From: Gregory P. Smith <68491+gpshead@users.noreply.github.com> Date: Sun, 16 Nov 2025 19:45:29 +0000 (-0800) Subject: [3.14] gh-137109: refactor warning about threads when forking (GH-141438) (GH-141614) X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=0d8fb0b8523769e24f249d19dab3781a10b50e8b;p=thirdparty%2FPython%2Fcpython.git [3.14] gh-137109: refactor warning about threads when forking (GH-141438) (GH-141614) This splits the OS API specific functionality to get the number of threads out from the fallback Python method and warning raising code itself. This way the OS APIs can be queried before we've run `os.register_at_fork(after_in_parent=...)` registered functions which themselves may (re)start threads that would otherwise be detected. This is best effort. If the OS APIs are either unavailable or fail, the warning generating code still falls back to looking at the Python threading state after the CPython interpreter world has been restarted and the after_in_parent calls have been made. The common case for most Linux and macOS environments should work today. This also lines up with the existing TODO refactoring, we may choose to expose this API to get the number of OS threads in the `os` module in the future. Note: This is a simplified backport that maintains the void return type for warn_about_fork_with_threads() and keeps PyErr_Clear() in the warning path, as the error handling changes from fd8f42d3d1 are not needed in 3.14. --- diff --git a/Misc/NEWS.d/next/Library/2025-11-12-01-49-03.gh-issue-137109.D6sq2B.rst b/Misc/NEWS.d/next/Library/2025-11-12-01-49-03.gh-issue-137109.D6sq2B.rst new file mode 100644 index 000000000000..32f4e39f6d5f --- /dev/null +++ b/Misc/NEWS.d/next/Library/2025-11-12-01-49-03.gh-issue-137109.D6sq2B.rst @@ -0,0 +1,5 @@ +The :mod:`os.fork` and related forking APIs will no longer warn in the +common case where Linux or macOS platform APIs return the number of threads +in a process and find the answer to be 1 even when a +:func:`os.register_at_fork` ``after_in_parent=`` callback (re)starts a +thread. diff --git a/Modules/posixmodule.c b/Modules/posixmodule.c index 9295324f6c1e..6e878e3de372 100644 --- a/Modules/posixmodule.c +++ b/Modules/posixmodule.c @@ -7975,53 +7975,19 @@ os_register_at_fork_impl(PyObject *module, PyObject *before, // running in the process. Best effort, silent if unable to count threads. // Constraint: Quick. Never overcounts. Never leaves an error set. // -// This should only be called from the parent process after +// This MUST only be called from the parent process after // PyOS_AfterFork_Parent(). static void -warn_about_fork_with_threads(const char* name) +warn_about_fork_with_threads( + const char* name, // Name of the API to use in the warning message. + const Py_ssize_t num_os_threads // Only trusted when >= 1. +) { // It's not safe to issue the warning while the world is stopped, because // other threads might be holding locks that we need, which would deadlock. assert(!_PyRuntime.stoptheworld.world_stopped); - // TODO: Consider making an `os` module API to return the current number - // of threads in the process. That'd presumably use this platform code but - // raise an error rather than using the inaccurate fallback. - Py_ssize_t num_python_threads = 0; -#if defined(__APPLE__) && defined(HAVE_GETPID) - mach_port_t macos_self = mach_task_self(); - mach_port_t macos_task; - if (task_for_pid(macos_self, getpid(), &macos_task) == KERN_SUCCESS) { - thread_array_t macos_threads; - mach_msg_type_number_t macos_n_threads; - if (task_threads(macos_task, &macos_threads, - &macos_n_threads) == KERN_SUCCESS) { - num_python_threads = macos_n_threads; - } - } -#elif defined(__linux__) - // Linux /proc/self/stat 20th field is the number of threads. - FILE* proc_stat = fopen("/proc/self/stat", "r"); - if (proc_stat) { - size_t n; - // Size chosen arbitrarily. ~60% more bytes than a 20th column index - // observed on the author's workstation. - char stat_line[160]; - n = fread(&stat_line, 1, 159, proc_stat); - stat_line[n] = '\0'; - fclose(proc_stat); - - char *saveptr = NULL; - char *field = strtok_r(stat_line, " ", &saveptr); - unsigned int idx; - for (idx = 19; idx && field; --idx) { - field = strtok_r(NULL, " ", &saveptr); - } - if (idx == 0 && field) { // found the 20th field - num_python_threads = atoi(field); // 0 on error - } - } -#endif + Py_ssize_t num_python_threads = num_os_threads; if (num_python_threads <= 0) { // Fall back to just the number our threading module knows about. // An incomplete view of the world, but better than nothing. @@ -8074,6 +8040,51 @@ warn_about_fork_with_threads(const char* name) PyErr_Clear(); } } + +// If this returns <= 0, we were unable to successfully use any OS APIs. +// Returns a positive number of threads otherwise. +static Py_ssize_t get_number_of_os_threads(void) +{ + // TODO: Consider making an `os` module API to return the current number + // of threads in the process. That'd presumably use this platform code but + // raise an error rather than using the inaccurate fallback. + Py_ssize_t num_python_threads = 0; +#if defined(__APPLE__) && defined(HAVE_GETPID) + mach_port_t macos_self = mach_task_self(); + mach_port_t macos_task; + if (task_for_pid(macos_self, getpid(), &macos_task) == KERN_SUCCESS) { + thread_array_t macos_threads; + mach_msg_type_number_t macos_n_threads; + if (task_threads(macos_task, &macos_threads, + &macos_n_threads) == KERN_SUCCESS) { + num_python_threads = macos_n_threads; + } + } +#elif defined(__linux__) + // Linux /proc/self/stat 20th field is the number of threads. + FILE* proc_stat = fopen("/proc/self/stat", "r"); + if (proc_stat) { + size_t n; + // Size chosen arbitrarily. ~60% more bytes than a 20th column index + // observed on the author's workstation. + char stat_line[160]; + n = fread(&stat_line, 1, 159, proc_stat); + stat_line[n] = '\0'; + fclose(proc_stat); + + char *saveptr = NULL; + char *field = strtok_r(stat_line, " ", &saveptr); + unsigned int idx; + for (idx = 19; idx && field; --idx) { + field = strtok_r(NULL, " ", &saveptr); + } + if (idx == 0 && field) { // found the 20th field + num_python_threads = atoi(field); // 0 on error + } + } +#endif + return num_python_threads; +} #endif // HAVE_FORK1 || HAVE_FORKPTY || HAVE_FORK #ifdef HAVE_FORK1 @@ -8108,10 +8119,12 @@ os_fork1_impl(PyObject *module) /* child: this clobbers and resets the import lock. */ PyOS_AfterFork_Child(); } else { + // Called before AfterFork_Parent in case those hooks start threads. + Py_ssize_t num_os_threads = get_number_of_os_threads(); /* parent: release the import lock. */ PyOS_AfterFork_Parent(); // After PyOS_AfterFork_Parent() starts the world to avoid deadlock. - warn_about_fork_with_threads("fork1"); + warn_about_fork_with_threads("fork1", num_os_threads); } if (pid == -1) { errno = saved_errno; @@ -8157,10 +8170,12 @@ os_fork_impl(PyObject *module) /* child: this clobbers and resets the import lock. */ PyOS_AfterFork_Child(); } else { + // Called before AfterFork_Parent in case those hooks start threads. + Py_ssize_t num_os_threads = get_number_of_os_threads(); /* parent: release the import lock. */ PyOS_AfterFork_Parent(); // After PyOS_AfterFork_Parent() starts the world to avoid deadlock. - warn_about_fork_with_threads("fork"); + warn_about_fork_with_threads("fork", num_os_threads); } if (pid == -1) { errno = saved_errno; @@ -9014,10 +9029,12 @@ os_forkpty_impl(PyObject *module) /* child: this clobbers and resets the import lock. */ PyOS_AfterFork_Child(); } else { + // Called before AfterFork_Parent in case those hooks start threads. + Py_ssize_t num_os_threads = get_number_of_os_threads(); /* parent: release the import lock. */ PyOS_AfterFork_Parent(); // After PyOS_AfterFork_Parent() starts the world to avoid deadlock. - warn_about_fork_with_threads("forkpty"); + warn_about_fork_with_threads("forkpty", num_os_threads); } if (pid == -1) { return posix_error();