On Tue, 2013-12-10 at 22:14 +0000, Peter Maydell wrote: > Googling again brought up this mailing list thread: > > http://www.pcisig.com/reflector/msg00459.html > > which includes what is supposedly a quote from the PCI 2.1 spec: > > # "Note: A Base Address register does not contain a valid > # address when it is equal to "0"" > > (I don't have access to the 2.1 version to check.) > > This text seems to have been removed from the 2.2 spec.
I have seen practical cases of both: - Systems where the FW sets up a BAR to 0 and considers it valid - Adapters that treat a BAR set to 0 as disabled There's no win here. However it makes sense for qemu not to treat 0 as a special value, it's not necessary. Cheers, Ben.