feat(db): 同步栈读写并发（_q 去锁）+ core.db 并发/跨loop 修复

auth_system._q：单语句读不再持 _DB_LOCK —— WAL 下每连接读一致性快照，读写互不阻塞；旧实现把全部读也串行化且被写挡住。写事务（_x 与 24 个 with _DB_LOCK 块，经扫描均为写事务）仍全程串行，无写-写竞争。 core.db 并发修复（由测试暴露的真实缺陷）： - WAL 切换竞态：并发连接同时执行 PRAGMA journal_mode=WAL 争排他锁 → 专用锁内串行切换 + 二次复查（busy_timeout 前置） - 跨事件循环锁：模块级 asyncio.Lock 绑定首个 loop，多 loop 场景报 "bound to a different event loop" → 锁按 loop 隔离测试：25/25 通过（新增「读不被写锁阻塞」红转绿、读+同步写者并发 WAL、异步并发读）； lint: ruff All checks passed；pyright 0 errors。验收：冒烟 10/10、奇偶校验 2/2、uvicorn 真实启动 Flask 兜底+FastAPI 双通。
2026-09-26 01:03:18 +08:00
1 parent 52ed8b54fb
commit c77ef40fd5
@@ -60,12 +60,17 @@ def get_conn() -> sqlite3.Connection:


 def _q(sql: str, params: tuple = ()) -> list[sqlite3.Row]:
-    with _DB_LOCK:
-        conn = get_conn()
-        try:
-            return conn.execute(sql, params).fetchall()
-        finally:
-            conn.close()
+    """单语句读：**不持锁**。
+
+    WAL 模式下每个连接读到一致性快照，读与写互不阻塞；旧实现对读也加 _DB_LOCK，
+    等于把全部读串行化并被写挡住。此处仅放开单语句读——写事务（_x 与各
+    `with _DB_LOCK` 块）仍全程串行，故不引入写-写竞争。
+    """
+    conn = get_conn()
+    try:
+        return conn.execute(sql, params).fetchall()
+    finally:
+        conn.close()


 def _x(sql: str, params: tuple = ()) -> int:
@@ -25,21 +25,50 @@ import aiosqlite
 DATA_DIR = Path(os.environ.get('SITE_DATA_DIR', 'data'))
 DB_PATH = DATA_DIR / 'site.db'

-# 写事务串行锁（惰性创建，避免跨事件循环绑定）
-_WRITE_LOCK: asyncio.Lock | None = None
+# 锁按事件循环隔离：asyncio.Lock 绑定创建它的 loop，跨 loop 复用会抛
+# "bound to a different event loop"（迁移期 FastAPI loop + 同步桥 loop + 测试各自
+# 的 loop 并存，必须分开持有）。
+_LOOP_LOCKS: dict[tuple[int, str], asyncio.Lock] = {}
+
+
+def _loop_lock(kind: str) -> asyncio.Lock:
+    key = (id(asyncio.get_running_loop()), kind)
+    lock = _LOOP_LOCKS.get(key)
+    if lock is None:
+        lock = asyncio.Lock()
+        _LOOP_LOCKS[key] = lock
+    return lock


 def _write_lock() -> asyncio.Lock:
-    global _WRITE_LOCK
-    if _WRITE_LOCK is None:
-        _WRITE_LOCK = asyncio.Lock()
-    return _WRITE_LOCK
+    """写事务串行锁（每事件循环一个）。"""
+    return _loop_lock('write')
+
+
+def _wal_lock() -> asyncio.Lock:
+    """WAL 切换专用锁（不能用 _write_lock：execute() 已持它再调
+    connect→_configure 会自锁死）。"""
+    return _loop_lock('wal')
+
+
+# 已确认切到 WAL 的库路径（首连设置一次；后续连接不再重复执行 PRAGMA 争排他锁）
+_WAL_READY_FOR: str | None = None


 async def _configure(conn: aiosqlite.Connection) -> None:
+    global _WAL_READY_FOR
    conn.row_factory = aiosqlite.Row
-    await conn.execute('PRAGMA journal_mode=WAL')  # 读写并发的前提
-    await conn.execute('PRAGMA busy_timeout=15000')  # 遇写锁自旋等待，而非立即报错
+    # busy_timeout 先设：让后续语句遇锁自旋等待，而不是立刻报 database is locked
+    await conn.execute('PRAGMA busy_timeout=15000')
+    key = str(DB_PATH)
+    if _WAL_READY_FOR != key:
+        # 切换日志模式需要排他锁，且 `_WAL_READY_FOR` 的检查不是原子的——
+        # 并发连接须在专用锁内串行完成「切换 + 标记」，并二次复查，否则批量并发
+        # 连接会同时执行 PRAGMA journal_mode=WAL 争锁（实测 database is locked）
+        async with _wal_lock():
+            if _WAL_READY_FOR != key:
+                await conn.execute('PRAGMA journal_mode=WAL')
+                _WAL_READY_FOR = key
    await conn.execute('PRAGMA synchronous=NORMAL')  # WAL 下兼顾安全与吞吐
    await conn.execute('PRAGMA foreign_keys=ON')

@@ -78,5 +78,9 @@ async def test_bridge_rejects_inside_loop():
    async def coro():
        return 1

-    with pytest.raises(RuntimeError):
-        async_bridge.run_sync(coro())
+    c = coro()
+    try:
+        with pytest.raises(RuntimeError):
+            async_bridge.run_sync(c)
+    finally:
+        c.close()  # run_sync 拒绝执行，需自行关闭以免「未 await」告警
@@ -0,0 +1,87 @@
+# -*- coding: utf-8 -*-
+"""auth_system 同步栈的读写并发验证：读不持锁、写仍串行（WAL）。"""
+import sqlite3
+import threading
+import time
+
+import pytest
+
+import auth_system
+
+
+@pytest.fixture()
+def temp_site(tmp_path, monkeypatch):
+    monkeypatch.setattr(auth_system, 'DATA_DIR', tmp_path)
+    monkeypatch.setattr(auth_system, 'DB_PATH', tmp_path / 'site.db')
+    monkeypatch.setattr(auth_system, 'BACKUP_DIR', tmp_path / 'backups')
+    monkeypatch.setattr(auth_system, 'SECRET_KEY_PATH', tmp_path / 'secret.key')
+    auth_system.init_db()
+    return tmp_path / 'site.db'
+
+
+def test_q_does_not_hold_lock(temp_site):
+    """_q 读不应获取 _DB_LOCK（否则读会被写挡住）。"""
+    # 让另一个线程持有写锁，读仍应在短时间返回（不被阻塞）
+    held = threading.Event()
+    release = threading.Event()
+
+    def holder():
+        with auth_system._DB_LOCK:
+            held.set()
+            release.wait(5)
+
+    t = threading.Thread(target=holder, daemon=True)
+    t.start()
+    held.wait(2)
+    try:
+        start = time.monotonic()
+        rows = auth_system._q('SELECT COUNT(*) AS n FROM users')
+        elapsed = time.monotonic() - start
+        assert rows is not None
+        assert elapsed < 1.0, f'读被写锁阻塞了 {elapsed:.2f}s（WAL 下不应发生）'
+    finally:
+        release.set()
+        t.join(2)
+
+
+def test_concurrent_reads_and_writes(temp_site):
+    """多读线程 + 一个写线程并发：无异常、结果一致。"""
+    errors = []
+    stop = threading.Event()
+
+    def writer():
+        try:
+            for i in range(50):
+                auth_system.set_setting(f'k{i}', str(i))
+        except Exception as e:
+            errors.append(('w', e))
+        finally:
+            stop.set()
+
+    def reader():
+        while not stop.is_set():
+            try:
+                auth_system.get_setting('k0')
+            except Exception as e:
+                errors.append(('r', e))
+
+    ws = threading.Thread(target=writer, daemon=True)
+    rs = [threading.Thread(target=reader, daemon=True) for _ in range(4)]
+    ws.start()
+    for r in rs:
+        r.start()
+    ws.join(20)
+    for r in rs:
+        r.join(2)
+
+    assert not errors, f'并发读写异常: {errors[:3]}'
+    assert auth_system.get_setting('k49') == '49'
+
+
+def test_wal_enabled(temp_site):
+    conn = sqlite3.connect(str(temp_site))
+    try:
+        mode = conn.execute('PRAGMA journal_mode').fetchone()[0]
+    finally:
+        conn.close()
+    assert str(mode).lower() == 'wal'