RDMA/core: Fix bogus WARN_ON during ib_unregister_device_queued()

ib_unregister_device_queued() can only be used by drivers using the new
dealloc_device callback flow, and it has a safety WARN_ON to ensure
drivers are using it properly.

However, if unregister and register are raced there is a special
destruction path that maintains the uniform error handling semantic of
'caller does ib_dealloc_device() on failure'. This requires disabling the
dealloc_device callback which triggers the WARN_ON.

Instead of using NULL to disable the callback use a special function
pointer so the WARN_ON does not trigger.

Fixes: d0899892edd0 ("RDMA/device: Provide APIs from the core code to help unregistration")
Link: https://lore.kernel.org/r/0-v1-a36d512e0a99+762-syz_dealloc_driver_jgg@nvidia.com
Reported-by: syzbot+4088ed905e4ae2b0e13b@syzkaller.appspotmail.com
Suggested-by: Hillf Danton <hdanton@sina.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

1 files changed, 8 insertions, 3 deletions

diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index 1335ed1f1e4a..40cf07129f66 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -1339,6 +1339,10 @@ out:
 	return ret;
 }
 
+static void prevent_dealloc_device(struct ib_device *ib_dev)
+{
+}
+
 /**
  * ib_register_device - Register an IB device with IB core
  * @device: Device to register
@@ -1409,11 +1413,11 @@ int ib_register_device(struct ib_device *device, const char *name)
 		 * possibility for a parallel unregistration along with this
 		 * error flow. Since we have a refcount here we know any
 		 * parallel flow is stopped in disable_device and will see the
-		 * NULL pointers, causing the responsibility to
+		 * special dealloc_driver pointer, causing the responsibility to
 		 * ib_dealloc_device() to revert back to this thread.
 		 */
 		dealloc_fn = device->ops.dealloc_driver;
-		device->ops.dealloc_driver = NULL;
+		device->ops.dealloc_driver = prevent_dealloc_device;
 		ib_device_put(device);
 		__ib_unregister_device(device);
 		device->ops.dealloc_driver = dealloc_fn;
@@ -1462,7 +1466,8 @@ static void __ib_unregister_device(struct ib_device *ib_dev)
 	 * Drivers using the new flow may not call ib_dealloc_device except
 	 * in error unwind prior to registration success.
 	 */
-	if (ib_dev->ops.dealloc_driver) {
+	if (ib_dev->ops.dealloc_driver &&
+	    ib_dev->ops.dealloc_driver != prevent_dealloc_device) {
 		WARN_ON(kref_read(&ib_dev->dev.kobj.kref) <= 1);
 		ib_dealloc_device(ib_dev);
 	}