nodejs / nodejs/node

buffer: UTF-16LE indexOf can hang on unaligned Buffer views

未关闭
#65,959 1 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

还没有人认领这个 Issue。

主要语言
JavaScript
星标
122k
派生
37.4k
平均合并
4 天 3 小时
30 天内合并 PR
272

描述

Version

v27.0.0-pre (b805fb5158a)

Platform
Darwin 24.6.0 arm64
Subsystem

buffer

What steps will reproduce the bug?

Buffer#indexOf() can hang when its receiver is an unaligned view. The child process timeout prevents the reproduction from hanging indefinitely.

'use strict';

const { spawnSync } = require('node:child_process');

for (const prefixLength of [0, 1]) {
  const source = `
    const data = Buffer.from('abc', 'utf16le');
    const backing = Buffer.alloc(data.length + ${prefixLength});
    data.copy(backing, ${prefixLength});

    const view = backing.subarray(${prefixLength});
    console.log(view.indexOf('a', 0, 'utf16le'));
  `;

  const result = spawnSync(process.execPath, ['-e', source], {
    encoding: 'utf8',
    timeout: 2000,
  });

  console.log({
    prefixLength,
    status: result.status,
    error: result.error?.code,
    stdout: result.stdout.trim(),
  });
}
How often does it reproduce? Is there a required condition?

It's consistent when the receiver view starts at an odd byte offset. The aligned control returns normally.

What is the expected behavior? Why is that the expected behavior?

Both searches should terminate and return 0.
The views contain the same bytes, and the UTF-16LE value begins at index 0 of each view. The result should not depend on the underlying memory address.

What do you see instead?

The aligned view returns 0. The view starting at byte offset 1 does not complete before the timeout:

{ prefixLength: 0, status: 0, error: undefined, stdout: '0' }
{ prefixLength: 1, status: null, error: 'ETIMEDOUT', stdout: '' }
Additional information

Buffer#indexOf() passes the receiver data to the 16-bit search, while nbytes aligns matched addresses down. For an odd-offset view, this can move before the start of the view and appears to cause the hang.

This is distinct from #26448, which concerns search-offset rounding, and remains reproducible after #65905, which affects UTF-16LE decoding rather than Buffer#indexOf().

Copying the view or using a Buffer needle avoids the hang. Both workarounds return 0 locally:

Buffer.from(view).indexOf('a', 0, 'utf16le');
view.indexOf(Buffer.from('a', 'utf16le'));

贡献指南

打开贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

调研方向

先运行提供的复现,然后检查 src/node_buffer.cc 中 1119-1124 行附近的 Buffer#indexOf,以及 deps/nbytes/include/nbytes.h 中 459-465 行附近的 16 位搜索对齐方式。跟踪具有奇数偏移量的视图如何传递给搜索,并为对齐和未对齐的视图添加回归覆盖。完成标准是两次 UTF-16LE 搜索都终止并返回 0。

由索引模型根据 Issue 内容生成。

评估

技术栈
cpp, javascript, node.js
领域
backend
Issue 类型
缺陷
难度
3/5
预计耗时
1-2 天
活跃度
活跃
描述清晰度
描述清楚
新手友好度
68/100

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。