On 2017/3/22 16:00, Arnd Bergmann wrote: > On Mar 22, 2017 04:27, "Zhou Wang" <wangzhou1@xxxxxxxxxxxxx <mailto:wangzhou1@xxxxxxxxxxxxx>> wrote: > > On 2017/3/21 23:48, Jingoo Han wrote: > > (+cc: Joao Pinto, Zhou Wang, Gabriele Paoloni) > > > > On Tuesday, March 21, 2017 10:32 AM, Arnd Bergmann wrote: > >> > >> Without PCI_HOST_COMMON support enabled, we get a link error: > >> > >> drivers/pci/dwc/built-in.o: In function `hisi_pcie_map_bus': > >> pcie-hisi.c:(.text+0x8860): undefined reference to `pci_ecam_map_bus' > >> drivers/pci/dwc/built-in.o: In function `hisi_pcie_almost_ecam_probe': > >> pcie-hisi.c:(.text+0x88b4): undefined reference to `pci_host_common_probe' > >> > >> This adds an explicit 'select', as the other users have. > > Hi Arnd, > > I wonder if we should select PCI_HOST_GENERIC here, like: > > diff --git a/drivers/pci/dwc/Kconfig b/drivers/pci/dwc/Kconfig > index dfb8a69..b25dce4 100644 > --- a/drivers/pci/dwc/Kconfig > +++ b/drivers/pci/dwc/Kconfig > @@ -89,6 +89,7 @@ config PCI_HISI > depends on PCI_MSI_IRQ_DOMAIN > select PCIEPORTBUS > select PCIE_DW_HOST > + select PCI_HOST_GENERIC > help > Say Y here if you want PCIe controller support on HiSilicon > Hip05 and Hip06 SoCs > > > I don't know how that would be better. PCI_HOST_GENERIC is > a standalone driver for machines that don't need any special > handling (e.g. SBSA or KVM), while PCI_HOST_COMMON is > the infrastructure used by both PCI_HISI and PCI_HOST_GENERIC. > > Selecting PCI_HOST_GENERIC would also work as we implicitly > get PCI_HOST_COMMON, but it would not be clear from reading > the Kconfig file what the purpose of doing that is. Yes, you are right. I just disabled PCI_HOST_GENERIC and enabled PCI_HOST_COMMON to test PCI hosts of HiSilicon and Thunder, a build error happened. We can fix this by your PCI: generic: stop hiding pci_host_common_probe declaration. And select PCI_HOST_COMMON here. Thanks for pointing this! Zhou. > > Arnd
