Re: [PATCH v2 2/3] PM / core: Add IN_BAND_WAKEUP driver flag

[Ulf Hansson <ulf.hansson@linaro.org>]

On 10 December 2017 at 11:16, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> Hi Rafael, Ulf,
>
> On Sun, Dec 10, 2017 at 3:30 AM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote:
>> On Monday, November 13, 2017 4:46:42 PM CET Ulf Hansson wrote:
>>> For some bus types and PM domains, it's not sufficient to only check the
>>> return value from device_may_wakeup(), to fully understand how to configure
>>> wakeup settings for the device during system suspend.
>>>
>>> In particular, sometimes the device may need to remain in its power state,
>>> in case the driver has configured it for in band IRQs, as to be able to
>>> generate wakeup signals. Therefore, define and document an IN_BAND_WAKEUP
>>> driver flag, to enable drivers to instruct bus types and PM domains about
>>> this setting.
>>>
>>> Of course, in case the bus type and PM domain has additional information
>>> about wakeup settings of the device, they may override the behaviour.
>>>
>>> Using the IN_BAND_WAKEUP driver flag for a device, may also affect how bus
>>> types and PM domains should treat the device's parent during system
>>> suspend. Therefore, in __device_suspend(), let the PM core propagate the
>>> wakeup setting by using a status flag in the struct dev_pm_info for the
>>> parent. This also makes it consistent with how the existing "wakeup_path"
>>> status flag is being assigned.
>>
>> I've been thinking about this quite a bit recently and my conclusion is that
>> the flag makes perfect sence (as it covers a valid use case), but I would
>> define it and design the handling of it a bit differently.
>
> I'm also still thinking about this, cfr. my recent silence w.r.t.
> these matters...
>
> On Renesas ARM SoCs (and at least some SH SoCs, but these are stuck in the
> pre-DT legacy clock domain), a device needs to be kept running if it is a
> wake-up source (e.g. WoL), or if it's part of the wake-up patch (e.g.
> irqchip). So in the absence of the GENPD_FLAG_ACTIVE_WAKEUP flag in the PM
> Domain driver, all individual drivers would need to set the IN_BAND_WAKEUP
> flag (but see the exception below)

Could you elaborate a bit about the wakeup-path? Let me start and you
can fill in.

In $subject patch, the PM core propagates the IN_BAND_WAKEUP flag of
the device to its parent device, via the ->power.wakeup_path_in_band
flag. That becomes additional new information, as we already has the
PM core to propagate the value of device_may_wakeup() (if true) to the
parent device, via the ->power.wakeup_path flag.

I assume that isn't sufficient, because the wakeup path isn't
reflected in the child/parent hierarchy?

So, I guess this is about drivers who deals with wakeup enabled
devices and which are consumers of other resources (irqchips for
example). As part of the wakup path, these resources also needs to be
kept running, right?

In your case, what other resources besides the irqchip, can be
considered as a part of the wakeup path, but also being outside the
parent/child hierarchy of the wakeup enabled device?

Note that, an important fact as of today, is that when
GENPD_FLAG_ACTIVE_WAKEUP is set for the genpd, genpd also checks the
->power.wakeup_path flag for the device. Both conditions must be true,
else genpd powers off the PM domain (and the device).

In other words, all devices being part of the wakeup path must either
have device_may_wakeup() to return true for it or the
->power.wakeup_path set, else genpd will power off the PM domain,
regardless of whether the GENPD_FLAG_ACTIVE_WAKEUP flag is set or not.

> Hence for this class of SoCs, this would duplicate the existing
> dev->power.wakeup and dev->power.wakeup_path flags, so that's why I would
> prefer using GENPD_FLAG_ACTIVE_WAKEUP instead (like we already do for the
> older SH-Mobile SoCs, but not for newer R-Car and RZ series).
> For other SoC families, the situation may be different.
>
> For us, the only exception are devices where the wakeup-source is not the
> device itself, but a GPIO, e.g. SDHI card detect. In such a case, only the
> device driver knows if the device is to be woken up through a dedicated CD
> line, or through a GPIO CD. So that would call for a method to opt-out,
> e.g. OUT_BAND_WAKEUP.

The main reason to why I chosen the IN_BAND_WAKEUP method, is because
genpd's *default* method is to power off the PM domain (and the
device) even if wakeup is enabled for the device. Hence, genpd treats
all wakeups as default being out-of-band IRQs.

Having an OUT_BAND_WAKEUP flag, would thus only make sense for those
genpds that has GENPD_FLAG_ACTIVE_WAKEUP set, I guess that is what you
are saying as well?

>
> To complicate matters, some drivers may be used on SoCs where the device
> needs to be kept running (clock and/or power domain), and on SoCs where the
> device is always running. This difference is typically handled by genpd,
> and the device driver may not even be aware. Of course the driver can just
> set IN_BAND_WAKEUP regardless, (else it has to check for the presence of
> clocks and/or power-domains properties itself, duplicating genpd
> core/driver code).
>
> So what about
>
>          if (IN_BAND_WAKEUP ||
>             (GENPD_FLAG_ACTIVE_WAKEUP && !OUT_BAND_WAKEUP)) {

We don't want to suspend the device in case of IN_BAND_WAKEUP, right!?

>                 ... suspend device...
>         }

Kind regards
Uffe