[PATCH 02/33] io-wq: have manager wait for all workers to exit

[Jens Axboe <axboe@kernel.dk>]

Instead of having to wait separately on workers and manager, just have
the manager wait on the workers. We use an atomic_t for the reference
here, as we need to start at 0 and allow increment from that. Since the
number of workers is naturally capped by the allowed nr of processes,
and that uses an int, there is no risk of overflow.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 fs/io-wq.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)

diff --git a/fs/io-wq.c b/fs/io-wq.c
index 965022fe9961..a61f867f98f4 100644
--- a/fs/io-wq.c
+++ b/fs/io-wq.c
@@ -120,6 +120,9 @@ struct io_wq {
 	refcount_t refs;
 	struct completion done;
 
+	atomic_t worker_refs;
+	struct completion worker_done;
+
 	struct hlist_node cpuhp_node;
 
 	pid_t task_pid;
@@ -189,7 +192,8 @@ static void io_worker_exit(struct io_worker *worker)
 	raw_spin_unlock_irq(&wqe->lock);
 
 	kfree_rcu(worker, rcu);
-	io_wq_put(wqe->wq);
+	if (atomic_dec_and_test(&wqe->wq->worker_refs))
+		complete(&wqe->wq->worker_done);
 }
 
 static inline bool io_wqe_run_queue(struct io_wqe *wqe)
@@ -648,14 +652,15 @@ static bool create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index)
 	init_completion(&worker->ref_done);
 	init_completion(&worker->started);
 
-	refcount_inc(&wq->refs);
+	atomic_inc(&wq->worker_refs);
 
 	if (index == IO_WQ_ACCT_BOUND)
 		pid = io_wq_fork_thread(task_thread_bound, worker);
 	else
 		pid = io_wq_fork_thread(task_thread_unbound, worker);
 	if (pid < 0) {
-		io_wq_put(wq);
+		if (atomic_dec_and_test(&wq->worker_refs))
+			complete(&wq->worker_done);
 		kfree(worker);
 		return false;
 	}
@@ -753,6 +758,9 @@ static int io_wq_manager(void *data)
 	} while (!test_bit(IO_WQ_BIT_EXIT, &wq->state));
 
 	io_wq_check_workers(wq);
+	/* we might not ever have created any workers */
+	if (atomic_read(&wq->worker_refs))
+		wait_for_completion(&wq->worker_done);
 	wq->manager = NULL;
 	io_wq_put(wq);
 	do_exit(0);
@@ -796,6 +804,7 @@ static int io_wq_fork_manager(struct io_wq *wq)
 	if (wq->manager)
 		return 0;
 
+	reinit_completion(&wq->worker_done);
 	clear_bit(IO_WQ_BIT_EXIT, &wq->state);
 	refcount_inc(&wq->refs);
 	current->flags |= PF_IO_WORKER;
@@ -1050,6 +1059,9 @@ struct io_wq *io_wq_create(unsigned bounded, struct io_wq_data *data)
 	init_completion(&wq->done);
 	refcount_set(&wq->refs, 1);
 
+	init_completion(&wq->worker_done);
+	atomic_set(&wq->worker_refs, 0);
+
 	ret = io_wq_fork_manager(wq);
 	if (!ret)
 		return wq;
-- 
2.30.1