Hi Bryan, On Tue, 21 Jan 2014, Bryan Wu wrote: > On Thu, Jan 16, 2014 at 3:04 PM, Guennadi Liakhovetski > <g.liakhovetski@xxxxxx> wrote: > > Hi Bryan, > > > > On Wed, 15 Jan 2014, Bryan Wu wrote: > > > >> Hi Guennadi, > >> > >> I'm working on upstream our Tegra soc_camera host driver. But found > >> the soc_camera framework is not fully supporting Device Tree probing, > >> am I wrong about that? > > > > Mostly correct, yes, currently soc-camera doesn't support device-tree > > probing. > > > >> While in upstream Tegra kernel, we only support > >> DT probing and there is no board files. > >> > >> Current soc_camera framework needs to put soc_camera_link information > >> in a board file and build up soc-camera-pdrv platform_device, then > >> finally register this soc-camera-pdrv platform_device. > >> > >> For the host driver, we can do DT probing but for i2c soc_camera > >> sensor driver I failed to find any DT probing in upstream kernel. So > >> how to do that without an board file but use DT for this whole thing? > >> > >> Can we use DT like this? > >> DTB file will pass those I2C, clock, regulator, GPIO information to > >> host driver. During host driver DT probing, we dynamically create > >> soc-camera-pdrv platform_device and soc_camera_link then register > >> them. Then the rest of the thing should be the same as None-DT > >> probing. > > > > I've worked on soc-camera DT in the past, this might be the last published > > version > > > > http://marc.info/?l=linux-sh&m=134875489304837&w=1 > > > > As you see, it's quite old. Since then a few things happened. Device tree > > support has been added to V4L2 (see > > Documentation/devicetree/bindings/media/video-interfaces.txt and other > > files in that directory for examples), it is based on asynchronous > > probing, which is also supported by the soc-camera core and some its > > host drivers (e.g. > > > > commit 4dbfd040757b8bf22f4ac17e80b39c068061a16c > > Author: Guennadi Liakhovetski <g.liakhovetski@xxxxxx> > > Date: Tue Jul 30 02:59:49 2013 -0300 > > > > [media] V4L2: mx3_camera: add support for asynchronous subdevice registration > > > > ). So, what you should do, is add asynchronous probing support to your > > driver, add DT support to the soc-camera core, add it to your drivers. > > Also see drivers/media/v4l2-core/v4l2-of.c for helper functions, you > > should be using. > > > > Great, Guennadi. I will start to help to add DT for soc-camera. > > But one more question is how to add DT support for soc-camera sensor > driver? For sensor driver, we need pass those regulator/gpio/clock > information for power on/off operations. If we add I2C device node in > DTS file to pass those settings to driver, driver will be > automatically loaded and start to probing. But I think loading > soc_camera sensor driver should be done by soc_camera core code during > host driver registering. Any suggestions to solve this problem? As I mentioned above, DT support in V4L2 is based on asynchronous probing patches, which allow exactly this - probing (I2C) sensors and (platform) hosts independently and asynchronously. If such a sensor is probed and some of its dependencies are missing (e.g. a master clock, normally provided by the camera host interface), the driver can return -EPROBE_DEFER and will be reprobed later, hopefully when resources are already there. After a successful probing such a sensor driver, supporting asynchronous probing, calls v4l2_async_register_subdev(). After the respective host registers with the async framework, its notifier will be called to inform it of your sensor presence. See drivers, listed by grep -rl v4l2_async_register_subdev drivers/media/i2c/ for examples. Thanks Guennadi --- Guennadi Liakhovetski, Ph.D. Freelance Open-Source Software Developer http://www.open-technology.de/ -- To unsubscribe from this list: send the line "unsubscribe linux-tegra" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
