Re: [PATCH 2/2] power: supply: Add Lenovo Yoga C630 EC driver

[Bjorn Andersson <andersson@kernel.org>]

On Fri, Sep 16, 2022 at 12:53:42AM +0300, Dmitry Baryshkov wrote:
> On Fri, 16 Sept 2022 at 00:25, Bjorn Andersson <andersson@kernel.org> wrote:
> >
> > On Tue, Sep 13, 2022 at 12:45:45PM +0200, Sebastian Reichel wrote:
> > > Hi,
> > >
> > > [+Cc Lee Jones, DRI devel]
> > >
> > > On Tue, Aug 09, 2022 at 10:05:00PM -0500, Bjorn Andersson wrote:
> > > > The Qualcomm Snapdragon-based Lenovo Yoga C630 has some sort of EC
> > > > providing AC-adapter and battery status, as well as USB Type-C altmode
> > > > notifications for Displayport operation.
> > > >
> > > > The Yoga C630 ships with Windows, where these operations primarily are
> > > > implemented in ACPI, but due to various issues with the hardware
> > > > representation therein it's not possible to run Linux on this
> > > > information. As such this is a best-effort re-implementation of these
> > > > operations, based on the register map expressed in ACPI and a fair
> > > > amount of trial and error.
> > > >
> > > > Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
> > > > ---
> > > > [...]
> > > > +   val = yoga_c630_ec_read8(ec, LENOVO_EC_BAT_ATTRIBUTES);
> > > > +   if (val < 0)
> > > > +           goto out_unlock;
> > > > +   ec->unit_ma = val & LENOVO_EC_BAT_ATTR_UNIT_IS_MA;
> > > > +   if (!ec->unit_ma)
> > > > +           ec->scale = 1000;
> > > > +   else
> > > > +           ec->scale = 1;
> > >
> > > Since I'm not sure how much of information was gained by reverse
> > > engineering: Is this really milliamps vs microamps and not milliamps
> > > vs milliwatt? SBS batteries usually report either mA or mW.
> > >
> >
> > Unfortunately I don't know the answer to this.
> >
> > > > [...]
> > > > +   case POWER_SUPPLY_PROP_SERIAL_NUMBER:
> > > > +           val->strval = "05072018";
> > > > +           break;
> > >
> > > why is this hardcoded? :)
> > >
> >
> > I don't know, but as Daniel suggests, it would make sense to just drop
> > it.
> >
> > > > [...]
> > > > +   device_for_each_child_node(dev, fwnode) {
> > > > +           ret = fwnode_property_read_u32(fwnode, "reg", &port);
> > > > +           if (ret < 0)
> > > > +                   continue;
> > > > +
> > > > +           /* Got multiple ports, but altmode is only possible on port 1 */
> > > > +           if (port != 1)
> > > > +                   continue;
> > > > +
> > > > +           ec->bridge.funcs = &yoga_c630_ec_bridge_funcs;
> > > > +           ec->bridge.of_node = to_of_node(fwnode);
> > > > +           ec->bridge.ops = DRM_BRIDGE_OP_HPD;
> > > > +           ec->bridge.type = DRM_MODE_CONNECTOR_USB;
> > > > +
> > > > +           ret = devm_drm_bridge_add(dev, &ec->bridge);
> > > > +           if (ret) {
> > > > +                   dev_err(dev, "failed to register drm bridge\n");
> > > > +                   fwnode_handle_put(fwnode);
> > > > +                   return ret;
> > > > +           }
> > >
> > > I wonder if DRM people want to see this in drivers/gpu/drm/bridge.
> > > Maybe it's better to make this a MFD driver?
> > >
> >
> > I did consider it, but it adds a whole bunch of boiler plate code
> > without a lot of benefit.
> >
> > There are a few other cases where I think it would make sense to have
> > drm bridges outside of drivers/gpu/drm, such as
> > drivers/usb/typec/altmodes/ and drivers/platform/chrome/...
> 
> What about a solution which might sound simpler than MFD: from your
> driver's probe() register a platform device which represents a drm
> bridge. Add drm_bridge driver for it into drivers/gpu/drm/bridges/.
> 

That is indeed simpler

But then rather than just calling drm_bridge_hpd_notify(status) we need
a new custom interface for this driver to call the other drive when
status changes and we waste 840 bytes just to track another struct
platform_device.

Perhaps that could be made generic, so we'd have "proxy_drm_bridge"
driver for drivers who implements a HPD drm_bridge but don't want to
implement that outside drivers/gpu/drm.

But this is just based on the assumption that the "DRM people" require
drm_bridges to only live in drivers/gpu/drm/. It's all one kernel and
the API is well defined, so I don't think there's a concrete problem to
solve.

Regards,
Bjorn