htl: Make pthread_cond_destroy wait for threads to be woken

This allows to reuse the storage after calling pthread_cond_destroy.

* sysdeps/htl/bits/types/struct___pthread_cond.h (__pthread_cond):
Replace unused struct __pthread_condimpl *__impl field with unsigned int
__wrefs.
(__PTHREAD_COND_INITIALIZER): Update accordingly.
* sysdeps/htl/pt-cond-timedwait.c (__pthread_cond_timedwait_internal):
Register as waiter in __wrefs field. On unregistering, wake any pending
pthread_cond_destroy.
* sysdeps/htl/pt-cond-destroy.c (__pthread_cond_destroy): Register wake
request in __wrefs.
* nptl/Makefile (tests): Move tst-cond20 tst-cond21 to...
* sysdeps/pthread/Makefile (tests): ... here.
* nptl/tst-cond20.c nptl/tst-cond21.c: Move to...
* sysdeps/pthread/tst-cond20.c sysdeps/pthread/tst-cond21.c: ... here.

nptl/Makefile

@@ -263,7 +263,7 @@ tests = tst-attr2 tst-attr3 tst-default-attr \
 	tst-mutexpi1 tst-mutexpi2 tst-mutexpi3 tst-mutexpi4 \
 	tst-mutexpi5 tst-mutexpi5a tst-mutexpi6 tst-mutexpi7 tst-mutexpi7a \
 	tst-mutexpi9 \
-	tst-cond20 tst-cond21 tst-cond22 tst-cond26 \
+	tst-cond22 tst-cond26 \
 	tst-robustpi1 tst-robustpi2 tst-robustpi3 tst-robustpi4 tst-robustpi5 \
 	tst-robustpi6 tst-robustpi7 tst-robustpi9 \
 	tst-rwlock2 tst-rwlock2a tst-rwlock2b tst-rwlock3 \

sysdeps/htl/bits/types/struct___pthread_cond.h

@@ -27,12 +27,12 @@ struct __pthread_cond
   __pthread_spinlock_t __lock;
   struct __pthread *__queue;
   struct __pthread_condattr *__attr;
-  struct __pthread_condimpl *__impl;
+  unsigned int __wrefs;
   void *__data;
 };
 
 /* Initializer for a condition variable.  */
 #define __PTHREAD_COND_INITIALIZER \
-  { __PTHREAD_SPIN_LOCK_INITIALIZER, NULL, NULL, NULL, NULL }
+  { __PTHREAD_SPIN_LOCK_INITIALIZER, NULL, NULL, 0, NULL }
 
 #endif /* bits/types/struct___pthread_cond.h */

sysdeps/htl/pt-cond-destroy.c

@@ -22,14 +22,26 @@
 int
 __pthread_cond_destroy (pthread_cond_t *cond)
 {
-  int ret = 0;
+  /* Set the wake request flag. */
+  unsigned int wrefs = atomic_fetch_or_acquire (&cond->__wrefs, 1);
 
   __pthread_spin_wait (&cond->__lock);
   if (cond->__queue)
-    ret = EBUSY;
+    {
+      __pthread_spin_unlock (&cond->__lock);
+      return EBUSY;
+    }
   __pthread_spin_unlock (&cond->__lock);
 
-  return ret;
+  while (wrefs >> 1 != 0)
+    {
+      gsync_wait (__mach_task_self (), (vm_offset_t) &cond->__wrefs, wrefs,
+		  0, 0, 0);
+      wrefs = atomic_load_acquire (&cond->__wrefs);
+    }
+  /* The memory the condvar occupies can now be reused.  */
+
+  return 0;
 }
 
 weak_alias (__pthread_cond_destroy, pthread_cond_destroy);

sysdeps/htl/pt-cond-timedwait.c

@@ -144,6 +144,10 @@ __pthread_cond_timedwait_internal (pthread_cond_t *cond,
   /* Release MUTEX before blocking.  */
   __pthread_mutex_unlock (mutex);
 
+  /* Increase the waiter reference count.  Relaxed MO is sufficient because
+     we only need to synchronize when decrementing the reference count.  */
+  atomic_fetch_add_relaxed (&cond->__wrefs, 2);
+
   /* Block the thread.  */
   if (abstime != NULL)
     err = __pthread_timedblock (self, abstime, clock_id);
@@ -178,6 +182,13 @@ __pthread_cond_timedwait_internal (pthread_cond_t *cond,
     }
   __pthread_spin_unlock (&cond->__lock);
 
+  /* If destruction is pending (i.e., the wake-request flag is nonzero) and we
+     are the last waiter (prior value of __wrefs was 1 << 1), then wake any
+     threads waiting in pthread_cond_destroy.  Release MO to synchronize with
+     these threads.  Don't bother clearing the wake-up request flag.  */
+  if ((atomic_fetch_add_release (&cond->__wrefs, -2)) == 3)
+    __gsync_wake (__mach_task_self (), (vm_offset_t) &cond->__wrefs, 0, 0);
+
   if (drain)
     __pthread_block (self);
 

sysdeps/mach/hurd/htl/pt-hurd-cond-timedwait.c

@@ -111,6 +111,10 @@ __pthread_hurd_cond_timedwait_internal (pthread_cond_t *cond,
       /* Release MUTEX before blocking.  */
       __pthread_mutex_unlock (mutex);
 
+  /* Increase the waiter reference count.  Relaxed MO is sufficient because
+     we only need to synchronize when decrementing the reference count.  */
+  atomic_fetch_add_relaxed (&cond->__wrefs, 2);
+
       /* Block the thread.  */
       if (abstime != NULL)
 	err = __pthread_timedblock (self, abstime, clock_id);
@@ -144,6 +148,13 @@ __pthread_hurd_cond_timedwait_internal (pthread_cond_t *cond,
 	__pthread_block (self);
     }
 
+  /* If destruction is pending (i.e., the wake-request flag is nonzero) and we
+     are the last waiter (prior value of __wrefs was 1 << 1), then wake any
+     threads waiting in pthread_cond_destroy.  Release MO to synchronize with
+     these threads.  Don't bother clearing the wake-up request flag.  */
+  if ((atomic_fetch_add_release (&cond->__wrefs, -2)) == 3)
+    __gsync_wake (__mach_task_self (), (vm_offset_t) &cond->__wrefs, 0, 0);
+
   /* Clear the hook, now that we are done blocking.  */
   ss->cancel_hook = NULL;
   /* Check the cancellation flag; we might have unblocked due to

sysdeps/pthread/Makefile

@@ -49,7 +49,7 @@ tests += tst-cnd-basic tst-mtx-trylock tst-cnd-broadcast \
 	 tst-cond1 tst-cond2 tst-cond3 tst-cond4 tst-cond5 tst-cond6 tst-cond7 \
 	 tst-cond8 tst-cond9 tst-cond10 tst-cond11 tst-cond12 tst-cond13 \
 	 tst-cond14 tst-cond15 tst-cond16 tst-cond17 tst-cond18 tst-cond19 \
-	 tst-cond23 tst-cond24 tst-cond25 tst-cond27 \
+	 tst-cond20 tst-cond21 tst-cond23 tst-cond24 tst-cond25 tst-cond27 \
 	 tst-cond-except \
 	 tst-join1 tst-join2 tst-join3 tst-join4 tst-join5 tst-join6 tst-join7 \
 	 tst-join8 tst-join9 tst-join10 tst-join11 tst-join12 tst-join13 \