On 13/03/2019 19:29, Srinivas Kandagatla wrote: > On 13/03/2019 11:45, Marc Gonzalez wrote: > >> We don't need to fudge the device class in qcom_pcie_rd_own_conf() >> because dw_pcie_setup_rc() already does the right thing: >> >> /* Program correct class for RC */ >> dw_pcie_wr_own_conf(pp, PCI_CLASS_DEVICE, 2, PCI_CLASS_BRIDGE_PCI); >> >> Signed-off-by: Marc Gonzalez <marc.w.gonzalez@xxxxxxx> >> --- >> Changes from v1 to v2: Completely drop qcom_pcie_rd_own_conf >> Srinivas, could you test this patch on apq/ipq8064? >> --- >> drivers/pci/controller/dwc/pcie-qcom.c | 17 ----------------- >> 1 file changed, 17 deletions(-) >> > I tried this patch on IFC6410 board which has APQ8064, and the pci does > not work anymore after applying this patch. > > non-working output of 'lspci -vvv' > https://paste.ubuntu.com/p/ndmdsW4zkJ/ and bootlog > https://paste.ubuntu.com/p/CqVzGymnq8/ > > working without patch : https://paste.ubuntu.com/p/TJm4hgjGW4/ > > I have not debugged it any further other than just testing the patch. > Let me know if you need any more dumps for more debug. Could you pastebin the output of lspci -vvv -n in the working case? I think that programming the device class might not work on APQ8064, and therefore that platform requires the quirk. So maybe I can add the quirk back, just for APQ8064? Regards.
