CUDA Memcheck Initialization failed (not profiler error)

Hi,
I’m facing an ambiguous internal memcheck error while calling cuda-memcheck on an application on the Jetson TX2 P3310. I’m using CUDA 10.0

It is not the common profiler error which requires COMPUTE_PROFILE or CUDA_PROFILE to be unset.

========= Internal Memcheck Error: Initialization failed
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame:/usr/lib/aarch64-linux-gnu/tegra/libcuda.so.1 (cuDevicePrimaryCtxRetain + 0x154) [0x1fd7d4]

What could be the possible cause? This crops up in every tool (memcheck, racecheck, initcheck)

1 Like

It turns out I can only run cuda-memcheck with sudo, is it possible to run without sudo?

You should not need to run with SUDO

Maybe this will help (from 11.04 - after upgrade gdb won't attach to process - Ask Ubuntu):
On distributions that use the Yama LSM (that includes Ubuntu and Chrome OS), process A can attach to process B only if A is an ancestor of B.

You will probably want to disable this feature by using

echo 0 | sudo tee /proc/sys/kernel/yama/ptrace_scope

To permanently allow it edit /etc/sysctl.d/10-ptrace.conf and change the line:

kernel.yama.ptrace_scope = 1
To read

kernel.yama.ptrace_scope = 0
For some background on why this change was made, see the Ubuntu wiki