Re: [RFC PATCH v1] PCI: qcom: Use quirk to override incorrect device class

From: Marc Gonzalez
Date: Tue Mar 12 2019 - 13:19:25 EST


On 12/03/2019 13:42, Stanimir Varbanov wrote:

> On 3/11/19 4:56 PM, Marc Gonzalez wrote:
>
>> Some chips report an incorrect device class. Override the incorrect
>> value using a quirk, instead of code in the read function.
>>
>> Signed-off-by: Marc Gonzalez <marc.w.gonzalez@xxxxxxx>
>> ---
>> FWIW, this quirk is no longer required on recent chips:
>> msm8996 (tested by Stanimir), msm8998 (tested by me), sdm845 (untested) are unaffected
>> apq/ipq8064 is affected => what is the device ID for these chips?
>> others?
>>
>> Stanimir added: "this will become a real problem (now we use the driver as RC)
>> when someone decide to use it as an endpoint"
>> ---
>> drivers/pci/controller/dwc/pcie-qcom.c | 14 ++++++--------
>> 1 file changed, 6 insertions(+), 8 deletions(-)
>>
>> diff --git a/drivers/pci/controller/dwc/pcie-qcom.c b/drivers/pci/controller/dwc/pcie-qcom.c
>> index 3de5510fd3d5..94da2c9c2ad5 100644
>> --- a/drivers/pci/controller/dwc/pcie-qcom.c
>> +++ b/drivers/pci/controller/dwc/pcie-qcom.c
>> @@ -1136,17 +1136,15 @@ static int qcom_pcie_rd_own_conf(struct pcie_port *pp, int where, int size,
>> {
>> struct dw_pcie *pci = to_dw_pcie_from_pp(pp);
>>
>> - /* the device class is not reported correctly from the register */
>> - if (where == PCI_CLASS_REVISION && size == 4) {
>> - *val = readl(pci->dbi_base + PCI_CLASS_REVISION);
>> - *val &= 0xff; /* keep revision id */
>> - *val |= PCI_CLASS_BRIDGE_PCI << 16;
>> - return PCIBIOS_SUCCESSFUL;
>> - }
>> -
>
> once you dropped the above snippet this function becomes absolutely
> useless so please delete it at all and also from qcom_pcie_dw_ops.

Good catch.

>> return dw_pcie_read(pci->dbi_base + where, size, val);
>> }
>>
>> +static void qcom_fixup_class(struct pci_dev *dev)
>> +{
>> + dev->class = PCI_CLASS_BRIDGE_PCI << 8;
>> +}
>> +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_QCOM, PCI_ANY_ID, qcom_fixup_class);
>
> I wonder, in case that dw_pcie_setup_rc() already has a write to
> PCI_CLASS_DEVICE configuration register to set it as a bridge do we
> still need to do the above fixup?

I don't know, I don't have an affected device. Unless the msm8998 /is/ affected,
and dw_pcie_setup_rc() actually fixes it?

Regards.