Semaphore functions
[RTAI schedulers modules]


Files

file  sem.c
 Semaphore functions.


Functions

void rt_typed_sem_init (SEM *sem, int value, int type)
 Initialize a specifically typed (counting, binary, resource) semaphore.

void rt_sem_init (SEM *sem, int value)
 Initialize a counting semaphore.

int rt_sem_delete (SEM *sem)
 Delete a semaphore.

int rt_sem_signal (SEM *sem)
 Signaling a semaphore.

int rt_sem_broadcast (SEM *sem)
 Signaling a semaphore.

int rt_sem_wait (SEM *sem)
 Take a semaphore.

int rt_sem_wait_if (SEM *sem)
 Take a semaphore, only if the calling task is not blocked.

int rt_sem_wait_until (SEM *sem, RTIME time)
 Wait a semaphore with timeout.

int rt_sem_wait_timed (SEM *sem, RTIME delay)
 Wait a semaphore with timeout.

int rt_sem_wait_barrier (SEM *sem)
 Wait on a semaphore barrier.

int rt_cond_signal (CND *cnd)
 Wait for a signal to a conditional variable.

int rt_cond_wait (CND *cnd, SEM *mutex)
 Wait for a signal to a conditional variable.

int rt_cond_wait_until (CND *cnd, SEM *mutex, RTIME time)
 Wait a semaphore with timeout.

int rt_cond_wait_timed (CND *cnd, SEM *mutex, RTIME delay)
 Wait a semaphore with timeout.

int rt_rwl_init (RWL *rwl)
 Initialize a multi readers single writer lock.

int rt_rwl_delete (RWL *rwl)
 destroys a multi readers single writer lock.

int rt_rwl_rdlock (RWL *rwl)
 acquires a multi readers single writer lock for reading.

int rt_rwl_rdlock_if (RWL *rwl)
 try to acquire a multi readers single writer lock just for reading.

int rt_rwl_rdlock_until (RWL *rwl, RTIME time)
 try to acquire a multi readers single writer lock for reading within an absolute deadline time.

int rt_rwl_rdlock_timed (RWL *rwl, RTIME delay)
 try to acquire a multi readers single writer lock for reading within a relative deadline time.

int rt_rwl_wrlock (RWL *rwl)
 acquires a multi readers single writer lock for wrtiting.

int rt_rwl_wrlock_if (RWL *rwl)
 acquires a multi readers single writer lock for writing.

int rt_rwl_wrlock_until (RWL *rwl, RTIME time)
 try to acquire a multi readers single writer lock for writing within an absolute deadline time.

int rt_rwl_wrlock_timed (RWL *rwl, RTIME delay)
 try to acquire a multi readers single writer lock for writing within a relative deadline time.

int rt_rwl_unlock (RWL *rwl)
 unlock an acquired multi readers single writer lock.

int rt_spl_init (SPL *spl)
 Initialize a spinlock.

int rt_spl_delete (SPL *spl)
 Initialize a spinlock.

int rt_spl_lock (SPL *spl)
 Acquire a spinlock.

int rt_spl_lock_if (SPL *spl)
 Acquire a spinlock without waiting.

int rt_spl_lock_timed (SPL *spl, unsigned long ns)
 Acquire a spinlock with timeout.

int rt_spl_unlock (SPL *spl)
 Release an owned spinlock.

SEM * _rt_typed_named_sem_init (unsigned long sem_name, int value, int type)
 Initialize a specifically typed (counting, binary, resource) semaphore identified by a name.

int rt_named_sem_delete (SEM *sem)
 Delete a semaphore initialized in named mode.

RWL * _rt_named_rwl_init (unsigned long rwl_name)
 Initialize a multi readers single writer lock identified by a name.

int rt_named_rwl_delete (RWL *rwl)
 Delete a multi readers single writer lock in named mode.

SPL * _rt_named_spl_init (unsigned long spl_name)
 Initialize a spinlock identified by a name.

int rt_named_spl_delete (SPL *spl)
 Delete a spinlock in named mode.


Function Documentation

RWL* _rt_named_rwl_init unsigned long  rwl_name  ) 
 

Initialize a multi readers single writer lock identified by a name.

_rt_named_rwl_init allocate and initializes a multi readers single writer lock (RWL) identified by name. Once the lock structure is allocated the initialization is as for rt_rwl_init. The function returns the handle pointing to the allocated multi readers single writer lock o structure, to be used as the usual lock address in all rwl based services. Named objects are useful for use among different processes, kernel/user space and in distributed applications, see netrpc.

Parameters:
rwl_name is the identifier associated with the returned object.
Since name can be a clumsy identifier, services are provided to convert 6 characters identifiers to unsigned long, and vice versa.

See also:
nam2num() and num2nam().
As for all the named initialization functions it must be remarked that only the very first call to initilize/create a named RTAI object does a real allocation of the object, any following call with the same name will just increase its usage count. In any case the function returns a pointer to the named object, or zero if in error.

Returns:
either a valid pointer or 0 if in error.

SPL* _rt_named_spl_init unsigned long  spl_name  ) 
 

Initialize a spinlock identified by a name.

_rt_named_spl_init allocate and initializes a spinlock (SPL) identified by name. Once the spinlock structure is allocated the initialization is as for rt_spl_init. The function returns the handle pointing to the allocated spinlock structure, to be used as the usual spinlock address in all spinlock based services. Named objects are useful for use among different processes and kernel/user space.

Parameters:
spl_name is the identifier associated with the returned object.
Since name can be a clumsy identifier, services are provided to convert 6 characters identifiers to unsigned long, and vice versa.

See also:
nam2num() and num2nam().
As for all the named initialization functions it must be remarked that only the very first call to initilize/create a named RTAI object does a real allocation of the object, any following call with the same name will just increase its usage count. In any case the function returns a pointer to the named object, or zero if in error.

Returns:
either a valid pointer or 0 if in error.

SEM* _rt_typed_named_sem_init unsigned long  sem_name,
int  value,
int  type
 

Initialize a specifically typed (counting, binary, resource) semaphore identified by a name.

_rt_typed_named_sem_init allocate and initializes a semaphore identified by name of type type. Once the semaphore structure is allocated the initialization is as for rt_typed_sem_init. The function returns the handle pointing to the allocated semaphore structure, to be used as the usual semaphore address in all semaphore based services. Named objects are useful for use among different processes, kernel/user space and in distributed applications, see netrpc.

Parameters:
sem_name is the identifier associated with the returned object.
value is the initial value of the semaphore, always set to 1 for a resource semaphore.
type is the semaphore type and queuing policy. It can be an OR a semaphore kind: CNT_SEM for counting semaphores, BIN_SEM for binary semaphores, RES_SEM for resource semaphores; and queuing policy: FIFO_Q, PRIO_Q for a fifo and priority queueing respectively. Resource semaphores will enforce a PRIO_Q policy anyhow.
Since name can be a clumsy identifier, services are provided to convert 6 characters identifiers to unsigned long, and vice versa.

See also:
nam2num() and num2nam().
See rt_typed_sem_init for further clues.

As for all the named initialization functions it must be remarked that only the very first call to initilize/create a named RTAI object does a real allocation of the object, any following call with the same name will just increase its usage count. In any case the function returns a pointer to the named object, or zero if in error.

Returns:
either a valid pointer or 0 if in error.

int rt_cond_signal CND *  cnd  )  [inline]
 

Wait for a signal to a conditional variable.

rt_cond_signal resumes one of the tasks that are waiting on the condition semaphore cnd. Nothing happens if no task is waiting on cnd, while it resumed the first queued task blocked on cnd, according to the queueing method set at rt_cond_init.

Parameters:
cnd points to the structure used in the call to rt_cond_init().
Returns:
0

int rt_cond_wait CND *  cnd,
SEM *  mtx
[inline]
 

Wait for a signal to a conditional variable.

rt_cond_wait atomically unlocks mtx (as for using rt_sem_signal) and waits for the condition semaphore cnd to be signaled. The task execution is suspended until the condition semaphore is signalled. Mtx must be obtained by the calling task, before calling rt_cond_wait is called. Before returning to the calling task rt_cond_wait reacquires mtx by calling rt_sem_wait.

Parameters:
cnd points to the structure used in the call to rt_cond_init().
mtx points to the structure used in the call to rt_sem_init().
Returns:
0 on succes, SEM_ERR in case of error.

int rt_cond_wait_timed CND *  cnd,
SEM *  mtx,
RTIME  delay
[inline]
 

Wait a semaphore with timeout.

rt_cond_wait_timed atomically unlocks mtx (as for using rt_sem_signal) and waits for the condition semaphore cnd to be signalled. The task execution is suspended until the condition semaphore is either signaled or a timeout expires. Mtx must be obtained by the calling task, before calling rt_cond_wait is called. Before returning to the calling task rt_cond_wait_until reacquires mtx by calling rt_sem_wait and returns a value to indicate if it has been signalled pr timedout.

Parameters:
cnd points to the structure used in the call to rt_cond_init().
mtx points to the structure used in the call to rt_sem_init().
delay is a realtive time values with respect to the current time, in timer count unit.
Returns:
0 if it was signaled, SEM_TIMOUT if a timeout occured, SEM_ERR if the task has been resumed because of any other action (likely cnd was deleted).

int rt_cond_wait_until CND *  cnd,
SEM *  mtx,
RTIME  time
[inline]
 

Wait a semaphore with timeout.

rt_cond_wait_until atomically unlocks mtx (as for using rt_sem_signal) and waits for the condition semaphore cnd to be signalled. The task execution is suspended until the condition semaphore is either signaled or a timeout expires. Mtx must be obtained by the calling task, before calling rt_cond_wait is called. Before returning to the calling task rt_cond_wait_until reacquires mtx by calling rt_sem_wait and returns a value to indicate if it has been signalled pr timedout.

Parameters:
cnd points to the structure used in the call to rt_cond_init().
mtx points to the structure used in the call to rt_sem_init().
time is an absolute value to the current time, in timer count unit.
Returns:
0 if it was signaled, SEM_TIMOUT if a timeout occured, SEM_ERR if the task has been resumed because of any other action (likely cnd was deleted).

int rt_named_rwl_delete RWL *  rwl  ) 
 

Delete a multi readers single writer lock in named mode.

rt_named_rwl_delete deletes a multi readers single writer lock previously created with _rt_named_rwl_init().

Parameters:
rwl points to the structure pointer returned by a corresponding call to rt_named_rwl_init.
As it is done by all the named allocation functions delete calls have just the effect of decrementing a usage count till the last is done, as that is the one the really frees the object.

Returns:
an int >=0 is returned upon success, SEM_ERR if it failed to delete the multi readers single writer lock, -EFAULT if the lock does not exist anymore.

int rt_named_sem_delete SEM *  sem  )  [inline]
 

Delete a semaphore initialized in named mode.

rt_named_sem_delete deletes a semaphore previously created with _rt_typed_named_sem_init().

Parameters:
sem points to the structure pointer returned by a corresponding call to _rt_typed_named_sem_init.
Any tasks blocked on this semaphore is returned in error and allowed to run when semaphore is destroyed. As it is done by all the named allocation functions delete calls have just the effect of decrementing a usage count till the last is done, as that is the one the really frees the object.

Returns:
an int >=0 is returned upon success, SEM_ERR if it failed to delete the semafore, -EFAULT if the semaphore does not exist anymore.

int rt_named_spl_delete SPL *  spl  ) 
 

Delete a spinlock in named mode.

rt_named_spl_delete deletes a spinlock previously created with _rt_named_spl_init().

Parameters:
spl points to the structure pointer returned by a corresponding call to rt_named_spl_init.
As it is done by all the named allocation functions delete calls have just the effect of decrementing a usage count till the last is done, as that is the one the really frees the object.

Returns:
an int >=0 is returned upon success, -EFAULT if the spinlock does not exist anymore.

int rt_rwl_delete RWL *  rwl  ) 
 

destroys a multi readers single writer lock.

rt_rwl_init destroys a multi readers single writer lock rwl.

Parameters:
rwl must point to an allocated RWL structure.
Returns:
0 if OK, SEM_ERR if anything went wrong.

int rt_rwl_init RWL *  rwl  ) 
 

Initialize a multi readers single writer lock.

rt_rwl_init initializes a multi readers single writer lock rwl.

Parameters:
rwl must point to an allocated RWL structure.
A multi readers single writer lock (RWL) is a synchronization mechanism that allows to have simultaneous read only access to an object, while only one task can have write access. A data set which is searched more frequently than it is changed can be usefully controlled by using an rwl. The lock acquisition policy is determined solely on the priority of tasks applying to own a lock.

Returns:
0 if always.

int rt_rwl_rdlock RWL *  rwl  ) 
 

acquires a multi readers single writer lock for reading.

rt_rwl_rdlock acquires a multi readers single writer lock rwl for reading. The calling task will block only if any writer owns the lock already or there are writers with higher priority waiting to acquire write access.

Parameters:
rwl must point to an allocated RWL structure.
Returns:
0 if OK, SEM_ERR if anything went wrong after being blocked.

int rt_rwl_rdlock_if RWL *  rwl  ) 
 

try to acquire a multi readers single writer lock just for reading.

rt_rwl_rdlock_if tries to acquire a multi readers single writer lock rwl for reading immediately, i.e. without blocking if a writer owns the lock or there are writers with higher priority waiting to acquire write access.

Parameters:
rwl must point to an allocated RWL structure.
Returns:
0 if the lock was acquired, -1 if the lock was already owned.

int rt_rwl_rdlock_timed RWL *  rwl,
RTIME  delay
 

try to acquire a multi readers single writer lock for reading within a relative deadline time.

rt_rwl_rdlock_timed tries to acquire a multi readers single writer lock rwl for reading, as for rt_rwl_rdlock, but timing out if the lock has not been acquired within an assigned deadline.

Parameters:
rwl must point to an allocated RWL structure.
delay is the time delay within which the lock must be acquired, in internal count units.
Returns:
0 if the lock was acquired, SEM_TIMOUT if the deadline expired without acquiring the lock, SEM_ERR in case something went wrong.

int rt_rwl_rdlock_until RWL *  rwl,
RTIME  time
 

try to acquire a multi readers single writer lock for reading within an absolute deadline time.

rt_rwl_rdlock_untill tries to acquire a multi readers single writer lock rwl for reading, as for rt_rwl_rdlock, but timing out if the lock has not been acquired within an assigned deadline.

Parameters:
rwl must point to an allocated RWL structure.
time is the time deadline, in internal count units.
Returns:
0 if the lock was acquired, SEM_TIMOUT if the deadline expired without acquiring the lock, SEM_ERR in case something went wrong.

int rt_rwl_unlock RWL *  rwl  ) 
 

unlock an acquired multi readers single writer lock.

rt_rwl_unlock unlocks an acquired multi readers single writer lock rwl. After releasing the lock any task waiting to acquire it will own the lock according to its priority, whether it is a reader or a writer, otherwise the lock will be fully unlocked.

Parameters:
rwl must point to an allocated RWL structure.
Returns:
0 always.

int rt_rwl_wrlock RWL *  rwl  ) 
 

acquires a multi readers single writer lock for wrtiting.

rt_rwl_rwlock acquires a multi readers single writer lock rwl for writing. The calling task will block if any other task, reader or writer, owns the lock already.

Parameters:
rwl must point to an allocated RWL structure.
Returns:
0 if OK, SEM_ERR if anything went wrong after being blocked.

int rt_rwl_wrlock_if RWL *  rwl  ) 
 

acquires a multi readers single writer lock for writing.

rt_rwl_wrlock_if try to acquire a multi readers single writer lock rwl for writing immediately, i.e without blocking if the lock is owned already.

Parameters:
rwl must point to an allocated RWL structure.
Returns:
0 if the lock was acquired, -1 if the lock was already owned.

int rt_rwl_wrlock_timed RWL *  rwl,
RTIME  delay
 

try to acquire a multi readers single writer lock for writing within a relative deadline time.

rt_rwl_wrlock_timed tries to acquire a multi readers single writer lock rwl for writing, as for rt_rwl_wrlock, timing out if the lock has not been acquired within an assigned deadline.

Parameters:
rwl must point to an allocated RWL structure.
delay is the time delay within which the lock must be acquired, in internal count units.
Returns:
0 if the lock was acquired, SEM_TIMOUT if the deadline expired without acquiring the lock, SEM_ERR in case something went wrong.

int rt_rwl_wrlock_until RWL *  rwl,
RTIME  time
 

try to acquire a multi readers single writer lock for writing within an absolute deadline time.

rt_rwl_rwlock_until tries to acquire a multi readers single writer lock rwl for writing, as for rt_rwl_rwlock, but timing out if the lock has not been acquired within an assigned deadline.

Parameters:
rwl must point to an allocated RWL structure.
time is the time deadline, in internal count units.
Returns:
0 if the lock was acquired, SEM_TIMOUT if the deadline expired without acquiring the lock, SEM_ERR in case something went wrong.

int rt_sem_broadcast SEM *  sem  )  [inline]
 

Signaling a semaphore.

rt_sem_broadcast signals an event to a semaphore that unblocks all tasks waiting on it. It is used as a support for RTAI proper conditional variables but can be of help in many other instances. After the broadcast the semaphore counts is set to zero, thus all tasks waiting on it will blocked.

Parameters:
sem points to the structure used in the call to rt_sem_init().
Returns:
0 always.

int rt_sem_delete SEM *  sem  )  [inline]
 

Delete a semaphore.

rt_sem_delete deletes a semaphore previously created with rt_sem_init().

Parameters:
sem points to the structure used in the corresponding call to rt_sem_init.
Any tasks blocked on this semaphore is returned in error and allowed to run when semaphore is destroyed.

Returns:
0 is returned upon success. A negative value is returned on failure as described below:
  • 0xFFFF: sem does not refer to a valid semaphore.
Note:
In principle 0xFFFF could theoretically be a usable semaphores events count, so it could be returned also under normal circumstances. It is unlikely you are going to count up to such number of events, in any case avoid counting up to 0xFFFF.

void rt_sem_init SEM *  sem,
int  value
 

Initialize a counting semaphore.

rt_sem_init initializes a counting fifo queueing semaphore sem.

A semaphore can be used for communication and synchronization among real time tasks.

Parameters:
sem must point to an allocated SEM structure.
value is the initial value of the semaphore.
Positive values of the semaphore variable show how many tasks can do a rt_sem_wait() call without blocking. Negative value of a semaphore shows how many tasks are blocked on the semaphore queue, waiting to be awaken by calls to rt_sem_signal().

Note:
RTAI counting semaphores assume that their counter will never exceed 0xFFFF, such a number being used to signal returns in error. Thus also the initial count value cannot be greater than 0xFFFF. This is an old legacy function. RTAI 24.1.xx has also rt_typed_sem_init(), allowing to choose among counting, binary and resource semaphores. Resource semaphores have priority inherithance.

int rt_sem_signal SEM *  sem  )  [inline]
 

Signaling a semaphore.

rt_sem_signal signals an event to a semaphore. It is typically called when the task leaves a critical region. The semaphore value is incremented and tested. If the value is not positive, the first task in semaphore's waiting queue is allowed to run. rt_sem_signal never blocks the caller task.

Parameters:
sem points to the structure used in the call to rt_sem_init().
Returns:
0 is returned upon success. A negative value is returned on failure as described below:
  • 0xFFFF: sem does not refer to a valid semaphore.
Note:
In principle 0xFFFF could theoretically be a usable semaphores events count, so it could be returned also under normal circumstances. It is unlikely you are going to count up to such number of events, in any case avoid counting up to 0xFFFF. See rt_sem_wait() notes for some curiosities.

int rt_sem_wait SEM *  sem  )  [inline]
 

Take a semaphore.

rt_sem_wait waits for a event to be signaled to a semaphore. It is typically called when a task enters a critical region. The semaphore value is decremented and tested. If it is still non-negative rt_sem_wait returns immediately. Otherwise the caller task is blocked and queued up. Queuing may happen in priority order or on FIFO base. This is determined by the compile time option SEM_PRIORD. In this case rt_sem_wait returns if:

  • The caller task is in the first place of the waiting queue and another task issues a rt_sem_signal() call;
  • An error occurs (e.g. the semaphore is destroyed);

Parameters:
sem points to the structure used in the call to rt_sem_init().
Returns:
the number of events already signaled upon success. A special value" as described below in case of a failure :
  • 0xFFFF: sem does not refer to a valid semaphore.
Note:
In principle 0xFFFF could theoretically be a usable semaphores events count, so it could be returned also under normal circumstances. It is unlikely you are going to count up to such number of events, in any case avoid counting up to 0xFFFF.
Just for curiosity: the original Dijkstra notation for rt_sem_wait was a "P" operation, and rt_sem_signal was a "V" operation. The name for P comes from the Dutch "prolagen", a combination of "proberen" (to probe) and "verlagen" (to decrement). Also from the word "passeren" (to pass).
The name for V comes from the Dutch "verhogen" (to increase) or "vrygeven" (to release). (Source: Daniel Tabak - Multiprocessors, Prentice Hall, 1990).
It should be also remarked that real time programming practitioners were using semaphores a long time before Dijkstra formalized P and V. "In Italian semaforo" means a traffic light, so that semaphores have an intuitive appeal and their use and meaning is easily understood.

int rt_sem_wait_barrier SEM *  sem  )  [inline]
 

Wait on a semaphore barrier.

rt_sem_wait_barrier is a gang waiting in that a task issuing such a request will be blocked till a number of tasks equal to the semaphore count set at rt_sem_init is reached.

Returns:
0 always.

int rt_sem_wait_if SEM *  sem  )  [inline]
 

Take a semaphore, only if the calling task is not blocked.

rt_sem_wait_if is a version of the semaphore wait operation is similar to rt_sem_wait() but it is never blocks the caller. If the semaphore is not free, rt_sem_wait_if returns immediately and the semaphore value remains unchanged.

Parameters:
sem points to the structure used in the call to rt_sem_init().
Returns:
the number of events already signaled upon success. A special value as described below in case of a failure:
  • 0xFFFF: sem does not refer to a valid semaphore.
Note:
In principle 0xFFFF could theoretically be a usable semaphores events count so it could be returned also under normal circumstances. It is unlikely you are going to count up to such number of events, in any case avoid counting up to 0xFFFF.

int rt_sem_wait_timed SEM *  sem,
RTIME  delay
[inline]
 

Wait a semaphore with timeout.

rt_sem_wait_timed, like rt_sem_wait_until(), is a timed version of the standard semaphore wait call. The semaphore value is decremented and tested. If it is still non-negative these functions return immediately. Otherwise the caller task is blocked and queued up. Queuing may happen in priority order or on FIFO base. This is determined by the compile time option SEM_PRIORD. In this case the function returns if:

  • The caller task is in the first place of the waiting queue and an other task issues a rt_sem_signal() call;
  • a timeout occurs;
  • an error occurs (e.g. the semaphore is destroyed);

In case of a timeout, the semaphore value is incremented before return.

Parameters:
sem points to the structure used in the call to rt_sem_init().
delay is an absolute value to the current time.
Returns:
the number of events already signaled upon success. A special value as described below in case of a failure:
  • 0xFFFF: sem does not refer to a valid semaphore.
Note:
In principle 0xFFFF could theoretically be a usable semaphores events count so it could be returned also under normal circumstances. It is unlikely you are going to count up to such number of events, in any case avoid counting up to 0xFFFF.

int rt_sem_wait_until SEM *  sem,
RTIME  time
[inline]
 

Wait a semaphore with timeout.

rt_sem_wait_until, like rt_sem_wait_timed() is a timed version of the standard semaphore wait call. The semaphore value is decremented and tested. If it is still non-negative these functions return immediately. Otherwise the caller task is blocked and queued up. Queuing may happen in priority order or on FIFO base. This is determined by the compile time option SEM_PRIORD. In this case the function returns if:

  • The caller task is in the first place of the waiting queue and an other task issues a rt_sem_signal call();
  • a timeout occurs;
  • an error occurs (e.g. the semaphore is destroyed);

In case of a timeout, the semaphore value is incremented before return.

Parameters:
sem points to the structure used in the call to rt_sem_init().
time is an absolute value to the current time.
Returns:
the number of events already signaled upon success. Aa special value" as described below in case of a failure:
  • 0xFFFF: sem does not refer to a valid semaphore.
Note:
In principle 0xFFFF could theoretically be a usable semaphores events count so it could be returned also under normal circumstances. It is unlikely you are going to count up to such number of events, in any case avoid counting up to 0xFFFF.

int rt_spl_delete SPL *  spl  ) 
 

Initialize a spinlock.

rt_spl_delete destroies a spinlock spl.

Parameters:
spl must point to an allocated SPL structure.
Returns:
0 if always.

int rt_spl_init SPL *  spl  ) 
 

Initialize a spinlock.

rt_spl_init initializes a spinlock spl.

Parameters:
spl must point to an allocated SPL structure.
A spinlock is an active wait synchronization mechanism useful for multi processors very short synchronization, when it is more efficient to wait at a meeting point instead of being suspended and the reactivated, as by using semaphores, to acquire ownership of any object. Spinlocks can be recursed once acquired, a recurring owner must care of unlocking as many times as he took the spinlock.

Returns:
0 if always.

int rt_spl_lock SPL *  spl  ) 
 

Acquire a spinlock.

rt_spl_lock acquires a spinlock spl.

Parameters:
spl must point to an allocated SPL structure.
rt_spl_lock spins on lock till it can be acquired. If a tasks asks for lock it owns already it will acquire it immediately but will have to care to unlock it as many times as it recursed the spinlock ownership.

Returns:
0 if always.

int rt_spl_lock_if SPL *  spl  ) 
 

Acquire a spinlock without waiting.

rt_spl_lock_if acquires a spinlock spl without waiting.

Parameters:
spl must point to an allocated SPL structure.
rt_spl_lock_if tries to acquire a spinlock but will not spin on it if it is owned already.

Returns:
0 if it succeeded, -1 if the lock was owned already.

int rt_spl_lock_timed SPL *  spl,
unsigned long  ns
 

Acquire a spinlock with timeout.

rt_spl_lock_timed acquires a spinlock spl, but waiting spinning only for an allowed time.

Parameters:
spl must point to an allocated SPL structure.
ns timeout
rt_spl_lock spins on lock till it can be acquired, as for rt_spl_lock, but only for an allowed time. If the spinlock cannot be acquired in time the functions returns in error. This function can be usefull either in itself or as a diagnosis toll during code development.

Returns:
0 if the spinlock was acquired, -1 if a timeout occured.

int rt_spl_unlock SPL *  spl  ) 
 

Release an owned spinlock.

rt_spl_lock releases an owned spinlock spl.

Parameters:
spl must point to an allocated SPL structure.
rt_spl_unlock releases an owned lock. The spinlock can remain locked and its ownership can remain with the task is the spinlock acquisition was recursed.

Returns:
0 if the function was used legally, -1 if a tasks tries to unlock a spinlock it does not own.

void rt_typed_sem_init SEM *  sem,
int  value,
int  type
 

Initialize a specifically typed (counting, binary, resource) semaphore.

rt_typed_sem_init initializes a semaphore sem of type type. A semaphore can be used for communication and synchronization among real time tasks. Negative value of a semaphore shows how many tasks are blocked on the semaphore queue, waiting to be awaken by calls to rt_sem_signal.

Parameters:
sem must point to an allocated SEM structure.
value is the initial value of the semaphore, always set to 1 for a resource semaphore.
type is the semaphore type and queuing policy. It can be an OR a semaphore kind: CNT_SEM for counting semaphores, BIN_SEM for binary semaphores, RES_SEM for resource semaphores; and queuing policy: FIFO_Q, PRIO_Q for a fifo and priority queueing respectively. Resource semaphores will enforce a PRIO_Q policy anyhow.
Counting semaphores can register up to 0xFFFE events. Binary semaphores do not count signalled events, their count will never exceed 1 whatever number of events is signaled to them. Resource semaphores are special binary semaphores suitable for managing resources. The task that acquires a resource semaphore becomes its owner, also called resource owner, since it is the only one capable of manipulating the resource the semaphore is protecting. The owner has its priority increased to that of any task blocking on a wait to the semaphore. Such a feature, called priority inheritance, ensures that a high priority task is never slaved to a lower priority one, thus allowing to avoid any deadlock due to priority inversion. Resource semaphores can be recursed, i.e. their task owner is not blocked by nested waits placed on an owned resource. The owner must insure that it will signal the semaphore, in reversed order, as many times as he waited on it. Note that that full priority inheritance is supported both for resource semaphores and inter task messages, for a singly owned resource. Instead it becomes an adaptive priority ceiling when a task owns multiple resources, including messages sent to him. In such a case in fact its priority is returned to its base one only when all such resources are released and no message is waiting for being received. This is a compromise design choice aimed at avoiding extensive searches for the new priority to be inherited across multiply owned resources and blocked tasks sending messages to him. Such a solution will be implemented only if it proves necessary. Note also that, to avoid deadlocks, a task owning a resource semaphore cannot be suspended. Any rt_task_suspend() posed on it is just registered. An owner task will go into suspend state only when it releases all the owned resources.

Note:
RTAI counting semaphores assume that their counter will never exceed 0xFFFF, such a number being used to signal returns in error. Thus also the initial count value cannot be greater than 0xFFFF. To be used only with RTAI24.x.xx (FIXME).


Generated on Sat Jul 24 19:36:21 2004 for RTAI API by doxygen 1.3.4