On Wed, Dec 15, 2021 at 07:11:53PM +0300, Dmitry Osipenko wrote:
> 15.12.2021 18:55, Thierry Reding пишет:
> > On Wed, Dec 01, 2021 at 02:23:07AM +0300, Dmitry Osipenko wrote:
> >> This series adds runtime PM support to Tegra drivers and enables core
> >> voltage scaling for Tegra20/30 SoCs, resolving overheating troubles.
> >>
> >> All patches in this series are interdependent and should go via Tegra tree
> >> for simplicity.
> > 
> > So these can be applied in any order without breaking anything?
> 
> Please notice that the word is *inter* dependent, not *in* dependent.
> 
> There is a build dependency for the patches. The first two "soc/tegra"
> must be applied first.

Okay, so I've separated the first two patches out into a separate stable
branch that I can share between the Tegra and drm/tegra trees to pull in
the build dependency and then I've applied the driver patches to those
two trees and I've verified that the two branches build correctly. I've
not done any runtime testing, but I'll trust you on that.

> The "soc/tegra: pmc: Enable core domain support for Tegra20 and Tegra30"
> *must* be the last applied patch if we want to preserve bisectability.
> The core voltage scaling can be enabled only once all the drivers got
> the power management support.
> 
> The rest could be applied out-of-order.

One last remaining question: I don't think I can apply that one patch if
it requires that all the others are enabled first because it would
basically create a circular dependency.

Can I pick up the final 7 patches (the DT ones) independently of that
one patch without things breaking? If so, one option we could try is to
wait for both Tegra and drm/tegra trees to get merged into v5.17-rc1 and
then send that one patch (which is only a 4-line diff) right after
v5.17-rc1 so that it makes it into v5.17-rc2. That avoids the circular
dependency and should get everything enabled for v5.17.

Do you see any problems with that?

Thierry