Tokio also provides an event loop, on which we will be able to execute some code (with futures) that will be executed when some I/O events happen, such as when the result of a socket read is ready. To do so, the event loop will register events on specific file descriptors that represent sockets. It registers these events using the aforementioned system calls and then waits for any of the registered events to happen. The file descriptors and the system calls are low-level stuff that we do not need to know to use tokio, but it is important to understand how it works at the lower level. For instance, epoll does not support regular files, so if you try to wait for an event to happen on a regular file, it could block even though we're using asynchronous I/O which should not block.