Re: [PATCH 4/4] riscv: dts: spacemit: k3-com260-ifx: Add rfkill nodes

[Shengyu Qu]

在 2026/10/13 2:32, Jennifer Berringer 写道:
> Hi,

...

> To my understanding, the pwrseq-pcie-m2 driver that uses this binding
> currently only supports systems that use UART for Bluetooth, and is
> needed for them because UART is a non-discoverable bus. Quoting the
> cover letter of the series that introduced this binding:
>
> "This series adds support for connectors that expose PCIe interface for
> WiFi and UART interface for BT. Other interfaces are left for future
> improvements." [1]
>
> K3 CoM260 Dev Kit ships with its E-key slot populated with an RTL8852BE
> module, which uses PCIe for WiFi and USB2 for BT. I verified that my
> patch series does make the rfkill command be able to disconnect and
> reconnect the WiFi and BT devices with this module. I also tested this
> patch set with an Intel AX210.

There are some M.2 E key cards that can be bought and is using UART BT,
for example: Radxa's wireless cards (except A8) are using SDIO+UART
interface on E key cards. Also Quectel has some QCNCM865A cards on sale
which uses PCIE+UART. To achieve best compatibility we can't just use
current implementation.

> In addition to PCIe and USB2, this board also has its M.2 E-key slot
> connected to sdio, uart6, and i2s3, but none of those are currently
> enabled in this dts and the RTL8852BE module does not use them.
>
> Several other existing device trees seem to use rfkill-gpio nodes to
> control W_DISABLE pins on M.2 E-key slots. Looking at rk3588 boards for
> example: OrangePi 5 Plus, ArmSoM W3, and Rock 5B all do so in the
> current tree.

That's due to the M.2 E key series just landed a few months ago and
those boards came much earlier. And that old implementation has created
some issues if we try to support UART cards while some users don't have
that card:

https://lore.kernel.org/all/20250220165051.1889055-1-wens@xxxxxxxxxx/

Best regards,
Shengyu

> Unlike those, the rfkill nodes in my patch aren't actually necessary for
> usable WiFi and BT on this board. The shutdown-gpios here are
> GPIO_ACTIVE_LOW and this merely allows using rfkill to block both radios
> via the W_DISABLE pins, so I'm not opposed to dropping this patch if
> anyone believes it should be dropped.
>
> [1] https://lore.kernel.org/lkml/20260326-pci-m2-e-v7-0-43324a7866e6@xxxxxxxxxxxxxxxx/

...

Attachment: OpenPGP_0xE3520CC91929C8E7.asc
Description: OpenPGP public key

Attachment: OpenPGP_signature.asc
Description: OpenPGP digital signature