Merge tag 'io_uring-5.15-2021-09-11' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:

 - Fix an off-by-one in a BUILD_BUG_ON() check. Not a real issue right
   now as we have plenty of flags left, but could become one. (Hao)

 - Fix lockdep issue introduced in this merge window (me)

 - Fix a few issues with the worker creation (me, Pavel, Qiang)

 - Fix regression with wq_has_sleeper() for IOPOLL (Pavel)

 - Timeout link error propagation fix (Pavel)

* tag 'io_uring-5.15-2021-09-11' of git://git.kernel.dk/linux-block:
  io_uring: fix off-by-one in BUILD_BUG_ON check of __REQ_F_LAST_BIT
  io_uring: fail links of cancelled timeouts
  io-wq: fix memory leak in create_io_worker()
  io-wq: fix silly logic error in io_task_work_match()
  io_uring: drop ctx->uring_lock before acquiring sqd->lock
  io_uring: fix missing mb() before waitqueue_active
  io-wq: fix cancellation on create-worker failure

fs/io-wq.c

@@ -709,6 +709,7 @@ static void create_worker_cont(struct callback_head *cb)
 		}
 		raw_spin_unlock(&wqe->lock);
 		io_worker_ref_put(wqe->wq);
+		kfree(worker);
 		return;
 	}
 
@@ -725,6 +726,7 @@ static void io_workqueue_create(struct work_struct *work)
 	if (!io_queue_worker_create(worker, acct, create_worker_cont)) {
 		clear_bit_unlock(0, &worker->create_state);
 		io_worker_release(worker);
+		kfree(worker);
 	}
 }
 
@@ -759,6 +761,7 @@ static bool create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index)
 	if (!IS_ERR(tsk)) {
 		io_init_new_worker(wqe, worker, tsk);
 	} else if (!io_should_retry_thread(PTR_ERR(tsk))) {
+		kfree(worker);
 		goto fail;
 	} else {
 		INIT_WORK(&worker->work, io_workqueue_create);
@@ -832,6 +835,11 @@ static void io_wqe_insert_work(struct io_wqe *wqe, struct io_wq_work *work)
 	wq_list_add_after(&work->list, &tail->list, &acct->work_list);
 }
 
+static bool io_wq_work_match_item(struct io_wq_work *work, void *data)
+{
+	return work == data;
+}
+
 static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work)
 {
 	struct io_wqe_acct *acct = io_work_get_acct(wqe, work);
@@ -844,7 +852,6 @@ static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work)
 	 */
 	if (test_bit(IO_WQ_BIT_EXIT, &wqe->wq->state) ||
 	    (work->flags & IO_WQ_WORK_CANCEL)) {
-run_cancel:
 		io_run_cancel(work, wqe);
 		return;
 	}
@@ -864,15 +871,22 @@ static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work)
 		bool did_create;
 
 		did_create = io_wqe_create_worker(wqe, acct);
-		if (unlikely(!did_create)) {
-			raw_spin_lock(&wqe->lock);
-			/* fatal condition, failed to create the first worker */
-			if (!acct->nr_workers) {
-				raw_spin_unlock(&wqe->lock);
-				goto run_cancel;
-			}
-			raw_spin_unlock(&wqe->lock);
+		if (likely(did_create))
+			return;
+
+		raw_spin_lock(&wqe->lock);
+		/* fatal condition, failed to create the first worker */
+		if (!acct->nr_workers) {
+			struct io_cb_cancel_data match = {
+				.fn		= io_wq_work_match_item,
+				.data		= work,
+				.cancel_all	= false,
+			};
+
+			if (io_acct_cancel_pending_work(wqe, acct, &match))
+				raw_spin_lock(&wqe->lock);
 		}
+		raw_spin_unlock(&wqe->lock);
 	}
 }
 
@@ -1122,7 +1136,7 @@ static bool io_task_work_match(struct callback_head *cb, void *data)
 {
 	struct io_worker *worker;
 
-	if (cb->func != create_worker_cb || cb->func != create_worker_cont)
+	if (cb->func != create_worker_cb && cb->func != create_worker_cont)
 		return false;
 	worker = container_of(cb, struct io_worker, create_work);
 	return worker->wqe->wq == data;
@@ -1143,9 +1157,14 @@ static void io_wq_exit_workers(struct io_wq *wq)
 
 	while ((cb = task_work_cancel_match(wq->task, io_task_work_match, wq)) != NULL) {
 		struct io_worker *worker;
+		struct io_wqe_acct *acct;
 
 		worker = container_of(cb, struct io_worker, create_work);
-		atomic_dec(&worker->wqe->acct[worker->create_index].nr_running);
+		acct = io_wqe_get_acct(worker);
+		atomic_dec(&acct->nr_running);
+		raw_spin_lock(&worker->wqe->lock);
+		acct->nr_workers--;
+		raw_spin_unlock(&worker->wqe->lock);
 		io_worker_ref_put(wq);
 		clear_bit_unlock(0, &worker->create_state);
 		io_worker_release(worker);

fs/io_uring.c

@@ -1482,6 +1482,8 @@ static void io_kill_timeout(struct io_kiocb *req, int status)
 	struct io_timeout_data *io = req->async_data;
 
 	if (hrtimer_try_to_cancel(&io->timer) != -1) {
+		if (status)
+			req_set_fail(req);
 		atomic_set(&req->ctx->cq_timeouts,
 			atomic_read(&req->ctx->cq_timeouts) + 1);
 		list_del_init(&req->timeout.list);
@@ -1619,8 +1621,11 @@ static void io_cqring_ev_posted(struct io_ring_ctx *ctx)
 
 static void io_cqring_ev_posted_iopoll(struct io_ring_ctx *ctx)
 {
+	/* see waitqueue_active() comment */
+	smp_mb();
+
 	if (ctx->flags & IORING_SETUP_SQPOLL) {
-		if (wq_has_sleeper(&ctx->cq_wait))
+		if (waitqueue_active(&ctx->cq_wait))
 			wake_up_all(&ctx->cq_wait);
 	}
 	if (io_should_trigger_evfd(ctx))
@@ -10550,7 +10555,14 @@ static int io_register_iowq_max_workers(struct io_ring_ctx *ctx,
 	if (ctx->flags & IORING_SETUP_SQPOLL) {
 		sqd = ctx->sq_data;
 		if (sqd) {
+			/*
+			 * Observe the correct sqd->lock -> ctx->uring_lock
+			 * ordering. Fine to drop uring_lock here, we hold
+			 * a ref to the ctx.
+			 */
+			mutex_unlock(&ctx->uring_lock);
 			mutex_lock(&sqd->lock);
+			mutex_lock(&ctx->uring_lock);
 			tctx = sqd->thread->io_uring;
 		}
 	} else {
@@ -10853,7 +10865,7 @@ static int __init io_uring_init(void)
 	BUILD_BUG_ON(SQE_VALID_FLAGS >= (1 << 8));
 
 	BUILD_BUG_ON(ARRAY_SIZE(io_op_defs) != IORING_OP_LAST);
-	BUILD_BUG_ON(__REQ_F_LAST_BIT >= 8 * sizeof(int));
+	BUILD_BUG_ON(__REQ_F_LAST_BIT > 8 * sizeof(int));
 
 	req_cachep = KMEM_CACHE(io_kiocb, SLAB_HWCACHE_ALIGN | SLAB_PANIC |
 				SLAB_ACCOUNT);