
> > - dev_err(info->cpu_dev, "invalid Vsram value: %d\n", pre_vsram);
> > - return pre_vsram;
> > - }
> > + vsram = min(new_vsram,
> > + pre_vproc + soc_data->min_volt_shift);
> >
> > - new_vsram = clamp(new_vproc + soc_data->min_volt_shift,
> > - soc_data->sram_min_volt, soc_data->sram_max_volt);
> > + if (vsram + VOLT_TOL >= soc_data->sram_max_volt) {
> > + vsram = soc_data->sram_max_volt;
> >
> > - do {
> > - if (pre_vproc <= new_vproc) {
> > - vsram = clamp(pre_vproc + soc_data->max_volt_shift,
> > - soc_data->sram_min_volt, new_vsram);
> > - ret = regulator_set_voltage(sram_reg, vsram,
> > - soc_data->sram_max_volt);
> > + /*
> > + * If the target Vsram hits the maximum voltage,
> > + * try to set the exact voltage value first.
> > + */
> > + ret = regulator_set_voltage(sram_reg, vsram,
> > + vsram);
> > + if (ret)
> > + ret = regulator_set_voltage(sram_reg,
> > + vsram - VOLT_TOL,
> > + vsram);
> >
> > - if (ret)
> > - return ret;
> > -
> > - if (vsram == soc_data->sram_max_volt ||
> > - new_vsram == soc_data->sram_min_volt)
> > vproc = new_vproc;
> > - else
> > + } else {
> > + ret = regulator_set_voltage(sram_reg, vsram,
> > + vsram + VOLT_TOL);
> > +
> > vproc = vsram - soc_data->min_volt_shift;
> > + }
> > + if (ret)
> > + return ret;
> >
> > ret = regulator_set_voltage(proc_reg, vproc,
> > - soc_data->proc_max_volt);
> > + vproc + VOLT_TOL);
> > if (ret) {
> > regulator_set_voltage(sram_reg, pre_vsram,
> > - soc_data->sram_max_volt);
> > + pre_vsram);
> > return ret;
> > }
> > - } else if (pre_vproc > new_vproc) {
> > + } while (vproc < new_vproc || vsram < new_vsram);
> > + } else if (pre_vproc > new_vproc) {
> > + /*
> > + * When scaling down voltages, Vsram and Vproc scale down step
> > + * by step. At each step, set Vproc to (Vsram - 200mV) first,
> > + * then set Vproc to (Vproc + 100mV).
> > + * Keep doing it until Vsram and Vproc hit target voltages.
> > + */
> > + do {
> > + pre_vproc = regulator_get_voltage(proc_reg);
> > + if (pre_vproc < 0) {
> > + dev_err(info->cpu_dev,
> > + "invalid Vproc value: %d\n", pre_vproc);
> > + return pre_vproc;
> > + }
> > + pre_vsram = regulator_get_voltage(sram_reg);
> > + if (pre_vsram < 0) {
> > + dev_err(info->cpu_dev,
> > + "invalid Vsram value: %d\n", pre_vsram);
> > + return pre_vsram;
> > + }
> > +
> > vproc = max(new_vproc,
> > pre_vsram - soc_data->max_volt_shift);
> > ret = regulator_set_voltage(proc_reg, vproc,
> > - soc_data->proc_max_volt);
> > + vproc + VOLT_TOL);
> > if (ret)
> > return ret;
> >
> > @@ -137,24 +183,32 @@ static int mtk_cpufreq_voltage_tracking(struct mtk_cpu_dvfs_info *info,
> > vsram = max(new_vsram,
> > vproc + soc_data->min_volt_shift);
> >
> > - ret = regulator_set_voltage(sram_reg, vsram,
> > - soc_data->sram_max_volt);
> > + if (vsram + VOLT_TOL >= soc_data->sram_max_volt) {
> > + vsram = soc_data->sram_max_volt;
> > +
> > + /*
> > + * If the target Vsram hits the maximum voltage,
> > + * try to set the exact voltage value first.
> > + */
> > + ret = regulator_set_voltage(sram_reg, vsram,
> > + vsram);
> > + if (ret)
> > + ret = regulator_set_voltage(sram_reg,
> > + vsram - VOLT_TOL,
> > + vsram);
> > + } else {
> > + ret = regulator_set_voltage(sram_reg, vsram,
> > + vsram + VOLT_TOL);
> > + }
> > +
> > if (ret) {
> > regulator_set_voltage(proc_reg, pre_vproc,
> > - soc_data->proc_max_volt);
> > + pre_vproc);
> > return ret;
> > }
> > - }
> > -
> > - pre_vproc = vproc;
> > - pre_vsram = vsram;
> > -
> > - if (--retry < 0) {
> > - dev_err(info->cpu_dev,
> > - "over loop count, failed to set voltage\n");
> > - return -EINVAL;
> > - }
> > - } while (vproc != new_vproc || vsram != new_vsram);
> > + } while (vproc > new_vproc + VOLT_TOL ||
> > + vsram > new_vsram + VOLT_TOL);
> > + }
> >
> > return 0;
> > }
> > @@ -250,8 +304,8 @@ static int mtk_cpufreq_set_target(struct cpufreq_policy *policy,
> > * If the new voltage or the intermediate voltage is higher than the
> > * current voltage, scale up voltage first.
> > */
> > - target_vproc = max(inter_vproc, vproc);
> > - if (pre_vproc <= target_vproc) {
> > + target_vproc = (inter_vproc > vproc) ? inter_vproc : vproc;
> > + if (pre_vproc < target_vproc) {
> > ret = mtk_cpufreq_set_voltage(info, target_vproc);
> > if (ret) {
> > dev_err(cpu_dev,
> > @@ -513,15 +567,6 @@ static int mtk_cpu_dvfs_info_init(struct mtk_cpu_dvfs_info *info, int cpu)
> > */
> > info->need_voltage_tracking = (info->sram_reg != NULL);
> >
> > - /*
> > - * We assume min voltage is 0 and tracking target voltage using
> > - * min_volt_shift for each iteration.
> > - * The vtrack_max is 3 times of expeted iteration count.
> > - */
> > - info->vtrack_max = 3 * DIV_ROUND_UP(max(info->soc_data->sram_max_volt,
> > - info->soc_data->proc_max_volt),
> > - info->soc_data->min_volt_shift);
> > -
> > return 0;
> >
> > out_disable_inter_clock:
> > --
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
next prev parent reply other threads:[~2022-12-02 19:47 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-09 13:35 Kernel Kernel bug caused by (cpufreq: mediatek: Refine mtk_cpufreq_voltage_tracking()) on Banana Pi R64 (MT7622) Nick
2022-11-09 19:40 ` Kernel Kernel bug caused by (cpufreq: mediatek: Refine mtk_cpufreq_voltage_tracking()) on Banana Pi R64 (MT7622) #forregzbot Thorsten Leemhuis
2022-11-10 11:26 ` Kernel Kernel bug caused by (cpufreq: mediatek: Refine mtk_cpufreq_voltage_tracking()) on Banana Pi R64 (MT7622) Matthias Brugger
2022-11-15 19:44 ` Nick
2022-12-01 15:08 ` Thorsten Leemhuis
2022-12-01 15:26 ` Daniel Golle
2022-12-01 15:39 ` Thorsten Leemhuis
2022-12-01 21:40 ` Nick
2022-12-02 5:27 ` Viresh Kumar
2022-12-02 8:57 ` AngeloGioacchino Del Regno
2022-12-02 9:19 ` AngeloGioacchino Del Regno
2022-12-02 9:43 ` Allen-KH Cheng (程冠勳)
2022-12-02 10:02 ` AngeloGioacchino Del Regno
2022-12-02 10:41 ` Thorsten Leemhuis
2022-12-02 10:47 ` Viresh Kumar
2022-12-02 10:51 ` Thorsten Leemhuis
2022-12-02 11:00 ` AngeloGioacchino Del Regno
2022-12-02 11:01 ` Viresh Kumar
2022-12-07 15:34 ` Nick
2022-12-19 12:21 ` Allen-KH Cheng (程冠勳)
2022-12-20 14:37 ` Nick
2022-12-21 9:24 ` Kernel Kernel bug caused by (cpufreq: mediatek: Refine mtk_cpufreq_voltage_tracking()) on Banana Pi R64 (MT7622) #forregzbot Thorsten Leemhuis
2022-12-02 12:25 ` Kernel Kernel bug caused by (cpufreq: mediatek: Refine mtk_cpufreq_voltage_tracking()) on Banana Pi R64 (MT7622) Nick
2022-12-02 5:26 ` [PATCH] Revert "cpufreq: mediatek: Refine mtk_cpufreq_voltage_tracking()" Viresh Kumar
2022-12-02 12:17 ` Rafael J. Wysocki
2022-12-02 19:46 ` Rafael J. Wysocki [this message]
2022-12-05 4:30 ` Viresh Kumar
2022-12-05 12:08 ` Rafael J. Wysocki
2022-12-05 23:30 ` Viresh Kumar
2022-12-02 12:47 ` Nick
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAJZ5v0hc8CsvqLKxi5iRq7iR0bkt25dRnLBd23mx-zdi2Sjgsw@mail.gmail.com \
--to=rafael@kernel.org \
--cc=daniel@makrotopia.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mediatek@lists.infradead.org \
--cc=linux-pm@vger.kernel.org \
--cc=matthias.bgg@gmail.com \
--cc=regressions@leemhuis.info \
--cc=stable@vger.kernel.org \
--cc=thomas.huehn@hs-nordhausen.de \
--cc=vincent.guittot@linaro.org \
--cc=vincent@systemli.org \
--cc=viresh.kumar@linaro.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).