| Message ID | 20200924110449.329523-1-ulf.hansson@linaro.org |
|---|---|
| Headers | show |
| Series | PM / Domains: Add power on/off notifiers for genpd | expand |
> Subject: Re: [PATCH v2 3/3] PM / Domains: Add support for PM domain on/off > notifiers for genpd > > On Fri, 25 Sep 2020 at 08:08, Peng Fan <peng.fan@nxp.com> wrote: > > > > Hi Ulf, > > > > > Subject: [PATCH v2 3/3] PM / Domains: Add support for PM domain > > > on/off notifiers for genpd > > > > > > A device may have specific HW constraints that must be obeyed to, > > > before its corresponding PM domain (genpd) can be powered off - and > > > vice verse at power on. These constraints can't be managed through > > > the regular runtime PM based deployment for a device, because the > > > access pattern for it, isn't always request based. In other words, > > > using the runtime PM callbacks to deal with the constraints doesn't work > for these cases. > > > > Could the notification be added before/after power on, and > > before/after power off? not just after power on and before power off? > > > > Our SoC has a requirement that before power on/off the specific > > module, the corresponding clk needs to be on to make sure the hardware > > async bridge could finish handshake. > > Thanks for your comments! > > May I ask, to be sure - does the clock correspond to the genpd provider or is it > a clock for the genpd consumer device? It is the clock for the genpd consumer device. > > If the former, couldn't the clock be managed from the ->power_on|off() > callbacks for the genpd provider? Sadly not former. Our current solution is to add a clock property to the power domain node(NXP ARM SIP based power domain driver), and when power_on/off, we enable/disable the clocks. But we are moving to use SCMI power domain, and leave the clock in Linux, Scmi power domain driver is a generic driver, and we are not able to mix clock parts in the driver which is bad. Your patch gives me a light that if we could use notification to let consumer device driver do the clock enable/disable for each power on and power off. That would be great. Thanks, Peng. > > > > > So we need clk_prepare_on/off to guard power on and power off as below: > > > > clk_prepare_on > > power on > > clk_prepare_off > > > > clk_prepare_on > > power off > > clk_prepare_off > > > > Thanks, > > Peng. > > > > [...] > > Kind regards > Uffe
On Fri, 25 Sep 2020 at 16:30, Peng Fan <peng.fan@nxp.com> wrote: > > > Subject: Re: [PATCH v2 3/3] PM / Domains: Add support for PM domain on/off > > notifiers for genpd > > > > On Fri, 25 Sep 2020 at 08:08, Peng Fan <peng.fan@nxp.com> wrote: > > > > > > Hi Ulf, > > > > > > > Subject: [PATCH v2 3/3] PM / Domains: Add support for PM domain > > > > on/off notifiers for genpd > > > > > > > > A device may have specific HW constraints that must be obeyed to, > > > > before its corresponding PM domain (genpd) can be powered off - and > > > > vice verse at power on. These constraints can't be managed through > > > > the regular runtime PM based deployment for a device, because the > > > > access pattern for it, isn't always request based. In other words, > > > > using the runtime PM callbacks to deal with the constraints doesn't work > > for these cases. > > > > > > Could the notification be added before/after power on, and > > > before/after power off? not just after power on and before power off? > > > > > > Our SoC has a requirement that before power on/off the specific > > > module, the corresponding clk needs to be on to make sure the hardware > > > async bridge could finish handshake. > > > > Thanks for your comments! > > > > May I ask, to be sure - does the clock correspond to the genpd provider or is it > > a clock for the genpd consumer device? > > It is the clock for the genpd consumer device. > > > > > If the former, couldn't the clock be managed from the ->power_on|off() > > callbacks for the genpd provider? > > Sadly not former. > > Our current solution is to add a clock property to the power domain node(NXP ARM SIP > based power domain driver), and when power_on/off, we enable/disable the clocks. > > But we are moving to use SCMI power domain, and leave the clock in Linux, > Scmi power domain driver is a generic driver, and we are not able to mix clock > parts in the driver which is bad. > > Your patch gives me a light that if we could use notification to let consumer > device driver do the clock enable/disable for each power on and power off. > That would be great. Thanks for clarifying. Let me re-spin the patch to address your points. [...] Kind regards Uffe
Rafael, On Thu, 24 Sep 2020 at 13:06, Ulf Hansson <ulf.hansson@linaro.org> wrote: > > Changes in v2: > - Improved error handling in patch3. > > A device may have specific HW constraints that must be obeyed to, before its > corresponding PM domain (genpd) can be powered off - and vice verse at power > on. These constraints can't be managed through the regular runtime PM based > deployment for a device, because the access pattern for it, isn't always > request based. In other words, using the runtime PM callbacks to deal with the > constraints doesn't work for these cases. > > For these reasons, this series introduces a power on/off notification mechanism > to genpd. To add/remove a notifier for a device, the device must already have > been attached to the genpd, which also means that it needs to be a part of the > PM domain topology. > > The intent is to allow these genpd power on/off notifiers to replace the need > for the existing CPU_CLUSTER_PM_ENTER|EXIT notifiers. For example, those would > otherwise be needed in psci_pd_power_off() in cpuidle-psci-domain.c, when > powering off the CPU cluster. > > Another series that enables drivers/soc/qcom/rpmh-rsc.c to make use of the new > genpd on/off notifiers, are soon to be posted. However, I would appreciate any > feedback on the approach taken, even before that series hits LKML. > > Kind regards > Ulf Hansson > > > Ulf Hansson (3): > PM / Domains: Rename power state enums for genpd > PM / Domains: Allow to abort power off when no ->power_off() callback > PM / Domains: Add support for PM domain on/off notifiers for genpd > > drivers/base/power/domain.c | 187 +++++++++++++++++++++++++++++------- > include/linux/pm_domain.h | 19 +++- > 2 files changed, 171 insertions(+), 35 deletions(-) > > -- > 2.25.1 > I will need to iterate patch 3, potentially even a couple of more times. As I expect patch 1 and patch2 to not get changed, may I suggest that you pick up those so we can move focus to patch3? Kind regards Uffe
On Mon, Sep 28, 2020 at 1:57 PM Ulf Hansson <ulf.hansson@linaro.org> wrote: > > Rafael, > > On Thu, 24 Sep 2020 at 13:06, Ulf Hansson <ulf.hansson@linaro.org> wrote: > > > > Changes in v2: > > - Improved error handling in patch3. > > > > A device may have specific HW constraints that must be obeyed to, before its > > corresponding PM domain (genpd) can be powered off - and vice verse at power > > on. These constraints can't be managed through the regular runtime PM based > > deployment for a device, because the access pattern for it, isn't always > > request based. In other words, using the runtime PM callbacks to deal with the > > constraints doesn't work for these cases. > > > > For these reasons, this series introduces a power on/off notification mechanism > > to genpd. To add/remove a notifier for a device, the device must already have > > been attached to the genpd, which also means that it needs to be a part of the > > PM domain topology. > > > > The intent is to allow these genpd power on/off notifiers to replace the need > > for the existing CPU_CLUSTER_PM_ENTER|EXIT notifiers. For example, those would > > otherwise be needed in psci_pd_power_off() in cpuidle-psci-domain.c, when > > powering off the CPU cluster. > > > > Another series that enables drivers/soc/qcom/rpmh-rsc.c to make use of the new > > genpd on/off notifiers, are soon to be posted. However, I would appreciate any > > feedback on the approach taken, even before that series hits LKML. > > > > Kind regards > > Ulf Hansson > > > > > > Ulf Hansson (3): > > PM / Domains: Rename power state enums for genpd > > PM / Domains: Allow to abort power off when no ->power_off() callback > > PM / Domains: Add support for PM domain on/off notifiers for genpd > > > > drivers/base/power/domain.c | 187 +++++++++++++++++++++++++++++------- > > include/linux/pm_domain.h | 19 +++- > > 2 files changed, 171 insertions(+), 35 deletions(-) > > > > -- > > 2.25.1 > > > > I will need to iterate patch 3, potentially even a couple of more times. > > As I expect patch 1 and patch2 to not get changed, may I suggest that > you pick up those so we can move focus to patch3? OK, [1-2/3] applied as 5.10 material with minor subject edits, thanks!