asyncio module state should be locked with mutexes in free-threaded builds

[Fidget-Spinner]

Feature or enhancement

Proposal:

Please see issue title.

Has this already been discussed elsewhere?

No response given

Links to previous discussion of this feature:

No response

Linked PRs

• gh-121621: Use PyMutex for writes to asyncio state #121622
• gh-121621: Move asyncio running loop to thread state #121695
• [3.13] gh-121621: Use PyMutex for writes to asyncio state (GH-121622) #121774
• [3.13] gh-121621: Move asyncio running loop to thread state (GH-121695) #121864
• gh-121621: Move asyncio freelist to thread state #121915
• gh-121621: Move asyncio_running_loop to private struct #121939
• [3.13] gh-121621: Move asyncio_running_loop to private struct (GH-121939) #121943
• gh-121621: Disable asyncio freelist in free-threaded build #122046
• [3.13] gh-121621: Disable asyncio freelist in free-threaded build (GH-122046) #122048

[colesbury]

I think we also want to make enter_task, leave_task, and swap_current_task behave atomically. For example, enter_task makes two dict API calls that individually lock the dictionary, but the combined operation isn't atomic.

I can think of two strategies for this:

1. Try to replace multiple API calls with a single call. For example, I think enter_task may be able to use PyDict_SetDefaultRef and leave_task may be able to use _PyDict_DelItemIf with some modifications.
2. Use a critical section to lock the dict and use the "lock held" variants.

[Fidget-Spinner]

Fixed.