Merge tag 'io_uring-5.14-2021-08-07' of git://git.kernel.dk/linux-block

Pull io_uring from Jens Axboe:
 "A few io-wq related fixes:

   - Fix potential nr_worker race and missing max_workers check from one
     path (Hao)

   - Fix race between worker exiting and new work queue (me)"

* tag 'io_uring-5.14-2021-08-07' of git://git.kernel.dk/linux-block:
  io-wq: fix lack of acct->nr_workers < acct->max_workers judgement
  io-wq: fix no lock protection of acct->nr_worker
  io-wq: fix race between worker exiting and activating free worker

fs/io-wq.c

@@ -130,6 +130,7 @@ struct io_cb_cancel_data {
 };
 
 static void create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index);
+static void io_wqe_dec_running(struct io_worker *worker);
 
 static bool io_worker_get(struct io_worker *worker)
 {
@@ -168,26 +169,21 @@ static void io_worker_exit(struct io_worker *worker)
 {
 	struct io_wqe *wqe = worker->wqe;
 	struct io_wqe_acct *acct = io_wqe_get_acct(worker);
-	unsigned flags;
 
 	if (refcount_dec_and_test(&worker->ref))
 		complete(&worker->ref_done);
 	wait_for_completion(&worker->ref_done);
 
-	preempt_disable();
-	current->flags &= ~PF_IO_WORKER;
-	flags = worker->flags;
-	worker->flags = 0;
-	if (flags & IO_WORKER_F_RUNNING)
-		atomic_dec(&acct->nr_running);
-	worker->flags = 0;
-	preempt_enable();
-
 	raw_spin_lock_irq(&wqe->lock);
-	if (flags & IO_WORKER_F_FREE)
+	if (worker->flags & IO_WORKER_F_FREE)
 		hlist_nulls_del_rcu(&worker->nulls_node);
 	list_del_rcu(&worker->all_list);
 	acct->nr_workers--;
+	preempt_disable();
+	io_wqe_dec_running(worker);
+	worker->flags = 0;
+	current->flags &= ~PF_IO_WORKER;
+	preempt_enable();
 	raw_spin_unlock_irq(&wqe->lock);
 
 	kfree_rcu(worker, rcu);
@@ -214,16 +210,20 @@ static bool io_wqe_activate_free_worker(struct io_wqe *wqe)
 	struct hlist_nulls_node *n;
 	struct io_worker *worker;
 
-	n = rcu_dereference(hlist_nulls_first_rcu(&wqe->free_list));
-	if (is_a_nulls(n))
-		return false;
-
-	worker = hlist_nulls_entry(n, struct io_worker, nulls_node);
-	if (io_worker_get(worker)) {
-		wake_up_process(worker->task);
+	/*
+	 * Iterate free_list and see if we can find an idle worker to
+	 * activate. If a given worker is on the free_list but in the process
+	 * of exiting, keep trying.
+	 */
+	hlist_nulls_for_each_entry_rcu(worker, n, &wqe->free_list, nulls_node) {
+		if (!io_worker_get(worker))
+			continue;
+		if (wake_up_process(worker->task)) {
 			io_worker_release(worker);
 			return true;
 		}
+		io_worker_release(worker);
+	}
 
 	return false;
 }
@@ -247,9 +247,18 @@ static void io_wqe_wake_worker(struct io_wqe *wqe, struct io_wqe_acct *acct)
 	ret = io_wqe_activate_free_worker(wqe);
 	rcu_read_unlock();
 
-	if (!ret && acct->nr_workers < acct->max_workers) {
+	if (!ret) {
+		bool do_create = false;
+
+		raw_spin_lock_irq(&wqe->lock);
+		if (acct->nr_workers < acct->max_workers) {
 			atomic_inc(&acct->nr_running);
 			atomic_inc(&wqe->wq->worker_refs);
+			acct->nr_workers++;
+			do_create = true;
+		}
+		raw_spin_unlock_irq(&wqe->lock);
+		if (do_create)
 			create_io_worker(wqe->wq, wqe, acct->index);
 	}
 }
@@ -271,9 +280,17 @@ static void create_worker_cb(struct callback_head *cb)
 {
 	struct create_worker_data *cwd;
 	struct io_wq *wq;
+	struct io_wqe *wqe;
+	struct io_wqe_acct *acct;
 
 	cwd = container_of(cb, struct create_worker_data, work);
-	wq = cwd->wqe->wq;
+	wqe = cwd->wqe;
+	wq = wqe->wq;
+	acct = &wqe->acct[cwd->index];
+	raw_spin_lock_irq(&wqe->lock);
+	if (acct->nr_workers < acct->max_workers)
+		acct->nr_workers++;
+	raw_spin_unlock_irq(&wqe->lock);
 	create_io_worker(wq, cwd->wqe, cwd->index);
 	kfree(cwd);
 }
@@ -635,6 +652,9 @@ static void create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index)
 		kfree(worker);
 fail:
 		atomic_dec(&acct->nr_running);
+		raw_spin_lock_irq(&wqe->lock);
+		acct->nr_workers--;
+		raw_spin_unlock_irq(&wqe->lock);
 		io_worker_ref_put(wq);
 		return;
 	}
@@ -650,9 +670,8 @@ static void create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index)
 	worker->flags |= IO_WORKER_F_FREE;
 	if (index == IO_WQ_ACCT_BOUND)
 		worker->flags |= IO_WORKER_F_BOUND;
-	if (!acct->nr_workers && (worker->flags & IO_WORKER_F_BOUND))
+	if ((acct->nr_workers == 1) && (worker->flags & IO_WORKER_F_BOUND))
 		worker->flags |= IO_WORKER_F_FIXED;
-	acct->nr_workers++;
 	raw_spin_unlock_irq(&wqe->lock);
 	wake_up_new_task(tsk);
 }