Re: [PATCH V2] PCI: Clear errors logged in Secondary Status Register

[Vidya Sagar <vidyas@nvidia.com>]

Hi Bjorn,
Sorry to bug you.
Is this change good to be accepted?

Thanks,
Vidya Sagar

On 01-04-2024 13:29, Vidya Sagar wrote:
> Hi Bjorn,
> Just checking on this thread.
> Is there anything else you want me to clarify on?
>
> Thanks,
> Vidya Sagar
>
> On 14-03-2024 06:09, Vidya Sagar wrote:
>>
>>
>> On 23-01-2024 04:30, Bjorn Helgaas wrote:
>>> External email: Use caution opening links or attachments
>>>
>>>
>>> On Tue, Jan 16, 2024 at 08:02:58PM +0530, Vidya Sagar wrote:
>>>> The enumeration process leaves the 'Received Master Abort' bit set in
>>>> the Secondary Status Register of the downstream port in the following
>>>> scenarios.
>>>>
>>>> (1) The device connected to the downstream port has ARI capability
>>>>      and that makes the kernel set the 'ARI Forwarding Enable' bit in
>>>>      the Device Control 2 Register of the downstream port. This
>>>>      effectively makes the downstream port forward the configuration
>>>>      requests targeting the devices downstream of it, even though they
>>>>      don't exist in reality. It causes the downstream devices return
>>>>      completions with UR set in the status in turn causing 'Received
>>>>      Master Abort' bit set.
>>>>
>>>>      In contrast, if the downstream device doesn't have ARI capability,
>>>>      the 'ARI Forwarding Enable' bit in the downstream port is not set
>>>>      and any configuration requests targeting the downstream devices
>>>>      that don't exist are terminated (section 6.13 of PCI Express Base
>>>>      6.0 spec) in the downstream port itself resulting in no change of
>>>>      the 'Received Master Abort' bit.
>>>>
>>>> (2) A PCIe switch is connected to the downstream port and when the
>>>>      enumeration flow tries to explore the presence of devices that
>>>>      don't really exist downstream of the switch, the downstream
>>>>      port receives the completions with UR set causing the 'Received
>>>>      Master Abort' bit set.
>>> Are these the only possible ways this error is logged?  I expected
>>> them to be logged when we enumerate below a Root Port that has nothing
>>> attached, for example.
>> In this case, there won't be any TLP sent downstream. I talked about this 
>> scenario in the
>> second paragraph of point (1) above.
>>> Does clearing them in pci_scan_bridge_extend() cover all ways this
>>> error might be logged during enumeration?  I can't remember whether
>>> all enumeration goes through this path.
>> So far in my testing, clearing it in pci_scan_bridge_extend() covers all the 
>> cases.
>>
>>>> Clear 'Received Master Abort' bit to keep the bridge device in a clean
>>>> state post enumeration.
>>>>
>>>> Signed-off-by: Vidya Sagar <vidyas@nvidia.com>
>>>> ---
>>>> V2:
>>>> * Changed commit message based on Bjorn's feedback
>>>>
>>>>   drivers/pci/probe.c | 3 +++
>>>>   1 file changed, 3 insertions(+)
>>>>
>>>> diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
>>>> index 795534589b98..640d2871b061 100644
>>>> --- a/drivers/pci/probe.c
>>>> +++ b/drivers/pci/probe.c
>>>> @@ -1470,6 +1470,9 @@ static int pci_scan_bridge_extend(struct pci_bus 
>>>> *bus, struct pci_dev *dev,
>>>>        }
>>>>
>>>>   out:
>>>> +     /* Clear errors in the Secondary Status Register */
>>>> +     pci_write_config_word(dev, PCI_SEC_STATUS, 0xffff);
>>>> +
>>>>        pci_write_config_word(dev, PCI_BRIDGE_CONTROL, bctl);
>>>>
>>>>        pm_runtime_put(&dev->dev);
>>>> -- 
>>>> 2.25.1
>>>>
>>
>