如何使用Valgrind memcheck工具进行C/C++的内存泄漏检测

jopen 11年前

系统编程中一个重要的方面就是有效地处理与内存相关的问题。你的工作越接近系统,你就需要面对越多的内存问题。有时这些问题非常琐碎,而更多时候它会演变成一个调试内存问题的恶梦。所以,在实践中会用到很多工具来调试内存问题。

在本文中,我们将讨论最流行的开源内存管理框架 VALGRIND。

摘自 Valgrind.org:

Valgrind是用于构建动态分析工具的探测框架。它包括一个工具集,每个工具执行某种类型的调试、分析或类似的任务,以帮助完善你的程序。Valgrind的架构是模块化的,所以可以容易地创建新的工具而又不会扰乱现有的结构。

许多有用的工具被作为标准而提供。

  1. Memcheck是一个内存错误检测器。它有助于使你的程序,尤其是那些用C和C++写的程序,更加准确。
  2. Cachegrind是一个缓存和分支预测分析器。它有助于使你的程序运行更快。
  3. Callgrind是一个调用图缓存生成分析器。它与Cachegrind的功能有重叠,但也收集Cachegrind不收集的一些信息。
  4. Helgrind是一个线程错误检测器。它有助于使你的多线程程序更加准确。
  5. DRD也是一个线程错误检测器。它和Helgrind相似,但使用不同的分析技术,所以可能找到不同的问题。
  6. Massif是一个堆分析器。它有助于使你的程序使用更少的内存。
  7. DHAT是另一种不同的堆分析器。它有助于理解块的生命期、块的使用和布局的低效等问题。
  8. SGcheck是一个实验工具,用来检测堆和全局数组的溢出。它的功能和Memcheck互补:SGcheck找到Memcheck无法找到的问题,反之亦然。
  9. BBV是个实验性质的SimPoint基本块矢量生成器。它对于进行计算机架构的研究和开发很有用处。

也有一些对大多数用户没有用的小工具:Lackey是演示仪器基础的示例工具;Nulgrind是一个最小化的Valgrind工具,不做分析或者操作,仅用于测试目的。

在这篇文章我们将关注“memcheck”工具。

使用 Valgrind Memcheck

memcheck工具的使用方式如下:

valgrind --tool=memcheck ./a.out

从上面的命令可以清楚的看到, 主要的命令是valgrind,而我们想使用的工具是通过'-tool'选项来指定的. 上面的‘a.out’指的是我们想使用memcheck运行的可执行文件.

该工具可以检测下列与内存相关的问题 :

  • 未释放内存的使用
  • 对释放后内存的读/写
  • 对已分配内存块尾部的读/写
  • 内存泄露
  • 不匹配的使用malloc/new/new[] 和 free/delete/delete[]
  • 重复释放内存

注意: 上面列出的并不很全面,但却包含了能被该工具检测到的很多普遍的问题.

让我们一个一个地对上面的场景进行讨论:

注意: 下面讨论的所有测试代码都应该使用gcc并且加上-g选项(用来在memcheck的输出中生成行号)进行编译. 就想我们之前讨论过的 C程序被编译成可执行文件, 它需要经历四个不同的阶段.

1. 使用未初始化的内存

Code :

#include <stdio.h>  #include <stdlib.h>     int main(void)  {      char *p;         char c = *p;         printf("\n [%c]\n",c);         return 0;  }

在上面的代码中,我们尝试使用未初始化的指针 ‘p’.

让我们运行Memcheck来看下结果.

$ valgrind --tool=memcheck ./val  ==2862== Memcheck, a memory error detector  ==2862== Copyright (C) 2002-2009, and GNU GPL'd, by Julian Seward et al.  ==2862== Using Valgrind-3.6.0.SVN-Debian and LibVEX; rerun with -h for copyright info  ==2862== Command: ./val  ==2862==  ==2862== Use of uninitialised value of size 8  ==2862==    at 0x400530: main (valgrind.c:8)  ==2862==    [#]  ==2862==  ==2862== HEAP SUMMARY:  ==2862==     in use at exit: 0 bytes in 0 blocks  ==2862==   total heap usage: 0 allocs, 0 frees, 0 bytes allocated  ==2862==  ==2862== All heap blocks were freed -- no leaks are possible  ==2862==  ==2862== For counts of detected and suppressed errors, rerun with: -v  ==2862== Use --track-origins=yes to see where uninitialized values come from  ==2862== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 4 from 4)

从上面的输出可以看到,Valgrind检测到了未初始化的变量,然后给出了警告(上面加粗的几行(译者注:貌似上面没有加粗的)).

2. 在内存被释放后进行读/写

Code :

#include <stdio.h>  #include <stdlib.h>     int main(void)  {      char *p = malloc(1);      *p = 'a';         char c = *p;         printf("\n [%c]\n",c);         free(p);      c = *p;      return 0;  }

上面的代码中,我们有一个释放了内存的指针 ‘p’ 然后我们又尝试利用指针获取值.

让我们运行memcheck来看一下Valgrind对这种情况是如何反应的.

$ valgrind --tool=memcheck ./val  ==2849== Memcheck, a memory error detector  ==2849== Copyright (C) 2002-2009, and GNU GPL'd, by Julian Seward et al.  ==2849== Using Valgrind-3.6.0.SVN-Debian and LibVEX; rerun with -h for copyright info  ==2849== Command: ./val  ==2849==      [a]  ==2849== Invalid read of size 1  ==2849==    at 0x400603: main (valgrind.c:30)  ==2849==  Address 0x51b0040 is 0 bytes inside a block of size 1 free'd  ==2849==    at 0x4C270BD: free (vg_replace_malloc.c:366)  ==2849==    by 0x4005FE: main (valgrind.c:29)  ==2849==  ==2849==  ==2849== HEAP SUMMARY:  ==2849==     in use at exit: 0 bytes in 0 blocks  ==2849==   total heap usage: 1 allocs, 1 frees, 1 bytes allocated  ==2849==  ==2849== All heap blocks were freed -- no leaks are possible  ==2849==  ==2849== For counts of detected and suppressed errors, rerun with: -v  ==2849== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 4 from 4)

从上面的输出内容可以看到,Valgrind检测到了无效的读取操作然后输出了警告 ‘Invalid read of size 1′.

另注,使用gdb来调试c程序.

3. 从已分配内存块的尾部进行读/写

Code :

#include <stdio.h>  #include <stdlib.h>     int main(void)  {      char *p = malloc(1);      *p = 'a';         char c = *(p+1);         printf("\n [%c]\n",c);         free(p);      return 0;  }

在上面的代码中,我们已经为‘p’分配了一个字节的内存,但我们在将值读取到 ‘c’中的时候使用的是地址p+1.

现在我们使用Valgrind运行上面的代码 :

$ valgrind --tool=memcheck ./val  ==2835== Memcheck, a memory error detector  ==2835== Copyright (C) 2002-2009, and GNU GPL'd, by Julian Seward et al.  ==2835== Using Valgrind-3.6.0.SVN-Debian and LibVEX; rerun with -h for copyright info  ==2835== Command: ./val  ==2835==  ==2835== Invalid read of size 1  ==2835==    at 0x4005D9: main (valgrind.c:25)  ==2835==  Address 0x51b0041 is 0 bytes after a block of size 1 alloc'd  ==2835==    at 0x4C274A8: malloc (vg_replace_malloc.c:236)  ==2835==    by 0x4005C5: main (valgrind.c:22)  ==2835==      []  ==2835==  ==2835== HEAP SUMMARY:  ==2835==     in use at exit: 0 bytes in 0 blocks  ==2835==   total heap usage: 1 allocs, 1 frees, 1 bytes allocated  ==2835==  ==2835== All heap blocks were freed -- no leaks are possible  ==2835==  ==2835== For counts of detected and suppressed errors, rerun with: -v  ==2835== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 4 from 4)

同样,该工具在这种情况下也检测到了无效的读取操作.

4. 内存泄露

Code:

#include <stdio.h>  #include <stdlib.h>     int main(void)  {      char *p = malloc(1);      *p = 'a';         char c = *p;         printf("\n [%c]\n",c);         return 0;  }

在这次的代码中, 我们申请了一个字节但是没有将它释放.现在让我们运行Valgrind看看会发生什么:

$ valgrind --tool=memcheck --leak-check=full ./val  ==2888== Memcheck, a memory error detector  ==2888== Copyright (C) 2002-2009, and GNU GPL'd, by Julian Seward et al.  ==2888== Using Valgrind-3.6.0.SVN-Debian and LibVEX; rerun with -h for copyright info  ==2888== Command: ./val  ==2888==      [a]  ==2888==  ==2888== HEAP SUMMARY:  ==2888==     in use at exit: 1 bytes in 1 blocks  ==2888==   total heap usage: 1 allocs, 0 frees, 1 bytes allocated  ==2888==  ==2888== 1 bytes in 1 blocks are definitely lost in loss record 1 of 1  ==2888==    at 0x4C274A8: malloc (vg_replace_malloc.c:236)  ==2888==    by 0x400575: main (valgrind.c:6)  ==2888==  ==2888== LEAK SUMMARY:  ==2888==    definitely lost: 1 bytes in 1 blocks  ==2888==    indirectly lost: 0 bytes in 0 blocks  ==2888==      possibly lost: 0 bytes in 0 blocks  ==2888==    still reachable: 0 bytes in 0 blocks  ==2888==         suppressed: 0 bytes in 0 blocks  ==2888==  ==2888== For counts of detected and suppressed errors, rerun with: -v  ==2888== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 4 from 4)

输出行(上面加粗的部分)显示,该工具能够检测到内存的泄露.

注意: 在这里我们增加了一个选项‘–leak-check=full’来得到内存泄露的详细细节.

5. 不匹配地使用malloc/new/new[] 和 free/delete/delete[]

Code:

#include <stdio.h>  #include <stdlib.h>  #include<iostream>     int main(void)  {      char *p = (char*)malloc(1);      *p = 'a';         char c = *p;         printf("\n [%c]\n",c);      delete p;      return 0;  }

上面的代码中,我们使用了malloc()来分配内存,但是使用了delete操作符来删除内存.

注意 : 使用g++来编译上面的代码,因为delete操作符是在C++中引进的,而要编译C++需要使用g++.

让我们运行来看一下 :

$ valgrind --tool=memcheck --leak-check=full ./val  ==2972== Memcheck, a memory error detector  ==2972== Copyright (C) 2002-2009, and GNU GPL'd, by Julian Seward et al.  ==2972== Using Valgrind-3.6.0.SVN-Debian and LibVEX; rerun with -h for copyright info  ==2972== Command: ./val  ==2972==      [a]  ==2972== Mismatched free() / delete / delete []  ==2972==    at 0x4C26DCF: operator delete(void*) (vg_replace_malloc.c:387)  ==2972==    by 0x40080B: main (valgrind.c:13)  ==2972==  Address 0x595e040 is 0 bytes inside a block of size 1 alloc'd  ==2972==    at 0x4C274A8: malloc (vg_replace_malloc.c:236)  ==2972==    by 0x4007D5: main (valgrind.c:7)  ==2972==  ==2972==  ==2972== HEAP SUMMARY:  ==2972==     in use at exit: 0 bytes in 0 blocks  ==2972==   total heap usage: 1 allocs, 1 frees, 1 bytes allocated  ==2972==  ==2972== All heap blocks were freed -- no leaks are possible  ==2972==  ==2972== For counts of detected and suppressed errors, rerun with: -v  ==2972== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 4 from 4)

从上面的输出可以看到 (加粗的行), Valgrind清楚的说明了‘不匹配的使用了free() / delete / delete []‘

你可以尝试在测试代码中使用'new'和'free'进行组合来看看Valgrind给出的结果是什么.

6. 两次释放内存

Code :

#include <stdio.h>  #include <stdlib.h>     int main(void)  {      char *p = (char*)malloc(1);      *p = 'a';         char c = *p;      printf("\n [%c]\n",c);      free(p);      free(p);      return 0;  }

在上面的代码中, 我们两次释放了'p'指向的内存. 现在让我们运行memcheck :

$ valgrind --tool=memcheck --leak-check=full ./val  ==3167== Memcheck, a memory error detector  ==3167== Copyright (C) 2002-2009, and GNU GPL'd, by Julian Seward et al.  ==3167== Using Valgrind-3.6.0.SVN-Debian and LibVEX; rerun with -h for copyright info  ==3167== Command: ./val  ==3167==      [a]  ==3167== Invalid free() / delete / delete[]  ==3167==    at 0x4C270BD: free (vg_replace_malloc.c:366)  ==3167==    by 0x40060A: main (valgrind.c:12)  ==3167==  Address 0x51b0040 is 0 bytes inside a block of size 1 free'd  ==3167==    at 0x4C270BD: free (vg_replace_malloc.c:366)  ==3167==    by 0x4005FE: main (valgrind.c:11)  ==3167==  ==3167==  ==3167== HEAP SUMMARY:  ==3167==     in use at exit: 0 bytes in 0 blocks  ==3167==   total heap usage: 1 allocs, 2 frees, 1 bytes allocated  ==3167==  ==3167== All heap blocks were freed -- no leaks are possible  ==3167==  ==3167== For counts of detected and suppressed errors, rerun with: -v  ==3167== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 4 from 4)

从上面的输出可以看到(加粗的行), 该功能检测到我们对同一个指针调用了两次释放内存操作.

在本文中,我们把注意力放在了内存管理框架Valgrind,然后使用memcheck(Valgrind框架提供的)工具来了解它是如何降低需要经常操作内存的程序员的负担的. 该工具能够检测到很多手动检测不到的与内存相关的问题