On 13 October 2015 at 07:31, Alexander Gordeev <agord...@redhat.com> wrote: > On Mon, Oct 12, 2015 at 10:03:03PM +0100, Peter Maydell wrote: >> On 12 October 2015 at 21:55, Alexander Gordeev <agord...@redhat.com> wrote: >> > Currently PCI IO address 0 is not allowed even though >> > the IO space starts from 0. As result, PCI IO is not >> > possible to use at all. >> >> I don't see any reason for us not to allow 0 IO addresses, >> but I'm not sure how your your conclusion follows. It >> should be entirely possible to map PCI IO to some other >> address than zero in the IO window, which is what I would >> have expected the guest to do. > > You are right - my changelog is incorrect. The rest of IO > space should be alright. > > However, as 0 IO address is exposed via "ranges" to the guest, > it must be usable - isn't it? So it either should be allowed > or the range should be different.
Depends on your point of view. If you take the Linus Torvalds view that 0 is always an invalid PCI address, then you'll never try to use it anyway. In any case, setting pci_allow_0_address is the right thing, so we can just change the commit message in this patch. Incidentally, why is this a property on the machine and not on the PCI controller device? thanks -- PMM