On Mon, Jun 7, 2021 at 4:14 PM Joerg Roedel <jroedel@xxxxxxx> wrote: > > On Mon, Jun 07, 2021 at 02:56:24PM +0200, Rafael J. Wysocki wrote: > > On Fri, Jun 4, 2021 at 7:09 PM Bjorn Helgaas <helgaas@xxxxxxxxxx> wrote: > > > If either "pcie_ports_disabled" or Linux doesn't support everything in > > > ACPI_PCIE_REQ_SUPPORT, we will never evaluate _OSC at all, so > > > the platform won't know that Linux has OSC_PCI_SEGMENT_GROUPS_SUPPORT, > > > OSC_PCI_HPX_TYPE_3_SUPPORT, OSC_PCI_EXT_CONFIG_SUPPORT, etc. > > > > Right. > > Thanks Bjorn and Rafael. So I think the important thing to do is to > issue at least one _OSC call even when Linux is not trying to take > control of anything. > > I look into a clean way to do this and get the kernel messages right. > One thing to change is probably only calculating 'control' if > !pcie_ports_disabled in negotiate_os_control(). Please also see https://lore.kernel.org/linux-acpi/93d783c4-4468-023b-193e-3fc6eca35445@xxxxxxxxxx/ for possible clashes etc.
