Re: [PATCH 3/5] power: sequencing: pcie-m2: Create auxiliary device for the M.2 BT interface

From: Manivannan Sadhasivam

Date: Mon Sep 21 2026 - 07:48:12 EST


On Mon, Sep 21, 2026 at 05:51:28AM -0400, Bartosz Golaszewski wrote:
> On Mon, 21 Sep 2026 11:08:58 +0200, Bartosz Golaszewski <brgl@xxxxxxxxxx> said:
> > On Mon, 21 Sep 2026 09:58:27 +0200, Manivannan Sadhasivam <mani@xxxxxxxxxx> said:
> >> On Wed, Sep 16, 2026 at 04:19:16AM -0500, Bartosz Golaszewski wrote:
> >>> On Tue, 15 Sep 2026 16:24:04 +0200, Manivannan Sadhasivam via B4 Relay
> >>> <devnull+manivannan.sadhasivam.oss.qualcomm.com@xxxxxxxxxx> said:
> >>> > From: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxxxxxxxx>
> >>> >
> >>> > Currently, the driver instantiates the BT interface of the WCN6855 and
> >>> > WCN7850 combo modules by creating a devicetree node at runtime using an
> >>> > of_changeset, with the compatible passed through the PCI ID table. But this
> >>> > method of creating a devicetree node just for matching with the driver is
> >>> > not recommended.
> >>> >
> >>> > So create an auxiliary device for the BT interface instead. The serdev
> >>> > transport is still allocated here and handed over to the Bluetooth driver
> >>> > through the auxiliary device, along with the power sequencing target used
> >>> > to power up the controller. Also extend the provider match so that the
> >>> > auxiliary device created for this connector is matched as its power
> >>> > sequencing consumer.
> >>> >
> >>> > Since the runtime node creation is now gone, select AUXILIARY_BUS in
> >>> > place of OF_DYNAMIC.
> >>> >
> >>> > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxxxxxxxx>
> >>> > ---
> >>>
> >>> That's some nice stuff, thanks for doing it!
> >>>
> >>> It looks like there's no build-time dependency on this, only run-time? Can I
> >>> take it independently?
> >>>
> >>
> >> Yeah sure!
> >>
> >
> > Nevermind that, it still needs is_controllable().
> >
>
> Patch 3/5 also needs the header patch 1/5 adds, so we'll need a different
> merge strategy here.
>

Oops... I replied without checking the patch order and trusting my memory :/
But even without the dependencies, merging just the pwrseq driver will break the
functionality. So a better strategy would be to merge the series in a single go
to the Bluetooth tree.

- Mani

--
மணிவண்ணன் சதாசிவம்