Other Parts Discussed in Thread: AM5728
Hi,TI
The system contains a RC+7 EP,connected by PCIE Swich.The master MCU is TI AM5728, EP0 is TI DSP C6657, and EP1-EP6 is the media chip of Hess.
Phenomenon of the problem:
The RC connects 7 EP through IDT bridge, and the RC dies after running for a period of time, with a low recurrence probability (more than 20 devices, randomly reappearing one each day).In case of an exception, the main control cannot be connected with the emulator, and the CPU is in the pipeline suspended state.According to the strong correlation between multi-party positioning and PCIE, reset the master pCIE module or reset the PCIE SWITCH, and the master control can be connected by the simulator.
The preliminary analysis:
1.Attempt to reset in the normal operation of the PCIE SWITCH, hang dead phenomenon is consistent with the phenomenon of baking machine, emersion record 100%, due to the reduction and PCIE SWITCH makes master sent PCIE TLP possible packet loss on the deal, so the suspected reason for the kao machine problems is due to the PCIE TLP package packet loss on the deal, consulting bridge piece, doubt is PCIE signal has a problem.
2.through to the PCIE SWITCH register contrast analysis, find that DSP along the PCIE downstream, surprise down the error, the error, may have just started, also may be running for a period of time did not appear, but this error is not a problem to 100% have master hung dead, we suspect is currently just met with surprise when communication down, lead to the RC hanging to death.
At present, we have tested the PCIE signal eye map between the bridge and 6657 by ourselves. The Sdk version we used is as follows:
The main problem: Please help to solve how to eliminate the surprise Down error.Whether in hardware or software