Re: [GIT PULL] PCI: Support for configurable PCI endpoint

From: Kishon Vijay Abraham I
Date: Wed Mar 29 2017 - 08:37:24 EST


Hi,

On Wednesday 29 March 2017 05:40 PM, Kishon Vijay Abraham I wrote:
> Hi Niklas,
>
> On Wednesday 29 March 2017 05:12 PM, Niklas Cassel wrote:
>> On 03/27/2017 11:44 AM, Kishon Vijay Abraham I wrote:
>>> Hi Bjorn,
>>>
>>> Please find the pull request for PCI endpoint support below. I've
>>> also included all the history here.
>>>
>>> Changes from v4:
>>> *) add #syscon-cells property and used of_parse_phandle_with_args
>>> to perform a configuration in syscon module (as suggested by
>>> Rob Herring)
>>> *) Remove unnecessary white space.
>>>
>>> Changes from v3:
>>> *) fixed a typo and adapted to https://lkml.org/lkml/2017/3/13/562.
>>>
>>> Changes from v2:
>>> *) changed the configfs structure as suggested by Christoph Hellwig. With
>>> this change the framework creates configfs entry for EP function driver
>>> and EP controller. Previously these entries have to be created by the
>>> the user. (Haven't changed the epc core or epf core except for invoking
>>> configfs APIs to create entries for EP function driver and EP controller.
>>> That's mostly because the EP function device can still be created by
>>> directly invoking the epf core API without using configfs).
>>> *) Now the user has to use configfs entry 'start' to start the link.
>>> This was previously done by the function driver. However in the case of
>>> multi function EP, the function driver shouldn't start the link.
>>>
>>> Changes from v1:
>>> *) The preparation patches for adding EP support is removed and is sent
>>> separately
>>> *) Added device ID for DRA74x/DRA72x and used it instead of
>>> using "PCI_ANY_ID"
>>> *) Added userguide for PCI endpoint test function
>>>
>>> Major Improvements from RFC:
>>> *) support multi-function devices (hw supported not virtual)
>>> *) Access host side buffers
>>> *) Raise MSI interrupts
>>> *) Add user space program to use the host side PCI driver
>>> *) Adapt all other users of designware to use the new design (only
>>> compile tested. Since I have only dra7xx boards, the new design
>>> has only been tested in dra7xx. I'd require the help of others
>>> to test the platforms they have access to).
>>>
>>> This series has been developed over 4.11-rc1 + [1]
>>> [1] -> https://lkml.org/lkml/2017/3/13/562
>>>
>>> Let me know if this has to be re-based to some of your branch.
>>>
>>> Thanks
>>> Kishon
>>>
>>> The following changes since commit 623e87fec8ab7867fb51b3079196bd10718a60ce:
>>>
>>> PCI: dwc: dra7xx: Push request_irq call to the bottom of probe (2017-03-22 20:35:30 +0530)
>>>
>>> are available in the git repository at:
>>>
>>> git://git.kernel.org/pub/scm/linux/kernel/git/kishon/pci-endpoint.git tags/pci-endpoint-for-4.12
>>>
>>> for you to fetch changes up to e98bf80074be4654faae42fe0f5a622a776b6fdd:
>>>
>>> ARM: DRA7: clockdomain: Change the CLKTRCTRL of CM_PCIE_CLKSTCTRL to SW_WKUP (2017-03-27 15:08:22 +0530)
>>>
>>> ----------------------------------------------------------------
>>
>> FWIW:
>> I've tested Kishon's tag pci-endpoint-for-4.12
>> and PCIe on artpec6 SoC is still working fine.
>
> Thanks for testing it.
>>
>> I also included the DRA7xx PCIe driver in my
>> kernel so that pcie-designware-ep.c gets built.
>>
>> My only worry is that the code in pcie-designware-ep.c
>> is not compile tested if DRA7xx is not selected
>> (as it is the only driver using PCIE_DW_EP at
>> the moment).
>
> yeah, we should plan to include COMPILE_TEST in all pci drivers but I guess
> there is some problem with non-ARM builds [1]. As Bjorn mentioned in the
> thread, we could add #ifdef ARM and then include COMPILE_TEST.

I think I misunderstood your concern. yeah, there is no direct way to compile
pcie-designware-ep.c without selecting DRA7xx.

Thanks
Kishon