Sched MC test cases fails even when sched_mc is not enabled at the kernel
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Linaro PMWG QA |
Fix Released
|
Medium
|
Hongbo Zhang |
Bug Description
In case SCHED_MC is not available at the kernel, the test suite will still fail and run the sched_mc related test cases:
make[1]: Entering directory `/root/
###
### sched_01:
### test the presence of the 'sched_
### https:/
###
sched_01.0: checking 'sched_
make[1]: *** [sched_01.log] Error 1
###
### sched_02:
### test the topology is implemented in the kernel
### https:/
###
sched_02.0/cpu0: checking topology is enabled... pass
sched_02.0/cpu1: checking topology is enabled... pass
###
### sched_03:
### test the topology files are present
### https:/
###
sched_03.0/cpu0: checking 'core_id' exists... pass
sched_03.1/cpu0: checking 'core_siblings' exists... pass
sched_03.2/cpu0: checking 'core_siblings_
sched_03.3/cpu0: checking 'physical_
sched_03.4/cpu0: checking 'thread_siblings' exists... pass
sched_03.5/cpu0: checking 'thread_
sched_03.0/cpu1: checking 'core_id' exists... pass
sched_03.1/cpu1: checking 'core_siblings' exists... pass
sched_03.2/cpu1: checking 'core_siblings_
sched_03.3/cpu1: checking 'physical_
sched_03.4/cpu1: checking 'thread_siblings' exists... pass
sched_03.5/cpu1: checking 'thread_
###
### sched_04:
### test the change are effective for the 'sched_
### https:/
###
sched_04.0: checking setting value to 0... fail
sched_04.1: checking setting value to 1... fail
sched_04.2: checking setting value to 2... fail
sched_04.3: checking setting invalid value to 3... pass
sched_04.4: checking setting invalid value to -1... pass
make[1]: *** [sched_04.log] Error 1
make[1]: Target `check' not remade because of errors.
make[1]: Leaving directory `/root/
make: *** [check] Error 2
This can be the expected behaviour, but I got the wrong impression from the logs, as it seems it's partially supported. 'Skipped' would probably be a better description.
Also, I'm confused if one single 'fail' should make 'make -k check' to get a !0 return code. At LAVA, if the test script returns !0, it'll think that the test case it self failed to run, and not that it produced failed results.
Changed in linaro-power-qa: | |
status: | New → In Progress |
Hongbo, Please make the sched_mc tests optional since the feature is about to be removed from the kernel in 3.5.