Gerd Hoffmann <kra...@redhat.com> writes: > On Wed, Oct 17, 2018 at 07:57:39AM +0200, Markus Armbruster wrote: >> Laine Stump <la...@redhat.com> writes: >> >> [...] >> > In the end, having a device that changed PCI ID depending on what kind >> > of slot it was plugged into was an idea "too clever for its own good", >> > should be avoided when new devices are added in the future, and we >> > should at least provide an alternative that doesn't do that for existing >> > devices. >> >> That means for each chameleon PCI/PCIe device: >> >> * create a pair of devices that can only go into one kind of slot >> * deprecate the chameleon > > I think virtio devices are the only ones which actually change the pci > id and have non-trivial differences. > > qemu-xhci can likewise be plugged into both pci and pcie slots. When > plugged into a pcie slot it'll have pcie endpoint capability. That is > the only difference though. Do you consider that a chameleon device?
Since the enticing simplicity of letting PCI device models go into PCIe slots as well is exactly what lured us into this mess, I do. I believe all we actually saved by creating chameleons was a bunch of QOM types. New, non-chameleon PCIe devices would have shared the actual device model code with the existing, non-chameleon PCI devices. Correcting the design mistake now involves yet another set of QOM types, for backward compatibility. A loss of simplicity we failed to consider properly back then was at the external interface. Traditionally, a qdev has a bus type (printed by -device help), and it can go into a slot provided by such a bus. But chameleon PCI qdevs can also go into a slot provided by a PCIe bus. The interface is now less regular, for no convincing reason.
