haiku/docs/user/support/Locker.dox

234 lines
6.5 KiB
Plaintext
Raw Normal View History

/*
* Copyright 2007-2013 Haiku, Inc. All rights reserved.
* Distributed under the terms of the MIT License.
*
* Authors:
* Niels Sascha Reedijk, niels.reedijk@gmail.com
*
* Corresponds to:
* headers/os/support/Locker.h rev 36218
* src/kits/support/Locker.cpp rev 32758
*/
/*!
\file Locker.h
\ingroup support
\ingroup libbe
\brief Provides locking class BLocker.
*/
/*!
\class BLocker
\ingroup support
\ingroup libbe
\brief Semaphore-type class for thread safety.
The BLocker interface is not merely a wrapper around a semaphore, but it
also has two advantages. First of all, it implements a benaphore.
A benaphore is in some ways more speed efficient,
because before it uses the internal semaphore, it first checks against a
variable that is only operated on with atomic operations. Setting a variable
is a lot more efficient than acquiring a semaphore, thus this type of locking
is much preferred.
It basically works as follows. Whenever you newly created BLocker object
receives a locking request, it atomically sets the benaphore variable to
\c 1. Then only additional calls from different threads will utilize the
semaphore. You can imagine that in many cases where you protect
of data that \em might be accessed by two or more concurrent threads, but
the chances of it happening being very small, the benaphore benefits the
most from it's speed.
The other feature of BLocker that improves basic semaphore handling is that
it allows for recursive locks. The following piece of code works with a
BLocker, but block inevitably with a semaphore. Let's pretend I call
\c Water():
\code
status_t
Flower::Grow(int length)
{
if (fLock->Lock()) {
fLength += length;
fLock->Unlock();
return B_OK;
} else {
return B_ERROR;
}
}
status_t
Flower::Water(int amount)
{
if (fLock->Lock()) {
status_t status = Grow(amount * 2);
fLock->Unlock();
return status;
} else {
return B_ERROR;
}
}
\endcode
This code would work because BLocker keeps track of the amount of lock
requests from the same thread. A normal semaphore would block in \c Grow()
because the semaphore would be acquired already. Please do make sure you
pair every Lock() with an Unlock() though, or you'll create a deadlock.
*/
/*!
\fn BLocker::BLocker()
\brief Constructor.
Create a new BLocker with the default name of some BLocker. This
BLocker will use the benaphore-style locking.
\note For debugging purposes, it's extremely convenient to actually give a
name to the object. In case of a deadlock, it's easier to track down which
BLocker object might have caused the problems.
\see BLocker(const char* name, bool benaphoreStyle) for all the options.
*/
/*!
\fn BLocker::BLocker(const char* name)
\brief Constructor.
Create a new BLocker with benaphore-style locking.
\param name A NULL-terminated string that contains the name of the
semaphore. Note that the length of the names are limited to
\c B_OS_NAME_LENGTH constant, which includes the \c \\0
character.
\see BLocker(const char* name, bool benaphoreStyle) for all the options.
*/
/*!
\fn BLocker::BLocker(bool benaphoreStyle)
\brief Constructor.
Creates a BLocker with the default name of <tt>some BLocker</tt>.
\note For debugging purposes, it's extremely convenient to actually give a
name to the object. In case of a deadlock, it's easier to track down
which BLocker object might have caused the problems.
\param benaphoreStyle If you pass \c true, the locker will be in benaphore
style (which is the default option for other constructors). If you
pass \c false, the object will completely rely on semaphores for
it's functioning.
\see BLocker(const char* name, bool benaphoreStyle) if you also want
to set a name.
*/
/*!
\fn BLocker::BLocker(const char* name, bool benaphoreStyle)
\brief Constructor.
\param name A NULL-terminated string that contains the name of the
semaphore. Note that the length of the names are limited to
\c B_OS_NAME_LENGTH constant, which includes the \c \\0
character.
\param benaphoreStyle If you pass \c true, the locker will be in benaphore
style (which is the default option for other constructors). If
you pass \c false, the object will completely rely on semaphores
for its functioning.
*/
/*!
\fn virtual BLocker::~BLocker()
\brief Destructor.
Release the internal semaphore. Because of this, any pending Lock() calls
from other threads be cancelled. The return code will be \c false for
those calls.
*/
/*!
\fn status_t BLocker::InitCheck() const
\brief Check whether the locker has properly initialized
\retval B_OK The semaphore has been properly initialized
\retval (negative) Any other error value that is related to semaphore
initialization
*/
/*!
\fn bool BLocker::Lock()
\brief Add a lock request and block on it until we get it.
\retval true Lock acquired successfully.
\retval false Failed to acquire the lock. Most probable cause is that the
object is deleted. This frees the semaphore and releases the
pending Lock() requests.
\see LockWithTimeout(bigtime_t timeout), Unlock()
*/
/*!
\fn status_t BLocker::LockWithTimeout(bigtime_t timeout)
\brief Add a lock request and block until we get it or until it times out.
\param timeout This is a timeout in microseconds (one millionth of a
second) relative to now.
\see Lock(), Unlock()
*/
/*!
\fn void BLocker::Unlock(void)
\brief Release the lock that's currently held.
*/
/*!
\fn thread_id BLocker::LockingThread(void) const
\brief Return the \c thread_id of the thread that's currently holding the
lock.
*/
/*!
\fn bool BLocker::IsLocked(void) const
\brief Check if the calling thread is actually holding the lock.
\retval true The thread from which this method is called from is currently
holding the lock.
\retval false The object is unlocked or the lock is held by another thread.
*/
/*!
\fn int32 BLocker::CountLocks(void) const
\brief Return the number of recursive locks that are currently held.
*/
/*!
\fn nt32 BLocker::CountLockRequests(void) const
\brief Return the number of threads with a pending lock request.
*/
/*!
\fn sem_id BLocker::Sem(void) const
\brief Return the sem_id of the semaphore this object holds.
\warning Like any other internal objects that the Haiku API might expose,
this semaphore id should in general be left alone. You should not use any
of the public low-level semaphore functions on this semaphore, because it
will harm the internal consistency of the object.
*/