Description of problem:

All of the mvapich2 benchmarks fail with RC134 or RC1 when tested on MLX5 ROCE devices with bonding.

This is a regression from RHEL-8.7.0, where no such issue was encountered and ALL BENCHMARKS PASSED.

Version-Release number of selected component (if applicable):

Clients: rdma-virt-03
Servers: rdma-virt-02

DISTRO=RHEL-8.8.0-20221120.2

+ [22-11-22 15:16:18] cat /etc/redhat-release
Red Hat Enterprise Linux release 8.8 Beta (Ootpa)

+ [22-11-22 15:16:18] uname -a
Linux rdma-virt-03.rdma.lab.eng.rdu2.redhat.com 4.18.0-438.el8.x86_64 #1 SMP Mon Nov 14 13:08:07 EST 2022 x86_64 x86_64 x86_64 GNU/Linux

+ [22-11-22 15:16:18] cat /proc/cmdline
BOOT_IMAGE=(hd0,msdos1)/vmlinuz-4.18.0-438.el8.x86_64 root=UUID=f786bce9-41a1-4145-96e0-8d6faa88314b ro intel_idle.max_cstate=0 processor.max_cstate=0 intel_iommu=on iommu=on console=tty0 rd_NO_PLYMOUTH crashkernel=auto resume=UUID=f5aad8ae-14bc-47b5-8ab5-7bbb30c30b4f console=ttyS1,115200n81

+ [22-11-22 15:16:18] rpm -q rdma-core linux-firmware
rdma-core-41.0-1.el8.x86_64
linux-firmware-20220726-110.git150864a4.el8.noarch

+ [22-11-22 15:16:18] tail /sys/class/infiniband/mlx5_0/fw_ver /sys/class/infiniband/mlx5_1/fw_ver /sys/class/infiniband/mlx5_bond_0/fw_ver
==> /sys/class/infiniband/mlx5_0/fw_ver <==
12.28.2006

==> /sys/class/infiniband/mlx5_1/fw_ver <==
12.28.2006

==> /sys/class/infiniband/mlx5_bond_0/fw_ver <==
14.32.1010

+ [22-11-22 15:16:18] lspci
+ [22-11-22 15:16:18] grep -i -e ethernet -e infiniband -e omni -e ConnectX
02:00.0 Ethernet controller: Broadcom Inc. and subsidiaries NetXtreme BCM5720 Gigabit Ethernet PCIe
02:00.1 Ethernet controller: Broadcom Inc. and subsidiaries NetXtreme BCM5720 Gigabit Ethernet PCIe
03:00.0 Ethernet controller: Broadcom Inc. and subsidiaries NetXtreme BCM5720 Gigabit Ethernet PCIe
03:00.1 Ethernet controller: Broadcom Inc. and subsidiaries NetXtreme BCM5720 Gigabit Ethernet PCIe
04:00.0 Infiniband controller: Mellanox Technologies MT27700 Family [ConnectX-4]
04:00.1 Infiniband controller: Mellanox Technologies MT27700 Family [ConnectX-4]
05:00.0 Ethernet controller: Mellanox Technologies MT27710 Family [ConnectX-4 Lx]
05:00.1 Ethernet controller: Mellanox Technologies MT27710 Family [ConnectX-4 Lx]

This same failures were observed in another MLX5 ROCE with bonding on rdma-dev-19/20 host pair, as well.

Installed:
mpitests-mvapich2-5.8-1.el8.x86_64 mvapich2-2.3.6-1.el8.x86_64

How reproducible:
100%

Steps to Reproduce:
1. bring up the RDMA hosts mentioned above with RHEL8.6 build
2. set up RDMA hosts for mvapich2 benchamrk tests
3. run one of the mvapich2 benchmark with "mpirun" command, as the following:

timeout --preserve-status --kill-after=5m 3m mpirun -hostfile /root/hfile_one_core -np 2 mpitests-IMB-MPI1 PingPong -time 1.5

Actual results:

+ [22-11-23 12:46:14] timeout --preserve-status --kill-after=5m 3m mpirun -hostfile /root/hfile_one_core -np 2 mpitests-IMB-MPI1 PingPong -time 1.5
[rdma-virt-03.rdma.lab.eng.rdu2.redhat.com:mpi_rank_1][rdma_find_network_type] Warning: Detected active HCAs in different subnets (mlx5_1 in subnet 1 and mlx5_0 in subnet 3). This can cause hangs. Plese use MV2_IBA_HCA to select appropriate HCAs or force rail sharing.

• • • buffer overflow detected ***: mpitests-IMB-MPI1 terminated
      [rdma-virt-03.rdma.lab.eng.rdu2.redhat.com:mpi_rank_1][error_sighandler] Caught error: Aborted (signal 6)

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= PID 59810 RUNNING AT 172.31.45.203
= EXIT CODE: 134
= CLEANING UP REMAINING PROCESSES
= YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
===================================================================================
[proxy:0:0@rdma-virt-02.rdma.lab.eng.rdu2.redhat.com] HYD_pmcd_pmip_control_cmd_cb (pm/pmiserv/pmip_cb.c:911): assert (!closed) failed
[proxy:0:0@rdma-virt-02.rdma.lab.eng.rdu2.redhat.com] HYDT_dmxu_poll_wait_for_event (tools/demux/demux_poll.c:76): callback returned error status
[proxy:0:0@rdma-virt-02.rdma.lab.eng.rdu2.redhat.com] main (pm/pmiserv/pmip.c:202): demux engine error waiting for event
YOUR APPLICATION TERMINATED WITH THE EXIT STRING: Aborted (signal 6)
This typically refers to a problem with your application.
Please see the FAQ page for debugging suggestions
+ [22-11-23 12:46:14] __MPI_check_result 134 mpitests-mvapich2 IMB-MPI1 PingPong mpirun /root/hfile_one_core

Expected results:
Normal completion of the benchmark tests with expected stat outputs

example normal result for the above benchmark with RHEL-8.7.0

+ [22-11-22 23:51:49] timeout --preserve-status --kill-after=5m 3m mpirun -hostfile /root/hfile_one_core -np 2 mpitests-IMB-MPI1 PingPong -time 1.5
[rdma-virt-03.rdma.lab.eng.rdu2.redhat.com:mpi_rank_1][rdma_find_network_type] Warning: Detected active HCAs in different subnets (mlx5_1 in subnet 1 and mlx5_0 in subnet 3). This can cause hangs. Plese use MV2_IBA_HCA to select appropriate HCAs or force rail sharing.
[rdma-virt-02.rdma.lab.eng.rdu2.redhat.com:mpi_rank_0][rdma_find_network_type] Warning: Detected active HCAs in different subnets (mlx5_1 in subnet 1 and mlx5_0 in subnet 3). This can cause hangs. Plese use MV2_IBA_HCA to select appropriate HCAs or force rail sharing.
[rdma-virt-02.rdma.lab.eng.rdu2.redhat.com:mpi_rank_0][rdma_param_handle_heterogeneity] All nodes involved in the job were detected to be homogeneous in terms of processors and interconnects. Setting MV2_HOMOGENEOUS_CLUSTER=1 can improve job startup performance on such systems. The following link has more details on enhancing job startup performance. http://mvapich.cse.ohio-state.edu/performance/job-startup/.
[rdma-virt-02.rdma.lab.eng.rdu2.redhat.com:mpi_rank_0][rdma_param_handle_heterogeneity] To suppress this warning, please set MV2_SUPPRESS_JOB_STARTUP_PERFORMANCE_WARNING to 1
#----------------------------------------------------------------

1. Intel(R) MPI Benchmarks 2021.3, MPI-1 part
   #----------------------------------------------------------------
2. Date : Tue Nov 22 23:51:49 2022
3. Machine : x86_64
4. System : Linux
5. Release : 4.18.0-425.3.1.el8.x86_64
6. Version : #1 SMP Fri Sep 30 11:45:06 EDT 2022
7. MPI Version : 3.1
8. MPI Thread Environment:

1. Calling sequence was:

1. mpitests-IMB-MPI1 PingPong -time 1.5

1. Minimum message length in bytes: 0
2. Maximum message length in bytes: 4194304
   #
3. MPI_Datatype : MPI_BYTE
4. MPI_Datatype for reductions : MPI_FLOAT
5. MPI_Op : MPI_SUM

1. List of Benchmarks to run:

1. PingPong

#---------------------------------------------------

1. Benchmarking PingPong
2. #processes = 2
   #---------------------------------------------------
   #bytes #repetitions t[usec] Mbytes/sec
   0 1000 1.11 0.00
   1 1000 1.18 0.84
   2 1000 1.18 1.70
   4 1000 1.19 3.36
   8 1000 1.17 6.84
   16 1000 1.22 13.10
   32 1000 1.22 26.13
   64 1000 1.23 52.03
   128 1000 1.29 99.10
   256 1000 1.67 153.13
   512 1000 1.74 293.45
   1024 1000 1.92 534.03
   2048 1000 2.28 897.07
   4096 1000 3.15 1302.10
   8192 1000 4.27 1918.73
   16384 1000 6.25 2619.48
   32768 1000 7.72 4245.28
   65536 640 11.85 5530.48
   131072 320 17.72 7396.96
   262144 160 27.29 9605.08
   524288 80 46.38 11305.31
   1048576 40 84.53 12404.59
   2097152 20 160.96 13029.32
   4194304 10 315.14 13309.26

1. All processes entering MPI_Finalize

Additional info: