[libcamera-devel] [PATCH 0/5] Enumerate CameraLens by following sensor's ancillary links

[Hanlin Chen]

Hi Daniel,

On Tue, Nov 30, 2021 at 7:03 AM Daniel Scally <djrscally at gmail.com> wrote:
>
> Hey Hanlin
>
> On 29/11/2021 11:42, Hanlin Chen wrote:
> > On Fri, Nov 26, 2021 at 8:48 PM Laurent Pinchart
> > <laurent.pinchart at ideasonboard.com> wrote:
> >> Hi Han-lin,
> >>
> >> On Fri, Nov 26, 2021 at 07:32:51PM +0800, Hanlin Chen wrote:
> >>> On Fri, Nov 26, 2021 at 2:44 PM Jean-Michel Hautbois wrote:
> >>>> On 26/11/2021 01:31, Daniel Scally wrote:
> >>>>> Hello All
> >>>>>
> >>>>> This series is an attempt at making the incoming VCM support a little more
> >>>>> agnostic, by following the new style of media links described in my series to
> >>>>> linux-media [0] to find VCMs connected to Sensors in libcamera. It's based on
> >>>>> top of a series from Han-Lin [1] and another from Kate Hsuan [2] (both of which
> >>>>> series are pretty neat, by the way)
> >>>> Thanks to all three of you for this great work !
> >>>>
> >>>> How can I test it ? Which kernel for SGo2 is usable (A branch would be
> >>>> great) ?
> >>> Many thanks for the great work, Daniel!
> >>> I just tested it on Chromebook with kernel v5.4, and it works perfectly.
> >> Does it work out of the box, without a need for any kernel or firmware
> >> change ? That would be great, it would certainly facilitate adoption.
> >>
> > I didn't change anything in the firmware or kernel, except for
> > Daniel's patches ;-).
>
>
> Excellent; glad it's working for you. Does that mean you guys are
> including some reference between the sensor and vcm device in your ACPI
> design then?
>
I guess so ;-|. It needs some time for the experts on our side to
confirm this and see how we should merge them to ChromeOS.
> >
> >>>>> The general principle of the new links is an entity to entity link which will
> >>>>> be connected by the kernel between a sensor's entity and an entity for a VCM
> >>>>> device, where those entities have a fwnode match based on the "lens-focus"
> >>>>> property against the sensor. These links are then discovered by libcamera and
> >>>>> followed to create an instance of the CameraLens class, replacing the matching
> >>>>> on driver/device names in Han-Lin's original series.
> >>>>>
> >>>>> With the CameraLens available to carry out the controlling of the VCM, I have
> >>>>> pushed the controls to the pipeline handler and removed both all of that
> >>>>> functionality (including the open()/close() of the VCM subdev) from Kate's
> >>>>> work instead.
> >>>>>
> >>>>> Thanks
> >>>>> Dan
> >>>>>
> >>>>> [0] https://patchwork.linuxtv.org/project/linux-media/list/?series=6792
> >>>>> [1] https://patchwork.libcamera.org/project/libcamera/list/?series=2743
> >>>>> [2] https://patchwork.libcamera.org/project/libcamera/list/?series=2750
> >>>>>
> >>>>> Daniel Scally (5):
> >>>>>    libcamera: Add support for ancillary links to MediaLink
> >>>>>    libcamera: media_device: Handle ancillary links in populateLinks()
> >>>>>    libcamera: ipu3-cio2: Discover VCMs through ancillary links
> >>>>>    ipa: ipu3: Send lens controls to pipeline handler
> >>>>>    ipa: ipu3: af: Remove v4l2 interaction from AF algorithm
> >>>>>
> >>>>>   include/libcamera/internal/media_object.h | 10 +++++
> >>>>>   include/linux/media.h                     |  1 +
> >>>>>   src/ipa/ipu3/algorithms/af.cpp            | 29 +------------
> >>>>>   src/ipa/ipu3/algorithms/af.h              |  3 --
> >>>>>   src/ipa/ipu3/ipu3.cpp                     |  4 ++
> >>>>>   src/libcamera/media_device.cpp            | 52 ++++++++++++++++-------
> >>>>>   src/libcamera/media_object.cpp            | 24 ++++++++++-
> >>>>>   src/libcamera/pipeline/ipu3/cio2.cpp      | 45 +++++++++++---------
> >>>>>   8 files changed, 101 insertions(+), 67 deletions(-)
> >>>>>
> >> --
> >> Regards,
> >>
> >> Laurent Pinchart