pm: Include disabled states in search inside pm_state_get

[nordic-krch]

pm_state_get() is used to get a PM state that can be forced. Disabled states (not included in automatic state selection) can also be forced thus they should be included in search inside pm_state_get().

[ceolin]

The change behavior looks good to me (thanks !). There is though some tests breaking because the implementation is assuming

cpus {
          power-states {

I am ok with it, since the other way would need to check per cpu. Can you please update failing tests ?

[tmleman]

@nordic-krch, thank you for the quick response and help in solving the problem. I have one more request for you, when you update this PR, please do a rebase. One of your patches has already been merged, so after the rebase, I will be able to prepare a PR for SOF pointing to this branch, which will trigger the execution of SOF's CI on these changes.

[tmleman]

The proposed solution is okay and solves my problem. However, I wasn't able to build them immediately with SOF:

zephyr/subsys/pm/state.c:64:2: warning: implicit declaration of function 'DT_N_S_cpus_S_power_states_FOREACH_CHILD' is invalid in C99 [-Wimplicit-function-declaration]
        DT_FOREACH_CHILD(DT_PATH(cpus, power_states), DEFINE_DISABLED_PM_STATE)
        ^
zephyr/include/zephyr/devicetree.h:3111:2: note: expanded from macro 'DT_FOREACH_CHILD'
        DT_CAT(node_id, _FOREACH_CHILD)(fn)
        ^
zephyr/include/zephyr/devicetree.h:5269:24: note: expanded from macro 'DT_CAT'
#define DT_CAT(a1, a2) a1 ## a2
                       ^
<scratch space>:102:1: note: expanded from here
DT_N_S_cpus_S_power_states_FOREACH_CHILD
^
zephyr/subsys/pm/state.c:64:48: error: use of undeclared identifier 'DEFINE_DISABLED_PM_STATE'
        DT_FOREACH_CHILD(DT_PATH(cpus, power_states), DEFINE_DISABLED_PM_STATE)
                                                      ^
zephyr/subsys/pm/state.c:90:26: error: invalid application of 'sizeof' to an incomplete type 'const struct pm_state_info []'
        for (uint8_t i = 0; i < ARRAY_SIZE(disabled_states); i++) {
                                ^~~~~~~~~~~~~~~~~~~~~~~~~~~
zephyr/include/zephyr/sys/util.h:122:38: note: expanded from macro 'ARRAY_SIZE'
        ((size_t) (IS_ARRAY(array) + (sizeof(array) / sizeof((array)[0]))))
                                            ^~~~~~~
1 warning and 2 errors generated.

Below are the changes I had to make:

diff --git a/subsys/pm/state.c b/subsys/pm/state.c
index d0e8ad3da92..86f935399b0 100644
--- a/subsys/pm/state.c
+++ b/subsys/pm/state.c
@@ -57,11 +57,11 @@ static const uint8_t states_per_cpu[] = {
 };

 #define DEFINE_DISABLED_PM_STATE(node) \
-       IF_ENABLED(DT_NODE_HAS_STATUS(node, disabled), (PM_STATE_INFO_DT_INIT(node),))
+       COND_CODE_1(DT_NODE_HAS_STATUS(node, disabled), (PM_STATE_INFO_DT_INIT(node),), ())

 /* Array with all states which are disabled but can be forced. */
 static const struct pm_state_info disabled_states[] = {
-       DT_FOREACH_CHILD(DT_PATH(cpus, power_states), DEFINE_DISABLED_PM_STATE)
+       DT_FOREACH_CHILD(DT_PATH(power_states), DEFINE_DISABLED_PM_STATE)
 };

 uint8_t pm_state_cpu_get_all(uint8_t cpu, const struct pm_state_info **states)

[nordic-krch]

So power-states in DT are defined in root and not in cpus?
I've updated patch to look for power states in /power_states or /cpus/power_states.

[tmleman]

So power-states in DT are defined in root and not in cpus?

Yes.

But from what I see, this is the case only for one platform. Initially, it was more widespread but was changed in pull request #60749. The ACE30 platform was integrated into the main branch later and this was not caught in the review. I made a PR that fixes this so that the structure is consistent across all platforms (#89492).

[nordic-krch]

@ceolin I've updated patch to include case when no power states are defined in DT.