[PATCH v2] device-dax: fix cdev leak

[Dan Williams]

If device_add() fails, cleanup the cdev. Otherwise, we leak a kobj_map()
with a stale device number.

As Jason points out, there is a small possibility that userspace has
opened and mapped the device in the time between cdev_add() and the
device_add() failure. We need a new kill_dax_dev() helper to invalidate
any established mappings.

Fixes: ba09c01d2fa8 ("dax: convert to the cdev api")
Cc: <stable@xxxxxxxxxxxxxxx>
Cc: Logan Gunthorpe <logang@xxxxxxxxxxxx>
Reported-by: Jason Gunthorpe <jgunthorpe@xxxxxxxxxxxxxxxxxxxx>
Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx>
---
v2: handle the potential of userspace establishing mappings
    between cdev_add() and device_add()

 drivers/dax/dax.c |   15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

diff --git a/drivers/dax/dax.c b/drivers/dax/dax.c
index ed758b74ddf0..b62a52fa3707 100644
--- a/drivers/dax/dax.c
+++ b/drivers/dax/dax.c
@@ -630,13 +630,10 @@ static void dax_dev_release(struct device *dev)
 	kfree(dax_dev);
 }
 
-static void unregister_dax_dev(void *dev)
+static void kill_dax_dev(struct dax_dev *dax_dev)
 {
-	struct dax_dev *dax_dev = to_dax_dev(dev);
 	struct cdev *cdev = &dax_dev->cdev;
 
-	dev_dbg(dev, "%s\n", __func__);
-
 	/*
 	 * Note, rcu is not protecting the liveness of dax_dev, rcu is
 	 * ensuring that any fault handlers that might have seen
@@ -648,6 +645,15 @@ static void unregister_dax_dev(void *dev)
 	synchronize_rcu();
 	unmap_mapping_range(dax_dev->inode->i_mapping, 0, 0, 1);
 	cdev_del(cdev);
+}
+
+static void unregister_dax_dev(void *dev)
+{
+	struct dax_dev *dax_dev = to_dax_dev(dev);
+
+	dev_dbg(dev, "%s\n", __func__);
+
+	kill_dax_dev(dax_dev);
 	device_unregister(dev);
 }
 
@@ -724,6 +730,7 @@ struct dax_dev *devm_create_dax_dev(struct dax_region *dax_region,
 	dev_set_name(dev, "dax%d.%d", dax_region->id, dax_dev->id);
 	rc = device_add(dev);
 	if (rc) {
+		kill_dax_dev(dax_dev);
 		put_device(dev);
 		return ERR_PTR(rc);
 	}