Why does default_idle_hook() lock deep sleep?

[amq]

Description

• Type: Question
• Related: RTX idle: sleep without locked deep sleep fix #5866

Question

According to the documentation, sleep is managed automatically, including deep sleep. However from what I see, deep sleep is locked by default_idle_hook() for non-tickless targets

mbed-os/rtos/TARGET_CORTEX/mbed_rtx_idle.cpp

Lines 220 to 228 in ce4d6c9

	static void default_idle_hook(void)
	{
	// critical section to complete sleep with locked deepsleep
	core_util_critical_section_enter();
	sleep_manager_lock_deep_sleep();
	sleep();
	sleep_manager_unlock_deep_sleep();
	core_util_critical_section_exit();
	}

What is the reason for this? What is the expected way of entering deep sleep?

I know that it is possible to use a custom default_idle_hook(), but before I do, I would like to understand it more.

[0xc0170]

This was intentionally added for backward compatibility. Previously only sleep was invoked, therefore deep sleep is locked to keep the same behavior. If you remove the deep sleep locking, some platform might potentially not wake up.

Does that answer your question?

[amq]

How is deep sleep meant to be used? Is it just meant for tickless targets or is Thread::attach_idle_hook(&sleep); also a good approach?

[0xc0170]

Yes, you can attach your own handler and overwrite the default behavior. Or enable tickless for your platform (there is feature sleep branch to actually address this, it contains specifications and tests, target additions will follow).