Isolate the faulthandler Module Between Interpreters

オープン

#101,509 2023年2月1日に作成コメント 1 件リアクション 2 件担当者 0 名 GitHub で見る

まだ誰も着手していません。

3.12 interpreter-core topic-subinterpreters type-feature

主要言語: Python
スター: 77.2k
フォーク: 36k
平均マージ: 1日 9時間
マージ済み PR（30日）: 558

説明

(See gh-100227.)

Currently the faulthandler module has some state in _PyRuntimeState, including objects, which is shared by all interpreters. Interpreters should be isolated from each other, for a variety of reasons (including the possibility of a per-interpreter GIL). Isolating the module will involve moving some of the state to PyInterpreterState (and, under per-interpreter GIL, guarding other state with a global lock).

Code Analysis:

(expand)

Signals

The module installs handlers for various signals.

(expand)

fatal:

SIGBUS - bus error
SIGILL - illegal instruction
SIGFPE - floating point exception
SIGABRT - abort
SIGSEGV - segfault

"user":

the 5 fatal signals
others (up to Py_NSIG)

handlers:

faulthandler_fatal_error(signum)
faulthandler_exc_handler() (on Windows)
faulthandler_user()

State

Fields

https://github.com/python/cpython/blob/main/Include/internal/pycore_faulthandler.h#L49-L86
https://github.com/python/cpython/blob/main/Include/internal/pycore_runtime.h#L146

raw

https://github.com/python/cpython/blob/62251c3da06eb4662502295697f39730565b1717/Include/internal/pycore_faulthandler.h#L49-L86
https://github.com/python/cpython/blob/62251c3da06eb4662502295697f39730565b1717/Include/internal/pycore_faulthandler.h#L36-L46

tables:

name	type	#ifdef	notes
`fatal_error`
. `enabled`	`bool`		"is faulthandler enabled?"
. `file`	*`PyObject `**		only for keeping `.fd` alive
. `fd`	`int`
. `all_threads`	`bool`
. `interp`	*`PyInterpreterState `**		show only its threads the one that enabled faulthandler
. `exc_handler`	`void *`	`MS_WINDOWS`

name	type	#ifdef	notes
`thread`	---
. `file`	`PyObject `*		only for keeping `.fd` alive
. `fd`	`int`		defaults to `sys.stderr`
. `timeout_us`	`PY_TIMEOUT_T`
. `repeat`	`bool`
. `interp`	*`PyInterpreterState `**		show only its threads the one that requested the info
. `exit`	`bool`
. `header`	`char *`
. `header_len`	`size_t`
. `cancel_event`	`PyThread_type_lock`
. `running`	`PyThread_type_lock`
name	type	#ifdef	notes
`user_signals`	*`struct faulthandler_user_signal `**	`FAULTHANDLER_USER`	a dynamically allocated array (`Py_NSIG`)
`user_signals[signum]`	---
. `enabled`	`bool`
. `file`	`PyObject *`
. `fd`	`int`
. `all_threads`	`bool`
. `chain`	`bool`
. `previous`	`_Py_sighandler_t`
. `interp`	*`PyInterpreterState `**		show only its threads the one that added the handler
name	type	#ifdef	notes
`old_stack`	`stack_t`	`FAULTHANDLER_USE_ALT_STACK`
`stack`	`stack_t`	`FAULTHANDLER_USE_ALT_STACK`

Usage

simple:

name	context	get	set
`fatal_error`
. `enabled`	module	`faulthandler_is_enabled()`	`faulthandler_disable_py()`
	signal	`faulthandler_fatal_error()`
	internal	`faulthandler_enable()` `faulthandler_disable()`	`faulthandler_enable()` `faulthandler_disable()`
. `file`	module	`faulthandler_traverse()`	`faulthandler_py_enable()` `faulthandler_traverse()`
	internal	`faulthandler_disable()`	`faulthandler_disable()`
. `fd`		module
	signal	`faulthandler_fatal_error()` `faulthandler_exc_handler()`
. `all_threads`	module		`faulthandler_py_enable()`
	signal	`faulthandler_fatal_error()` `faulthandler_exc_handler()`
. `interp`	module		`faulthandler_py_enable()`
	internal	`faulthandler_fatal_error()` `faulthandler_exc_handler()`
. `exc_handler`	internal	`faulthandler_enable()` `faulthandler_disable()`	`faulthandler_enable()` `faulthandler_disable()`

name	context	get	set
`thread`
. `file`	module	`faulthandler_dump_traceback_later()` `faulthandler_traverse()`	`faulthandler_dump_traceback_later()` `faulthandler_traverse()`
	internal	`cancel_dump_traceback_later()`	`cancel_dump_traceback_later()`
. `fd`
module		`faulthandler_dump_traceback_later()`
	thread	`faulthandler_thread()`
. `timeout_us`	module		`faulthandler_dump_traceback_later()`
	thread	`faulthandler_thread()`
. `repeat`	module		`faulthandler_dump_traceback_later()`
	thread	`faulthandler_thread()`
. `interp`	module		`faulthandler_dump_traceback_later()`
	thread	`faulthandler_thread()`
. `exit`	module		`faulthandler_dump_traceback_later()`
	thread	`faulthandler_thread()`
. `header`	module		`faulthandler_dump_traceback_later()`
	thread	`faulthandler_thread()`
	internal	`cancel_dump_traceback_later()`	`cancel_dump_traceback_later()`
. `header_len`	module		`faulthandler_dump_traceback_later()`
	thread	`faulthandler_thread()`
. `cancel_event`	module	`faulthandler_dump_traceback_later()`	`faulthandler_dump_traceback_later()`
	thread	`faulthandler_thread()`
	internal	`cancel_dump_traceback_later()`
. `running`	module	`faulthandler_dump_traceback_later()`	`faulthandler_dump_traceback_later()`
	thread	`faulthandler_thread()`
	internal	`cancel_dump_traceback_later()`
name	context	get	set
`user_signals`	module	`faulthandler_register_py()` `faulthandler_unregister_py()`	`faulthandler_register_py()` `faulthandler_traverse()`
	C-API	`_PyFaulthandler_Fini()`	`_PyFaulthandler_Fini()`
	signal	`faulthandler_user()`
`user_signals[signum]`
. `enabled`	module	`faulthandler_register_py()`	`faulthandler_register_py()`
	signal	`faulthandler_user()`
	internal	`faulthandler_unregister()`	`faulthandler_unregister()`
. `file`	module	`faulthandler_register_py()` `faulthandler_traverse()`	`faulthandler_register_py()` `faulthandler_traverse()`
	internal	`faulthandler_unregister()`	`faulthandler_unregister()`
. `fd`	module		`faulthandler_register_py()`
	signal	`faulthandler_user()`
	internal		`faulthandler_unregister()`
. `all_threads`	module		`faulthandler_register_py()`
	signal	`faulthandler_user()`
. `chain`	module		`faulthandler_register_py()`
	signal	`faulthandler_user()`
. `previous`	module		`faulthandler_register_py()`
	signal	`faulthandler_user()`
	internal	`faulthandler_unregister()`
. `interp`	module		`faulthandler_register_py()`
	signal	`faulthandler_user()`
name	context	get	set
`stack`	C-API	`_PyFaulthandler_Init()` `_PyFaulthandler_Fini()`
	internal	`faulthandler_enable()` `faulthandler_register()` `faulthandler_allocate_stack()`
`old_stack`	C-API	`_PyFaulthandler_Fini()`

modify data in complex fields:

name	context	calls
`fatal_error .fd`	signal	`faulthandler_fatal_error()` -> `faulthandler_dump_traceback()` `faulthandler_fatal_error()` -> `_Py_DumpExtensionModules()` `faulthandler_exc_handler()` `faulthandler_exc_handler()` -> `_Py_DumpHexadecimal()` `faulthandler_exc_handler()` -> `faulthandler_dump_traceback()`
`thread.fd`	thread	`faulthandler_thread()` -> `_Py_write_noraise()` `faulthandler_thread()` -> `_Py_DumpTracebackThreads()`
`thread .cancel_event`	module	`faulthandler_dump_traceback_later()`
	C-API	`_PyFaulthandler_Fini()`
	thread	`faulthandler_thread()`
	internal	`cancel_dump_traceback_later()`
`thread .running`	module	`faulthandler_dump_traceback_later()`
	thread	`faulthandler_thread()`
	internal	`cancel_dump_traceback_later()`
`user_signals[signum] .fd`	signal	`faulthandler_user()` -> `faulthandler_dump_traceback()`
`stack`	C-API	`_PyFaulthandler_Init()` `_PyFaulthandler_Fini()`
	internal	`faulthandler_allocate_stack()`
`old_stack`	internal	`faulthandler_allocate_stack()` -> `sigaltstack()`

コントリビューションガイド

コントリビューションガイドを開く

はじめの一歩

issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
リポジトリをフォークし、ブランチを切って変更します。
issue 番号を参照したプルリクエストを送ります。

調査の方向性

まず Include/internal/pycore_faulthandler.h と Include/internal/pycore_runtime.h から始め、次に _PyFaulthandler_Init()、_PyFaulthandler_Fini()、faulthandler_enable()、faulthandler_user() など、列挙されているエントリポイントを追跡します。状態テーブルにある各共有フィールドを、そのインタープリター単位またはグローバルな所有範囲に対応付け、シグナルハンドラーの動作を損なうことなく faulthandler の状態が分離されていることを確認します。

索引モデルが issue の本文から書いたものです。

評価

技術スタック: c, python
領域: operating-systems
issue の種類: リファクタリング
難易度: 5/5
見積もり時間: 1週間以上
活発さ: 停滞
明瞭さ: おおむね明確
初心者へのやさしさ: 25/100