[PATCH 0/9] i.MX8MP power-domains part 1 and GPU support

[Lucas Stach]

Hi Abel,

Am Mittwoch, dem 19.01.2022 um 16:38 +0200 schrieb Abel Vesa:
> On 22-01-19 14:40:18, Lucas Stach wrote:
> > Hi all,
> > 
> > this series starts adding the power-domain control for the i.MX8MP
> > SoC. The GPCv2 support is complete (at least from going over the RM,
> > TF-A and experience with other i.MX8M* SoCs), but not all
> > power-domains are usable right now. Currently only the HSIO
> > (USB and PCIe) and GPU power domains are enabled.
> > 
> > Other power domains (MEDIA, VPU, HDMI, AUDIO) can be added when the
> > blk-ctrl driver support for those domains is ready, which is still
> > work in progress at the moment. As my priorities are shifting to
> > other things for a while, I wanted to push out the part that is
> > usable now and enables more functionality on the i.MX8MP.
> > 
> 
> Great effort! Thanks for working on this!
> 
> I started doing it myself a couple of months ago. I did the media and
> hdmi blk-ctrls. The audio blk-ctrl is the one that got me stuck since it
> has PLLs in it and they need to be part of the clock tree somehow.
> 
> Let me know if you want me to send the hdmi and media blk-ctrls.
> I'll try to rebase them on top of this patchset.

That would certainly be very helpful!

The HSIO one also has a PLL that can optionally be used as a reference
for the USB and PCIe PHYs. I think it should be doable to integrate
them in the clock tree. We need some additional smarts to save/restore
the clock state when the *MIX domain powers down/up.

Other than that I think we need to add a rule that those blk-ctrl
clocks can only be prepared/enabled when the power domain is already up
to avoid circling back into the clock framework via the GPC, but I
guess that's a reasonable rule for the peripheral drivers to adhere to.
Just always runtime resume the peripheral before enabling any clocks,
or possibly even just enable the clocks in the runtime resume callback
of the driver.

Regards,
Lucas