On Fri, May 07, 2021 at 11:51:20AM +0200, Rafael J. Wysocki wrote: > On Fri, May 7, 2021 at 12:07 AM Lukas Wunner <lukas@xxxxxxxxx> wrote: > > With s2idle, the machine isn't suspended via ACPI, so the AML code > > which powers the controller off isn't executed. The dance to prepare > > the controller for power-off consequently isn't necessary but rather > > harmful. > > > > To get the same power savings as with ACPI suspend, the controller > > needs to be powered off via runtime suspend. > > I'm not quite sure why runtime PM needs to be involved. > > The controller suspend can happen in the system-wide suspend code path > directly. Sorry, my comments were unclear. What I meant to say is that the Thunderbolt runtime PM patches allow us to power the Thunderbolt controller up or down from the OS and we can re-use that code in the system-wide suspend code path if s2idle is used. Without s2idle, power to the controller is cut by transitioning to ACPI S3 or S4, which disables the controller's power rails without OS involvement. Thanks, Lukas
