JPEG encoding reports "Surface not registered" after hundreds of images on Jetpack 5.x

[smilerlee]

We use NvJPEGEncoder to encode images (resolution: 1920x1080, color format: NVBUF_COLOR_FORMAT_YUV420):

NVJPGGetSurfPinHandle 91: NVJPGGetSurfPinHandle : Surface not registered
NVJPGPushSurfFalconMethodRelocShift 369: Surface not registered
tvmrNvjpgJPEGEncFeedFrame_safe 583: NVJPGPushSurfFalconMethodRelocShift failed for SET_CUR_PIC_CHROMA_U with error 0x4
NVJPGGetSurfPinHandle 91: NVJPGGetSurfPinHandle : Surface not registered
NVJPGPushSurfFalconMethodRelocShift 369: Surface not registered
tvmrNvjpgJPEGEncFeedFrame_safe 590: NVJPGPushSurfFalconMethodRelocShift failed for SET_CUR_PIC_CHROMA_V with error 0x4
NvRmHost1xStreamEndClass: Requested number of operations in StreamBegin not yet pushed
NvRmHost1xStreamEnd: StreamEndClass failed, err = 8
pushNVJPGStreamEnd 153: Stream end failed
tvmrNvjpgJPEGEncFeedFrame_safe 597: Call to pushNVJPGStreamEnd failed with error 0x1
tvmrNvjpgJPEGEncFeedFrame_safe 605: Stream flush failed with error 0x8

Thread 103 "mediapipe/44073" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0xfffedffec900 (LWP 440733)]
0x0000ffffe8ef01c4 in jpeg_write_raw_data () from /usr/lib/aarch64-linux-gnu/tegra/libnvjpeg.so
(gdb) bt
#0  0x0000ffffe8ef01c4 in jpeg_write_raw_data () at /usr/lib/aarch64-linux-gnu/tegra/libnvjpeg.so
#1  0x0000ffffe9d53148 in NvJPEGEncoder::encodeFromFd(int, J_COLOR_SPACE, unsigned char**, unsigned long&, int) (this=
    0xffff4c0034e0, fd=277, color_space=JCS_YCbCr, out_buf=0xfffedffeb120, out_buf_size=@0xfffedffeb130: 1555200, quality=90)
    at /home/work/workspace/jetson/jetson_multimedia_api_35.1.0/samples/common/classes/NvJpegEncoder.cpp:115
...

Any idea for this problem?

[DaneLLL]

Hi,
The latest release is 5.1.1. Please check if you can run 05_jpeg_encode sample to reproduce the failure. If yes, please share us the command and we will set up and try to replicate the failure.

[smilerlee]

The sample is ok, we encounter this problem only after correctly encoding hundreds of images, any idea what might be related to?

[DaneLLL]

Hi,
It might be a bug in low-level and we would need to reporduce it and check further. Would need your help to check if it can be reproduced on Jetpack 5.1.1. And share us a method to reproduce it.

[smilerlee]

Hi, thanks for responding, we might need some time to prepare the reproducible method even through it’s easy in our product, I’ll let you know if we’re ready.