On 03/30/2017 04:02 AM, Russell King - ARM Linux wrote:
This fails at step 1. The removal of the frame interval support now means my setup script fails when trying to set the frame interval on the camera: Enumerating pads and links Setting up format SRGGB8_1X8 816x616 on pad imx219 0-0010/0 Format set: SRGGB8_1X8 816x616 Setting up frame interval 1/25 on pad imx219 0-0010/0 Frame interval set: 1/25 Setting up format SRGGB8_1X8 816x616 on pad imx6-mipi-csi2/0 Format set: SRGGB8_1X8 816x616 Setting up frame interval 1/25 on pad imx6-mipi-csi2/0 Unable to set frame interval: Inappropriate ioctl for device (-25)Unable to setup formats: Inappropriate ioctl for device (25) This is because media-ctl tries to propagate it from the imx219 source pad to the csi2 sink pad, and the csi2 now fails that ioctl.
I assume you're using Philipp's frame interval patches to media-ctl. Can you make the frame interval propagation optional in those patches? I.e. don't error-out with a failure code if the ioctl returns ENOTTY. Steve
This makes media-ctl return a failure code, which means that it's not possible for a script to determine whether the failure was due to the camera setup or something else. So, we have to assume that the whole command failed. This is completely broken, and I'm even more convinced that those arguing for this behaviour really have not thought it through well enough before demanding that this code was removed. As far as I'm concerned, the end result is completely broken and unusable. I'm going to be merging the frame interval support back into my test tree, because that's the only sane thing to do. If v4l2 people want to object to having frame interval support present for all subdevs, then _they_ need to make sure that the rest of their software conforms to what they're telling people to do.
_______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
