On Mon, Feb 7, 2022 at 9:39 AM gregkh <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> On Mon, Feb 07, 2022 at 08:29:40AM +0000, Tony Huang 黃懷厚 wrote:
> > >
> > > As discussed before, I would suggest leaving out all custom attributes for now,
> > > and first hooking up the driver to all the in-kernel subsystems.
> > >
> > > The mailbox0 register data definitely feels like an implementation detail, not
> > > something that should be exposed to user space as an interface.
> > >
> > > For standby mode, this would normally be handled by the power management
> > > subsystem in the kernel. not a custom interface. From your earlier description,
> > > I assume this interface puts the main CPU into standby mode, not the IOP,
> > > right?
> > >
> > > CPU standby is handled by the cpuidle subsystem, so you need a driver in
> > > drivers/cpuidle/ to replace your sysfs attribute.
> > > If you plan to hook up the driver to multiple subsystems, keeping a generic
> > > driver file is ok, so you'll end up with two driver modules, with one of them
> > > calling into the other, using
> > > EXPORT_SYMBOL() to link between them.
> > >
> >
> > The purpose of adding sysfs is only for users to debug.
> > So this is not needed?
>
> If this is only for debugging, please put it in debugfs and not sysfs.
I don't think that works for the idle mode interface, as this is the only
thing the driver really does at the moment.
In a previous review round, I already asked for the driver to implement
at least two in-kernel interfaces before any custom user interface
is added.
Arnd
