Hi,
I'm working on an NXP i.MX6-based system running Linux kernel 4.9.88. I'm using a custom camera application that interfaces with V4L2 and the IPU (CSI/IC).
After running for some time, the application crashes, and I see the following messages in the kernel log:
Shortly after this, the application exits with code 255 (checked with "echo $?"). Also, in some cases, the message send_sig_info error appears in dmesg.
From what I can tell:
The camera pipeline works for a while (changed between 2 and 7.5 hours) and then stops.
The mxc_v4l_dqueue() function appears to timeout, likely due to no frames being dequeued from the IPU buffer.
This eventually leads to the application terminating.
My questions:
What exactly does enc_counter 0 indicate in this context?
What could cause the mxc_v4l_dqueue() timeout — sensor failure, IPU issue, or buffer mismanagement?
Are there known stability issues in the IPU or V4L2 stack in kernel 4.9.88?
Is there any recommended workaround or patch for this issue?
Any guidance or suggestions would be greatly appreciated.
Thanks,
Hello,
A1.
IPU_INT_STAT_10 means AXI read access error, one can check programming requirements provided for AXIR units in sect.37.4.2.4.1 Handling real time channels,
37.4.2.5, Table 37-15 i.MX6DQ Reference Manual
A2. The que wait to much for buffer fullfil check your camera and check the dtb
A3. No
A4.
You need to add the following line when defining your capture device:
VideoCapture cap("imxv4l2videosrc device=\"/dev/video0\" ! videoconvert ! appsink")
This worked with the Analogue Camera Adapter. For you webcam, you may need different videosources. The different video sources can be found with the command gst-inspect-1.0 | grep v4l2 on the EVK.
Regards