PM policy: Separate default policy and events

[bjarki-andreasen]

The default policy currently directly references the private variable next_event from policy_events.c:

zephyr/subsys/pm/policy/policy_default.c

Line 12 in e1389fa

extern struct pm_policy_event *next_event;

to then convert the cycle of said event (if exists) to a cycle in the future:

zephyr/subsys/pm/policy/policy_default.c

Lines 33 to 52 in e1389fa

	if ((next_event) && (next_event->value_cyc >= 0)) {
	uint32_t cyc_curr = k_cycle_get_32();
	int64_t cyc_evt = next_event->value_cyc - cyc_curr;
	/* event happening after cycle counter max value, pad */
	if (next_event->value_cyc <= cyc_curr) {
	cyc_evt += UINT32_MAX;
	}
	if (cyc_evt > 0) {
	/* if there's no system wakeup event always wins,
	* otherwise, who comes earlier wins
	*/
	if (cyc < 0) {
	cyc = cyc_evt;
	} else {
	cyc = MIN(cyc, cyc_evt);
	}
	}
	}

something policy_events.c already implements quite similarly and exposes through pm_policy_next_event_ticks():

zephyr/subsys/pm/policy/policy_events.c

Lines 56 to 69 in e1389fa

	int32_t pm_policy_next_event_ticks(void)
	{
	int32_t cyc_evt = -1;
	if ((next_event) && (next_event->value_cyc > 0)) {
	cyc_evt = next_event->value_cyc - k_cycle_get_32();
	cyc_evt = MAX(0, cyc_evt);
	BUILD_ASSERT(CONFIG_SYS_CLOCK_HW_CYCLES_PER_SEC >= CONFIG_SYS_CLOCK_TICKS_PER_SEC,
	"HW Cycles per sec should be greater that ticks per sec");
	return k_cyc_to_ticks_floor32(cyc_evt);
	}
	return -1;
	}

Additionally, the implementation of pm_policy_next_state() in policy_default.c already gets the nearest kernel tick, wherein the next event has already been accounted for, see implementation of pm_system_suspend():

zephyr/subsys/pm/pm.c

Lines 155 to 156 in e1389fa

	events_ticks = pm_policy_next_event_ticks();
	ticks = ticks_expiring_sooner(kernel_ticks, events_ticks);

zephyr/subsys/pm/pm.c

Line 165 in e1389fa

info = pm_policy_next_state(id, ticks);

This commit removes the redundant and layer violating computation of the tick of the next event from policy_default.c and updates the test test_pm_policy_events to not use default policy to determine if pm_policy_next_event_ticks() is correct.

This PR updates the test suite to actually be able to detect the incorrect argument handling of pm_policy_event_register() fixed with PR #80556 so this PR includes said fix, to be dropped from here when fix is merged :)

[bjarki-andreasen]

added test for pm_policy_event_update()

[ceolin]

Thanks again, same thing again, recent change makes subsystem to account for event out of the policy but the policy was not updated. My bad.

It looks legit.