Re: [PATCH v3] media: ipu-bridge: do not use the CVS device lookup for IVSC

[Junjie Cao <junjie.cao@intel.com>]

On Sun, 13 Sep 2026 17:37:49 +0800, Junjie Cao wrote:
> Its module is also called intel_cvs. A vision-drivers DKMS build
> installs under updates/, which depmod searches before kernel/, so a
> CONFIG_VIDEO_INTEL_CVS=m kernel still binds the vendor module.

Arch's kernel is not such a kernel. config.x86_64 of linux
7.2.4.arch1 has I2C, ACPI, VIDEO_DEV=m, IPU_BRIDGE=m,
INTEL_VSC=m, VIDEO_INTEL_IPU6=m and VIDEO_INTEL_IPU7=m, and no
VIDEO_INTEL_CVS line at all: the option sits in the
"Miscellaneous helper chips" menu, which
MEDIA_HIDE_ANCILLARY_SUBDRV hides on the default
MEDIA_SUBDRV_AUTOSELECT=y without EXPERT, and nothing selects
or implies it. olddefconfig with that config on current
mainline writes no symbol; with EXPERT=y it shows up as
"is not set".

So on a stock Arch kernel, and per
https://github.com/omacom/omarchy/issues/10948 on linux-ptl,
the DKMS module is the only intel_cvs there is, and without it
nothing clears the sensor's _DEP on INTC10E1. Either way the
camera is gone on 7.2. Patch: "media: i2c: cvs: Do not hide
the option in the ancillary menu", sent separately; with it
that config gets intel_cvs=m on the next olddefconfig.

For the DA16260: apply the patch, or set EXPERT=y so the
option is visible, then build 7.2.x with
CONFIG_VIDEO_INTEL_CVS=m, uninstall the vision-drivers DKMS
module for that kernel, boot. Without one of the two
olddefconfig drops the =m again. The spk-id check from my
previous mail applies once the in-tree driver binds.