On Mon, 28 Feb 2022 08:39:57 +0100 Gerd Hoffmann <kra...@redhat.com> wrote:
> Hi, > > > This where I wasn't comfortable with idea of calling random PCIe code > > chunks and thought about chaining callbacks so that > > pcie_cap_slot_[pre_]plug_cb() would do necessary PCIe steps > > and acpi_pcihp_device_[pre_]plug_cb() do ACPI specific things not > > intruding on each other, but that requires telling PCIe code that > > it should not issue native hotplug event to guest. > > I think with both acpi and pcie hotplug being active it surely makes > sense that both are in sync when it comes to device state. So acpihp > updating pcie slot state (power control, maybe also device presence) > looks sane to me. > > Not sure whenever it would be better to call into pcie code or just > update the pci config space bits directly to make sure pcie doesn't > take unwanted actions like sending out events. If changing power state is preferred over disabling power control, I can respin series with what Michael has suggested earlier and drop 2/4 as not necessary. I'll wait for a day to see if there would more ideas/suggestions > take care, > Gerd >