Prepare for radical source tree reorganization. zack/build-layout-experiment

All top-level files and directories are moved into a temporary storage
directory, REORG.TODO, except for files that will certainly still
exist in their current form at top level when we're done (COPYING,
COPYING.LIB, LICENSES, NEWS, README), all old ChangeLog files (which
are moved to the new directory OldChangeLogs, instead), and the
generated file INSTALL (which is just deleted; in the new order, there
will be no generated files checked into version control).

1 files changed, 571 insertions, 0 deletions

diff --git a/REORG.TODO/sysdeps/pthread/timer_routines.c b/REORG.TODO/sysdeps/pthread/timer_routines.c
new file mode 100644
index 0000000000..1c0988a4ba
--- /dev/null
+++ b/REORG.TODO/sysdeps/pthread/timer_routines.c
@@ -0,0 +1,571 @@
+/* Helper code for POSIX timer implementation on NPTL.
+   Copyright (C) 2000-2017 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Kaz Kylheku <kaz@ashi.footprints.net>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public License as
+   published by the Free Software Foundation; either version 2.1 of the
+   License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; see the file COPYING.LIB.  If
+   not, see <http://www.gnu.org/licenses/>.  */
+
+#include <assert.h>
+#include <errno.h>
+#include <pthread.h>
+#include <stddef.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sysdep.h>
+#include <time.h>
+#include <unistd.h>
+#include <sys/syscall.h>
+
+#include "posix-timer.h"
+#include <nptl/pthreadP.h>
+
+
+/* Number of threads used.  */
+#define THREAD_MAXNODES	16
+
+/* Array containing the descriptors for the used threads.  */
+static struct thread_node thread_array[THREAD_MAXNODES];
+
+/* Static array with the structures for all the timers.  */
+struct timer_node __timer_array[TIMER_MAX];
+
+/* Global lock to protect operation on the lists.  */
+pthread_mutex_t __timer_mutex = PTHREAD_MUTEX_INITIALIZER;
+
+/* Variable to protext initialization.  */
+pthread_once_t __timer_init_once_control = PTHREAD_ONCE_INIT;
+
+/* Nonzero if initialization of timer implementation failed.  */
+int __timer_init_failed;
+
+/* Node for the thread used to deliver signals.  */
+struct thread_node __timer_signal_thread_rclk;
+
+/* Lists to keep free and used timers and threads.  */
+static struct list_head timer_free_list;
+static struct list_head thread_free_list;
+static struct list_head thread_active_list;
+
+
+#ifdef __NR_rt_sigqueueinfo
+extern int __syscall_rt_sigqueueinfo (int, int, siginfo_t *);
+#endif
+
+
+/* List handling functions.  */
+static inline void
+list_append (struct list_head *list, struct list_head *newp)
+{
+  newp->prev = list->prev;
+  newp->next = list;
+  list->prev->next = newp;
+  list->prev = newp;
+}
+
+static inline void
+list_insbefore (struct list_head *list, struct list_head *newp)
+{
+  list_append (list, newp);
+}
+
+/*
+ * Like list_unlink_ip, except that calling it on a node that
+ * is already unlinked is disastrous rather than a noop.
+ */
+
+static inline void
+list_unlink (struct list_head *list)
+{
+  struct list_head *lnext = list->next, *lprev = list->prev;
+
+  lnext->prev = lprev;
+  lprev->next = lnext;
+}
+
+static inline struct list_head *
+list_first (struct list_head *list)
+{
+  return list->next;
+}
+
+static inline struct list_head *
+list_null (struct list_head *list)
+{
+  return list;
+}
+
+static inline struct list_head *
+list_next (struct list_head *list)
+{
+  return list->next;
+}
+
+static inline int
+list_isempty (struct list_head *list)
+{
+  return list->next == list;
+}
+
+
+/* Functions build on top of the list functions.  */
+static inline struct thread_node *
+thread_links2ptr (struct list_head *list)
+{
+  return (struct thread_node *) ((char *) list
+				 - offsetof (struct thread_node, links));
+}
+
+static inline struct timer_node *
+timer_links2ptr (struct list_head *list)
+{
+  return (struct timer_node *) ((char *) list
+				- offsetof (struct timer_node, links));
+}
+
+
+/* Initialize a newly allocated thread structure.  */
+static void
+thread_init (struct thread_node *thread, const pthread_attr_t *attr, clockid_t clock_id)
+{
+  if (attr != NULL)
+    thread->attr = *attr;
+  else
+    {
+      pthread_attr_init (&thread->attr);
+      pthread_attr_setdetachstate (&thread->attr, PTHREAD_CREATE_DETACHED);
+    }
+
+  thread->exists = 0;
+  INIT_LIST_HEAD (&thread->timer_queue);
+  pthread_cond_init (&thread->cond, 0);
+  thread->current_timer = 0;
+  thread->captured = pthread_self ();
+  thread->clock_id = clock_id;
+}
+
+
+/* Initialize the global lists, and acquire global resources.  Error
+   reporting is done by storing a non-zero value to the global variable
+   timer_init_failed.  */
+static void
+init_module (void)
+{
+  int i;
+
+  INIT_LIST_HEAD (&timer_free_list);
+  INIT_LIST_HEAD (&thread_free_list);
+  INIT_LIST_HEAD (&thread_active_list);
+
+  for (i = 0; i < TIMER_MAX; ++i)
+    {
+      list_append (&timer_free_list, &__timer_array[i].links);
+      __timer_array[i].inuse = TIMER_FREE;
+    }
+
+  for (i = 0; i < THREAD_MAXNODES; ++i)
+    list_append (&thread_free_list, &thread_array[i].links);
+
+  thread_init (&__timer_signal_thread_rclk, 0, CLOCK_REALTIME);
+}
+
+
+/* This is a handler executed in a child process after a fork()
+   occurs.  It reinitializes the module, resetting all of the data
+   structures to their initial state.  The mutex is initialized in
+   case it was locked in the parent process.  */
+static void
+reinit_after_fork (void)
+{
+  init_module ();
+  pthread_mutex_init (&__timer_mutex, 0);
+}
+
+
+/* Called once form pthread_once in timer_init. This initializes the
+   module and ensures that reinit_after_fork will be executed in any
+   child process.  */
+void
+__timer_init_once (void)
+{
+  init_module ();
+  pthread_atfork (0, 0, reinit_after_fork);
+}
+
+
+/* Deinitialize a thread that is about to be deallocated.  */
+static void
+thread_deinit (struct thread_node *thread)
+{
+  assert (list_isempty (&thread->timer_queue));
+  pthread_cond_destroy (&thread->cond);
+}
+
+
+/* Allocate a thread structure from the global free list.  Global
+   mutex lock must be held by caller.  The thread is moved to
+   the active list. */
+struct thread_node *
+__timer_thread_alloc (const pthread_attr_t *desired_attr, clockid_t clock_id)
+{
+  struct list_head *node = list_first (&thread_free_list);
+
+  if (node != list_null (&thread_free_list))
+    {
+      struct thread_node *thread = thread_links2ptr (node);
+      list_unlink (node);
+      thread_init (thread, desired_attr, clock_id);
+      list_append (&thread_active_list, node);
+      return thread;
+    }
+
+  return 0;
+}
+
+
+/* Return a thread structure to the global free list.  Global lock
+   must be held by caller.  */
+void
+__timer_thread_dealloc (struct thread_node *thread)
+{
+  thread_deinit (thread);
+  list_unlink (&thread->links);
+  list_append (&thread_free_list, &thread->links);
+}
+
+
+/* Each of our threads which terminates executes this cleanup
+   handler. We never terminate threads ourselves; if a thread gets here
+   it means that the evil application has killed it.  If the thread has
+   timers, these require servicing and so we must hire a replacement
+   thread right away.  We must also unblock another thread that may
+   have been waiting for this thread to finish servicing a timer (see
+   timer_delete()).  */
+
+static void
+thread_cleanup (void *val)
+{
+  if (val != NULL)
+    {
+      struct thread_node *thread = val;
+
+      /* How did the signal thread get killed?  */
+      assert (thread != &__timer_signal_thread_rclk);
+
+      pthread_mutex_lock (&__timer_mutex);
+
+      thread->exists = 0;
+
+      /* We are no longer processing a timer event.  */
+      thread->current_timer = 0;
+
+      if (list_isempty (&thread->timer_queue))
+	__timer_thread_dealloc (thread);
+      else
+	(void) __timer_thread_start (thread);
+
+      pthread_mutex_unlock (&__timer_mutex);
+
+      /* Unblock potentially blocked timer_delete().  */
+      pthread_cond_broadcast (&thread->cond);
+    }
+}
+
+
+/* Handle a timer which is supposed to go off now.  */
+static void
+thread_expire_timer (struct thread_node *self, struct timer_node *timer)
+{
+  self->current_timer = timer; /* Lets timer_delete know timer is running. */
+
+  pthread_mutex_unlock (&__timer_mutex);
+
+  switch (__builtin_expect (timer->event.sigev_notify, SIGEV_SIGNAL))
+    {
+    case SIGEV_NONE:
+      break;
+
+    case SIGEV_SIGNAL:
+#ifdef __NR_rt_sigqueueinfo
+      {
+	siginfo_t info;
+
+	/* First, clear the siginfo_t structure, so that we don't pass our
+	   stack content to other tasks.  */
+	memset (&info, 0, sizeof (siginfo_t));
+	/* We must pass the information about the data in a siginfo_t
+           value.  */
+	info.si_signo = timer->event.sigev_signo;
+	info.si_code = SI_TIMER;
+	info.si_pid = timer->creator_pid;
+	info.si_uid = getuid ();
+	info.si_value = timer->event.sigev_value;
+
+	INLINE_SYSCALL (rt_sigqueueinfo, 3, info.si_pid, info.si_signo, &info);
+      }
+#else
+      if (pthread_kill (self->captured, timer->event.sigev_signo) != 0)
+	{
+	  if (pthread_kill (self->id, timer->event.sigev_signo) != 0)
+	    abort ();
+        }
+#endif
+      break;
+
+    case SIGEV_THREAD:
+      timer->event.sigev_notify_function (timer->event.sigev_value);
+      break;
+
+    default:
+      assert (! "unknown event");
+      break;
+    }
+
+  pthread_mutex_lock (&__timer_mutex);
+
+  self->current_timer = 0;
+
+  pthread_cond_broadcast (&self->cond);
+}
+
+
+/* Thread function; executed by each timer thread. The job of this
+   function is to wait on the thread's timer queue and expire the
+   timers in chronological order as close to their scheduled time as
+   possible.  */
+static void
+__attribute__ ((noreturn))
+thread_func (void *arg)
+{
+  struct thread_node *self = arg;
+
+  /* Register cleanup handler, in case rogue application terminates
+     this thread.  (This cannot happen to __timer_signal_thread, which
+     doesn't invoke application callbacks). */
+
+  pthread_cleanup_push (thread_cleanup, self);
+
+  pthread_mutex_lock (&__timer_mutex);
+
+  while (1)
+    {
+      struct list_head *first;
+      struct timer_node *timer = NULL;
+
+      /* While the timer queue is not empty, inspect the first node.  */
+      first = list_first (&self->timer_queue);
+      if (first != list_null (&self->timer_queue))
+	{
+	  struct timespec now;
+
+	  timer = timer_links2ptr (first);
+
+	  /* This assumes that the elements of the list of one thread
+	     are all for the same clock.  */
+	  clock_gettime (timer->clock, &now);
+
+	  while (1)
+	    {
+	      /* If the timer is due or overdue, remove it from the queue.
+		 If it's a periodic timer, re-compute its new time and
+		 requeue it.  Either way, perform the timer expiry. */
+	      if (timespec_compare (&now, &timer->expirytime) < 0)
+		break;
+
+	      list_unlink_ip (first);
+
+	      if (__builtin_expect (timer->value.it_interval.tv_sec, 0) != 0
+		  || timer->value.it_interval.tv_nsec != 0)
+		{
+		  timer->overrun_count = 0;
+		  timespec_add (&timer->expirytime, &timer->expirytime,
+				&timer->value.it_interval);
+		  while (timespec_compare (&timer->expirytime, &now) < 0)
+		    {
+		      timespec_add (&timer->expirytime, &timer->expirytime,
+				    &timer->value.it_interval);
+		      if (timer->overrun_count < DELAYTIMER_MAX)
+			++timer->overrun_count;
+		    }
+		  __timer_thread_queue_timer (self, timer);
+		}
+
+	      thread_expire_timer (self, timer);
+
+	      first = list_first (&self->timer_queue);
+	      if (first == list_null (&self->timer_queue))
+		break;
+
+	      timer = timer_links2ptr (first);
+	    }
+	}
+
+      /* If the queue is not empty, wait until the expiry time of the
+	 first node.  Otherwise wait indefinitely.  Insertions at the
+	 head of the queue must wake up the thread by broadcasting
+	 this condition variable.  */
+      if (timer != NULL)
+	pthread_cond_timedwait (&self->cond, &__timer_mutex,
+				&timer->expirytime);
+      else
+	pthread_cond_wait (&self->cond, &__timer_mutex);
+    }
+  /* This macro will never be executed since the while loop loops
+     forever - but we have to add it for proper nesting.  */
+  pthread_cleanup_pop (1);
+}
+
+
+/* Enqueue a timer in wakeup order in the thread's timer queue.
+   Returns 1 if the timer was inserted at the head of the queue,
+   causing the queue's next wakeup time to change. */
+
+int
+__timer_thread_queue_timer (struct thread_node *thread,
+			    struct timer_node *insert)
+{
+  struct list_head *iter;
+  int athead = 1;
+
+  for (iter = list_first (&thread->timer_queue);
+       iter != list_null (&thread->timer_queue);
+        iter = list_next (iter))
+    {
+      struct timer_node *timer = timer_links2ptr (iter);
+
+      if (timespec_compare (&insert->expirytime, &timer->expirytime) < 0)
+	  break;
+      athead = 0;
+    }
+
+  list_insbefore (iter, &insert->links);
+  return athead;
+}
+
+
+/* Start a thread and associate it with the given thread node.  Global
+   lock must be held by caller.  */
+int
+__timer_thread_start (struct thread_node *thread)
+{
+  int retval = 1;
+
+  assert (!thread->exists);
+  thread->exists = 1;
+
+  if (pthread_create (&thread->id, &thread->attr,
+		      (void *(*) (void *)) thread_func, thread) != 0)
+    {
+      thread->exists = 0;
+      retval = -1;
+    }
+
+  return retval;
+}
+
+
+void
+__timer_thread_wakeup (struct thread_node *thread)
+{
+  pthread_cond_broadcast (&thread->cond);
+}
+
+
+/* Compare two pthread_attr_t thread attributes for exact equality.
+   Returns 1 if they are equal, otherwise zero if they are not equal
+   or contain illegal values.  This version is NPTL-specific for
+   performance reason.  One could use the access functions to get the
+   values of all the fields of the attribute structure.  */
+static int
+thread_attr_compare (const pthread_attr_t *left, const pthread_attr_t *right)
+{
+  struct pthread_attr *ileft = (struct pthread_attr *) left;
+  struct pthread_attr *iright = (struct pthread_attr *) right;
+
+  return (ileft->flags == iright->flags
+	  && ileft->schedpolicy == iright->schedpolicy
+	  && (ileft->schedparam.sched_priority
+	      == iright->schedparam.sched_priority)
+	  && ileft->guardsize == iright->guardsize
+	  && ileft->stackaddr == iright->stackaddr
+	  && ileft->stacksize == iright->stacksize
+	  && ((ileft->cpuset == NULL && iright->cpuset == NULL)
+	      || (ileft->cpuset != NULL && iright->cpuset != NULL
+		  && ileft->cpusetsize == iright->cpusetsize
+		  && memcmp (ileft->cpuset, iright->cpuset,
+			     ileft->cpusetsize) == 0)));
+}
+
+
+/* Search the list of active threads and find one which has matching
+   attributes.  Global mutex lock must be held by caller.  */
+struct thread_node *
+__timer_thread_find_matching (const pthread_attr_t *desired_attr,
+			      clockid_t desired_clock_id)
+{
+  struct list_head *iter = list_first (&thread_active_list);
+
+  while (iter != list_null (&thread_active_list))
+    {
+      struct thread_node *candidate = thread_links2ptr (iter);
+
+      if (thread_attr_compare (desired_attr, &candidate->attr)
+	  && desired_clock_id == candidate->clock_id)
+	return candidate;
+
+      iter = list_next (iter);
+    }
+
+  return NULL;
+}
+
+
+/* Grab a free timer structure from the global free list.  The global
+   lock must be held by the caller.  */
+struct timer_node *
+__timer_alloc (void)
+{
+  struct list_head *node = list_first (&timer_free_list);
+
+  if (node != list_null (&timer_free_list))
+    {
+      struct timer_node *timer = timer_links2ptr (node);
+      list_unlink_ip (node);
+      timer->inuse = TIMER_INUSE;
+      timer->refcount = 1;
+      return timer;
+    }
+
+  return NULL;
+}
+
+
+/* Return a timer structure to the global free list.  The global lock
+   must be held by the caller.  */
+void
+__timer_dealloc (struct timer_node *timer)
+{
+  assert (timer->refcount == 0);
+  timer->thread = NULL;	/* Break association between timer and thread.  */
+  timer->inuse = TIMER_FREE;
+  list_append (&timer_free_list, &timer->links);
+}
+
+
+/* Thread cancellation handler which unlocks a mutex.  */
+void
+__timer_mutex_cancel_handler (void *arg)
+{
+  pthread_mutex_unlock (arg);
+}