hyperthreading in ubuntu_ltp_stable failed on J-ibm with node onibi

[Po-Hsu Lin]

This failure was hinted without a proper bug report. It seems this is specific to node onibi.

Test log:
 Failed test cases : smt_smp_affinity smt_smp_enabled
 startup='Tue Feb 21 01:51:52 2023'
 smt_smp_enabled 0 TINFO : Begin: SMT/SMP Enabled
 smt_smp_enabled 1 TCONF : ltpapicmd.c:188: Enable:/sys/devices/system/cpu/cpuX/online
 smt_smp_enabled 0 TINFO : End: SMT/SMP Enabled
 tag=smt_smp_enabled stime=1676944312 dur=0 exit=exited stat=1 core=no cu=0 cs=0
 startup='Tue Feb 21 01:51:52 2023'
 smt_smp_affinity 0 TINFO : Begin: SMT/SMP Affinity
 smt_smp_affinity 1 TCONF : ltpapicmd.c:188: cpuX:offline
 smt_smp_affinity 0 TINFO : End: SMT/SMP Affinity
 tag=smt_smp_affinity stime=1676944312 dur=0 exit=exited stat=1 core=no cu=0 cs=0

Nothing interesting in dmesg:
[104477.578371] LTP: starting smt_smp_enabled (smt_smp_enabled.sh)
[104477.588795] LTP: starting smt_smp_affinity (smt_smp_affinity.sh)

With "set -x" added to test test script (smt_smp_enabled.sh, smt_smp_affinity.sh) it can help us to identify the failing point:

smt_smp_enabled test (smt_smp_enabled.sh):
<<<test_start>>>
tag=smt_smp_enabled stime=1682415438
cmdline="smt_smp_enabled.sh"
contacts=""
analysis=exit
<<<test_output>>>
+ tst_resm TINFO Begin: SMT/SMP Enabled
smt_smp_enabled 0 TINFO : Begin: SMT/SMP Enabled
+ [ -f ./ht_enabled ]
+ ./ht_enabled
+ ret_value=0
+ [ 0 -ne 0 ]
+ tst_ncpus
+ no_of_processor=4
+ tst_ncpus_conf
+ no_of_cpu=32
+ [ 4 -lt 32 ]
+ test_op=Enable:/sys/devices/system/cpu/cpuX/online
+ tst_resm TCONF Enable:/sys/devices/system/cpu/cpuX/online
smt_smp_enabled 1 TCONF : ltpapicmd.c:188: Enable:/sys/devices/system/cpu/cpuX/online
+ tst_resm TINFO End: SMT/SMP Enabled
smt_smp_enabled 0 TINFO : End: SMT/SMP Enabled
+ TFAILCNT=1
+ exit 1

smt_smp_affinity test (smt_smp_affinity.sh):
<<<test_output>>>
incrementing stop
smt_smp_affinity 0 TINFO : Begin: SMT/SMP Affinity
+ [ -f ./ht_enabled ]
+ ./ht_enabled
+ ret_value=0
+ [ 0 -ne 0 ]
+ tst_ncpus
+ no_of_processor=4
+ tst_ncpus_conf
+ no_of_cpu=32
+ [ 4 -lt 32 ]
+ tst_resm TCONF cpuX:offline
smt_smp_affinity 1 TCONF : ltpapicmd.c:188: cpuX:offline
+ tst_resm TINFO End: SMT/SMP Affinity
smt_smp_affinity 0 TINFO : End: SMT/SMP Affinity
+ TFAILCNT=1
+ exit 1

The failure is because no_of_processor < no_of_cpu.

This issue still exists with the latest LTP upstream.

[Po-Hsu Lin (cypressyew)]

With more investigation on this node, it seems the hyper-threading is probably not enabled in BIOS.
ubuntu@onibi:~/ltp$ cat /sys/devices/system/cpu/smt/active
0
ubuntu@onibi:~/ltp$ sudo dmidecode -t processor | grep Count
 Core Count: 4
 Thread Count: 4

And the console is not properly configured, I was unable to check this fact in BIOS with ipmi commands.