Synchronization Primitives

ove/sync.h provides five portable synchronization primitives: a non-recursive mutex, a recursive mutex, a counting semaphore, a binary event, and a condition variable. Each maps to the equivalent native object on every supported backend (FreeRTOS, Zephyr, NuttX, POSIX) with no virtual dispatch and minimal runtime overhead over the native API.

All primitives require CONFIG_OVE_SYNC. When it is not set, every function is replaced by a static inline stub that returns OVE_ERR_NOT_SUPPORTED.

Primitive Comparison

graph TD
    Q{"What do you need?"}

    Q -->|"Exclusive access<br/>to a shared resource"| M["Mutex<br/><small>ove_mutex_lock / ove_mutex_unlock</small>"]
    Q -->|"Same thread re-enters<br/>a critical section"| RM["Recursive Mutex<br/><small>ove_recursive_mutex_lock</small>"]
    Q -->|"Count available slots<br/>or signal across threads"| SEM["Counting Semaphore<br/><small>ove_sem_take / ove_sem_give</small>"]
    Q -->|"Simple one-shot<br/>notification, incl. from ISR"| EVT["Binary Event<br/><small>ove_event_wait / ove_event_signal</small>"]
    Q -->|"Wait for a condition<br/>associated with shared state"| CV["Condition Variable<br/><small>ove_condvar_wait / ove_condvar_signal</small>"]

Primitive	Use when	ISR-safe signal	Recursive
Mutex	Protecting a shared resource from concurrent access	No	No
Recursive mutex	Re-entrant critical sections; same thread locks multiple times	No	Yes
Semaphore	Counting available resources; producer/consumer slot tracking	No (use a binary event for ISR signalling)	N/A
Binary event	Simple one-shot wakeup; hardware interrupt notifies a task	Yes (`ove_event_signal_from_isr`)	N/A
Condition variable	Waiting for an arbitrary predicate guarded by a mutex	No	N/A

ISR safety

Only the binary event has an ISR-safe producer. ove_sem_give() and the mutex / condvar operations are not callable from an interrupt context, even though FreeRTOS, Zephyr, and NuttX each have native ISR variants for some of these — the portable layer doesn't expose them because the semantics differ across backends.

For ISR → thread handoff:

One-bit wakeup → ove_event_signal_from_isr() (this header).
Bitmask wakeup → ove_eventgroup_set_bits_from_isr() (in ove/eventgroup.h).
Queue a message → ove_queue_send_from_isr() (in ove/queue.h).

Mutex

A non-recursive mutex serializes access to a shared resource. A thread that already holds the mutex must not call ove_mutex_lock() again — this causes a deadlock.

sequenceDiagram
    participant A as Thread A
    participant M as Mutex
    participant B as Thread B

    A->>M: mutex_lock
    activate M
    Note over A,M: Thread A holds the mutex

    B->>M: mutex_lock
    Note over B,M: Thread B blocks, mutex is taken

    A->>M: mutex_unlock
    deactivate M
    Note over A,M: Mutex released

    M-->>B: Thread B unblocked
    activate M
    Note over B,M: Thread B now holds the mutex

    B->>M: mutex_unlock
    deactivate M

static ove_mutex_t shared_mutex;

void init(void)
{
    ove_mutex_create(&shared_mutex);
}

void update_shared_state(void)
{
    if (ove_mutex_lock(shared_mutex, OVE_MS(100)) != OVE_OK)
        return;  /* timed out after 100 ms */

    /* --- critical section --- */
    shared_value++;
    /* ------------------------ */

    ove_mutex_unlock(shared_mutex);
}

Recursive Mutex

A recursive mutex allows the same thread to lock it multiple times without deadlocking. Each ove_recursive_mutex_lock() must be balanced by a matching ove_recursive_mutex_unlock(). The mutex is only fully released — and made available to other threads — when the lock count returns to zero.

sequenceDiagram
    participant T as Thread (recursive caller)
    participant RM as Recursive Mutex

    T->>RM: lock(mtx) count 0->1
    activate RM
    T->>RM: lock(mtx) count 1->2
    Note over T,RM: Same thread re-enters, no deadlock

    T->>RM: unlock(mtx) count 2->1
    T->>RM: unlock(mtx) count 1->0
    deactivate RM
    Note over T,RM: Mutex fully released

Counting Semaphore

A counting semaphore tracks a pool of max tokens. ove_sem_take() consumes one token (blocks if count is zero); ove_sem_give() returns one token. Useful for producer/consumer patterns and for limiting concurrent access to a resource pool.

sequenceDiagram
    participant P as Producer Thread
    participant S as Semaphore (max=4)
    participant C as Consumer Thread

    Note over S: initial count = 0

    P->>S: sem_give() count 0->1
    P->>S: sem_give() count 1->2

    C->>S: sem_take() count 2->1
    Note over C: consumes item

    C->>S: sem_take() count 1->0
    Note over C: consumes item

    C->>S: sem_take()
    Note over C,S: Consumer blocks, count is 0

    P->>S: sem_give() count 0->1
    S-->>C: Consumer unblocked

#define QUEUE_SLOTS 8

static ove_sem_t slots_used;
static ove_sem_t slots_free;

void sync_init(void)
{
    ove_sem_create(&slots_used, 0,          QUEUE_SLOTS);
    ove_sem_create(&slots_free, QUEUE_SLOTS, QUEUE_SLOTS);
}

void producer(void *arg)
{
    for (;;) {
        ove_sem_take(slots_free, OVE_WAIT_FOREVER);
        enqueue_item(produce_item());
        ove_sem_give(slots_used);
    }
}

void consumer(void *arg)
{
    for (;;) {
        ove_sem_take(slots_used, OVE_WAIT_FOREVER);
        process_item(dequeue_item());
        ove_sem_give(slots_free);
    }
}

Binary Event

A binary event provides a lightweight one-bit signal. The waiting task blocks until the event is signalled; the signal is consumed on wakeup (auto-reset). ove_event_signal_from_isr() is the ISR-safe variant — it may trigger an immediate context switch to a higher-priority waiter when the interrupt exits.

sequenceDiagram
    participant ISR as Hardware ISR
    participant E as Binary Event
    participant T as Waiting Task

    T->>E: event_wait, blocks
    Note over T,E: Task blocks

    ISR->>E: event_signal_from_isr
    Note over E: Event signalled, waiter woken

    E-->>T: Task resumes
    Note over T: Event auto-reset, consumed

    T->>T: handle hardware data

static ove_event_t data_ready_evt;

void init(void)
{
    ove_event_create(&data_ready_evt);
}

/* Called from ISR context */
void DMA_IRQHandler(void)
{
    ove_event_signal_from_isr(data_ready_evt);
}

static void processing_entry(void *arg)
{
    for (;;) {
        ove_event_wait(data_ready_evt, OVE_WAIT_FOREVER);
        process_dma_buffer();
    }
}

Condition Variable

A condition variable lets a thread atomically release a mutex and sleep until an arbitrary condition becomes true. ove_condvar_wait() releases mtx on entry and re-acquires it before returning, whether the wakeup was due to ove_condvar_signal(), ove_condvar_broadcast(), or a timeout.

Always recheck the guarded condition in a while loop — spurious wakeups are possible on some backends.

sequenceDiagram
    participant W as Waiter Thread
    participant S as Signaller Thread
    participant CV as Condvar
    participant M as Mutex

    W->>M: mutex_lock
    activate M

    W->>CV: condvar_wait - releases mtx atomically
    deactivate M
    Note over W,CV: Waiter sleeps

    S->>M: mutex_lock
    activate M
    Note over S: modify shared state

    S->>CV: condvar_signal
    S->>M: mutex_unlock
    deactivate M

    CV-->>W: Waiter unblocked, mtx re-acquired
    activate M
    Note over W,M: Waiter re-checks condition

    W->>M: mutex_unlock
    deactivate M

static ove_mutex_t  state_mutex;
static ove_condvar_t state_ready_cv;
static bool          data_available = false;

void sync_init(void)
{
    ove_mutex_create(&state_mutex);
    ove_condvar_create(&state_ready_cv);
}

void producer_entry(void *arg)
{
    for (;;) {
        prepare_data();

        ove_mutex_lock(state_mutex, OVE_WAIT_FOREVER);
        data_available = true;
        ove_condvar_signal(state_ready_cv);
        ove_mutex_unlock(state_mutex);
    }
}

void consumer_entry(void *arg)
{
    ove_mutex_lock(state_mutex, OVE_WAIT_FOREVER);
    while (!data_available) {
        /* spurious-wakeup safe: always recheck in a while loop */
        ove_condvar_wait(state_ready_cv, state_mutex, OVE_WAIT_FOREVER);
    }
    data_available = false;
    consume_data();
    ove_mutex_unlock(state_mutex);
}

API Reference

_create() / _destroy() are gated behind OVE_HEAP_SYNC and unavailable in zero-heap mode. For zero-heap apps (or code that compiles in both modes), use _init() with caller-supplied storage or the OVE_*_DEFINE_STATIC() macros.

Mutex

Function	Description
`ove_mutex_init(mtx, storage)`	Initialise a non-recursive mutex from caller-supplied static storage.
`ove_mutex_deinit(mtx)`	Release resources; does not free static storage.
`ove_mutex_create(mtx)`	Allocate and initialise a mutex (heap mode only; gated by `OVE_HEAP_SYNC`).
`ove_mutex_destroy(mtx)`	Destroy and free a mutex created with `ove_mutex_create()` (heap mode only).
`ove_mutex_lock(mtx, timeout_ns)`	Acquire the mutex. Blocks up to `timeout_ns` ns; pass `OVE_WAIT_FOREVER` to block indefinitely. Returns `OVE_ERR_TIMEOUT` if the deadline expires.
`ove_mutex_unlock(mtx)`	Release the mutex. Must be called by the thread that acquired it.

Recursive Mutex

Function	Description
`ove_recursive_mutex_init(mtx, storage)`	Initialise a recursive mutex from static storage.
`ove_recursive_mutex_create(mtx)`	Allocate and initialise a recursive mutex (heap mode only; gated by `OVE_HEAP_SYNC`).
`ove_recursive_mutex_destroy(mtx)`	Destroy and free a recursive mutex (heap mode only).
`ove_recursive_mutex_lock(mtx, timeout_ns)`	Acquire one level of the recursive mutex. May be called multiple times by the same thread.
`ove_recursive_mutex_unlock(mtx)`	Release one level. The mutex is fully released when the count reaches zero.

Semaphore

Function	Description
`ove_sem_init(sem, storage, initial, max)`	Initialise a counting semaphore from static storage with the given initial and maximum count.
`ove_sem_deinit(sem)`	Release resources; does not free static storage.
`ove_sem_create(sem, initial, max)`	Allocate and initialise a semaphore (heap mode only; gated by `OVE_HEAP_SYNC`).
`ove_sem_destroy(sem)`	Destroy and free a semaphore created with `ove_sem_create()` (heap mode only).
`ove_sem_take(sem, timeout_ns)`	Decrement the count. Blocks up to `timeout_ns` ns if count is zero. Returns `OVE_ERR_TIMEOUT` on expiry.
`ove_sem_give(sem)`	Increment the count, potentially unblocking a waiting thread. Safe to call from normal thread context.

Binary Event

Function	Description
`ove_event_init(evt, storage)`	Initialise a binary event from static storage. Starts in the unsignalled state.
`ove_event_deinit(evt)`	Release resources; does not free static storage.
`ove_event_create(evt)`	Allocate and initialise an event (heap mode only; gated by `OVE_HEAP_SYNC`).
`ove_event_destroy(evt)`	Destroy and free an event created with `ove_event_create()` (heap mode only).
`ove_event_wait(evt, timeout_ns)`	Block until the event is signalled or timeout expires. The event is auto-reset (consumed) on success.
`ove_event_signal(evt)`	Signal the event from thread context, unblocking one waiter.
`ove_event_signal_from_isr(evt)`	ISR-safe signal; may trigger a context switch after the interrupt exits.

Condition Variable

Function	Description
`ove_condvar_init(cv, storage)`	Initialise a condition variable from static storage.
`ove_condvar_deinit(cv)`	Release resources; does not free static storage.
`ove_condvar_create(cv)`	Allocate and initialise a condition variable (heap mode only; gated by `OVE_HEAP_SYNC`).
`ove_condvar_destroy(cv)`	Destroy and free a condition variable created with `ove_condvar_create()` (heap mode only).
`ove_condvar_wait(cv, mtx, timeout_ns)`	Atomically release `mtx` and sleep. Re-acquires `mtx` before returning.
`ove_condvar_signal(cv)`	Wake one waiting thread. Signal is lost if no thread is waiting.
`ove_condvar_broadcast(cv)`	Wake all threads waiting on the condition variable.

Allocation Strategies

The same two strategies available for threads apply here:

graph TD
    subgraph heap["Heap mode (`OVE_HEAP_SYNC` set)"]
        direction LR
        C["ove_mutex_create"] --> CH["allocates from RTOS heap"]
        D["ove_mutex_destroy"] --> CH
    end

    subgraph static["Explicit Static Storage"]
        direction LR
        I["ove_mutex_init"] --> IS["caller owns storage<br/>works in loops / arrays / structs"]
    end

Use _create() / _destroy() for the simplest code — it works identically in both heap and zero-heap mode. Use _init() / _deinit() when you need to manage an array of primitives or embed them inside a struct.

Deadlock Avoidance

graph LR
    subgraph Bad["Circular wait - DEADLOCK"]
        direction LR
        A1["Thread A<br/>holds Mutex 1<br/>waits for Mutex 2"]
        B1["Thread B<br/>holds Mutex 2<br/>waits for Mutex 1"]
        A1 -- "blocked by" --> B1
        B1 -- "blocked by" --> A1
    end

    subgraph Good["Consistent ordering - safe"]
        direction LR
        A2["Thread A<br/>lock Mutex 1<br/>then Mutex 2"]
        B2["Thread B<br/>lock Mutex 1<br/>then Mutex 2"]
        A2 -- "same order" --> B2
    end

Rules to follow:

Use a timeout. Never pass OVE_WAIT_FOREVER when acquiring more than one mutex simultaneously. A bounded timeout_ns surfaces deadlocks as OVE_ERR_TIMEOUT instead of hangs.
Consistent lock ordering. When multiple mutexes are required, always acquire them in the same global order across all threads. This eliminates circular waits.
Prefer a single mutex. If possible, protect all shared state with one mutex rather than composing several fine-grained locks.
Do not call ove_mutex_lock() recursively. The non-recursive mutex will deadlock if the holding thread calls lock again. Use ove_recursive_mutex_lock() when re-entrancy is needed.
Keep critical sections short. Release the mutex before calling any blocking API.

Kconfig Options

Option	Default	Description
`CONFIG_OVE_SYNC`	`y`	Enable the synchronization subsystem. When disabled, all functions are replaced with stubs returning `OVE_ERR_NOT_SUPPORTED`.

Header	Contents
`ove/sync.h`	Mutex, recursive mutex, semaphore, binary event, and condition variable — init/deinit, create/destroy (heap mode), and operation variants (lock/unlock, take/give, wait/signal, broadcast, signal-from-ISR). Static-init macros are declared in `ove/storage.h`.

Synchronization Primitives

Primitive Comparison

ISR safety

Mutex

Recursive Mutex

Counting Semaphore

Binary Event

Condition Variable

API Reference

Mutex

Recursive Mutex

Semaphore

Binary Event

Condition Variable

Allocation Strategies

Deadlock Avoidance

Kconfig Options

Header