========= COMPUTE-SANITIZER
========= Program hit CUDA_ERROR_OPERATING_SYSTEM (error 304) due to “OS call failed or operation not supported on this OS” on CUDA API call to cuCtxCreate_v2.
========= Saved host backtrace up to driver entry point at error
========= Host Frame: [0x2cccaf]
========= in /lib64/libcuda.so.1
========= Host Frame: [0x38abe]
========= in ./libavutil.so.58
========= Host Frame:av_hwdevice_ctx_create [0x351cf]
========= in ./libavutil.so.58
========= Host Frame:/date/cyq/osdProject/ffmpegserver/ffmpegserver.cpp:570:ffmpeg::hw_decoder_init(AVCodecContext*&, AVHWDeviceType, AVBufferRef*&) [0x3397b]
========= in ./libosd.so
========= Host Frame:/date/cyq/osdProject/ffmpegserver/ffmpegserver.cpp:103:ffmpeg::createDed(char const*) [0x324dc]
cuda_12.2.2_535.104.05_linux.run
NVIDIA-Linux-x86_64-535.129.03.run
ffmpeg6.1
t4*4
centos7
I encountered an error while running the following code, and I noticed that the libnvcuvid library crashed. This type of situation is common when I have multiple cards on a single machine, but it is normal when I have multiple cards on a single machine
：if ((err = av_hwdevice_ctx_create(&hw_device_ctx, type,
(to_string(ff_deviceId)).c_str(), NULL, 0)) < 0)
Hi,
Did you seek help from FFmpeg community to isolate if this is issue in FFmpeg code or not? Can you share detailed steps with command line to use and hw/sw configuration you use for us to try reproduce this issue internally. Please also attach nvidia-bug-report and core dump file after issue is reproduced.