Re: [PATCH v2 4/4] PCI: Add runtime PM support for PCIe ports

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Wed, Apr 13, 2016 at 11:08:37AM +0200, Andreas Noever wrote:
> > <- Now, my understanding is that Macs do not use ACPI hotplug but
> > instead this is all native, correct? When you have the controller
> > exposed all the time, of course you can get hotplug events and handle
> > them in the driver.
> >
> > However, problem arises when enumeration and configuration is actually
> > done in BIOS SMI handler, like in normal non-Mac PCs. If the port is in
> > D3 the handler is not smart enough to move it back to D0 and then
> > re-enumerate ports. It just gives up.
> Is this issue specific to the "ACPI/SMI Thunderbolt" implementations
> used in non-Mac PCs or is PCI hotplug in general done through SMI/ACPI
> instead of the native pciehp port driver? Wouldn't it make more sense
> to only disable runtime suspend on non-Mac thunderbolt ports (or ports
> using ACPI hotplug, if that is detectable).

I think it is specific to ACPI hotplug implementations (non-Mac PCs). We
should be able to detect which one is used. ACPI _OSC method is used to
allow/disallow native PCIe hotplug so I think we use that information
along with the slot capabilities to decide whether we enable or disable
runtime PM.