gh-126688: Support Fork Parent and Child With Different Thread ID #127121
+69
−10
Conversation
This file contains hidden or 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
gpshead
reviewed
Nov 23, 2024
| @@ -694,6 +713,9 @@ PyOS_AfterFork_Child(void) | |||
| goto fatal_error; | |||
| } | |||
|
|
|||
| _PyRuntime.os_fork.parent = (struct _os_fork_parent){0}; | |||
| PyMutex_Unlock(&_PyRuntime.os_fork.mutex); | |||
There was a problem hiding this comment.
this might need to be done before _PyThreadState_DeleteList above to handle the very silly case of someone forking from a destructor and deadlocking? ("why is that even reasonable to support?" is valid...)
| @@ -678,7 +698,6 @@ PyOS_AfterFork_Child(void) | |||
| _PyEval_StartTheWorldAll(&_PyRuntime); | |||
| _PyThreadState_DeleteList(list); | |||
|
|
|||
| _PyImport_ReInitLock(tstate->interp); | |||
There was a problem hiding this comment.
add a comment mentioning which call above is now responsible for doing this.
| @@ -515,6 +516,7 @@ _PyRuntimeState_ReInitThreads(_PyRuntimeState *runtime) | |||
| for (PyInterpreterState *interp = runtime->interpreters.head; | |||
| interp != NULL; interp = interp->next) | |||
| { | |||
| _PyImport_ReInitLock(interp, parent); | |||
There was a problem hiding this comment.
why is this only being done when the GIL is disabled instead of always as it was in the past?
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Outdated suggestions cannot be applied.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.
gh-126692 took care of one situation where we were assuming the thread ID did not change across fork. However, there is at least one other, in
PyOS_AfterFork_Child(). This PR addresses that case along with making the parent's thread ID available to the child.Here's a breakdown:
_PyRuntimeState.os_fork.mutexto allow only one instance of forking to happen at once (in part, to protect the rest of_PyRuntimeState.os_fork)_PyRuntimeState.os_fork.parent.tidto make the parent's thread ID available to the child_PyRecursiveMutex_at_fork_reinit(), which makes use of the parent thread ID_PyRecursiveMutex_at_fork_reinit()in_PyImport_ReInitLock()I've also moved the
_PyImport_ReInitLock()call fromPyOS_AfterFork_Child()to_PyRuntimeState_ReInitThreads(), to be more consistent about where we reinitialize locks after forking. If there are any objections, I don't mind dropping that part.One motivation I have here is that I'd like to use a
_PyRecursiveMutexelsewhere and need to be able to correctly reinitialize after forking. That requires knowing the parent's thread ID, to decide if the forking thread held the lock or not. (This wasn't a problem for the import lock, which is a_PyRecursiveMutex, since we always acquire and hold that lock while forking.)@gpshead, I'd be particularly interested in your thoughts on this.