1. 程式人生 > >記錄一次gdb debug經歷

記錄一次gdb debug經歷

目錄

  • 問題描述
  • 檢視core檔案
  • 使用gdb檢視core檔案
  • 總結

問題描述

今天在寫程式碼時,執行時奔潰了。segment fault,而且是在程式退出main()函式後,才報的。
唯一的資訊是:Segmentation fault (core dumped)
簡直是一頭霧水。

檢視core檔案

系統預設是不會生成core檔案的,ulimit -c unlimited把core檔案設為無限大。

使用gdb檢視core檔案

gdb ./example/sudoku_batch_test core


提示如下:

Program terminated with signal SIGSEGV, Segmentation fault.
#0  __GI___libc_free (mem=0x313030303030300a) at malloc.c:2951
2951    malloc.c: No such file or directory.
(gdb) 

可以確定崩潰發生在malloc.c中。但是提示沒有malloc.c的原始碼。

首先安裝glibc的符號表,命令如下:
sudo apt-get install libc6-dbg

再來是安裝glibc的原始檔,命令如下:
sudo apt-get source libc6-dev


安裝完畢後在當前目錄下會多出一個glibc-2.23資料夾,該資料夾包含了glibc的原始碼。

原始碼準備就緒後,接著上面,在gdb命令提示符下輸入:
directory glibc-2.23/malloc/將glibc-2.23/malloc/設為gdb原始碼搜尋目錄。結果如下:

warning: core file may not match specified executable file.
[New LWP 24491]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Core was generated by `./example/sudoku_batch_test ../example/test1000 127.0.0.1 1'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0  __GI___libc_free (mem=0x313030303030300a) at malloc.c:2951
2951    malloc.c: No such file or directory.
(gdb) directory glibc-2.23/malloc/
Source directories searched: /root/work/melon/build/glibc-2.23/malloc:$cdir:$cwd
(gdb) 

現在我們就可以在gdb中檢視崩潰處的原始碼了,執行list

(gdb) l
warning: Source file is more recent than executable.
2946      if (mem == 0)                              /* free(0) has no effect */
2947        return;
2948    
2949      p = mem2chunk (mem);
2950    
2951      if (chunk_is_mmapped (p))                       /* release mmapped memory. */
2952        {
2953          /* see if the dynamic brk/mmap threshold needs adjusting */
2954          if (!mp_.no_dyn_threshold
2955              && p->size > mp_.mmap_threshold
(gdb) 

雖然知道了崩潰發生在2951行,但是貌似沒有更多有效的資訊。這時我想到了是不是可以看下函式的呼叫棧,或許會有資訊。
接著執行backtrace(或者bt):

(gdb) bt
#0  __GI___libc_free (mem=0x313030303030300a) at malloc.c:2951
#1  0x000000000048bc9d in melon::Coroutine::~Coroutine (this=0x1fc9120, __in_chrg=<optimized out>)
    at /root/work/melon/src/Coroutine.cpp:56
#2  0x000000000048d099 in std::_Sp_counted_ptr<melon::Coroutine*, (__gnu_cxx::_Lock_policy)2>::_M_dispose (
    this=0x1fc8190) at /usr/include/c++/5/bits/shared_ptr_base.h:374
#3  0x00000000004630f1 in std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release (this=0x1fc8190)
    at /usr/include/c++/5/bits/shared_ptr_base.h:150
#4  0x0000000000461f32 in std::__shared_count<(__gnu_cxx::_Lock_policy)2>::~__shared_count (this=0x7f07f4ff1770, 
    __in_chrg=<optimized out>) at /usr/include/c++/5/bits/shared_ptr_base.h:659
#5  0x00000000004749ed in std::__shared_ptr<melon::Coroutine, (__gnu_cxx::_Lock_policy)2>::~__shared_ptr (
    this=0x7f07f4ff1768, __in_chrg=<optimized out>) at /usr/include/c++/5/bits/shared_ptr_base.h:925
#6  0x0000000000474a39 in std::shared_ptr<melon::Coroutine>::~shared_ptr (this=0x7f07f4ff1768, 
    __in_chrg=<optimized out>) at /usr/include/c++/5/bits/shared_ptr.h:93
#7  0x00007f07f40915ff in __GI___call_tls_dtors () at cxa_thread_atexit_impl.c:155
#8  0x00007f07f4090f27 in __run_exit_handlers (status=0, listp=0x7f07f441b5f8 <__exit_funcs>, 
    run_list_atexit=run_list_atexit@entry=true) at exit.c:40
#9  0x00007f07f4091045 in __GI_exit (status=<optimized out>) at exit.c:104
#10 0x00007f07f4077837 in __libc_start_main (main=0x45f1c4 <main(int, char**)>, argc=4, argv=0x7ffcfb2ab218, 
    init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7ffcfb2ab208)
    at ../csu/libc-start.c:325
#11 0x000000000045ec89 in _start ()

這下問題找到了,首先線上程結束或者程式執行結束會呼叫__GI___call_tls_dtors函式來析構執行緒本地儲存。我確實用了thread_local關鍵字修飾Coroutine::Ptr變數。
#1 0x000000000048bc9d in melon::Coroutine::~Coroutine可知在melon::Coroutine類的解構函式中呼叫了free()導致奔潰。
這下問題基本明確了,我在Coroutine解構函式中會釋放stack_這個指標,

 53 Coroutine::~Coroutine() {
 54     LOG_DEBUG << "destroy coroutine:" << name_;
 55     if (stack_) {
 56         free(stack_);
 57     }
 58 }

有兩個建構函式,其中一個如下:

 39 Coroutine::Coroutine()
 40     :c_id_(++t_coroutine_id),
 41     name_("Main-" + std::to_string(c_id_)),
 42     cb_(nullptr),
 43     state_(CoroutineState::INIT) {
 44 
 45     if (getcontext(&context_)) {
 46         LOG_ERROR << "getcontext: errno=" << errno
 47                 << " error string:" << strerror(errno);
 58     }
 59 }

因為大意犯了個非常低階的錯誤,這個建構函式沒有正確初始化statck_指標,將statck_初始化為nullptr後,問題就解決了。

總結

遇到這類問題,一般用gdb檢視core檔案都能定位到崩潰的位置,如果不是直接引發的,可以檢視函式呼叫棧,一般都能找到問題原