How is this issue impacting you?
Application crash
Share Your Debug Logs
I found a possible null-check typo in ibdevx QP initialization code. After registering db_umem, the code appears to check wq_umem instead of db_umem.
Location
File:
src/modules/transport/ibdevx/ibdevx.cpp
Code snippet:
db_umem = mlx5dv_devx_umem_reg(context, dbr_buf, NVSHMEMT_IBDEVX_DBSIZE, 0);
NVSHMEMI_NULL_ERROR_JMP(wq_umem, status, NVSHMEMX_ERROR_INTERNAL, out,
"cannot register dbr buf for qpair.\n");
The second line is handling the result of db_umem registration, but checks wq_umem.
This might hide a db_umem == NULL failure path and lead to later undefined behavior.
Steps to Reproduce the Issue
No response
NVSHMEM Version
3.2.5
Your platform details
No response
Error Message & Behavior
No response