Re: [PATCH 0/4] scmi: Bypass set fwnode to address devlink issue

[Peng Fan <peng.fan@xxxxxxxxxxx>]

On Fri, Dec 27, 2024 at 06:06:24PM +0100, Linus Walleij wrote:
>On Wed, Dec 25, 2024 at 9:21 AM Peng Fan (OSS) <peng.fan@xxxxxxxxxxx> wrote:
>
>> Current scmi drivers not work well with devlink. This patchset is a
>> retry to address the issue in [1] which was a few months ago.
>>
>> Current scmi devices are not created from device tree, they are created
>> from a scmi_device_id entry of each driver with the protocol matches
>> with the fwnode reg value, this means there could be multiple devices created
>> for one fwnode, but the fwnode only has one device pointer.
>>
>> This patchset is to do more checking before setting the device fwnode.
>>
>> This may looks like hack, but seems no better way to make scmi works
>> well with devlink.
>>
>> [1]: https://lore.kernel.org/arm-scmi/CAGETcx8m48cy-EzP6_uoGN7KWsQw=CfZWQ-hNUzz_7LZ0voG8A@xxxxxxxxxxxxxx/
>
>Please drive any devlink-related patches by Saravana Kannan, he's pretty
>much the only person I trust to know how to do devlinks right.

Quote Saravana's conclution[1] here:
"The best fw_devlink could do is just not enforce any dependencies if
there is more than one device instantiated for a given supplier DT
node."

So I think for systems using scmi could not rely on devlink to build
supplier/consumer to make driver probe in order.

[1]https://lore.kernel.org/arm-scmi/CAGETcx8m48cy-EzP6_uoGN7KWsQw=CfZWQ-hNUzz_7LZ0voG8A@xxxxxxxxxxxxxx/

Thanks,
Peng
>
>Yours,
>Linus Walleij