gh-115999: Make list, tuple and range iteration more thread-safe. #128637
+200
−33
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
concurrent iteration. (This is prep work for enabling specialization of FOR_ITER in free-threaded builds.) The basic premise is: - Iterating over a shared _iterable_ (list, tuple or range) should be safe, not involve data races, and behave like iteration normally does. - Using a shared _iterator_ should not crash or involve data races, and should only produce items regular iteration would produce. It is _not_ guaranteed to produce all items, or produce each item only once. Providing stronger guarantees is possible for some of these iterators, but it's not always straight-forward and can significantly hamper the common case. Since iterators in general aren't shared between threads, and it's simply impossible to concurrently use many iterators (like generators), better to make sharing iterators without explicit synchronization clearly wrong. Specific issues fixed in order to make the tests pass: - List iteration could occasionally crash when a shared list wasn't already marked as shared when reallocated. - Tuple iteration could occasionally crash when the iterator's reference to the tuple was cleared on exhaustion. Like with list iteration, in free-threaded builds we can't safely and efficiently clear the iterator's reference to the iterable (doing it safely would mean extra, slow refcount operations), so just keep the iterable reference around. - Fast range iterators (for integers that fit in C longs) shared between threads would sometimes produce values past the end of the range, because the iterators use two bits of state that we can't efficiently update atomically. Rewriting the iterators to have a single bit of state is possible, but probably means more math for each iteration and may not be worth it. - Long range iterators (for other numbers) shared between threads would crash catastrophically in a variety of ways. This now uses a critical section. Rewriting this to be more efficient is probably possible, but since it deals with arbitrary Python objects it's difficult to get right. There seem to be no more exising races in list_get_item_ref, so drop it from the tsan suppression list.
corona10
reviewed
Jan 8, 2025
| Py_INCREF(r->len); | ||
| return r->len; | ||
| PyObject *len; | ||
| Py_BEGIN_CRITICAL_SECTION(r); |
There was a problem hiding this comment.
Out of curiosity, using atomic operation is not enough for this operation?
(Maybe with _Py_TryIncref?)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Make list, tuple and range iteration more thread-safe, and actually test concurrent iteration. (This is prep work for enabling specialization of FOR_ITER in free-threaded builds.) The basic premise is:
Iterating over a shared iterable (list, tuple or range) should be safe, not involve data races, and behave like iteration normally does.
Using a shared iterator should not crash or involve data races, and should only produce items regular iteration would produce. It is not guaranteed to produce all items, or produce each item only once.
Providing stronger guarantees is possible for some of these iterators, but it's not always straight-forward and can significantly hamper the common case. Since iterators in general aren't shared between threads, and it's simply impossible to concurrently use many iterators (like generators), better to make sharing iterators without explicit synchronization clearly wrong.
Specific issues fixed in order to make the tests pass:
List iteration could occasionally crash when a shared list wasn't already marked as shared when reallocated.
Tuple iteration could occasionally crash when the iterator's reference to the tuple was cleared on exhaustion. Like with list iteration, in free-threaded builds we can't safely and efficiently clear the iterator's reference to the iterable (doing it safely would mean extra, slow refcount operations), so just keep the iterable reference around.
Fast range iterators (for integers that fit in C longs) shared between threads would sometimes produce values past the end of the range, because the iterators use two pieces of state that we can't efficiently update atomically. Rewriting the iterators to have a single piece of state is possible, but probably means more math for each iteration and may not be worth it.
Long range iterators (for other numbers) shared between threads would crash catastrophically in a variety of ways. This now uses a critical section. Rewriting this to be more efficient is probably possible, but since it deals with arbitrary Python objects it's difficult to get right.
There seem to be no more existing races in list_get_item_ref, so drop it from the tsan suppression list.
--disable-gilbuilds #115999