On Tue, Nov 03 2020 at 11:41, Marc Zyngier wrote: > On 2020-11-03 10:31, Thomas Gleixner wrote: > We can do that, although I worried that it isn't 100% reliable: > > pci_host_probe() ends up calling pci_add_device(), and will start > probing devices if the endpoint drivers have already registered > with the core code, long before the flag gets set.
Bah, you're right. What a mess.
