On J722s (AEN), single-cam app shows frozen video with IMX390 RCM running at 60fps after ~2000 frames.

Single-cam app works well with IMX390 at 30fps, but stalls after around 2000 frames with IMX390 running at 60fps (AEN EVM + Fusion1 + IMX390 RCM)

Tests attempted for camera capture

• CSIRx Standalone test (No RTOS or Linux) – passing
  • This is Jahnavi’s test case modified for single channel at 60 FPS – Passing
  • Ran this for 6000 frames at 60 FPS and worked well (single cam app fails at ~2000 frames @ 60 FPS)
• Standalone capture node test (coupled with sensor driver) – passing
  • This is the video_io conformance test modified for single channel at 60 FPS. Ran for 6000 Frames @ 60 FPS
• With these 2 tests above, capture node seems fine for receiving 60fps output from IMX390

[^outputs.zip]

^{Further Debug Details:}

• The VISS pipeline doesn’t stop and no stall event in the HTS, it keeps processing the garbage data. None of the VISS, MSC or LDC configuration changes when this error event is encountered.
• The garbage data starts to stream when the below error is hit:
• [MCU2_0] 790.385886 s: RTOS DPL ERROR: Semaphore wait failed. Timeout expired.

• The input and the output frame list buffers in vhwa_m2mVissProcessReq. In a working scenario, the input buffer is expected to roll within 4 addresses. When the error hits, the input buffer address change.
• In this setup the correct input frame list buffer addresses are: 0xc1245000, 0xc0017000, 0xc0a25000, 0xc0e35000.
• When the error event has occurred, the input buffer address changes to 0xc283f000 and it stays consistent.