Hi everybody, I'm trying to implement proper suspend/resume in the OMAP3 ISP driver and ran into an issue that I'd like to discuss here. The OMAP3 ISP has a bad tendency to crash when the input video stream is stopped asynchronously to the ISP. For that reason the ISP must be suspended before the connected video sensors. As they sensors and ISP are connected to different busses (I2C and platform), the suspend/resume order is not guaranteed. My first issue is to model that dependency and instruct the kernel to suspend the sensors and ISP in the right order. The second issue comes from the fact that the sensor clock is (or least can be) supplied by the ISP. In order to suspend the sensor properly the clock needs to be present, and the ISP can't thus be fully suspended, which conflicts with the first requirement. I've thought about splitting the ISP suspend operation in prepare() and suspend(). prepare() would stop the video stream in the ISP, and suspend() would really suspend the ISP, disabling the clocks. The dev_pm_ops documentation states that prepare() must not touch the hardware state, so that's probably not a very good solution. I could split it using suspend()/suspend_noirq(), but it might be an abuse of suspend_noirq(). The same problem occurs during resume as well, as the ISP clock must be restarted before resuming the sensor, which in turn should be resumed before the ISP video stream is restarted. Thoughts and advices would be appreciated on both issues. -- Regards, Laurent Pinchart -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
