semaphore: unblock waiters when the front waiter cancels

When `Release`, if the remaining tokens are not enough for the front waiter, no waiters will be notified.

So if the canceled waiter is the front one, it should try to notify following waiters if there are remaining tokens.

I found this bug when implementing a cancelable rwmutex based on semaphore:

https://github.com/zhiqiangxu/util/blob/master/mutex/crwmutex.go

This bug can be verified by this test:

https://github.com/zhiqiangxu/util/blob/master/mutex/crwmutex_test.go#L43

Change-Id: Id8564976bd375a82c4fbc6cb08b0bb83118a346c
GitHub-Last-Rev: 29b6ff26bf779d23239cfe64a395378a9e41d1fc
GitHub-Pull-Request: golang/sync#10
Reviewed-on: https://go-review.googlesource.com/c/sync/+/223418
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>

semaphore/semaphore.go

@@ -67,7 +67,12 @@ func (s *Weighted) Acquire(ctx context.Context, n int64) error {
 			// fix up the queue, just pretend we didn't notice the cancelation.
 			err = nil
 		default:
+			isFront := s.waiters.Front() == elem
 			s.waiters.Remove(elem)
+			// If we're at the front and there're extra tokens left, notify other waiters.
+			if isFront && s.size > s.cur {
+				s.notifyWaiters()
+			}
 		}
 		s.mu.Unlock()
 		return err
@@ -97,6 +102,11 @@ func (s *Weighted) Release(n int64) {
 		s.mu.Unlock()
 		panic("semaphore: released more than held")
 	}
+	s.notifyWaiters()
+	s.mu.Unlock()
+}
+
+func (s *Weighted) notifyWaiters() {
 	for {
 		next := s.waiters.Front()
 		if next == nil {
@@ -123,5 +133,4 @@ func (s *Weighted) Release(n int64) {
 		s.waiters.Remove(next)
 		close(w.ready)
 	}
-	s.mu.Unlock()
 }

semaphore/semaphore_test.go

@@ -169,3 +169,34 @@ func TestLargeAcquireDoesntStarve(t *testing.T) {
 	sem.Release(n)
 	wg.Wait()
 }
+
+// translated from https://github.com/zhiqiangxu/util/blob/master/mutex/crwmutex_test.go#L43
+func TestAllocCancelDoesntStarve(t *testing.T) {
+	sem := semaphore.NewWeighted(10)
+
+	// Block off a portion of the semaphore so that Acquire(_, 10) can eventually succeed.
+	sem.Acquire(context.Background(), 1)
+
+	// In the background, Acquire(_, 10).
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	go func() {
+		sem.Acquire(ctx, 10)
+	}()
+
+	// Wait until the Acquire(_, 10) call blocks.
+	for sem.TryAcquire(1) {
+		sem.Release(1)
+		runtime.Gosched()
+	}
+
+	// Now try to grab a read lock, and simultaneously unblock the Acquire(_, 10) call.
+	// Both Acquire calls should unblock and return, in either order.
+	go cancel()
+
+	err := sem.Acquire(context.Background(), 1)
+	if err != nil {
+		t.Fatalf("Acquire(_, 1) failed unexpectedly: %v", err)
+	}
+	sem.Release(1)
+}