Pitfalls of using std::sync::Mutex with the multithreaded runtime

[samueltardieu]

The documentation of tokio::sync::Mutex contains:

Contrary to popular belief, it is okay and often preferred to use the ordinary Mutex from the standard library in asynchronous code.

This includes a well-articulated explanation of why this is the case. I concur with this explanation as long as the mutex is used within a single-threaded asynchronous runtime.

However, using std::sync::Mutex with a multithreaded executor may cause another thread to be temporarily blocked when the mutex is held. If this thread is running an asynchronous executor, it will be unable to execute other asynchronous tasks during this time, and other threads may not be available to execute these tasks through work stealing.

If asynchronous tasks accessing this mutex are dispersed across the executor threads, there is a risk that multiple executor threads could become blocked. Shouldn't this be highlighted?

[Darksonn]

Ultimately, it's perfectly fine to use std::sync::Mutex from a multi-threaded runtime. In fact, lots of things in tokio::sync work that way internally. As long as you don't hold the mutex for very long, there's no problem.