| Message ID | 20210616053335.4181780-1-hsinyi@chromium.org |
|---|---|
| State | New |
| Headers | show |
| Series | opp: of: Allow lazy-linking of required-opps to non genpd | expand |
On 16-06-21, 13:33, Hsin-Yi Wang wrote: > Don't limit required_opp_table to genpd only. One possible use case is > cpufreq based devfreq governor, which can use required-opps property to > derive devfreq from cpufreq. > > Suggested-by: Chanwoo Choi <cw00.choi@samsung.com> > Signed-off-by: Hsin-Yi Wang <hsinyi@chromium.org> > --- > This is tested with the non genpd case mt8183-cci with passive > governor[1]. > [1] https://patchwork.kernel.org/project/linux-mediatek/patch/1616499241-4906-2-git-send-email-andrew-sh.cheng@mediatek.com/ > --- > drivers/opp/of.c | 20 +------------------- > 1 file changed, 1 insertion(+), 19 deletions(-) > > diff --git a/drivers/opp/of.c b/drivers/opp/of.c > index aa75a1caf08a3..9573facce53a5 100644 > --- a/drivers/opp/of.c > +++ b/drivers/opp/of.c > @@ -201,17 +201,6 @@ static void _opp_table_alloc_required_tables(struct opp_table *opp_table, > lazy = true; > continue; > } > - > - /* > - * We only support genpd's OPPs in the "required-opps" for now, > - * as we don't know how much about other cases. Error out if the > - * required OPP doesn't belong to a genpd. > - */ > - if (!required_opp_tables[i]->is_genpd) { > - dev_err(dev, "required-opp doesn't belong to genpd: %pOF\n", > - required_np); > - goto free_required_tables; > - } > } > > /* Let's do the linking later on */ > @@ -379,13 +368,6 @@ static void lazy_link_required_opp_table(struct opp_table *new_table) > struct dev_pm_opp *opp; > int i, ret; > > - /* > - * We only support genpd's OPPs in the "required-opps" for now, > - * as we don't know much about other cases. > - */ > - if (!new_table->is_genpd) > - return; > - > mutex_lock(&opp_table_lock); > > list_for_each_entry_safe(opp_table, temp, &lazy_opp_tables, lazy) { > @@ -873,7 +855,7 @@ static struct dev_pm_opp *_opp_add_static_v2(struct opp_table *opp_table, > return ERR_PTR(-ENOMEM); > > ret = _read_opp_key(new_opp, opp_table, np, &rate_not_available); > - if (ret < 0 && !opp_table->is_genpd) { > + if (ret < 0) { > dev_err(dev, "%s: opp key field not found\n", __func__); > goto free_opp; > } How are you setting the required OPPs ? I mean when someone tries to change frequency of a device which has some required-opps, who is configuring those required OPPs ?
On Wed, Jun 16, 2021 at 1:58 PM Viresh Kumar <viresh.kumar@linaro.org> wrote: > > On 16-06-21, 13:33, Hsin-Yi Wang wrote: > > Don't limit required_opp_table to genpd only. One possible use case is > > cpufreq based devfreq governor, which can use required-opps property to > > derive devfreq from cpufreq. > > > > Suggested-by: Chanwoo Choi <cw00.choi@samsung.com> > > Signed-off-by: Hsin-Yi Wang <hsinyi@chromium.org> > > --- > > This is tested with the non genpd case mt8183-cci with passive > > governor[1]. > > [1] https://patchwork.kernel.org/project/linux-mediatek/patch/1616499241-4906-2-git-send-email-andrew-sh.cheng@mediatek.com/ > > --- > > drivers/opp/of.c | 20 +------------------- > > 1 file changed, 1 insertion(+), 19 deletions(-) > > > > diff --git a/drivers/opp/of.c b/drivers/opp/of.c > > index aa75a1caf08a3..9573facce53a5 100644 > > --- a/drivers/opp/of.c > > +++ b/drivers/opp/of.c > > @@ -201,17 +201,6 @@ static void _opp_table_alloc_required_tables(struct opp_table *opp_table, > > lazy = true; > > continue; > > } > > - > > - /* > > - * We only support genpd's OPPs in the "required-opps" for now, > > - * as we don't know how much about other cases. Error out if the > > - * required OPP doesn't belong to a genpd. > > - */ > > - if (!required_opp_tables[i]->is_genpd) { > > - dev_err(dev, "required-opp doesn't belong to genpd: %pOF\n", > > - required_np); > > - goto free_required_tables; > > - } > > } > > > > /* Let's do the linking later on */ > > @@ -379,13 +368,6 @@ static void lazy_link_required_opp_table(struct opp_table *new_table) > > struct dev_pm_opp *opp; > > int i, ret; > > > > - /* > > - * We only support genpd's OPPs in the "required-opps" for now, > > - * as we don't know much about other cases. > > - */ > > - if (!new_table->is_genpd) > > - return; > > - > > mutex_lock(&opp_table_lock); > > > > list_for_each_entry_safe(opp_table, temp, &lazy_opp_tables, lazy) { > > @@ -873,7 +855,7 @@ static struct dev_pm_opp *_opp_add_static_v2(struct opp_table *opp_table, > > return ERR_PTR(-ENOMEM); > > > > ret = _read_opp_key(new_opp, opp_table, np, &rate_not_available); > > - if (ret < 0 && !opp_table->is_genpd) { > > + if (ret < 0) { > > dev_err(dev, "%s: opp key field not found\n", __func__); > > goto free_opp; > > } > > How are you setting the required OPPs ? I mean when someone tries to > change frequency of a device which has some required-opps, who is > configuring those required OPPs ? > When cpufreq changes, the (cpufreq based) passive governor will calculate a target devfreq based on that, and the device governor (mt8183-cci-devfreq) will change the actual opp of the device. The required-opp is set in the cpufreq table: cpufreq_opp_table: opp_table0 { compatible = "operating-points-v2"; opp-shared; ... opp0_01 { opp-hz = /bits/ 64 <910000000>; opp-microvolt = <687500>; required-opps = <&opp2_01>; }; ... }; devfreq_opp_table: opp_table2 { compatible = "operating-points-v2"; opp-shared; ... opp2_01: opp-338000000 { opp-hz = /bits/ 64 <338000000>; opp-microvolt = <687500>; }; ... }; > -- > viresh
On 16-06-21, 17:47, Chanwoo Choi wrote: > On 6/16/21 4:55 PM, Viresh Kumar wrote: > > On 16-06-21, 13:33, Hsin-Yi Wang wrote: > >> Don't limit required_opp_table to genpd only. One possible use case is > >> cpufreq based devfreq governor, which can use required-opps property to > >> derive devfreq from cpufreq. > >> > >> Suggested-by: Chanwoo Choi <cw00.choi@samsung.com> > >> Signed-off-by: Hsin-Yi Wang <hsinyi@chromium.org> > >> --- > >> This is tested with the non genpd case mt8183-cci with passive > >> governor[1]. > >> [1] https://patchwork.kernel.org/project/linux-mediatek/patch/1616499241-4906-2-git-send-email-andrew-sh.cheng@mediatek.com/ > >> --- > >> drivers/opp/of.c | 20 +------------------- > >> 1 file changed, 1 insertion(+), 19 deletions(-) > >> > >> diff --git a/drivers/opp/of.c b/drivers/opp/of.c > >> index aa75a1caf08a3..9573facce53a5 100644 > >> --- a/drivers/opp/of.c > >> +++ b/drivers/opp/of.c > >> @@ -201,17 +201,6 @@ static void _opp_table_alloc_required_tables(struct opp_table *opp_table, > >> lazy = true; > >> continue; > >> } > >> - > >> - /* > >> - * We only support genpd's OPPs in the "required-opps" for now, > >> - * as we don't know how much about other cases. Error out if the > >> - * required OPP doesn't belong to a genpd. > >> - */ > >> - if (!required_opp_tables[i]->is_genpd) { > >> - dev_err(dev, "required-opp doesn't belong to genpd: %pOF\n", > >> - required_np); > >> - goto free_required_tables; > >> - } > >> } > >> > >> /* Let's do the linking later on */ > >> @@ -379,13 +368,6 @@ static void lazy_link_required_opp_table(struct opp_table *new_table) > >> struct dev_pm_opp *opp; > >> int i, ret; > >> > >> - /* > >> - * We only support genpd's OPPs in the "required-opps" for now, > >> - * as we don't know much about other cases. > >> - */ > >> - if (!new_table->is_genpd) > >> - return; > >> - > >> mutex_lock(&opp_table_lock); > >> > >> list_for_each_entry_safe(opp_table, temp, &lazy_opp_tables, lazy) { > >> @@ -873,7 +855,7 @@ static struct dev_pm_opp *_opp_add_static_v2(struct opp_table *opp_table, > >> return ERR_PTR(-ENOMEM); > >> > >> ret = _read_opp_key(new_opp, opp_table, np, &rate_not_available); > >> - if (ret < 0 && !opp_table->is_genpd) { > >> + if (ret < 0) { > >> dev_err(dev, "%s: opp key field not found\n", __func__); > >> goto free_opp; > >> } > > > > Plus this and few changes to commit log. > > > > diff --git a/drivers/opp/core.c b/drivers/opp/core.c > > index e366218d6736..b335c077f215 100644 > > --- a/drivers/opp/core.c > > +++ b/drivers/opp/core.c > > @@ -893,6 +893,16 @@ static int _set_required_opps(struct device *dev, > > if (!required_opp_tables) > > return 0; > > > > + /* > > + * We only support genpd's OPPs in the "required-opps" for now, as we > > + * don't know much about other use cases. Error out if the required OPP > > + * doesn't belong to a genpd. > > + */ > > + if (unlikely(!required_opp_tables[0]->is_genpd)) { > > + dev_err(dev, "required-opps don't belong to a genpd\n"); > > + return -ENOENT; > > + } > > + > > If you add this checking statement, I think that > when using dev_pm_opp_set_rate with required-opp property, it will be failed. Yes, that is exactly what I am trying to do here. Hsin already confirmed that you guys won't use this API, isn't ? The point here is that the _set_required_opps() function only updates the performance state of genpds today. So it won't work for you guys anyway.
On Wed, Jun 16, 2021 at 5:09 PM Viresh Kumar <viresh.kumar@linaro.org> wrote: > > On 16-06-21, 17:47, Chanwoo Choi wrote: > > On 6/16/21 4:55 PM, Viresh Kumar wrote: > > > On 16-06-21, 13:33, Hsin-Yi Wang wrote: > > >> Don't limit required_opp_table to genpd only. One possible use case is > > >> cpufreq based devfreq governor, which can use required-opps property to > > >> derive devfreq from cpufreq. > > >> > > >> Suggested-by: Chanwoo Choi <cw00.choi@samsung.com> > > >> Signed-off-by: Hsin-Yi Wang <hsinyi@chromium.org> > > >> --- > > >> This is tested with the non genpd case mt8183-cci with passive > > >> governor[1]. > > >> [1] https://patchwork.kernel.org/project/linux-mediatek/patch/1616499241-4906-2-git-send-email-andrew-sh.cheng@mediatek.com/ > > >> --- > > >> drivers/opp/of.c | 20 +------------------- > > >> 1 file changed, 1 insertion(+), 19 deletions(-) > > >> > > >> diff --git a/drivers/opp/of.c b/drivers/opp/of.c > > >> index aa75a1caf08a3..9573facce53a5 100644 > > >> --- a/drivers/opp/of.c > > >> +++ b/drivers/opp/of.c > > >> @@ -201,17 +201,6 @@ static void _opp_table_alloc_required_tables(struct opp_table *opp_table, > > >> lazy = true; > > >> continue; > > >> } > > >> - > > >> - /* > > >> - * We only support genpd's OPPs in the "required-opps" for now, > > >> - * as we don't know how much about other cases. Error out if the > > >> - * required OPP doesn't belong to a genpd. > > >> - */ > > >> - if (!required_opp_tables[i]->is_genpd) { > > >> - dev_err(dev, "required-opp doesn't belong to genpd: %pOF\n", > > >> - required_np); > > >> - goto free_required_tables; > > >> - } > > >> } > > >> > > >> /* Let's do the linking later on */ > > >> @@ -379,13 +368,6 @@ static void lazy_link_required_opp_table(struct opp_table *new_table) > > >> struct dev_pm_opp *opp; > > >> int i, ret; > > >> > > >> - /* > > >> - * We only support genpd's OPPs in the "required-opps" for now, > > >> - * as we don't know much about other cases. > > >> - */ > > >> - if (!new_table->is_genpd) > > >> - return; > > >> - > > >> mutex_lock(&opp_table_lock); > > >> > > >> list_for_each_entry_safe(opp_table, temp, &lazy_opp_tables, lazy) { > > >> @@ -873,7 +855,7 @@ static struct dev_pm_opp *_opp_add_static_v2(struct opp_table *opp_table, > > >> return ERR_PTR(-ENOMEM); > > >> > > >> ret = _read_opp_key(new_opp, opp_table, np, &rate_not_available); > > >> - if (ret < 0 && !opp_table->is_genpd) { > > >> + if (ret < 0) { > > >> dev_err(dev, "%s: opp key field not found\n", __func__); > > >> goto free_opp; > > >> } > > > > > > Plus this and few changes to commit log. > > > > > > diff --git a/drivers/opp/core.c b/drivers/opp/core.c > > > index e366218d6736..b335c077f215 100644 > > > --- a/drivers/opp/core.c > > > +++ b/drivers/opp/core.c > > > @@ -893,6 +893,16 @@ static int _set_required_opps(struct device *dev, > > > if (!required_opp_tables) > > > return 0; > > > > > > + /* > > > + * We only support genpd's OPPs in the "required-opps" for now, as we > > > + * don't know much about other use cases. Error out if the required OPP > > > + * doesn't belong to a genpd. > > > + */ > > > + if (unlikely(!required_opp_tables[0]->is_genpd)) { > > > + dev_err(dev, "required-opps don't belong to a genpd\n"); > > > + return -ENOENT; > > > + } > > > + > > > > If you add this checking statement, I think that > > when using dev_pm_opp_set_rate with required-opp property, it will be failed. > > Yes, that is exactly what I am trying to do here. Hsin already > confirmed that you guys won't use this API, isn't ? we used clk_set_rate() and regulator_set_voltage(). > > The point here is that the _set_required_opps() function only updates > the performance state of genpds today. So it won't work for you guys > anyway. > > -- > viresh
On 6/16/21 6:09 PM, Viresh Kumar wrote: > On 16-06-21, 17:47, Chanwoo Choi wrote: >> On 6/16/21 4:55 PM, Viresh Kumar wrote: >>> On 16-06-21, 13:33, Hsin-Yi Wang wrote: >>>> Don't limit required_opp_table to genpd only. One possible use case is >>>> cpufreq based devfreq governor, which can use required-opps property to >>>> derive devfreq from cpufreq. >>>> >>>> Suggested-by: Chanwoo Choi <cw00.choi@samsung.com> >>>> Signed-off-by: Hsin-Yi Wang <hsinyi@chromium.org> >>>> --- >>>> This is tested with the non genpd case mt8183-cci with passive >>>> governor[1]. >>>> [1] https://patchwork.kernel.org/project/linux-mediatek/patch/1616499241-4906-2-git-send-email-andrew-sh.cheng@mediatek.com/ >>>> --- >>>> drivers/opp/of.c | 20 +------------------- >>>> 1 file changed, 1 insertion(+), 19 deletions(-) >>>> >>>> diff --git a/drivers/opp/of.c b/drivers/opp/of.c >>>> index aa75a1caf08a3..9573facce53a5 100644 >>>> --- a/drivers/opp/of.c >>>> +++ b/drivers/opp/of.c >>>> @@ -201,17 +201,6 @@ static void _opp_table_alloc_required_tables(struct opp_table *opp_table, >>>> lazy = true; >>>> continue; >>>> } >>>> - >>>> - /* >>>> - * We only support genpd's OPPs in the "required-opps" for now, >>>> - * as we don't know how much about other cases. Error out if the >>>> - * required OPP doesn't belong to a genpd. >>>> - */ >>>> - if (!required_opp_tables[i]->is_genpd) { >>>> - dev_err(dev, "required-opp doesn't belong to genpd: %pOF\n", >>>> - required_np); >>>> - goto free_required_tables; >>>> - } >>>> } >>>> >>>> /* Let's do the linking later on */ >>>> @@ -379,13 +368,6 @@ static void lazy_link_required_opp_table(struct opp_table *new_table) >>>> struct dev_pm_opp *opp; >>>> int i, ret; >>>> >>>> - /* >>>> - * We only support genpd's OPPs in the "required-opps" for now, >>>> - * as we don't know much about other cases. >>>> - */ >>>> - if (!new_table->is_genpd) >>>> - return; >>>> - >>>> mutex_lock(&opp_table_lock); >>>> >>>> list_for_each_entry_safe(opp_table, temp, &lazy_opp_tables, lazy) { >>>> @@ -873,7 +855,7 @@ static struct dev_pm_opp *_opp_add_static_v2(struct opp_table *opp_table, >>>> return ERR_PTR(-ENOMEM); >>>> >>>> ret = _read_opp_key(new_opp, opp_table, np, &rate_not_available); >>>> - if (ret < 0 && !opp_table->is_genpd) { >>>> + if (ret < 0) { >>>> dev_err(dev, "%s: opp key field not found\n", __func__); >>>> goto free_opp; >>>> } >>> >>> Plus this and few changes to commit log. >>> >>> diff --git a/drivers/opp/core.c b/drivers/opp/core.c >>> index e366218d6736..b335c077f215 100644 >>> --- a/drivers/opp/core.c >>> +++ b/drivers/opp/core.c >>> @@ -893,6 +893,16 @@ static int _set_required_opps(struct device *dev, >>> if (!required_opp_tables) >>> return 0; >>> >>> + /* >>> + * We only support genpd's OPPs in the "required-opps" for now, as we >>> + * don't know much about other use cases. Error out if the required OPP >>> + * doesn't belong to a genpd. >>> + */ >>> + if (unlikely(!required_opp_tables[0]->is_genpd)) { >>> + dev_err(dev, "required-opps don't belong to a genpd\n"); >>> + return -ENOENT; >>> + } >>> + >> >> If you add this checking statement, I think that >> when using dev_pm_opp_set_rate with required-opp property, it will be failed. > > Yes, that is exactly what I am trying to do here. Hsin already > confirmed that you guys won't use this API, isn't ? > > The point here is that the _set_required_opps() function only updates > the performance state of genpds today. So it won't work for you guys > anyway. The devfreq driver(exynos-bus.c) has used the dev_pm_opp_set_rate() and used the passive governor without the required-opp property. I have a plan to use the required-opp property between devfreq drivers (exynos-bus.c) with dev_pm_opp_set_rate(). I'll support them on later if this approach doesn't break the any rule of required-opp property. -- Best Regards, Chanwoo Choi Samsung Electronics
On 17-06-21, 10:13, Chanwoo Choi wrote: > The devfreq driver(exynos-bus.c) has used the dev_pm_opp_set_rate() > and used the passive governor without the required-opp property. Which is fine. > I have a plan to use the required-opp property > between devfreq drivers (exynos-bus.c) with dev_pm_opp_set_rate(). > > I'll support them on later if this approach doesn't break the any > rule of required-opp property. You will be required to make some changes in core for that I am afraid. It won't work automatically. -- viresh
On 17-06-21, 13:09, Chanwoo Choi wrote: > On 6/17/21 12:33 PM, Viresh Kumar wrote: > > On 17-06-21, 10:13, Chanwoo Choi wrote: > >> The devfreq driver(exynos-bus.c) has used the dev_pm_opp_set_rate() > >> and used the passive governor without the required-opp property. > > > > Which is fine. > > > >> I have a plan to use the required-opp property > >> between devfreq drivers (exynos-bus.c) with dev_pm_opp_set_rate(). > >> > >> I'll support them on later if this approach doesn't break the any > >> rule of required-opp property. > > > > You will be required to make some changes in core for that I am > > afraid. It won't work automatically. > > Do you think that better to use clk_enable/regulator_enable directly > instead of dev_pm_opp_set_rate() for using required-opp property? No. All I am saying is that the OPP core won't work for your use case today and may need some updates. -- viresh
On 6/17/21 12:33 PM, Viresh Kumar wrote: > On 17-06-21, 10:13, Chanwoo Choi wrote: >> The devfreq driver(exynos-bus.c) has used the dev_pm_opp_set_rate() >> and used the passive governor without the required-opp property. > > Which is fine. > >> I have a plan to use the required-opp property >> between devfreq drivers (exynos-bus.c) with dev_pm_opp_set_rate(). >> >> I'll support them on later if this approach doesn't break the any >> rule of required-opp property. > > You will be required to make some changes in core for that I am > afraid. It won't work automatically. Do you think that better to use clk_enable/regulator_enable directly instead of dev_pm_opp_set_rate() for using required-opp property? -- Best Regards, Chanwoo Choi Samsung Electronics
On 6/17/21 1:00 PM, Viresh Kumar wrote: > On 17-06-21, 13:09, Chanwoo Choi wrote: >> On 6/17/21 12:33 PM, Viresh Kumar wrote: >>> On 17-06-21, 10:13, Chanwoo Choi wrote: >>>> The devfreq driver(exynos-bus.c) has used the dev_pm_opp_set_rate() >>>> and used the passive governor without the required-opp property. >>> >>> Which is fine. >>> >>>> I have a plan to use the required-opp property >>>> between devfreq drivers (exynos-bus.c) with dev_pm_opp_set_rate(). >>>> >>>> I'll support them on later if this approach doesn't break the any >>>> rule of required-opp property. >>> >>> You will be required to make some changes in core for that I am >>> afraid. It won't work automatically. >> >> Do you think that better to use clk_enable/regulator_enable directly >> instead of dev_pm_opp_set_rate() for using required-opp property? > > No. All I am saying is that the OPP core won't work for your use case > today and may need some updates. OK. I'll update them for this case. -- Best Regards, Chanwoo Choi Samsung Electronics
diff --git a/drivers/opp/of.c b/drivers/opp/of.c index aa75a1caf08a3..9573facce53a5 100644 --- a/drivers/opp/of.c +++ b/drivers/opp/of.c @@ -201,17 +201,6 @@ static void _opp_table_alloc_required_tables(struct opp_table *opp_table, lazy = true; continue; } - - /* - * We only support genpd's OPPs in the "required-opps" for now, - * as we don't know how much about other cases. Error out if the - * required OPP doesn't belong to a genpd. - */ - if (!required_opp_tables[i]->is_genpd) { - dev_err(dev, "required-opp doesn't belong to genpd: %pOF\n", - required_np); - goto free_required_tables; - } } /* Let's do the linking later on */ @@ -379,13 +368,6 @@ static void lazy_link_required_opp_table(struct opp_table *new_table) struct dev_pm_opp *opp; int i, ret; - /* - * We only support genpd's OPPs in the "required-opps" for now, - * as we don't know much about other cases. - */ - if (!new_table->is_genpd) - return; - mutex_lock(&opp_table_lock); list_for_each_entry_safe(opp_table, temp, &lazy_opp_tables, lazy) { @@ -873,7 +855,7 @@ static struct dev_pm_opp *_opp_add_static_v2(struct opp_table *opp_table, return ERR_PTR(-ENOMEM); ret = _read_opp_key(new_opp, opp_table, np, &rate_not_available); - if (ret < 0 && !opp_table->is_genpd) { + if (ret < 0) { dev_err(dev, "%s: opp key field not found\n", __func__); goto free_opp; }
Don't limit required_opp_table to genpd only. One possible use case is cpufreq based devfreq governor, which can use required-opps property to derive devfreq from cpufreq. Suggested-by: Chanwoo Choi <cw00.choi@samsung.com> Signed-off-by: Hsin-Yi Wang <hsinyi@chromium.org> --- This is tested with the non genpd case mt8183-cci with passive governor[1]. [1] https://patchwork.kernel.org/project/linux-mediatek/patch/1616499241-4906-2-git-send-email-andrew-sh.cheng@mediatek.com/ --- drivers/opp/of.c | 20 +------------------- 1 file changed, 1 insertion(+), 19 deletions(-)