On Wed, May 09, 2018 at 12:51:53PM +0800, Shawn Guo wrote: > Hi Bjorn, > > On Tue, May 08, 2018 at 07:56:58AM -0500, Bjorn Helgaas wrote: > ... > > > + return ret; > > > + } > > > + } > > > + > > > + pci->pp.root_bus_nr = -1; > > > > Setting root_bus_nr looks like an unrelated change that should be in a > > separate patch. > > > > But I'm not sure why you need to set root_bus_nr at all, since > > dw_pcie_host_init() always sets it. > > > > Some other callers of dw_pcie_host_init() do set it: > > > > exynos_add_pcie_port() > > imx6_add_pcie_port() > > armada8k_add_pcie_port() > > artpec6_add_pcie_port() > > dw_plat_add_pcie_port() > > histb_pcie_probe() > > qcom_pcie_probe() > > spear13xx_add_pcie_port() > > > > But I don't see *why* any of these need to set it. If they don't need to > > set it, they shouldn't. > > Mostly it's a blind copy of unnecessary code. I tested histb driver > by dropping the line, and did not see anything broken. I will cook up > a series to remove the code from all above drivers, and copy > corresponding driver owner to comment.
Thanks, Shawn, I really appreciate that! > > And it would be nice if histb and qcom followed the structure and naming > > conventions of the other drivers, i.e., they should have > > histb_add_pcie_port() and qcom_add_pcie_port(). > > I can create a patch for histb driver, but will leave qcom one to > Stanimir to decide. Sounds good, thanks again!