posix: implement pthread_attr_getinheritsched() and pthread_attr_setinheritsched()

[moonlight83340]

This is part of the See #51211 (RFC #51211).

pthread_attr_setinheritsched() and pthread_attr_getinheritsched() are required as part of _POSIX_THREAD_PRIORITY_SCHEDULING Option Group.

For more information, please refer to https://pubs.opengroup.org/onlinepubs/9699919799/functions/pthread_attr_setinheritsched.html

Fixes #66968
Fixes #66966

[moonlight83340]

What should I set on initial value in pthread_attr_init() ?
https://pubs.opengroup.org/onlinepubs/9699919799/functions/pthread_attr_setinheritsched.html
I may haven't take a good look at it but I see no mention 😓

[ycsin]

What should I set on initial value in pthread_attr_init() ?

It seems to be "implementation-defined" according to the spec:

When a process is created, its single thread has a scheduling policy and associated attributes equal to the policy and attributes of the process. The default scheduling contention scope value is implementation-defined. The default values of other scheduling attributes are implementation-defined.

I did a quick search on google for "default value of inheritsched" and seems like (based on my observation in the search result) most Linux OSes have the default value of PTHREAD_INHERIT_SCHED:

• QNX Neutrino ("The default value of the thread inherit scheduling attribute is PTHREAD_INHERIT_SCHED.")
• IBM AIX ("The default value of the inheritsched attribute is PTHREAD_INHERIT_SCHED")
• Oracle Solaris ("might change from PTHREAD_EXPLICIT_SCHED to PTHREAD_INHERIT_SCHED in a future Solaris release")

The Apache NuttX however, has the default value of PTHREAD_EXPLICIT_SCHED

[moonlight83340]

closed by error 😓

[cfriedt]

Wow @moonlight83340 - you have been very busy!

Thanks - I'm taking a look now.

[cfriedt]

@moonlight83340 - nice work 👍

I think there may be some additional tests to write, and likely some additional work on the implementation of pthread_create()

Tests:

1. Verify that the scheduling attributes of the parent thread (policy, priority, ..) are inherited by a child thread when INHERIT_SCHED is used. This may be tricky, because currently Zephyr threads are not pthreads. It might be necessary to extrapolate equivalent pthread attributes (policy, priority) from the running k_thread and compare those.

2. Verify that the scheduling attributes specified by the pthread_attr_t are passed to the child thread.

For both of those, you should be able to create a static helper function in the same file. Something like this?

static void test_pthread_attr_set_inheritsched_common(int inheritsched);

[moonlight83340]

@moonlight83340 - nice work 👍

I think there may be some additional tests to write, and likely some additional work on the implementation of pthread_create()

Tests:

1. Verify that the scheduling attributes of the parent thread (policy, priority, ..) are inherited by a child thread when INHERIT_SCHED is used. This may be tricky, because currently Zephyr threads are not pthreads. It might be necessary to extrapolate equivalent pthread attributes (policy, priority) from the running `k_thread` and compare those.

2. Verify that the scheduling attributes specified by the pthread_attr_t are passed to the child thread.

For both of those, you should be able to create a static helper function in the same file. Something like this?

static void test_pthread_attr_set_inheritsched_common(int inheritsched);

Still working on the test part not sure on how I will do that but I think I got it for the pthread_create() part.

I have done something like that :

	struct posix_thread *self = NULL;

	K_SPINLOCK(&pthread_pool_lock) {
		self = to_posix_thread(pthread_self());
		if (self == NULL) {
			K_SPINLOCK_BREAK;
		}
		if (self_attr.inheritsched == PTHREAD_INHERIT_SCHED) {
			t->attr.priority = self->attr.priority;
			t->attr.schedpolicy = self->attr.schedpolicy;
			t->attr.inheritsched = self->attr.inheritsched;
		}
	}

[moonlight83340]

Not sure on the error, I will check that when I have time.

[cfriedt]

@moonlight83340 - can you please update this PR?

[moonlight83340]

@moonlight83340 - can you please update this PR?

I will take a look on the week ! ^^'

[moonlight83340]

Assertion failed at CMAKE_SOURCE_DIR/src/pthread_attr.c:541: test_pthread_attr_set_inheritsched_parent_fn: pthread_create(&child, NULL, test_pthread_attr_set_inheritsched_child_fn, NULL) is non-zero

I think, I need to check on before the thread creation to found why the thread can't create another thread.

[cfriedt]

It's close - a suggestion to maybe simplify things

[cfriedt]

@moonlight83340 - can you apply this patch on top of your current commit? (please also fixup the code and test commits seperately)
https://pastebin.com/a7x4GXmg

[cfriedt]

@moonlight83340 - fixup commits are somewhat special. Check this out. Also, the Signed-off-by line is automatically added to commit messages with git commit --signoff or git commit -s.

See also Contribution Workflow.

[moonlight83340]

@moonlight83340 - fixup commits are somewhat special. Check this out. Also, the Signed-off-by line is automatically added to commit messages with git commit --signoff or git commit -s.

See also Contribution Workflow.

I had never used fixup in git command, so thank you !

[moonlight83340]

@moonlight83340 - fixup commits are somewhat special. Check this out. Also, the Signed-off-by line is automatically added to commit messages with git commit --signoff or git commit -s.

See also Contribution Workflow.

Should I rebase ? git rebase --autosquash --interactive base

[cfriedt]

@moonlight83340 - you should also rebase on main if you haven't already. This PR is a bit behind.

git checkout main
git fetch -p && git pull
git checkout getinheritsched_setinheritsched
git rebase main

For reference, my rebased version of this PR shows over 500 commits and nearly 2000 files changed
moonlight83340/zephyr@getinheritsched_setinheritsched...cfriedt:zephyr:blah

[cfriedt]

@moonlight83340 - this would be a more accurate test, I think.

https://pastebin.com/1mm733Lw

[cfriedt]

LGTM - thanks 👍

[moonlight83340]

LGTM - thanks 👍

Thanks to you, I will carefully check your patch.
Hope to get the chance to continue my contribution !