nvmet: Fix possible infinite loop triggered on hot namespace removal

commit e4fcf07c upstream.

When removing a namespace we delete it from the subsystem namespaces
list with list_del_init which allows us to know if it is enabled or
not.

The problem is that list_del_init initialize the list next and does
not respect the RCU list-traversal we do on the IO path for locating
a namespace. Instead we need to use list_del_rcu which is allowed to
run concurrently with the _rcu list-traversal primitives (keeps list
next intact) and guarantees concurrent nvmet_find_naespace forward
progress.

By changing that, we cannot rely on ns->dev_link for knowing if the
namspace is enabled, so add enabled indicator entry to nvmet_ns for
that.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Solganik Alexander <sashas@lightbitslabs.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/nvme/target/configfs.c

@@ -271,7 +271,7 @@ static ssize_t nvmet_ns_device_path_store(struct config_item *item,
 
 	mutex_lock(&subsys->lock);
 	ret = -EBUSY;
-	if (nvmet_ns_enabled(ns))
+	if (ns->enabled)
 		goto out_unlock;
 
 	kfree(ns->device_path);
@@ -307,7 +307,7 @@ static ssize_t nvmet_ns_device_nguid_store(struct config_item *item,
 	int ret = 0;
 
 	mutex_lock(&subsys->lock);
-	if (nvmet_ns_enabled(ns)) {
+	if (ns->enabled) {
 		ret = -EBUSY;
 		goto out_unlock;
 	}
@@ -339,7 +339,7 @@ CONFIGFS_ATTR(nvmet_ns_, device_nguid);
 
 static ssize_t nvmet_ns_enable_show(struct config_item *item, char *page)
 {
-	return sprintf(page, "%d\n", nvmet_ns_enabled(to_nvmet_ns(item)));
+	return sprintf(page, "%d\n", to_nvmet_ns(item)->enabled);
 }
 
 static ssize_t nvmet_ns_enable_store(struct config_item *item,

drivers/nvme/target/core.c

@@ -264,7 +264,7 @@ int nvmet_ns_enable(struct nvmet_ns *ns)
 	int ret = 0;
 
 	mutex_lock(&subsys->lock);
-	if (!list_empty(&ns->dev_link))
+	if (ns->enabled)
 		goto out_unlock;
 
 	ns->bdev = blkdev_get_by_path(ns->device_path, FMODE_READ | FMODE_WRITE,
@@ -309,6 +309,7 @@ int nvmet_ns_enable(struct nvmet_ns *ns)
 	list_for_each_entry(ctrl, &subsys->ctrls, subsys_entry)
 		nvmet_add_async_event(ctrl, NVME_AER_TYPE_NOTICE, 0, 0);
 
+	ns->enabled = true;
 	ret = 0;
 out_unlock:
 	mutex_unlock(&subsys->lock);
@@ -325,11 +326,11 @@ void nvmet_ns_disable(struct nvmet_ns *ns)
 	struct nvmet_ctrl *ctrl;
 
 	mutex_lock(&subsys->lock);
-	if (list_empty(&ns->dev_link)) {
-		mutex_unlock(&subsys->lock);
-		return;
-	}
-	list_del_init(&ns->dev_link);
+	if (!ns->enabled)
+		goto out_unlock;
+
+	ns->enabled = false;
+	list_del_rcu(&ns->dev_link);
 	mutex_unlock(&subsys->lock);
 
 	/*
@@ -351,6 +352,7 @@ void nvmet_ns_disable(struct nvmet_ns *ns)
 
 	if (ns->bdev)
 		blkdev_put(ns->bdev, FMODE_WRITE|FMODE_READ);
+out_unlock:
 	mutex_unlock(&subsys->lock);
 }
 

drivers/nvme/target/nvmet.h

@@ -47,6 +47,7 @@ struct nvmet_ns {
 	loff_t			size;
 	u8			nguid[16];
 
+	bool			enabled;
 	struct nvmet_subsys	*subsys;
 	const char		*device_path;
 
@@ -61,11 +62,6 @@ static inline struct nvmet_ns *to_nvmet_ns(struct config_item *item)
 	return container_of(to_config_group(item), struct nvmet_ns, group);
 }
 
-static inline bool nvmet_ns_enabled(struct nvmet_ns *ns)
-{
-	return !list_empty_careful(&ns->dev_link);
-}
-
 struct nvmet_cq {
 	u16			qid;
 	u16			size;