On Wed, May 23, 2012 at 1:52 PM, Jason Baron <jbaron@xxxxxxxxxx> wrote: > Ok, I've also verified that quirk approach resolves the issue for me as > well. It might not be a bad solution, especially if we anticipate moving > to pciehp. We can also always change bridge vendor id to not hit the > quirk, if we choose to implement more complex guest, or acpi fixes. The > defaults for i/o and mem seem pretty sane too: > > > #define DEFAULT_HOTPLUG_IO_SIZE (256) > #define DEFAULT_HOTPLUG_MEM_SIZE (2*1024*1024) > > And further can be set on the guest command line via: > > /* pci=hpmemsize=nnM,hpiosize=nn can override this */ > > So I don't see any down-side really to the quirk approach (which is a > quite simple 1-line patch to the guest), even if it is not a long-term > solution. Good. please check attached patch, it should auto set that bit for apciphp when complex slot are supported by platform. next, we still need to extend probe_resource() to support mmio/pref_mmio extension. that is for when you have a lot slots on hotplug bridge. Thanks Yinghai
Attachment:
acpi_slot_hotplug_bridge_detection.patch
Description: Binary data
