Hi,
I'm working on r48p0 ddk with a custom embedded device.
We are currently running a 3,000-cycle power on/off test on an embedded device.
During this process, we are encountering an issue where, roughly once every 100 cycles, the message "No mali devices found" appears, and EGL functions fail when running an application.
At the time the application is launched, /dev/mali0 has already been probed and the device node exists. However, is it possible that this error occurs because user space is attempting to access the device too early? Structurally, is such a scenario possible?
Interestingly, the issue disappears if we add a short sleep or udevadm before launching the application.
Is there any known history related to this issue?
Thanks.