|
Does this happen on both Windows and Linux hosts?
Is it a pattern sensitivity? ie is there a consistency to the failure, or the data sent at failure.
Does it occur after a uniform number of packets? ie a packet count vs time to fail.
Are you failing to clear a pending interrupt, or pending error condition, at the point of failure?
Can you use a JTAG debugger on your device to inspect the state of the system/peripherals before you power cycle it, and how does that compare to the state when it is working normally? ie break execution before it fails, compare/contrast to after it fails.
|