Re: [PATCH v5 3/3] Documentation: power: rt9471: Document exported sysfs entries

[ChiYuan Huang]

Sebastian Reichel <sebastian.reichel@xxxxxxxxxxxxx> 於 2022年10月2日 週日 凌晨4:58寫道：
>
> Hi,
>
> On Mon, Sep 19, 2022 at 09:11:09AM +0800, ChiYuan Huang wrote:
> > Sebastian Reichel <sebastian.reichel@xxxxxxxxxxxxx> 於 2022年9月17日 週六 上午9:19寫道：
> > > On Thu, Sep 15, 2022 at 12:30:15AM +0800, cy_huang wrote:
> > > > From: ChiYuan Huang <cy_huang@xxxxxxxxxxx>
> > > >
> > > > Document the settings exported by rt9471 charger driver through sysfs entries:
> > > > - sysoff_enable
> > > > - port_detect_enable
> > > >
> > > > Signed-off-by: ChiYuan Huang <cy_huang@xxxxxxxxxxx>
> > > > ---
> > > > Since v5:
> > > > - Recover all the change in sysfs-class-power.
> > > > - New a sysfs-class-power-rt9471 file.
> > > > - Remove 'charge_term_enable' sysfs entry, directly integrate it in
> > > >   'charge_term_current' power supply property control.
> > > >
> > > > ---
> > > >  Documentation/ABI/testing/sysfs-class-power-rt9471 | 29 ++++++++++++++++++++++
> > > >  1 file changed, 29 insertions(+)
> > > >  create mode 100644 Documentation/ABI/testing/sysfs-class-power-rt9471
> > > >
> > > > diff --git a/Documentation/ABI/testing/sysfs-class-power-rt9471 b/Documentation/ABI/testing/sysfs-class-power-rt9471
> > > > new file mode 100644
> > > > index 00000000..ad5b049
> > > > --- /dev/null
> > > > +++ b/Documentation/ABI/testing/sysfs-class-power-rt9471
> > > > @@ -0,0 +1,29 @@
> > > > +What:                /sys/class/power_supply/rt9471-*/sysoff_enable
> > > > +Date:                Oct 2022
> > > > +KernelVersion:       6.1
> > > > +Contact:     ChiYuan Huang <cy_huang@xxxxxxxxxxx>
> > > > +Description:
> > > > +             This entry allows enabling the sysoff mode of rt9471 charger devices.
> > > > +             If enabled and the input is removed, the internal battery FET is turned
> > > > +             off to reduce the leakage from the BAT pin. See device datasheet for details.
> > > > +             It's commonly used when the product enter shipping stage.
> > > > +
> > > > +             Access: Read, Write
> > > > +             Valid values:
> > > > +             - 1: enabled
> > > > +             - 0: disabled
> > >
> > > I still fail to see why this needs to be controllable at runtime.
> > > This looks like a hardware property. Are there any known products,
> > > which need this disabled?
> > It's just a switch, actually 'disabled' is not needed.
> > For the enabled case, mostly used in below scenarios
> > 1. Online testing, USB IN -> Factory testing -> write 1 to enable ->
> > USB out -> immediately VSYS off -> pack
> > 2. Offline testing no vbus -> Factory testing -> write 1 to enable ->
> > immediately VSYS off -> pack
> >
> > The 'disable" can use to cancel the shipping mode in case 1 before USB out.
> > It's more like the testing.
> >
> > Like as you said, shipping BATFET_OFF is all the hardware behavior.
> > To leave this mode after VSYS off, there're three ways
> > 1. power key pressed
> > 2. VBUS IN
> > 3. control BATFET_OFF to 0 (But it need SOC to be alive, at the time,
> > VSYS off, no one can execute this I2C command)
>
> If factory testing and preperation is the only use case, I don't
> think exposing this in sysfs and creating userspace ABI is worth
> it. Just tell factory to use i2c-dev and poke the correct registers.
>
I agree your comment if there's only this case will use it.

So I ask our HW members about this.
They said there's still one case I didn't consider about.
It's the dual charger scenario.
If the charging process is entering CV mode, the slave charger is no
need to join the charging.
Then in common case, slave charger need to minimize the battery leakage.
And the BATFET_OFF is needed to lower the battery leakage.

They think this sysfs entry is needed. Can this persuade you?

> > If what you care is no need to mention 'disable', then just remove it.
> > It's fine.
> > >
> > > > +What:                /sys/class/power_supply/rt9471-*/port_detect_enable
> > > > +Date:                Oct 2022
> > > > +KernelVersion:       6.1
> > > > +Contact:     ChiYuan Huang <cy_huang@xxxxxxxxxxx>
> > > > +Description:
> > > > +             This entry allows enabling the USB BC12 port detect function of rt9471 charger
> > > > +             devices. If enabled and VBUS is inserted, device will start to do the BC12
> > > > +             port detect and report the usb port type when port detect is done. See
> > > > +             datasheet for details. Normally controlled when TypeC/USBPD port integrated.
> > > > +
> > > > +             Access: Read, Write
> > > > +             Valid values:
> > > > +             - 1: enabled
> > > > +             - 0: disabled
> > >
> > > So basically this depends on the hardware integration (e.g. it
> > > should be disabled when power source is a DC barrel jack instead
> > > of USB) and is not supposed to change at all during runtime? Then
> > > the information wether it needs to be enabled should be derived
> > > from the device tree.
> >
> > It's a switching charger integrates OTG boost.
> > For the case 'DC Jack', there's no need to use this kind of product.
> >
> > With typec integration, at most time, it still need bc12 to be enabled
> > by default. Just in some case, like as power role swap (SNK -> SRC -> SNK),
> > to automatically identify the USB port, this may interrupt USB communication.
> >
> > So as my understanding, keep it enabled by default, just in some case,
> > it my need to control at runtime.
>
> This should be part of the description. You can drop the sentence
> "Normally controlled when TypeC/USBPD port integrated.", since
> that's hard to comprehend. Instead add the information that this
> is supposed to be always enabled, but can be disabled to avoid
> USB link loss (?) when doing a USB PD role swap.
>
Thanks for the comment.

I'll rewrite it as below
'It's supposed to be always enabled, but can be disabled to avoid usb
link interruption especially when doing a USBPD 'power' role swap.'

Sorry, due to the long reply period, I already sent v6/v7/v8 to fix
some coding like as missing header and irq wakeup check.
Actually only missing header change is needed.
Please ignore these noise.

And after the sysfs reviewing is finished, you can just review the revision v9.
Thanks.
> -- Sebastian