Merge tag 'io_uring-6.13-20241220' of git://git.kernel.dk/linux

torvalds · torvalds · commit 7c05bd92305d · 2024-12-20T13:32:43.000-08:00
Pull io_uring fixes from Jens Axboe:

 - Fix for a file ref leak for registered ring fds

 - Turn the -&gt;timeout_lock into a raw spinlock, as it nests under the
   io-wq lock which is a raw spinlock as it's called from the scheduler
   side

 - Limit ring resizing to DEFER_TASKRUN for now. We will broaden this in
   the future, but for now, ensure that it's only feasible on rings with
   a single user

 - Add sanity check for io-wq enqueuing

* tag 'io_uring-6.13-20241220' of git://git.kernel.dk/linux:
  io_uring: check if iowq is killed before queuing
  io_uring/register: limit ring resizing to DEFER_TASKRUN
  io_uring: Fix registered ring file refcount leak
  io_uring: make ctx-&gt;timeout_lock a raw spinlock
diff --git a/include/linux/io_uring.h b/include/linux/io_uring.h
@@ -15,10 +15,8 @@ bool io_is_uring_fops(struct file *file);
 
 static inline void io_uring_files_cancel(void)
 {
-	if (current->io_uring) {
-		io_uring_unreg_ringfd();
+	if (current->io_uring)
 		__io_uring_cancel(false);
-	}
 }
 static inline void io_uring_task_cancel(void)
 {
diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
@@ -345,7 +345,7 @@ struct io_ring_ctx {
 
 	/* timeouts */
 	struct {
-		spinlock_t		timeout_lock;
+		raw_spinlock_t		timeout_lock;
 		struct list_head	timeout_list;
 		struct list_head	ltimeout_list;
 		unsigned		cq_last_tm_flush;
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
@@ -215,9 +215,9 @@ bool io_match_task_safe(struct io_kiocb *head, struct io_uring_task *tctx,
 		struct io_ring_ctx *ctx = head->ctx;
 
 		/* protect against races with linked timeouts */
-		spin_lock_irq(&ctx->timeout_lock);
+		raw_spin_lock_irq(&ctx->timeout_lock);
 		matched = io_match_linked(head);
-		spin_unlock_irq(&ctx->timeout_lock);
+		raw_spin_unlock_irq(&ctx->timeout_lock);
 	} else {
 		matched = io_match_linked(head);
 	}
@@ -333,7 +333,7 @@ static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)
 	init_waitqueue_head(&ctx->cq_wait);
 	init_waitqueue_head(&ctx->poll_wq);
 	spin_lock_init(&ctx->completion_lock);
-	spin_lock_init(&ctx->timeout_lock);
+	raw_spin_lock_init(&ctx->timeout_lock);
 	INIT_WQ_LIST(&ctx->iopoll_list);
 	INIT_LIST_HEAD(&ctx->io_buffers_comp);
 	INIT_LIST_HEAD(&ctx->defer_list);
@@ -498,10 +498,10 @@ static void io_prep_async_link(struct io_kiocb *req)
 	if (req->flags & REQ_F_LINK_TIMEOUT) {
 		struct io_ring_ctx *ctx = req->ctx;
 
-		spin_lock_irq(&ctx->timeout_lock);
+		raw_spin_lock_irq(&ctx->timeout_lock);
 		io_for_each_link(cur, req)
 			io_prep_async_work(cur);
-		spin_unlock_irq(&ctx->timeout_lock);
+		raw_spin_unlock_irq(&ctx->timeout_lock);
 	} else {
 		io_for_each_link(cur, req)
 			io_prep_async_work(cur);
@@ -514,7 +514,11 @@ static void io_queue_iowq(struct io_kiocb *req)
 	struct io_uring_task *tctx = req->tctx;
 
 	BUG_ON(!tctx);
-	BUG_ON(!tctx->io_wq);
+
+	if ((current->flags & PF_KTHREAD) || !tctx->io_wq) {
+		io_req_task_queue_fail(req, -ECANCELED);
+		return;
+	}
 
 	/* init ->work of the whole link before punting */
 	io_prep_async_link(req);
@@ -3214,6 +3218,7 @@ __cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd)
 
 void __io_uring_cancel(bool cancel_all)
 {
+	io_uring_unreg_ringfd();
 	io_uring_cancel_generic(cancel_all, NULL);
 }
 
diff --git a/io_uring/register.c b/io_uring/register.c
@@ -414,6 +414,9 @@ static int io_register_resize_rings(struct io_ring_ctx *ctx, void __user *arg)
 	if (ctx->flags & IORING_SETUP_SINGLE_ISSUER &&
 	    current != ctx->submitter_task)
 		return -EEXIST;
+	/* limited to DEFER_TASKRUN for now */
+	if (!(ctx->flags & IORING_SETUP_DEFER_TASKRUN))
+		return -EINVAL;
 	if (copy_from_user(&p, arg, sizeof(p)))
 		return -EFAULT;
 	if (p.flags & ~RESIZE_FLAGS)
diff --git a/io_uring/timeout.c b/io_uring/timeout.c
@@ -74,10 +74,10 @@ static void io_timeout_complete(struct io_kiocb *req, struct io_tw_state *ts)
 	if (!io_timeout_finish(timeout, data)) {
 		if (io_req_post_cqe(req, -ETIME, IORING_CQE_F_MORE)) {
 			/* re-arm timer */
-			spin_lock_irq(&ctx->timeout_lock);
+			raw_spin_lock_irq(&ctx->timeout_lock);
 			list_add(&timeout->list, ctx->timeout_list.prev);
 			hrtimer_start(&data->timer, timespec64_to_ktime(data->ts), data->mode);
-			spin_unlock_irq(&ctx->timeout_lock);
+			raw_spin_unlock_irq(&ctx->timeout_lock);
 			return;
 		}
 	}
@@ -109,7 +109,7 @@ __cold void io_flush_timeouts(struct io_ring_ctx *ctx)
 	u32 seq;
 	struct io_timeout *timeout, *tmp;
 
-	spin_lock_irq(&ctx->timeout_lock);
+	raw_spin_lock_irq(&ctx->timeout_lock);
 	seq = ctx->cached_cq_tail - atomic_read(&ctx->cq_timeouts);
 
 	list_for_each_entry_safe(timeout, tmp, &ctx->timeout_list, list) {
@@ -134,7 +134,7 @@ __cold void io_flush_timeouts(struct io_ring_ctx *ctx)
 		io_kill_timeout(req, 0);
 	}
 	ctx->cq_last_tm_flush = seq;
-	spin_unlock_irq(&ctx->timeout_lock);
+	raw_spin_unlock_irq(&ctx->timeout_lock);
 }
 
 static void io_req_tw_fail_links(struct io_kiocb *link, struct io_tw_state *ts)
@@ -200,9 +200,9 @@ void io_disarm_next(struct io_kiocb *req)
 	} else if (req->flags & REQ_F_LINK_TIMEOUT) {
 		struct io_ring_ctx *ctx = req->ctx;
 
-		spin_lock_irq(&ctx->timeout_lock);
+		raw_spin_lock_irq(&ctx->timeout_lock);
 		link = io_disarm_linked_timeout(req);
-		spin_unlock_irq(&ctx->timeout_lock);
+		raw_spin_unlock_irq(&ctx->timeout_lock);
 		if (link)
 			io_req_queue_tw_complete(link, -ECANCELED);
 	}
@@ -238,11 +238,11 @@ static enum hrtimer_restart io_timeout_fn(struct hrtimer *timer)
 	struct io_ring_ctx *ctx = req->ctx;
 	unsigned long flags;
 
-	spin_lock_irqsave(&ctx->timeout_lock, flags);
+	raw_spin_lock_irqsave(&ctx->timeout_lock, flags);
 	list_del_init(&timeout->list);
 	atomic_set(&req->ctx->cq_timeouts,
 		atomic_read(&req->ctx->cq_timeouts) + 1);
-	spin_unlock_irqrestore(&ctx->timeout_lock, flags);
+	raw_spin_unlock_irqrestore(&ctx->timeout_lock, flags);
 
 	if (!(data->flags & IORING_TIMEOUT_ETIME_SUCCESS))
 		req_set_fail(req);
@@ -285,9 +285,9 @@ int io_timeout_cancel(struct io_ring_ctx *ctx, struct io_cancel_data *cd)
 {
 	struct io_kiocb *req;
 
-	spin_lock_irq(&ctx->timeout_lock);
+	raw_spin_lock_irq(&ctx->timeout_lock);
 	req = io_timeout_extract(ctx, cd);
-	spin_unlock_irq(&ctx->timeout_lock);
+	raw_spin_unlock_irq(&ctx->timeout_lock);
 
 	if (IS_ERR(req))
 		return PTR_ERR(req);
@@ -330,7 +330,7 @@ static enum hrtimer_restart io_link_timeout_fn(struct hrtimer *timer)
 	struct io_ring_ctx *ctx = req->ctx;
 	unsigned long flags;
 
-	spin_lock_irqsave(&ctx->timeout_lock, flags);
+	raw_spin_lock_irqsave(&ctx->timeout_lock, flags);
 	prev = timeout->head;
 	timeout->head = NULL;
 
@@ -345,7 +345,7 @@ static enum hrtimer_restart io_link_timeout_fn(struct hrtimer *timer)
 	}
 	list_del(&timeout->list);
 	timeout->prev = prev;
-	spin_unlock_irqrestore(&ctx->timeout_lock, flags);
+	raw_spin_unlock_irqrestore(&ctx->timeout_lock, flags);
 
 	req->io_task_work.func = io_req_task_link_timeout;
 	io_req_task_work_add(req);
@@ -472,12 +472,12 @@ int io_timeout_remove(struct io_kiocb *req, unsigned int issue_flags)
 	} else {
 		enum hrtimer_mode mode = io_translate_timeout_mode(tr->flags);
 
-		spin_lock_irq(&ctx->timeout_lock);
+		raw_spin_lock_irq(&ctx->timeout_lock);
 		if (tr->ltimeout)
 			ret = io_linked_timeout_update(ctx, tr->addr, &tr->ts, mode);
 		else
 			ret = io_timeout_update(ctx, tr->addr, &tr->ts, mode);
-		spin_unlock_irq(&ctx->timeout_lock);
+		raw_spin_unlock_irq(&ctx->timeout_lock);
 	}
 
 	if (ret < 0)
@@ -572,7 +572,7 @@ int io_timeout(struct io_kiocb *req, unsigned int issue_flags)
 	struct list_head *entry;
 	u32 tail, off = timeout->off;
 
-	spin_lock_irq(&ctx->timeout_lock);
+	raw_spin_lock_irq(&ctx->timeout_lock);
 
 	/*
 	 * sqe->off holds how many events that need to occur for this
@@ -611,7 +611,7 @@ int io_timeout(struct io_kiocb *req, unsigned int issue_flags)
 	list_add(&timeout->list, entry);
 	data->timer.function = io_timeout_fn;
 	hrtimer_start(&data->timer, timespec64_to_ktime(data->ts), data->mode);
-	spin_unlock_irq(&ctx->timeout_lock);
+	raw_spin_unlock_irq(&ctx->timeout_lock);
 	return IOU_ISSUE_SKIP_COMPLETE;
 }
 
@@ -620,7 +620,7 @@ void io_queue_linked_timeout(struct io_kiocb *req)
 	struct io_timeout *timeout = io_kiocb_to_cmd(req, struct io_timeout);
 	struct io_ring_ctx *ctx = req->ctx;
 
-	spin_lock_irq(&ctx->timeout_lock);
+	raw_spin_lock_irq(&ctx->timeout_lock);
 	/*
 	 * If the back reference is NULL, then our linked request finished
 	 * before we got a chance to setup the timer
@@ -633,7 +633,7 @@ void io_queue_linked_timeout(struct io_kiocb *req)
 				data->mode);
 		list_add_tail(&timeout->list, &ctx->ltimeout_list);
 	}
-	spin_unlock_irq(&ctx->timeout_lock);
+	raw_spin_unlock_irq(&ctx->timeout_lock);
 	/* drop submission reference */
 	io_put_req(req);
 }
@@ -668,15 +668,15 @@ __cold bool io_kill_timeouts(struct io_ring_ctx *ctx, struct io_uring_task *tctx
 	 * timeout_lockfirst to keep locking ordering.
 	 */
 	spin_lock(&ctx->completion_lock);
-	spin_lock_irq(&ctx->timeout_lock);
+	raw_spin_lock_irq(&ctx->timeout_lock);
 	list_for_each_entry_safe(timeout, tmp, &ctx->timeout_list, list) {
 		struct io_kiocb *req = cmd_to_io_kiocb(timeout);
 
 		if (io_match_task(req, tctx, cancel_all) &&
 		    io_kill_timeout(req, -ECANCELED))
 			canceled++;
 	}
-	spin_unlock_irq(&ctx->timeout_lock);
+	raw_spin_unlock_irq(&ctx->timeout_lock);
 	spin_unlock(&ctx->completion_lock);
 	return canceled != 0;
 }

Original file line number	Diff line number	Diff line change
`@@ -15,10 +15,8 @@ bool io_is_uring_fops(struct file *file);`
`15`	`15`
`16`	`16`	`static inline void io_uring_files_cancel(void)`
`17`	`17`	`{`
`18`		`- if (current->io_uring) {`
`19`		`- io_uring_unreg_ringfd();`
	`18`	`+ if (current->io_uring)`
`20`	`19`	`__io_uring_cancel(false);`
`21`		`- }`
`22`	`20`	`}`
`23`	`21`	`static inline void io_uring_task_cancel(void)`
`24`	`22`	`{`