Condition Wait Semantics

It is important to note that when pthread_cond_wait () and pthread_cond_timedwait () return without error, the associated predicate may still be false. Similarly, when pthread_cond_timedwait () returns with the timeout error, the associated predicate may be true due to an unavoidable race between the expiration of the timeout and the predicate state change.

The application needs to recheck the predicate on any return because it cannot be sure there is another thread waiting on the thread to handle the signal, and if there is not then the signal is lost. The burden is on the application to check the predicate.

Some implementations, particularly on a multi-processor, may sometimes cause multiple threads to wake up when the condition variable is signaled simultaneously on different processors.

In general, whenever a condition wait returns, the thread has to re-evaluate the predicate associated with the condition wait to determine whether it can safely proceed, should wait again, or should declare a timeout. A return from the wait does not imply that the associated predicate is either true or false.

It is thus recommended that a condition wait be enclosed in the equivalent of a "while loop" that checks the predicate.

Timed Wait Semantics

An absolute time measure was chosen for specifying the timeout parameter for two reasons. First, a relative time measure can be easily implemented on top of a function that specifies absolute time, but there is a race condition associated with specifying an absolute timeout on top of a function that specifies relative timeouts. For example, assume that clock_gettime(3) returns the current time and cond_relative_timed_wait () uses relative timeouts:

clock_gettime(CLOCK_REALTIME, &now)
reltime = sleep_til_this_absolute_time -now;
cond_relative_timed_wait(c, m, &reltime);

If the thread is preempted between the first statement and the last statement, the thread blocks for too long. Blocking, however, is irrelevant if an absolute timeout is used. An absolute timeout also need not be recomputed if it is used multiple times in a loop, such as that enclosing a condition wait.

For cases when the system clock is advanced discontinuously by an operator, it is expected that implementations process any timed wait expiring at an intervening time as if that time had actually occurred.

Cancellation and Condition Wait

A condition wait, whether timed or not, is a cancellation point. That is, the functions pthread_cond_wait () or pthread_cond_timedwait () are points where a pending (or concurrent) cancellation request is noticed. The reason for this is that an indefinite wait is possible at these points-whatever event is being waited for, even if the program is totally correct, might never occur; for example, some input data being awaited might never be sent. By making condition wait a cancellation point, the thread can be canceled and perform its cancellation cleanup handler even though it may be stuck in some indefinite wait.

A side effect of acting on a cancellation request while a thread is blocked on a condition variable is to re-acquire the mutex before calling any of the cancellation cleanup handlers. This is done in order to ensure that the cancellation cleanup handler is executed in the same state as the critical code that lies both before and after the call to the condition wait function. This rule is also required when interfacing to POSIX threads from languages, such as Ada or C++, which may choose to map cancellation onto a language exception; this rule ensures that each exception handler guarding a critical section can always safely depend upon the fact that the associated mutex has already been locked regardless of exactly where within the critical section the exception was raised. Without this rule, there would not be a uniform rule that exception handlers could follow regarding the lock, and so coding would become very cumbersome.

Therefore, since some statement has to be made regarding the state of the lock when a cancellation is delivered during a wait, a definition has been chosen that makes application coding most convenient and error free.

When acting on a cancellation request while a thread is blocked on a condition variable, the implementation is required to ensure that the thread does not consume any condition signals directed at that condition variable if there are any other threads waiting on that condition variable. This rule is specified in order to avoid deadlock conditions that could occur if these two independent requests (one acting on a thread and the other acting on the condition variable) were not processed independently.

Performance of Mutexes and Condition Variables

Mutexes are expected to be locked only for a few instructions. This practice is almost automatically enforced by the desire of programmers to avoid long serial regions of execution (which would reduce total effective parallelism).

When using mutexes and condition variables, one tries to ensure that the usual case is to lock the mutex, access shared data, and unlock the mutex. Waiting on a condition variable should be a relatively rare situation. For example, when implementing a read-write lock, code that acquires a read-lock typically needs only to increment the count of readers (under mutual-exclusion) and return. The calling thread would actually wait on the condition variable only when there is already an active writer. So the efficiency of a synchronization operation is bounded by the cost of mutex lock/unlock and not by condition wait. Note that in the usual case there is no context switch.

This is not to say that the efficiency of condition waiting is unimportant. Since there needs to be at least one context switch per Ada rendezvous, the efficiency of waiting on a condition variable is important. The cost of waiting on a condition variable should be little more than the minimal cost for a context switch plus the time to unlock and lock the mutex.

Features of Mutexes and Condition Variables

It had been suggested that the mutex acquisition and release be decoupled from condition wait. This was rejected because it is the combined nature of the operation that, in fact, facilitates realtime implementations. Those implementations can atomically move a high-priority thread between the condition variable and the mutex in a manner that is transparent to the caller. This can prevent extra context switches and provide more deterministic acquisition of a mutex when the waiting thread is signaled. Thus, fairness and priority issues can be dealt with directly by the scheduling discipline. Furthermore, the current condition wait operation matches existing practice.

Scheduling Behavior of Mutexes and Condition Variables

Synchronization primitives that attempt to interfere with scheduling policy by specifying an ordering rule are considered undesirable. Threads waiting on mutexes and condition variables are selected to proceed in an order dependent upon the scheduling policy rather than in some fixed order (for example, FIFO or priority). Thus, the scheduling policy determines which thread(s) are awakened and allowed to proceed.

Timed Condition Wait

The pthread_cond_timedwait () function allows an application to give up waiting for a particular condition after a given amount of time. An example of its use follows:

(void) pthread_mutex_lock(&t.mn);
        t.waiters++;
    clock_gettime(CLOCK_REALTIME, &ts);
    ts.tv_sec += 5;
    rc = 0;
    while (! mypredicate(&t) && rc == 0)
        rc = pthread_cond_timedwait(&t.cond, &t.mn, &ts);
    t.waiters--;
    if (rc == 0) setmystate(&t);
(void) pthread_mutex_unlock(&t.mn);

By making the timeout parameter absolute, it does not need to be recomputed each time the program checks its blocking predicate. If the timeout was relative, it would have to be recomputed before each call. This would be especially difficult since such code would need to take into account the possibility of extra wakeups that result from extra broadcasts or signals on the condition variable that occur before either the predicate is true or the timeout is due.