process-shared barrier support, based on discussion with bdonlan

this implementation is rather heavy-weight, but it's the first
solution i've found that's actually correct. all waiters actually wait
twice at the barrier so that they can synchronize exit, and they hold
a "vm lock" that prevents changes to virtual memory mappings (and
blocks pthread_barrier_destroy) until all waiters are finished
inspecting the barrier.

thus, it is safe for any thread to destroy and/or unmap the barrier's
memory as soon as pthread_barrier_wait returns, without further
synchronization.

1 files changed, 6 insertions, 0 deletions

diff --git a/src/thread/pthread_barrier_destroy.c b/src/thread/pthread_barrier_destroy.c
index 2898c41a..f82664d5 100644
--- a/src/thread/pthread_barrier_destroy.c
+++ b/src/thread/pthread_barrier_destroy.c
@@ -1,6 +1,12 @@
 #include "pthread_impl.h"
 
+void __vm_lock(int), __vm_unlock(void);
+
 int pthread_barrier_destroy(pthread_barrier_t *b)
 {
+	if (b->_b_limit < 0) {
+		__vm_lock(-1);
+		__vm_unlock();
+	}
 	return 0;
 }