config: runtime: tests: rt-tests: Rename PRIORITY -> RT_PRIORITY

[sebastianas]

The cyclictest command is run a low task priority. From the output in the test I see
| / # /lava-22804428/bin/lava-test-runner /lava-22804428/0 | + export TESTRUN_ID=0_rt-tests-cyclictest
| + TESTRUN_ID=0_rt-tests-cyclictest
| + cd /lava-22804428/0/tests/0_rt-tests-cyclictest | ++ cat uuid
| + UUID=22804428_1.1.4.1
| + set +x
| <LAVA_SIGNAL_STARTRUN 0_rt-tests-cyclictest 22804428_1.1.4.1> | + cd ./automated/linux/cyclictest/
| + ./cyclictest.sh -D 540s -p 48 -i 1000 -t 2 -a 0 -h '' -w hackbench | Received signal: 0_rt-tests-cyclictest 22804428_1.1.4.1 | Starting test lava.0_rt-tests-cyclictest (22804428_1.1.4.1) | Skipping test definition patterns.
| # /dev/cpu_dma_latency set to 0us
| T: 0 ( 437) P:48 I:1000 C: 540000 Min: 4 Act: 5 Avg: 18 Max: 217
| T: 1 ( 438) P:48 I:1500 C: 360000 Min: 4 Act: 9 Avg: 17 Max: 340
| t0-min-latency pass 4 us
| t0-avg-latency pass 18.29 us
| t0-max-latency pass 217 us
| t1-min-latency pass 4 us
| t1-avg-latency pass 17.69 us
| t1-max-latency pass 340 us
| cyclictest pass

and the test definition has
| name: rt-tests-cyclictest
| parameters:
…
| PRIORITY: 48
…
| status: finished
| priority: 48
| timeouts:

I think that PRIORITY and priority arguments are getting confused and instead of the cylictest priority it is using the scheduling priority.

Rename PRIORITY to RT_PRIORITY

[bhcopeland]

RT_PRIORITY needs to stay as PRIORITY. But the other than that this makes sense

PRIORITY: {{ rt_priority|default('98') }}

As per: https://github.com/Linaro/test-definitions/blob/master/automated/linux/cyclictest/cyclictest.sh

[sebastianas]

I see. I did open kernelci/test-definitions#25. Was this the wrong place or should I additionally open one against the Linaro repository? Having both as RT_PRIORITY does make it look more consistent.

[bhcopeland]

Yes sorry, please open in the Linaro repo. I will mark the KCI one as archived.

[sebastianas]

Done: Linaro/test-definitions#642

[bhcopeland]

Great thank you, I will give it a review 👍

[sashalevin]

The test-definitions side of this (Linaro/test-definitions#642) was merged on 09-23, but this one wasn't, so KernelCI still passes PRIORITY and the tests fall back to their own defaults. For rt-migrate-test that's 96, the value #2667 was trying to avoid, and it's been failing on pretty much every PREEMPT_RT board since: https://lore.kernel.org/linux-next/179038983337.594168.573709144911248204@ec7b70f3a45c/

Could we get this merged?

[bhcopeland]

Merged, and I noticed a typo which I've fixed here: #3226

Feel free to review, since it's going to change the default behaviour from 0 to 0-1.

Thanks!