In the words of the author:

  o The polling mechanism for I/O readiness was changed from
    select() to poll().  In additon, a wrapped version of poll()
    is now provided.

  o The wrapped select routine now converts each fd_set to a
    poll array so that the thread scheduler doesn't have to
    perform a bitwise search for selected fds each time file
    descriptors are polled for I/O readiness.

  o The thread scheduler was modified to use a new queue (_workq)
    for threads that need work.  Threads waiting for I/O readiness
    and spinblocks are added to the work queue in addition to the
    waiting queue.  This reduces the time spent forming/searching
    the array of file descriptors being polled.

  o The waiting queue (_waitingq) is now maintained in order of
    thread wakeup time.  This allows the thread scheduler to
    find the nearest wakeup time by looking at the first thread
    in the queue instead of searching the entire queue.

  o Removed file descriptor locking for select/poll routines.  An
    application should not rely on the threads library for providing
    this locking; if necessary, the application should use mutexes
    to protect selecting/polling of file descriptors.

  o Retrieve and use the kernel clock rate/resolution at startup
    instead of hardcoding the clock resolution to 10 msec (tested
    with kernel running at 1000 HZ).

  o All queues have been changed to use queue.h macros.  These
    include the queues of all threads, dead threads, and threads
    waiting for file descriptor locks.

  o Added reinitialization of the GC mutex and condition variable
    after a fork.  Also prevented reallocation of the ready queue
    after a fork.

  o Prevented the wrapped close routine from closing the thread
    kernel pipes.

  o Initialized file descriptor table for stdio entries at thread
    init.

  o Provided additional flags to indicate to what queues threads
    belong.

  o Moved TAILQ initialization for statically allocated mutex and
    condition variables to after the spinlock.

  o Added dispatching of signals to pthread_kill.  Removing the
    dispatching of signals from thread activation broke sigsuspend
    when pthread_kill was used to send a signal to a thread.

  o Temporarily set the state of a thread to PS_SUSPENDED when it
    is first created and placed in the list of threads so that it
    will not be accidentally scheduled before becoming a member
    of one of the scheduling queues.

  o Change the signal handler to queue signals to the thread kernel
    pipe if the scheduling queues are protected.  When scheduling
    queues are unprotected, signals are then dequeued and handled.

  o Ensured that all installed signal handlers block the scheduling
    signal and that the scheduling signal handler blocks all
    other signals.  This ensures that the signal handler is only
    interruptible for and by non-scheduling signals.  An atomic
    lock is used to decide which instance of the signal handler
    will handle pending signals.

  o Removed _lock_thread_list and _unlock_thread_list as they are
    no longer used to protect the thread list.

  o Added missing RCS IDs to modified files.

  o Added checks for appropriate queue membership and activity when
    adding, removing, and searching the scheduling queues.  These
    checks add very little overhead and are enabled when compiled
    with _PTHREADS_INVARIANTS defined.  Suggested and implemented
    by Tor Egge with some modification by me.

  o Close a race condition in uthread_close.  (Tor Egge)

  o Protect the scheduling queues while modifying them in
    pthread_cond_signal and _thread_fd_unlock.  (Tor Egge)

  o Ensure that when a thread gets a mutex, the mutex is on that
    threads list of owned mutexes.  (Tor Egge)

  o Set the kernel-in-scheduler flag in _thread_kern_sched_state
    and _thread_kern_sched_state_unlock to prevent a scheduling
    signal from calling the scheduler again.  (Tor Egge)

  o Don't use TAILQ_FOREACH macro while searching the waiting
    queue for threads in a sigwait state, because a change of
    state destroys the TAILQ link.  It is actually safe to do
    so, though, because once a sigwaiting thread is found, the
    loop ends and the function returns.  (Tor Egge)

  o When dispatching signals to threads, make the thread inherit
    the signal deferral flag of the currently running thread.
    (Tor Egge)

Submitted by: Daniel Eischen <eischen@vigrid.com> and
              Tor Egge <Tor.Egge@fast.no>

1 files changed, 120 insertions, 72 deletions

diff --git a/lib/libkse/thread/thr_mutex.c b/lib/libkse/thread/thr_mutex.c
index fa9c8cf..c967b46 100644
--- a/lib/libkse/thread/thr_mutex.c
+++ b/lib/libkse/thread/thr_mutex.c
@@ -29,6 +29,7 @@
  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  * SUCH DAMAGE.
  *
+ * $Id$
  */
 #include <stdlib.h>
 #include <errno.h>
@@ -39,6 +40,25 @@
 #include <pthread.h>
 #include "pthread_private.h"
 
+#if defined(_PTHREADS_INVARIANTS)
+#define _MUTEX_INIT_LINK(m) 		do {		\
+	(m)->m_qe.tqe_prev = NULL;			\
+	(m)->m_qe.tqe_next = NULL;			\
+} while (0)
+#define _MUTEX_ASSERT_IS_OWNED(m)	do {		\
+	if ((m)->m_qe.tqe_prev == NULL)			\
+		PANIC("mutex is not on list");		\
+} while (0)
+#define _MUTEX_ASSERT_NOT_OWNED(m)	do {		\
+	if (((m)->m_qe.tqe_prev != NULL) ||		\
+	    ((m)->m_qe.tqe_next != NULL))		\
+		PANIC("mutex is on list");		\
+} while (0)
+#else
+#define _MUTEX_INIT_LINK(m)
+#define _MUTEX_ASSERT_IS_OWNED(m)
+#define _MUTEX_ASSERT_NOT_OWNED(m)
+#endif
 
 /*
  * Prototypes
@@ -55,6 +75,34 @@ static inline void	mutex_queue_enq(pthread_mutex_t, pthread_t);
 
 static spinlock_t static_init_lock = _SPINLOCK_INITIALIZER;
 
+/* Reinitialize a mutex to defaults. */
+int
+_mutex_reinit(pthread_mutex_t * mutex)
+{
+	int ret = 0;
+
+	if (mutex == NULL)
+		ret = EINVAL;
+	else if (*mutex == NULL)
+		ret = pthread_mutex_init(mutex, NULL);
+	else {
+		/*
+		 * Initialize the mutex structure:
+		 */
+		(*mutex)->m_type = PTHREAD_MUTEX_DEFAULT;
+		(*mutex)->m_protocol = PTHREAD_PRIO_NONE;
+		TAILQ_INIT(&(*mutex)->m_queue);
+		(*mutex)->m_owner = NULL;
+		(*mutex)->m_data.m_count = 0;
+		(*mutex)->m_flags = MUTEX_FLAGS_INITED;
+		(*mutex)->m_refcount = 0;
+		(*mutex)->m_prio = 0;
+		(*mutex)->m_saved_prio = 0;
+		_MUTEX_INIT_LINK(*mutex);
+		memset(&(*mutex)->lock, 0, sizeof((*mutex)->lock));
+	}
+	return (ret);
+}
 
 int
 pthread_mutex_init(pthread_mutex_t * mutex,
@@ -138,6 +186,7 @@ pthread_mutex_init(pthread_mutex_t * mutex,
 				else
 					pmutex->m_prio = 0;
 				pmutex->m_saved_prio = 0;
+				_MUTEX_INIT_LINK(pmutex);
 				memset(&pmutex->lock, 0, sizeof(pmutex->lock));
 				*mutex = pmutex;
 			} else {
@@ -147,7 +196,7 @@ pthread_mutex_init(pthread_mutex_t * mutex,
 		}
 	}
 	/* Return the completion status: */
-	return (ret);
+	return(ret);
 }
 
 int
@@ -177,6 +226,7 @@ pthread_mutex_destroy(pthread_mutex_t * mutex)
 			 * Free the memory allocated for the mutex
 			 * structure:
 			 */
+			_MUTEX_ASSERT_NOT_OWNED(*mutex);
 			free(*mutex);
 
 			/*
@@ -222,28 +272,24 @@ pthread_mutex_trylock(pthread_mutex_t * mutex)
 	 */
 	else if (*mutex != NULL || (ret = init_static(mutex)) == 0) {
 		/*
+		 * Defer signals to protect the scheduling queues from
+		 * access by the signal handler:
+		 */
+		_thread_kern_sig_defer();
+
+		/* Lock the mutex structure: */
+		_SPINLOCK(&(*mutex)->lock);
+
+		/*
 		 * If the mutex was statically allocated, properly
 		 * initialize the tail queue.
 		 */
 		if (((*mutex)->m_flags & MUTEX_FLAGS_INITED) == 0) {
 			TAILQ_INIT(&(*mutex)->m_queue);
+			_MUTEX_INIT_LINK(*mutex);
 			(*mutex)->m_flags |= MUTEX_FLAGS_INITED;
 		}
 
-		/*
-		 * Guard against being preempted by a scheduling signal.
-		 * To support priority inheritence mutexes, we need to
-		 * maintain lists of mutex ownerships for each thread as
-		 * well as lists of waiting threads for each mutex.  In
-		 * order to propagate priorities we need to atomically
-		 * walk these lists and cannot rely on a single mutex
-		 * lock to provide protection against modification.
-		 */
-		_thread_kern_sched_defer();
-
-		/* Lock the mutex structure: */
-		_SPINLOCK(&(*mutex)->lock);
-
 		/* Process according to mutex type: */
 		switch ((*mutex)->m_protocol) {
 		/* Default POSIX mutex: */
@@ -254,6 +300,7 @@ pthread_mutex_trylock(pthread_mutex_t * mutex)
 				(*mutex)->m_owner = _thread_run;
 
 				/* Add to the list of owned mutexes: */
+				_MUTEX_ASSERT_NOT_OWNED(*mutex);
 				TAILQ_INSERT_TAIL(&_thread_run->mutexq,
 				    (*mutex), m_qe);
 			} else if ((*mutex)->m_owner == _thread_run)
@@ -282,6 +329,7 @@ pthread_mutex_trylock(pthread_mutex_t * mutex)
 				    _thread_run->inherited_priority;
 
 				/* Add to the list of owned mutexes: */
+				_MUTEX_ASSERT_NOT_OWNED(*mutex);
 				TAILQ_INSERT_TAIL(&_thread_run->mutexq,
 				    (*mutex), m_qe);
 			} else if ((*mutex)->m_owner == _thread_run)
@@ -317,6 +365,7 @@ pthread_mutex_trylock(pthread_mutex_t * mutex)
 				    (*mutex)->m_prio;
 
 				/* Add to the list of owned mutexes: */
+				_MUTEX_ASSERT_NOT_OWNED(*mutex);
 				TAILQ_INSERT_TAIL(&_thread_run->mutexq,
 				    (*mutex), m_qe);
 			} else if ((*mutex)->m_owner == _thread_run)
@@ -337,10 +386,10 @@ pthread_mutex_trylock(pthread_mutex_t * mutex)
 		_SPINUNLOCK(&(*mutex)->lock);
 
 		/*
-		 * Renable preemption and yield if a scheduling signal
-		 * arrived while in the critical region:
+		 * Undefer and handle pending signals, yielding if
+		 * necessary:
 		 */
-		_thread_kern_sched_undefer();
+		_thread_kern_sig_undefer();
 	}
 
 	/* Return the completion status: */
@@ -361,28 +410,24 @@ pthread_mutex_lock(pthread_mutex_t * mutex)
 	 */
 	else if (*mutex != NULL || (ret = init_static(mutex)) == 0) {
 		/*
+		 * Defer signals to protect the scheduling queues from
+		 * access by the signal handler:
+		 */
+		_thread_kern_sig_defer();
+
+		/* Lock the mutex structure: */
+		_SPINLOCK(&(*mutex)->lock);
+
+		/*
 		 * If the mutex was statically allocated, properly
 		 * initialize the tail queue.
 		 */
 		if (((*mutex)->m_flags & MUTEX_FLAGS_INITED) == 0) {
 			TAILQ_INIT(&(*mutex)->m_queue);
 			(*mutex)->m_flags |= MUTEX_FLAGS_INITED;
+			_MUTEX_INIT_LINK(*mutex);
 		}
 
-		/*
-		 * Guard against being preempted by a scheduling signal.
-		 * To support priority inheritence mutexes, we need to
-		 * maintain lists of mutex ownerships for each thread as
-		 * well as lists of waiting threads for each mutex.  In
-		 * order to propagate priorities we need to atomically
-		 * walk these lists and cannot rely on a single mutex
-		 * lock to provide protection against modification.
-		 */
-		_thread_kern_sched_defer();
-
-		/* Lock the mutex structure: */
-		_SPINLOCK(&(*mutex)->lock);
-
 		/* Process according to mutex type: */
 		switch ((*mutex)->m_protocol) {
 		/* Default POSIX mutex: */
@@ -392,6 +437,7 @@ pthread_mutex_lock(pthread_mutex_t * mutex)
 				(*mutex)->m_owner = _thread_run;
 
 				/* Add to the list of owned mutexes: */
+				_MUTEX_ASSERT_NOT_OWNED(*mutex);
 				TAILQ_INSERT_TAIL(&_thread_run->mutexq,
 				    (*mutex), m_qe);
 
@@ -419,12 +465,6 @@ pthread_mutex_lock(pthread_mutex_t * mutex)
 
 				/* Lock the mutex structure again: */
 				_SPINLOCK(&(*mutex)->lock);
-
-				/*
-				 * This thread is no longer waiting for
-				 * the mutex:
-				 */
-				_thread_run->data.mutex = NULL;
 			}
 			break;
 
@@ -449,6 +489,7 @@ pthread_mutex_lock(pthread_mutex_t * mutex)
 				    (*mutex)->m_prio;
 
 				/* Add to the list of owned mutexes: */
+				_MUTEX_ASSERT_NOT_OWNED(*mutex);
 				TAILQ_INSERT_TAIL(&_thread_run->mutexq,
 				    (*mutex), m_qe);
 
@@ -481,12 +522,6 @@ pthread_mutex_lock(pthread_mutex_t * mutex)
 
 				/* Lock the mutex structure again: */
 				_SPINLOCK(&(*mutex)->lock);
-
-				/*
-				 * This thread is no longer waiting for
-				 * the mutex:
-				 */
-				_thread_run->data.mutex = NULL;
 			}
 			break;
 
@@ -519,6 +554,7 @@ pthread_mutex_lock(pthread_mutex_t * mutex)
 				    (*mutex)->m_prio;
 
 				/* Add to the list of owned mutexes: */
+				_MUTEX_ASSERT_NOT_OWNED(*mutex);
 				TAILQ_INSERT_TAIL(&_thread_run->mutexq,
 				    (*mutex), m_qe);
 			} else if ((*mutex)->m_owner == _thread_run)
@@ -556,12 +592,6 @@ pthread_mutex_lock(pthread_mutex_t * mutex)
 				 */
 				ret = _thread_run->error;
 				_thread_run->error = 0;
-
-				/*
-				 * This thread is no longer waiting for
-				 * the mutex:
-				 */
-				_thread_run->data.mutex = NULL;
 			}
 			break;
 
@@ -576,10 +606,10 @@ pthread_mutex_lock(pthread_mutex_t * mutex)
 		_SPINUNLOCK(&(*mutex)->lock);
 
 		/*
-		 * Renable preemption and yield if a scheduling signal
-		 * arrived while in the critical region:
+		 * Undefer and handle pending signals, yielding if
+		 * necessary:
 		 */
-		_thread_kern_sched_undefer();
+		_thread_kern_sig_undefer();
 	}
 
 	/* Return the completion status: */
@@ -683,15 +713,10 @@ mutex_unlock_common(pthread_mutex_t * mutex, int add_reference)
 		ret = EINVAL;
 	} else {
 		/*
-		 * Guard against being preempted by a scheduling signal.
-		 * To support priority inheritence mutexes, we need to
-		 * maintain lists of mutex ownerships for each thread as
-		 * well as lists of waiting threads for each mutex.  In
-		 * order to propagate priorities we need to atomically
-		 * walk these lists and cannot rely on a single mutex
-		 * lock to provide protection against modification.
+		 * Defer signals to protect the scheduling queues from
+		 * access by the signal handler:
 		 */
-		_thread_kern_sched_defer();
+		_thread_kern_sig_defer();
 
 		/* Lock the mutex structure: */
 		_SPINLOCK(&(*mutex)->lock);
@@ -723,8 +748,10 @@ mutex_unlock_common(pthread_mutex_t * mutex, int add_reference)
 				(*mutex)->m_data.m_count = 0;
 
 				/* Remove the mutex from the threads queue. */
+				_MUTEX_ASSERT_IS_OWNED(*mutex);
 				TAILQ_REMOVE(&(*mutex)->m_owner->mutexq,
 				    (*mutex), m_qe);
+				_MUTEX_INIT_LINK(*mutex);
 
 				/*
 				 * Get the next thread from the queue of
@@ -738,6 +765,19 @@ mutex_unlock_common(pthread_mutex_t * mutex, int add_reference)
 					 */
 					PTHREAD_NEW_STATE((*mutex)->m_owner,
 					    PS_RUNNING);
+
+					/*
+					 * Add the mutex to the threads list of
+					 * owned mutexes:
+					 */
+					TAILQ_INSERT_TAIL(&(*mutex)->m_owner->mutexq,
+					    (*mutex), m_qe);
+
+					/*
+					 * The owner is no longer waiting for
+					 * this mutex:
+					 */
+					(*mutex)->m_owner->data.mutex = NULL;
 				}
 			}
 			break;
@@ -784,8 +824,10 @@ mutex_unlock_common(pthread_mutex_t * mutex, int add_reference)
 				_thread_run->priority_mutex_count--;
 
 				/* Remove the mutex from the threads queue. */
+				_MUTEX_ASSERT_IS_OWNED(*mutex);
 				TAILQ_REMOVE(&(*mutex)->m_owner->mutexq,
 				    (*mutex), m_qe);
+				_MUTEX_INIT_LINK(*mutex);
 
 				/*
 				 * Get the next thread from the queue of threads
@@ -891,8 +933,10 @@ mutex_unlock_common(pthread_mutex_t * mutex, int add_reference)
 				_thread_run->priority_mutex_count--;
 
 				/* Remove the mutex from the threads queue. */
+				_MUTEX_ASSERT_IS_OWNED(*mutex);
 				TAILQ_REMOVE(&(*mutex)->m_owner->mutexq,
-					    (*mutex), m_qe);
+				    (*mutex), m_qe);
+				_MUTEX_INIT_LINK(*mutex);
 
 				/*
 				 * Enter a loop to find a waiting thread whose
@@ -913,6 +957,11 @@ mutex_unlock_common(pthread_mutex_t * mutex, int add_reference)
 					(*mutex)->m_owner->error = EINVAL;
 					PTHREAD_NEW_STATE((*mutex)->m_owner,
 					    PS_RUNNING);
+					/*
+					 * The thread is no longer waiting for
+					 * this mutex:
+					 */
+					(*mutex)->m_owner->data.mutex = NULL;
 				}
 
 				/* Check for a new owner: */
@@ -978,10 +1027,10 @@ mutex_unlock_common(pthread_mutex_t * mutex, int add_reference)
 		_SPINUNLOCK(&(*mutex)->lock);
 
 		/*
-		 * Renable preemption and yield if a scheduling signal
-		 * arrived while in the critical region:
+		 * Undefer and handle pending signals, yielding if
+		 * necessary:
 		 */
-		_thread_kern_sched_undefer();
+		_thread_kern_sig_undefer();
 	}
 
 	/* Return the completion status: */
@@ -990,11 +1039,11 @@ mutex_unlock_common(pthread_mutex_t * mutex, int add_reference)
 
 
 /*
- * This function is called when a change in base priority occurs
- * for a thread that is thread holding, or waiting for, a priority
- * protection or inheritence mutex.  A change in a threads base
- * priority can effect changes to active priorities of other threads
- * and to the ordering of mutex locking by waiting threads.
+ * This function is called when a change in base priority occurs for
+ * a thread that is holding or waiting for a priority protection or
+ * inheritence mutex.  A change in a threads base priority can effect
+ * changes to active priorities of other threads and to the ordering
+ * of mutex locking by waiting threads.
  *
  * This must be called while thread scheduling is deferred.
  */
@@ -1231,8 +1280,7 @@ mutex_rescan_owned (pthread_t pthread, pthread_mutex_t mutex)
 		 * If this thread is in the priority queue, it must be
 		 * removed and reinserted for its new priority.
 	 	 */
-		if ((pthread != _thread_run) &&
-		    (pthread->state == PS_RUNNING)) {
+		if (pthread->flags & PTHREAD_FLAGS_IN_PRIOQ) {
 			/*
 			 * Remove the thread from the priority queue
 			 * before changing its priority: