Re: Re: [RFC 8/9] PCI/pwrctl: add PCI power control core code

[Manivannan Sadhasivam <mani@kernel.org>]

On Fri, Feb 09, 2024 at 10:04:33AM +0100, Lukas Wunner wrote:
> On Wed, Feb 07, 2024 at 05:26:16PM +0100, Bartosz Golaszewski wrote:
> > On Fri, Feb 2, 2024 at 5:52???PM Bjorn Andersson <andersson@kernel.org> wrote:
> > > On Fri, Feb 02, 2024 at 10:11:42AM +0100, Bartosz Golaszewski wrote:
> > > > I was also thinking about pci_pwrctl_device_ready() or
> > > > pci_pwrctl_device_prepared().
> > >
> > > I like both of these.
> > >
> > > I guess the bigger question is how the flow would look like in the event
> > > that we need to power-cycle the attached PCIe device, e.g. because
> > > firmware has gotten into a really bad state.
> > >
> > > Will we need an operation that removes the device first, and then cut
> > > the power, or do we cut the power and then call unprepared()?
> > 
> > How would the core be notified about this power-cycle from the PCI
> > subsystem? I honestly don't know. Is there a notifier we could
> > subscribe to? Is the device unbound and rebound in such case?
> 
> To power-manage the PCI device for runtime PM (suspend to D3cold)
> or system sleep, you need to amend:
> 
>   platform_pci_power_manageable()
>   platform_pci_set_power_state()
>   platform_pci_get_power_state()
>   platform_pci_refresh_power_state()
>   platform_pci_choose_state()
> 
> E.g. platform_pci_power_manageable() would check for presence of a
> regulator in the DT and platform_pci_set_power_state() would disable
> or enable the regulator.
> 

This will work if the sole control of the resources lies in these platform_*()
APIs. But in reality, the controller drivers are the ones controlling the power
supply to the devices  and with this series, the control would be shifted partly
to pwrctl driver.

I think what we need is to call in the callbacks of the drivers in a hierarchial
order.

- Mani

> To reset the device by power cycling it, amend pci_reset_fn_methods[]
> to provide a reset method which disables and re-enables the regulator.
> Then you can choose that reset method via sysfs and power-cycle the
> device.  The PCI core will also automatically use that reset method
> if there's nothing else available (e.g. if no Secondary Bus Reset
> is available because the device has siblings or children, or if FLR
> is not supported).
> 
> Thanks,
> 
> Lukas

-- 
மணிவண்ணன் சதாசிவம்