Am Wed, 20 Nov 2024 10:39:38 +0200 schrieb Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>: > Hi Andreas, > > On Wed, Nov 20, 2024 at 08:54:06AM +0100, Andreas Kemnade wrote: > > Am Tue, 19 Nov 2024 09:02:22 +0200 schrieb Laurent Pinchart: > > > On Mon, Nov 18, 2024 at 08:00:25PM +0100, Andreas Kemnade wrote: > > > > Am Mon, 2 Sep 2024 10:42:31 +0200 schrieb Hans Verkuil : > > > > > > > > > The omap4 camera driver has seen no progress since forever, and > > > > > now OMAP4 support has also been dropped from u-boot (1). So it is > > > > > time to retire this driver. > > > > > > > > Argumenting with OMAP4 support in U-Boot is silly. That indicates that > > > > there is no movement in keeping u-boot uptodate. Bootloader > > > > development/updating is more risky especially if not done by the vendor, > > > > good chances to brick something. And the bootloader might need > > > > signing. So that argument is done nothing. > > > > > > > > Better arguments would be to check if someone has something cooking and > > > > feels not comfortable yet to climb Mount Upstream. > > > > > > > > A good place to ask would be the omap platform > > > > list: linux-omap@xxxxxxxxxxxxxxx > > > > > > > > I get still devicetrees for omap4 devices to review. So there is some > > > > activity with omap4. If you look at postmarketOS you see also some > > > > activity. > > > > > > > > And also someone ported the driver to devicetree support: > > > > https://github.com/iridia-ulb/meta-builderbot/blob/master/recipes-kernel/linux/linux-stable-4.16/0008-omap4iss-Fix-multiple-bugs-and-use-device-tree.patch > > > > > > > > So the situation is not that simple. I am still evaluating it because I > > > > myself have a device with omap4 and camera. > > > > > > Have you tested the camera recently ? The omap4iss driver has been > > > unmaintained in mainline for a very, very long time, and I would be > > > surprised if it worked. > > > > No, I have not tested it. I only have a bitrot out of tree driver for my > > camera which was probably never used with omap. Vendor system seems to > > handle camera via the m3 processor in a closed-source firmware blob. So > > what is the overall picture: > > > > Which omap4 devices have cameras? What is the status of the sensor > > driver? Known working/Mainline/Out-of-tree/none? Datasheet for > > sensor available? > > The question is whether omap4iss can be tested together with a > > known-working camera sensor driver. That would make things a lot easier. > > > > BT200 has a camera without mainline sensor driver. > > Droid4 has also a camera. What is the status of the sensor driver? > > What about the samsung-espresso tablets? And the xyboards? > > > > Pandaboard camera module? If have a pandaboard I use for reference, but > > no camera there. > > I used to work on it using a Pandaboard and an MT9P031 camera module, > from Leopard Imaging if I recall correctly. > Hmm, that would require also soldering the connector and it seems like no standard thing, so I would need to create an adaptor. At least I do not find anything on ebay which looks like it can be attached to the Pandaboard. > > > If someone is interested in taking over maintainership and improving the > > > driver to get it out of drivers/staging/ to drivers/media/, the removal > > > can certainly be reverted. drivers/staging/ is not a place where drivers > > > are left to bitrot, it's meant for active development of code not fully > > > ready for mainline yet. > > > > I guess the way to start is to revert the remove and then update the > > above-mentioned devicetree support patch. I have no feeling how complex > > that whole task would be. > > That would be a first step, yes. After that, completing resizer support > will be a good task candidate, followed by exposing the ISP parameters > to userspace, and adding support for it to libcamera. > Well, I guess before adding the resizer stuff, I want to see a picture influenced by what is in front of the camera of a device I really care about (!= pandaboard) and then adjusting things to the point where format configurations are solved. At some point more could also jump in to screw at sensor chip drivers. Regards, Andreas
