Hello, I've been unable to get two USB CPIA cameras to work reliably on the same machine at more or less the same time. Either camera (/dev/video0 or /dev/video1) will work fine by itself when the other camera is not being used--the problem arises when trying to use them at or near the same time. While it is possible to open /dev/video0 and /dev/video1 simultaneously, if one device is in the middle of a VIDIOCMCAPTURE then opening the other device fails with a "device or resource busy error". If I coordinate the device openings to work around this problem (that is, not starting captures until both devices are successfully opened), then things fail if one card tries to do a VIDIOCMCAPTURE while the other is already doing a capture. I haven't found any V4L api docs anywhere that indicate that doing two captures at once on separate devices from separate processes is not permitted. I have experimented with protecting the actual capture with a shared memory semaphore so two processes never try to do a capture at the same time. This seems to work mostly, though there are enough lingering problems that I suspect that this may not be a proper solution. Does anyone have a recommended approach? Has anyone else gotten two USB cams working on the same machine reliably? If so, under what kernel/v4l driver version? I am on kernel 2.4.0. The cameras are generic USB CPIA cameras made by Ezonics. I have used a variety of test programs to explore the error behavior, but my main test environment is two "motion" processes (, each process controlling a camera. I have seen the same behavior with two concurrent gqcam processes as well, but it usually takes longer to manifest the problem with gqcam since gqcam spends a lot of time waiting between successive VIDIOCMCAPTURE calls and therefore the probability of both processes trying to do a VIDIOCMCAPTURE at the same time is lower. Eventually, however, an error will occur. Any suggestions will be appreciated. Rich
