[PATCH v2 12/12] [UNTESTED] riscv: dts: starfive: beaglev-starlight: Enable gmac

[Cristian Ciocaltea]

On 11/17/23 13:19, Cristian Ciocaltea wrote:
> On 11/17/23 11:12, Geert Uytterhoeven wrote:
>> Hi Cristian,
>>
>> On Fri, Nov 17, 2023 at 9:59 AM Cristian Ciocaltea
>> <cristian.ciocaltea at collabora.com> wrote:
>>> On 11/17/23 10:49, Cristian Ciocaltea wrote:
>>>> On 11/17/23 10:37, Geert Uytterhoeven wrote:
>>>>> On Thu, Nov 16, 2023 at 6:55 PM Conor Dooley <conor at kernel.org> wrote:
>>>>>> On Thu, Nov 16, 2023 at 03:15:46PM +0200, Cristian Ciocaltea wrote:
>>>>>>> On 10/30/23 00:53, Cristian Ciocaltea wrote:
>>>>>>>> On 10/29/23 20:46, Andrew Lunn wrote:
>>>>>>>>> On Sun, Oct 29, 2023 at 06:27:12AM +0200, Cristian Ciocaltea wrote:
>>>>>>>>>> The BeagleV Starlight SBC uses a Microchip KSZ9031RNXCA PHY supporting
>>>>>>>>>> RGMII-ID.
>>>>>>>>>>
>>>>>>>>>> TODO: Verify if manual adjustment of the RX internal delay is needed. If
>>>>>>>>>> yes, add the mdio & phy sub-nodes.
>>>>>>>>>
>>>>>>>>> Please could you try to get this tested. It might shed some light on
>>>>>>>>> what is going on here, since it is a different PHY.
>>>>>>>>
>>>>>>>> Actually, this is the main reason I added the patch. I don't have access
>>>>>>>> to this board, so it would be great if we could get some help with testing.
>>>>>>>
>>>>>>> @Emil, @Conor: Any idea who might help us with a quick test on the
>>>>>>> BeagleV Starlight board?
>>>>>>
>>>>>> I don't have one & I am not sure if Emil does. Geert (CCed) should have
>>>>>
>>>>> I believe Esmil has.
>>>>>
>>>>>> one though. Is there a specific test you need to have done?
>>>>>
>>>>> I gave it a try, on top of latest renesas-drivers[1].
>>>
>>> [...]
>>>
>>>>>
>>>>> Looks like it needs more non-coherent support before we can test
>>>>> Ethernet.
>>>>
>>>> Hi Geert,
>>>>
>>>> Thanks for taking the time to test this!
>>>>
>>>> Could you please check if the following are enabled in your kernel config:
>>>>
>>>>   CONFIG_DMA_GLOBAL_POOL
>>>>   CONFIG_RISCV_DMA_NONCOHERENT
>>>>   CONFIG_RISCV_NONSTANDARD_CACHE_OPS
>>>>   CONFIG_SIFIVE_CCACHE
>>
>> CONFIG_DMA_GLOBAL_POOL and CONFIG_RISCV_NONSTANDARD_CACHE_OPS were
>> indeed no longer enabled, as they cannot be enabled manually.
>>
>> After cherry-picking commit e14ad9ff67fd51dc ("riscv: errata: Add
>> StarFive JH7100 errata") in esmil/visionfive these options become
>> enabled. Now it gets a bit further, but still lots of CCACHE DataFail
>> errors.
> 
> Right, there is an open question [2] in PATCH v2 08/12 if this patch
> should have been part of Emil's ccache series or I will send it in v3
> of my series.
> 
> [2]: https://lore.kernel.org/lkml/4f661818-1585-41d8-a305-96fd359bc8b8@collabora.com/
> 
>>> Also please note the series requires the SiFive Composable Cache
>>> controller patches provided by Emil [1].
>>>
>>> [1]: https://lore.kernel.org/all/20231031141444.53426-1-emil.renner.berthing@canonical.com/
>>
>> That series does not contain any Kconfig changes, so there must be
>> other missing dependencies?
> 
> There shouldn't be any additional Kconfig changes or dependencies as 
> those patches just extend an already existing driver. There were some 
> changes in v2, but they are still compatible with this series (I've 
> retested that to make sure).
> 
> My tree is based on next-20231024, so I'm going to rebase it onto
> next-20231117, to exclude the possibility of a regression somewhere.
> 
> I will also test with renesas-drivers.

I verified with both trees and didn't notice any issues with my 
VisionFive board, so I don't really understand why BeagleV Starlight 
shows a different behavior.

For reference, please see [3] which contains all required patches 
applied on top of next-20231117. The top-most one 9d36dec7e6da ("riscv: 
dts: starfive: Add JH7100 MMC nodes") is optional, I added it to extend 
a bit the test suite (SD-card card access also works fine).

[3]: https://gitlab.collabora.com/cristicc/linux-next/-/tree/visionfive-eth

For configuring the kernel, I used:

  $ make [...] defconfig
  $ scripts/config --enable CONFIG_NONPORTABLE --enable ERRATA_STARFIVE_JH7100

I also noticed a warning message right before building starts, but it 
doesn't seem to be harmful:

WARNING: unmet direct dependencies detected for DMA_GLOBAL_POOL
  Depends on [n]: !ARCH_HAS_DMA_SET_UNCACHED [=n] && !DMA_DIRECT_REMAP [=y]
  Selected by [y]:
  - ERRATA_STARFIVE_JH7100 [=y] && ARCH_STARFIVE [=y] && NONPORTABLE [=y]

Thanks,
Cristian