Dne 13.5.2017 v 08:40 Petr Cvek napsal(a): > The second problem seems to be a same problem. When playing/encoding the data > from the sensor (with or without previous fix) and calling (probably) > anything on v4l2 the drivers stops in a same way. I discovered it by trying > to use the CONFIG_VIDEO_ADV_DEBUG interface to realtime poking the sensor. > > You should be able to recreate that by starting the stream (mplayer or > ffmpeg) and run: > v4l2-ctl -n > or reading registers, running v4l2-compliance etc... >
I get it now. The second problem is likely in pxac_fops_camera_release(). At the closing of the device the function calls sensor_call → s_power and sets the sensor to OFF regardless if somebody is using it somewhere else. There should be reference counter (as was in the original soc_camera_close() function). This bug makes impossible to debug pxa camera and its sensors. cheers, Petr
