Rust中使用jemalloc动态开启内存分析报错，求解决方案

我想用jemalloc检测Rust程序的内存泄漏，需要生成dump文件并通过外部接口动态控制生成。操作步骤如下：

1. 在Cargo.toml中添加依赖：

[package]
name = "rust-learn"
version = "0.1.0"
edition = "2018"

[dependencies]
tokio = { version = "1.17.0", features = ["full"] }
actix-web = "4"
serde = { version = "1.0.64", features = ["derive"] }
serde_json = "1.0.64"
log4rs = "1.2.0"
jemallocator = "0.5.4"
jemalloc-ctl = "0.5.4"

[dependencies.jemalloc-sys]
version = "0.5.4"
features = ["stats", "profiling", "unprefixed_malloc_on_supported_platforms"]

[profile.release]
debug = true

[build-dependencies]

2. 在main.rs中编写控制分析开关的HTTP接口：

use actix_web::{get, post, web, App, HttpResponse, HttpServer, Responder};
use jemalloc_ctl::{AsName, Access};

#[global_allocator]
static ALLOC: jemallocator::Jemalloc = jemallocator::Jemalloc;
const PROF_ACTIVE: &'static [u8] = b"prof.active\0";

#[get("/active")]
async fn hello() -> impl Responder {
    let name = PROF_ACTIVE.name();
    name.write(true).expect("Should succeed to set prof");
    HttpResponse::Ok().body("Hello, world!")
}

#[actix_web::main]
async fn main() -> std::io::Result<()> {
    HttpServer::new(|| {
        App::new()
            .service(hello)
    })
    .bind("127.0.0.1:8080")?
    .run()
    .await
}

3. 调用接口时出现错误：
   执行curl http://127.0.0.1:8080/active返回curl: (52) Empty reply from server，服务端报错：

thread 'actix-rt|system:0|arbiter:1' panicked at src/main.rs:17:22:
Should succeed to set prof: `name` or `mib` specifies an unknown/invalid value.
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

环境信息：macOS 13.3.1，Rust 1.73。尝试在.cargo/config.toml添加MALLOC_CONF="prof:true"，问题仍未解决。

错误核心原因是jemalloc的profiling功能未在初始化阶段启用，导致运行时无法访问prof.active配置项。以下是具体修复步骤：

1. 确保jemalloc-sys的profiling特性正确激活 #

在jemalloc-sys依赖中添加default-features = false，避免默认特性覆盖自定义配置：

[dependencies.jemalloc-sys]
version = "0.5.4"
features = ["stats", "profiling", "unprefixed_malloc_on_supported_platforms"]
default-features = false

2. 在程序启动时启用profiling #

.cargo/config.toml的环境变量可能未正确传递，建议通过启动命令或代码初始化配置：

方式一：启动时手动设置环境变量 #

运行程序时执行：

MALLOC_CONF="prof:true,prof_prefix:jeprof" cargo run --release

• prof:true：启用profiling功能
• prof_prefix:jeprof：指定dump文件前缀，方便后续分析

方式二：代码中初始化配置 #

在main函数开头添加初始化代码，确保profiling在启动时生效：

use jemalloc_ctl::config;

#[actix_web::main]
async fn main() -> std::io::Result<()> {
    // 初始化jemalloc profiling配置
    let _ = config::prof::true_().write(true);
    let _ = config::prof::prefix::write("jeprof");

    HttpServer::new(|| {
        App::new()
            .service(hello)
    })
    .bind("127.0.0.1:8080")?
    .run()
    .await
}

3. 修正prof.active的访问方式 #

使用jemalloc_ctl提供的类型安全API替代字节数组构造，避免配置项名称错误：

use jemalloc_ctl::prof;

#[get("/active")]
async fn hello() -> impl Responder {
    // 激活profiling
    prof::active::write(true).expect("Failed to activate profiling");
    // 立即生成dump文件
    prof::dump::write(true).expect("Failed to generate dump");
    HttpResponse::Ok().body("Profiling activated and dump generated!")
}

4. 验证dump文件生成 #

调用接口后，程序运行目录会生成类似jeprof.<pid>.0.f.heap的dump文件，使用jeprof工具分析：

jeprof --show_bytes ./target/release/rust-learn jeprof.<pid>.0.f.heap

内容的提问来源于stack exchange，提问作者Dolphin