Files
pytorch/aten/tools/valgrind.sup
Nikita Shulga efeb988518 Suppress "ioctl points to uninitialised" check (#48187)
Summary:
libcuda.so from CUDA-11.1 makes ioctl() that valgrind's memcheck tool considers dangerous
Instruct valgrind to suppress that check

Fixes false positives reported in https://app.circleci.com/pipelines/github/pytorch/pytorch/240774/workflows/d4c66de8-f13b-47a2-ae62-2ec1bbe0664b/jobs/9026496

Pull Request resolved: https://github.com/pytorch/pytorch/pull/48187

Reviewed By: janeyx99

Differential Revision: D25059850

Pulled By: malfet

fbshipit-source-id: 982df5860524482b0fcb2bfc6bb490fb06694cf6
2020-11-18 18:45:46 -08:00

52 lines
705 B
Plaintext

{
<startup>
Memcheck:Cond
fun:index
fun:expand_dynamic_string_token
fun:_dl_map_object
fun:map_doit
fun:_dl_catch_error
fun:handle_ld_preload
...
}
{
Cond_cuda
Memcheck:Cond
obj:*libcuda.so*
...
}
{
Value8_cuda
Memcheck:Value8
obj:*libcuda.so*
...
}
{
ignore_cuda_ioctl_param_points_to_uninitialised_bytes
Memcheck:Param
ioctl(generic)
fun:ioctl
obj:*libcuda.so*
...
}
{
ignore_libomp_setaffinity_check
Memcheck:Param
sched_setaffinity(mask)
fun:syscall
obj:*/lib*/libomp.so*
...
}
{
ignore_memmove_source_destination_overlap
Memcheck:Overlap
fun:__memcpy_chk
fun:cpuinfo_linux_parse_cpulist
...
}