From: Jens Axboe <axboe@kernel.dk>
Date: Mon, 15 Feb 2021 21:35:43 +0000 (-0700)
Subject: io-wq: fix worker creation
X-Git-Url: https://git.kernel.dk/?a=commitdiff_plain;h=refs%2Fheads%2Fio_uring-worker;p=linux-2.6-block.git

io-wq: fix worker creation

Don't fork a new worker under preempt disable and task being non-runnable.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---

diff --git a/fs/io-wq.c b/fs/io-wq.c
index 78fa38d3b9c9..66da0fc6bc5a 100644
--- a/fs/io-wq.c
+++ b/fs/io-wq.c
@@ -218,7 +218,7 @@ static bool io_wqe_activate_free_worker(struct io_wqe *wqe)
  * We need a worker. If we find a free one, we're good. If not, attempt to
  * create a new one.
  */
-static void io_wqe_wake_worker(struct io_wqe *wqe, struct io_wqe_acct *acct)
+static bool io_wqe_wake_worker(struct io_wqe *wqe, struct io_wqe_acct *acct)
 {
 	bool ret;
 
@@ -233,7 +233,9 @@ static void io_wqe_wake_worker(struct io_wqe *wqe, struct io_wqe_acct *acct)
 	rcu_read_unlock();
 
 	if (!ret && acct->nr_workers < acct->max_workers)
-		create_io_worker(wqe, acct);
+		return false;
+
+	return true;
 }
 
 static void io_wqe_inc_running(struct io_wqe *wqe, struct io_worker *worker)
@@ -243,13 +245,15 @@ static void io_wqe_inc_running(struct io_wqe *wqe, struct io_worker *worker)
 	atomic_inc(&acct->nr_running);
 }
 
-static void io_wqe_dec_running(struct io_wqe *wqe, struct io_worker *worker)
+static bool io_wqe_dec_running(struct io_wqe *wqe, struct io_worker *worker)
 	__must_hold(wqe->lock)
 {
 	struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker);
 
 	if (atomic_dec_and_test(&acct->nr_running) && io_wqe_run_queue(wqe))
-		io_wqe_wake_worker(wqe, acct);
+		return io_wqe_wake_worker(wqe, acct);
+
+	return true;
 }
 
 static void io_worker_start(struct io_wqe *wqe, struct io_worker *worker)
@@ -590,6 +594,7 @@ void io_wq_worker_sleeping(struct task_struct *tsk)
 {
 	struct io_worker *worker = tsk->pf_io_worker;
 	struct io_wqe *wqe = worker->wqe;
+	bool did_wake;
 
 	if (!(worker->flags & IO_WORKER_F_UP))
 		return;
@@ -599,8 +604,25 @@ void io_wq_worker_sleeping(struct task_struct *tsk)
 	worker->flags &= ~IO_WORKER_F_RUNNING;
 
 	raw_spin_lock_irq(&wqe->lock);
-	io_wqe_dec_running(wqe, worker);
+	did_wake = io_wqe_dec_running(wqe, worker);
 	raw_spin_unlock_irq(&wqe->lock);
+
+	if (!did_wake) {
+		struct io_wqe_acct *acct;
+		long state = current->state;
+
+		if (worker->flags & IO_WORKER_F_BOUND)
+			acct = &wqe->acct[IO_WQ_ACCT_BOUND];
+		else
+			acct = &wqe->acct[IO_WQ_ACCT_UNBOUND];
+
+		/*
+		 * EEK
+		 */
+		__set_current_state(TASK_RUNNING);
+		create_io_worker(wqe, acct);
+		current->state = state;
+	}
 }
 
 static inline bool io_wqe_need_worker(struct io_wqe *wqe, int index)
@@ -723,8 +745,12 @@ static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work)
 	raw_spin_unlock_irqrestore(&wqe->lock, flags);
 
 	if ((work_flags & IO_WQ_WORK_CONCURRENT) ||
-	    !atomic_read(&acct->nr_running))
-		io_wqe_wake_worker(wqe, acct);
+	    !atomic_read(&acct->nr_running)) {
+		bool did_wake = io_wqe_wake_worker(wqe, acct);
+
+		if (!did_wake)
+			create_io_worker(wqe, acct);
+	}
 }
 
 void io_wq_enqueue(struct io_wq *wq, struct io_wq_work *work)
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index ff74fca39ed2..a34235495ffa 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -5118,12 +5118,13 @@ static inline void sched_submit_work(struct task_struct *tsk)
 	 * requires it.
 	 */
 	if (task_flags & (PF_WQ_WORKER | PF_IO_WORKER)) {
-		preempt_disable();
-		if (task_flags & PF_WQ_WORKER)
+		if (task_flags & PF_WQ_WORKER) {
+			preempt_disable();
 			wq_worker_sleeping(tsk);
-		else
+			preempt_enable_no_resched();
+		} else {
 			io_wq_worker_sleeping(tsk);
-		preempt_enable_no_resched();
+		}
 	}
 
 	if (tsk_is_pi_blocked(tsk))