Merge ~narahuang/plainbox-provider-checkbox/+git/plainbox-provider-checkbox:SLP_S0-in-s2-idle-check into plainbox-provider-checkbox:master

[Nara Huang]

[Sylvain Pineau (sylvain-pineau)]

Do you want the test to fail if /sys/kernel/debug/pmc_core/slp_s0_residency_usec is not present?

We could check this path in a resource job and mark the test as skipped in such case.

The new requirements would then be something like:

requires:
 cpuinfo.type == "GenuineIntel"
 pmc_core.slp_s0 == "supported"

pmc_core being the name of this new resource job.

What do you think?

review: Needs Information

[Nara Huang (narahuang)]

> Do you want the test to fail if
> /sys/kernel/debug/pmc_core/slp_s0_residency_usec is not present?
>
> We could check this path in a resource job and mark the test as skipped in
> such case.
>
> The new requirements would then be something like:
>
> requires:
> cpuinfo.type == "GenuineIntel"
> pmc_core.slp_s0 == "supported"
>
> pmc_core being the name of this new resource job.
>
> What do you think?
That would be great!
I can study how to write the resource job...

[Jonathan Cave (jocave)]

If I could chip having just done a quick bit of research about what pmc_core is... I thing the resource job could produce output like:

supported: True <- test based on /sys/kernel/debug/pmc_core directory
slp_s0_residency_usec: 10 <- contents of the slp_s0_residency_usec file
<filename>: <value> <- name of other files in dir and their contents

This should provide useful features for any future test we might want to write. On devices that do not support this feature, just:

supported: False

[Nara Huang (narahuang)]

Checked with HWE khfeng, if the system node is not there, that's also an issue.

So I would like to keep the existence check, but change log message for tester to file a bug on that case.

[Sylvain Pineau (sylvain-pineau)]

OK for me but let's land it after the deb release (0.49 of p-p-c)

[Jonathan Cave (jocave)]

This isn't included in any test plans so shouldnt do any harm - but I don't understand yet how it will be used because I don't believe this feature is supported on all Intel kernels.

review: Needs Information

[Nara Huang (narahuang)]

This is from HWE's request, please refer to bug
https://bugs.launchpad.net/plainbox-provider-checkbox/+bug/1838409

[Sylvain Pineau (sylvain-pineau)]

See below (suggestion to rely on fwts to catch problems that only occur after multiple cycles)

[Sylvain Pineau (sylvain-pineau)]

It's not clear to me how it will actually work. It seems assuming that by default the suspend action will use S2idle rather than S3. not all systems/releases have this.

review: Disapprove

Unmerged commits

6cf801f... by Nara Huang on 2019-09-20

Add job to check Intel SoC hits SLP_S0 in S2 idle

diff --git a/units/power-management/jobs.pxu b/units/power-management/jobs.pxu
index 904c2f0..ed2e986 100644
--- a/units/power-management/jobs.pxu
+++ b/units/power-management/jobs.pxu
@@ -318,3 +318,28 @@ requires: cpuinfo.platform in ('i386', 'x86_64', 'ppc64el', 'pSeries')
 _description: Check to see if CONFIG_NO_HZ is set in the kernel (this is just a simple regression check)
 command:
  zgrep 'CONFIG_NO_HZ=y' /snap/{kernel}/current/config-`uname -r` >/dev/null 2>&1 || ( echo "WARNING: Tickless Idle is NOT set" >&2 && exit 1 )
+
+id: power-management/slp_s0-in-S2-idle
+category_id: com.canonical.plainbox::power-management
+unit: job
+plugin: shell
+estimated_duration: 0.6
+requires: cpuinfo.type == "GenuineIntel"
+user: root
+command:
+if [ ! -f /sys/kernel/debug/pmc_core/slp_s0_residency_usec ]; then
+    echo "SLP S0 system node is missing"
+    exit 1
+fi
+OUTPUT=$(cat /sys/kernel/debug/pmc_core/slp_s0_residency_usec)
+if [ $OUTPUT -eq 0 ]; then
+    echo "Didn't reach deepest power saving mode"
+    exit 1
+else
+    echo "Reached deepest power saving mode"
+    exit 0
+fi
+_summary:
+ Check system can reach deepest power saving mode with S2 idle.
+_description:
+ Check system can reach deepest power saving mode with S2 idle.