On Fri, 2017-03-03 at 12:37 +1100, David Gibson wrote:
> Hm, you have a point. I'll try to see how difficult it
> would be to teach PCI devices about the quirkiness in
> bus naming when it comes to PHBs.
An update: I'm intending to squeeze the device tree fix to advertise
extended config space into qemu-2.9. I'm not going to include the
naming change - I'll see how your investigation here turns out before
proceeding with that.
Sorry for not keeping you up to date, here's the situation:
I got the code working with the current naming scheme and
it's not even too big a kludge, so dropping that patch is
definitely the right thing to do.
Unfortunately I've uncovered a few other issues while
writing more test cases, that's why I haven't yet posted
a respin.
--
Andrea Bolognani / Red Hat / Virtualization