fuse: alloc pqueue before installing fch in fuse_dev

Prior to this patchset, fuse_dev (containing fuse_pqueue) was allocated on
mount.  But now fuse_dev is allocated when opening /dev/fuse, even though
the queues are not needed at that time.

Delay allocation of the pqueue (4k worth of list_head) just before mounting
or cloning a device.

Various distributions (e.g. Debian/Fedora) configure /dev/fuse as world
writable, so the pqueue allocation should be deferred to a privileged
operation (mount) to prevent unprivileged userspace from consuming pinned
kernel memory.

[Li Wang: fix kernel NULL pointer dereference in fuse_uring_add_to_pq()]
[Fix race in fuse_dev_release()]

Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>

fs/fuse/dev.c

@@ -329,6 +329,7 @@ void fuse_chan_release(struct fuse_chan *fch)
 void fuse_chan_free(struct fuse_chan *fch)
 {
 	WARN_ON(!list_empty(&fch->devices));
+	kfree(fch->pq_prealloc);
 	kfree(fch);
 }
 EXPORT_SYMBOL_GPL(fuse_chan_free);
@@ -355,15 +356,30 @@ struct fuse_chan *fuse_chan_new(void)
 }
 EXPORT_SYMBOL_GPL(fuse_chan_new);
 
+struct list_head *fuse_pqueue_alloc(void)
+{
+	struct list_head *pq = kzalloc_objs(struct list_head, FUSE_PQ_HASH_SIZE);
+
+	if (pq) {
+		for (int i = 0; i < FUSE_PQ_HASH_SIZE; i++)
+			INIT_LIST_HEAD(&pq[i]);
+	}
+	return pq;
+}
+
 struct fuse_chan *fuse_dev_chan_new(void)
 {
-	struct fuse_chan *fch = fuse_chan_new();
+	struct fuse_chan *fch __free(kfree) = fuse_chan_new();
 	if (!fch)
 		return NULL;
 
+	fch->pq_prealloc = fuse_pqueue_alloc();
+	if (!fch->pq_prealloc)
+		return NULL;
+
 	fuse_iqueue_init(&fch->iq, &fuse_dev_fiq_ops, NULL);
 
-	return fch;
+	return no_free_ptr(fch);
 }
 EXPORT_SYMBOL_GPL(fuse_dev_chan_new);
 
@@ -404,39 +420,42 @@ void fuse_chan_io_uring_enable(struct fuse_chan *fch)
 
 void fuse_pqueue_init(struct fuse_pqueue *fpq)
 {
-	unsigned int i;
-
 	spin_lock_init(&fpq->lock);
-	for (i = 0; i < FUSE_PQ_HASH_SIZE; i++)
-		INIT_LIST_HEAD(&fpq->processing[i]);
 	INIT_LIST_HEAD(&fpq->io);
 	fpq->connected = 1;
+	fpq->processing = NULL;
 }
 
-struct fuse_dev *fuse_dev_alloc(void)
+static struct fuse_dev *fuse_dev_alloc_no_pq(void)
 {
 	struct fuse_dev *fud;
-	struct list_head *pq;
 
 	fud = kzalloc_obj(struct fuse_dev);
 	if (!fud)
 		return NULL;
 
 	refcount_set(&fud->ref, 1);
-	pq = kzalloc_objs(struct list_head, FUSE_PQ_HASH_SIZE);
-	if (!pq) {
-		kfree(fud);
-		return NULL;
-	}
-
-	fud->pq.processing = pq;
 	fuse_pqueue_init(&fud->pq);
 
 	return fud;
 }
+
+struct fuse_dev *fuse_dev_alloc(void)
+{
+	struct fuse_dev *fud __free(kfree) = fuse_dev_alloc_no_pq();
+	if (!fud)
+		return NULL;
+
+	fud->pq.processing = fuse_pqueue_alloc();
+	if (!fud->pq.processing)
+		return NULL;
+
+	return no_free_ptr(fud);
+}
 EXPORT_SYMBOL_GPL(fuse_dev_alloc);
 
-void fuse_dev_install(struct fuse_dev *fud, struct fuse_chan *fch)
+static void fuse_dev_install_with_pq(struct fuse_dev *fud, struct fuse_chan *fch,
+				     struct list_head *pq)
 {
 	struct fuse_chan *old_fch;
 
@@ -454,20 +473,33 @@ void fuse_dev_install(struct fuse_dev *fud, struct fuse_chan *fch)
 		 *  - it was set to disconneted
 		 */
 		fch->connected = 0;
+		kfree(pq);
 	} else {
+		if (pq) {
+			WARN_ON(fud->pq.processing);
+			fud->pq.processing = pq;
+		}
 		list_add_tail(&fud->entry, &fch->devices);
 		fuse_conn_get(fch->conn);
 		wake_up_all(&fuse_dev_waitq);
 	}
 	spin_unlock(&fch->lock);
 }
+
+void fuse_dev_install(struct fuse_dev *fud, struct fuse_chan *fch)
+{
+	struct list_head *pq = fch->pq_prealloc;
+
+	fch->pq_prealloc = NULL;
+	fuse_dev_install_with_pq(fud, fch, pq);
+}
 EXPORT_SYMBOL_GPL(fuse_dev_install);
 
 struct fuse_dev *fuse_dev_alloc_install(struct fuse_chan *fch)
 {
 	struct fuse_dev *fud;
 
-	fud = fuse_dev_alloc();
+	fud = fuse_dev_alloc_no_pq();
 	if (!fud)
 		return NULL;
 
@@ -1631,7 +1663,7 @@ static ssize_t fuse_dev_do_read(struct fuse_dev *fud, struct file *file,
 
 static int fuse_dev_open(struct inode *inode, struct file *file)
 {
-	struct fuse_dev *fud = fuse_dev_alloc();
+	struct fuse_dev *fud = fuse_dev_alloc_no_pq();
 
 	if (!fud)
 		return -ENOMEM;
@@ -2204,20 +2236,21 @@ int fuse_dev_release(struct inode *inode, struct file *file)
 		unsigned int i;
 		bool last;
 
+		/* Make sure fuse_dev_install_with_pq() has finished */
+		spin_lock(&fch->lock);
 		spin_lock(&fpq->lock);
 		WARN_ON(!list_empty(&fpq->io));
 		for (i = 0; i < FUSE_PQ_HASH_SIZE; i++)
 			list_splice_init(&fpq->processing[i], &to_end);
 		spin_unlock(&fpq->lock);
 
-		fuse_dev_end_requests(&to_end);
-
-		spin_lock(&fch->lock);
 		list_del(&fud->entry);
 		/* Are we the last open device? */
 		last = list_empty(&fch->devices);
 		spin_unlock(&fch->lock);
 
+		fuse_dev_end_requests(&to_end);
+
 		if (last) {
 			WARN_ON(fch->iq.fasync != NULL);
 			fuse_chan_abort(fch, false);
@@ -2244,6 +2277,7 @@ static long fuse_dev_ioctl_clone(struct file *file, __u32 __user *argp)
 {
 	int oldfd;
 	struct fuse_dev *fud, *new_fud;
+	struct list_head *pq;
 
 	if (get_user(oldfd, argp))
 		return -EFAULT;
@@ -2267,7 +2301,11 @@ static long fuse_dev_ioctl_clone(struct file *file, __u32 __user *argp)
 	if (fuse_dev_chan_get(new_fud))
 		return -EINVAL;
 
-	fuse_dev_install(new_fud, fud->chan);
+	pq = fuse_pqueue_alloc();
+	if (!pq)
+		return -ENOMEM;
+
+	fuse_dev_install_with_pq(new_fud, fud->chan, pq);
 
 	return 0;
 }

fs/fuse/dev_uring.c

@@ -281,7 +281,7 @@ static struct fuse_ring_queue *fuse_uring_create_queue(struct fuse_ring *ring,
 	queue = kzalloc_obj(*queue, GFP_KERNEL_ACCOUNT);
 	if (!queue)
 		return NULL;
-	pq = kzalloc_objs(struct list_head, FUSE_PQ_HASH_SIZE);
+	pq = fuse_pqueue_alloc();
 	if (!pq) {
 		kfree(queue);
 		return NULL;
@@ -299,8 +299,8 @@ static struct fuse_ring_queue *fuse_uring_create_queue(struct fuse_ring *ring,
 	INIT_LIST_HEAD(&queue->fuse_req_bg_queue);
 	INIT_LIST_HEAD(&queue->ent_released);
 
-	queue->fpq.processing = pq;
 	fuse_pqueue_init(&queue->fpq);
+	queue->fpq.processing = pq;
 
 	spin_lock(&fch->lock);
 	if (ring->queues[qid]) {

fs/fuse/fuse_dev_i.h

@@ -247,6 +247,9 @@ struct fuse_chan {
 	/* Maximum number of pages that can be used in a single request */
 	unsigned int max_pages;
 
+	/* Before being installed into fud, contains the preallocated pq array*/
+	struct list_head *pq_prealloc;
+
 	/** Connection aborted via sysfs, respond with ECONNABORTED on device I/O */
 	bool abort_with_err;
 
@@ -396,6 +399,8 @@ struct fuse_dev *fuse_dev_alloc(void);
 
 int fuse_dev_release(struct inode *inode, struct file *file);
 
+struct list_head *fuse_pqueue_alloc(void);
+
 /**
  * Initialize the fuse processing queue
  */