Valgrind with CUDA: noted but unhandled ioctl with no size/direction hints

Hi everyone.

I'm trying Valgrind on a simple CUDA program for CPU memory check (It
worked for GPU memcheck when using cuda-memcheck):

```
#include <stdlib.h>

int main()
{
  void *dp = NULL;
  cudaMalloc(&dp, 1024);
  cudaFree(dp);

  void *p = NULL;
  p = malloc(1024);
  free(p);

  cudaDeviceReset();
  return 0;
}
```

The program hangs with many `noted but unhandled ioctl with no
size/direction hints` warnings.

```
$ nvcc prog.cu
$ valgrind ./a.out
==61517== Memcheck, a memory error detector
==61517== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==61517== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info
==61517== Command: ./a.out
==61517==
==61517== Warning: noted but unhandled ioctl 0x30000001 with no
size/direction hints.
==61517==    This could cause spurious value errors to appear.
==61517==    See README_MISSING_SYSCALL_OR_IOCTL for guidance on
writing a proper wrapper.
==61517== Warning: noted but unhandled ioctl 0x27 with no size/direction hints.
==61517==    This could cause spurious value errors to appear.
==61517==    See README_MISSING_SYSCALL_OR_IOCTL for guidance on
writing a proper wrapper.
==61517== Warning: noted but unhandled ioctl 0x7ff with no size/direction hints.
==61517==    This could cause spurious value errors to appear.
==61517==    See README_MISSING_SYSCALL_OR_IOCTL for guidance on
writing a proper wrapper.
==61517== Warning: noted but unhandled ioctl 0x25 with no size/direction hints.
==61517==    This could cause spurious value errors to appear.
==61517==    See README_MISSING_SYSCALL_OR_IOCTL for guidance on
writing a proper wrapper.
==61517== Warning: noted but unhandled ioctl 0x37 with no size/direction hints.
==61517==    This could cause spurious value errors to appear.
==61517==    See README_MISSING_SYSCALL_OR_IOCTL for guidance on
writing a proper wrapper.
==61517== Warning: noted but unhandled ioctl 0x17 with no size/direction hints.
==61517==    This could cause spurious value errors to appear.
==61517==    See README_MISSING_SYSCALL_OR_IOCTL for guidance on
writing a proper wrapper.
==61517== Warning: set address range perms: large range [0x200000000,
0x300200000) (noaccess)
==61517== Warning: set address range perms: large range [0x7798000,
0x27797000) (noaccess)
==61517== Warning: set address range perms: large range
[0x10006000000, 0x10106000000) (noaccess)
==61517== Warning: noted but unhandled ioctl 0x19 with no size/direction hints.
==61517==    This could cause spurious value errors to appear.
==61517==    See README_MISSING_SYSCALL_OR_IOCTL for guidance on
writing a proper wrapper.
==61517== Warning: set address range perms: large range
[0x10106000000, 0x10206000000) (noaccess)
==61517== Warning: noted but unhandled ioctl 0x21 with no size/direction hints.
==61517==    This could cause spurious value errors to appear.
==61517==    See README_MISSING_SYSCALL_OR_IOCTL for guidance on
writing a proper wrapper.
==61517== Warning: noted but unhandled ioctl 0x1b with no size/direction hints.
==61517==    This could cause spurious value errors to appear.
==61517==    See README_MISSING_SYSCALL_OR_IOCTL for guidance on
writing a proper wrapper.
```

I'm also follwing this question[0] to generate Valgrind suppressions,
the program hangs again with same warnings. How can I use Valgrind
with CUDA correctly? Thanks in advance.

I'm using Ubuntu 18.04 + CUDA 9.1 + Valgrind 3.13.0.

Best regards,
Jiading Guo

[0]: https://stackoverflow.com/questions/20593450/valgrind-and-cuda-are-reported-leaks-real