sys/pal/unix/sync/mutex: Fix Mutex::new() on NuttX

[no1wudi]

PTHREAD_MUTEX_INITIALIZER is highly configurable on NuttX, making it difficult to use a hardcoded initializer from libc crate in Rust way.

Instead, call pthread_mutex_init() to initialize the mutex with default attributes, ensuring consistent behavior with PTHREAD_MUTEX_INITIALIZER.

• Added conditional compilation for NuttX target
• Replaced PTHREAD_MUTEX_INITIALIZER with pthread_mutex_init() for NuttX
• Ensured consistent mutex initialization behavior across platforms

[rustbot]

r? @workingjubilee

rustbot has assigned @workingjubilee.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[bjorn3]

NuttX allows moving a mutex after calling pthread_mutex_init? POSIX makes that UB as the mutex could eg be self-referential.

[no1wudi]

Yes, in this aspect, the situation is somewhat different. When initializing a mutex, NuttX only sets the flags:

int pthread_mutex_init(FAR pthread_mutex_t *mutex, FAR const pthread_mutexattr_t *attr) { int status; sinfo("mutex=%p attr=%p\n", mutex, attr); if (!mutex) { return EINVAL; } /* Initialize the mutex like a semaphore with initial count = 1 */ status = mutex_init(&mutex->mutex); if (status < 0) { return -status; } /* Were attributes specified? If so, use them */ #ifdef CONFIG_PTHREAD_MUTEX_TYPES mutex->type = attr ? attr->type : PTHREAD_MUTEX_DEFAULT; #endif #ifndef CONFIG_PTHREAD_MUTEX_UNSAFE mutex->flink = NULL; # ifdef CONFIG_PTHREAD_MUTEX_BOTH mutex->flags = attr && attr->robust == PTHREAD_MUTEX_ROBUST ? _PTHREAD_MFLAGS_ROBUST : 0; # else mutex->flags = 0; # endif #endif #if defined(CONFIG_PRIORITY_INHERITANCE) || defined(CONFIG_PRIORITY_PROTECT) if (attr) { status = mutex_set_protocol(&mutex->mutex, attr->proto); if (status < 0) { mutex_destroy(&mutex->mutex); return -status; } # ifdef CONFIG_PRIORITY_PROTECT if (attr->proto == PTHREAD_PRIO_PROTECT) { status = mutex_setprioceiling(&mutex->mutex, attr->ceiling, NULL); if (status < 0) { mutex_destroy(&mutex->mutex); return -status; } } # endif } #endif return 0; }

[bjorn3]

Is this a documented guarantee or does NuttX reserve the right to change the implementation in such a way that this becomes UB?

[no1wudi]

Unfortunately I can't find such a document, I think I can add it to the test cases on the relevant side and incorporate it into the NuttX side to remind developers and prevent accidental changes to this behavior.

[workingjubilee]

I indeed do not think we can realistically accept this PR until NuttX documents it's actually okay.