On Fri, 2013-04-05 at 14:11 -0600, Bjorn Helgaas wrote: > > I think sparc has the same issue in its own copy of > of_create_pci_dev(). > > Of course, both of_create_pci_dev() implementations are basically > copies of the generic pci_setup_device() that most arches use. That's > the reason why I wish sparc and powerpc had used config space > accessors that hid the OF mangling internally so they could use the > generic pci_setup_device() instead of cloning it. I disagree :-) I want the config space accessors to actually do the config space access (it might be necessary for some reasons, if anything for diagnostic). Also one of the reasons we create devices that way originally iirc, is that on older pre-PCIe setups, we could have cases of a bridge showing up at function N > 0 without anything at function 0. We are also not allowed to mess with bridge BARs on old EADS bridges, and similar issues where the hypervisor can get upset. A "filtering" config space code would be a lot messier than just creating them like we do. However we could/should probably make the code more common between powerpc and sparc and maybe move the bulk of it to a generic place more easily grepped by the PCI folks. > Of course, they don't, and that's too much work for fixing this issue, > but if anybody wanted to work on that, I think it would be an > interesting project. > > But what if you set dev->dev.type in alloc_pci_dev()? I think if you > did that, you wouldn't need to export "pci_dev_type," and it should > fix this for both powerpc and sparc. Cheers, Ben. -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html