获取崩溃时的调用栈和生成dump文件,然后自动重启

首先要说明的是：

 linux 下 比较方便可以得到 崩溃时的调用栈，win下 比较难办
 

1. linux 获取调用栈

代码奉上：
  1 #include <execinfo.h> //在头文件"execinfo.h"中声明了三个函数用于获取当前线程的函数调用堆栈
  2 #include <fcntl.h>
  3 #include <sys/resource.h>
  4 
  5 #include <signal.h>
  6 static int _core_dump_signals[] = {
  7     SIGABRT, SIGFPE, SIGILL, SIGQUIT, SIGSEGV,
  8     SIGTRAP, SIGSYS, SIGBUS, SIGXCPU, SIGXFSZ
  9 #ifdef SIGEMT
 10     ,SIGEMT
 11 #endif
 12 };
 13 //安装崩溃时//打印当前线程的函数调用堆栈
 14 void installTrace()
 15 {
 16     struct sigaction act;
 17     size_t i;
 18     memset(&act, 0, sizeof(act));
 19 
 20     act.sa_handler = &CLog::LogTrace;
 21     sigfillset(&act.sa_mask);
 22 
 23     for (i = 0; i < (sizeof(_core_dump_signals) / sizeof((_core_dump_signals)[0])); i++)
 24         sigaction(_core_dump_signals[i], &act, NULL);
 25     /* unblock all the signals, because if the current process is
 26     * spawned in the previous signal handler, all the signals are
 27     * blocked. In order to make it sense of signals, we should
 28     * unblock them. Certainly, you should call this function as
 29     * early as possible. :) */
 30     sigprocmask(SIG_UNBLOCK, &act.sa_mask, NULL);
 31 
 32     //sigemptyset(&myAction.sa_mask);
 33     //myAction.sa_flags = SA_RESTART | SA_SIGINFO;
 34     //LOG(LOGLV_WARN, "Log::installTrace!!!\n");
 35 }
 36 //#include <sys/wait.h>
 37 ////等待fok的子进程信号,防止僵尸进程
 38 //void fok_chld_wait(int signo) {
 39 //    pid_t pid;
 40 //    int stat;
 41 //    while ((pid = waitpid(-1, &stat, WNOHANG)) > 0) {
 42 //        printf("child %d exit\n", pid);
 43 //    }
 44 //    return;
 45 //}
 46 ////设置等待fok的子进程信号
 47 //int set_fok_chld_wait_signal() {
 48 //    struct sigaction act, oact;
 49 //    act.sa_handler = fok_chld_wait;
 50 //    sigemptyset(&act.sa_mask);
 51 //    act.sa_flags = 0;
 52 //
 53 //    if (sigaction(SIGCHLD, &act, &oact) < 0) {
 54 //        return -1;
 55 //    }
 56 //    return 0;
 57 //}
 58 //重新启动进程
 59 static void _restart(void)
 60 {
 61     char buf[4096];
 62     char exe[512];
 63     char **argv;
 64     int  fd, len;
 65     size_t argv_size, i;
 66     char *ptr, *ptr_end;
 67     struct rlimit limit;
 68 
 69     /* find the file executable. */
 70     len = readlink("/proc/self/exe", exe, sizeof(exe) - 1);
 71     if (len == -1 || len == sizeof(exe) - 1)
 72         _exit(EXIT_FAILURE);
 73     exe[len] = '\0';
 74 
 75     /* generate the variable argv. */
 76     fd = open("/proc/self/cmdline", O_RDONLY);
 77     if (fd == -1)
 78         _exit(EXIT_FAILURE);
 79     len = read(fd, buf, sizeof(buf));
 80     if (len == -1 || len == sizeof(buf))
 81         _exit(EXIT_FAILURE);
 82     buf[len] = '\0';
 83     argv_size = 16;
 84     argv = (char**)malloc(sizeof(char*) * argv_size);
 85     if (argv == NULL)
 86         _exit(EXIT_FAILURE);
 87     for (i = 0, ptr = buf, ptr_end = buf + len;
 88         ptr < ptr_end; i++, ptr += strlen(ptr) + 1) {
 89         if (i >= argv_size - 1) {
 90             argv_size <<= 1;
 91             argv = (char**)realloc(argv, sizeof(char*) * argv_size);
 92             if (argv == NULL)
 93                 _exit(EXIT_FAILURE);
 94         }
 95         argv[i] = ptr;
 96     }
 97     argv[i] = NULL;
 98 
 99     /* close all the file descriptors except of stdin/stdout/stderr. */
100     getrlimit(RLIMIT_NOFILE, &limit);
101     for (i = 3; i < limit.rlim_cur; i++)
102         close(i);
103 
104     execvp(exe, argv);
105 
106     /* exit if it fails to restart self. */
107     _exit(EXIT_FAILURE);
108 }
109 //打印当前线程的函数调用堆栈
110 void LogTrace(int signo)
111 {
112     //set_fok_chld_wait_signal();
113     //父进程不关心子进程什么时候结束，那么可以用signal(SIGCHLD, SIG_IGN)通知内核，自己对子进程的结束不感兴趣，
114     //那么子进程结束后，内核会回收， 并不再给父进程发送信号。
115     //或用sigaction函数为SIGCHLD设置SA_NOCLDWAIT，这样子进程结束后，就不会进入僵死状态
116     struct sigaction sa;
117     sa.sa_handler = SIG_IGN;
118     sa.sa_flags = SA_NOCLDWAIT;
119     sigemptyset(&sa.sa_mask);
120     sigaction(SIGCHLD, &sa, NULL);
121 
122     void *callstacks[20];
123     if (!_pLogFile)
124         Log.ChangeFile();
125 
126     if (_pLogFile)
127     {
128         std::lock_guard<std::mutex> lock{ _csLog };
129         struct timeval  tv;
130         struct tm       *p;
131         gettimeofday(&tv, NULL);
132         p = localtime(&tv.tv_sec);
133 
134         size_t size = sizeof(callstacks);
135         int fd = fileno(_pLogFile);
136         fprintf(_pLogFile, "-----------------------------------0 SIG=%d-%s  err=%d-%s  tid=%d  %d-%02d-%02d %02d:%02d:%02d.%06d\n"
137             , signo, strsignal(signo), errno, strerror(errno), gettid(), p->tm_year + 1900, p->tm_mon + 1, p->tm_mday, p->tm_hour, p->tm_min, (int)p->tm_sec, (int)tv.tv_usec );
138         fflush(_pLogFile);
139         size = backtrace(callstacks, size);
140         backtrace_symbols_fd(callstacks, size, fd);
141         fprintf(_pLogFile, "-----------------------------------1 callstack=20/%d\n", (int)size);
142         fflush(_pLogFile);
143         //定位源代码行
144         char cmd[264] = "addr2line -f -e ";
145         char buff1[256] = { 0 }, buff2[1024] = { 0 }, buff3[256] = { 0 };
146         char* prog = cmd + 16;
147         readlink("/proc/self/exe", prog, 247);// 获取进程的完整路径
148         int leftlen = strlen(cmd);
149         prog = cmd + leftlen;
150         leftlen = 263 - leftlen;
151         for (size_t i = 0; i < size; ++i)
152         {
153             snprintf(prog, leftlen, " %p\n", callstacks[i]);
154             FILE* fp = popen(cmd, "r");
155             if (fp != NULL)
156             {
157                 if (NULL != fgets(buff1, 255, fp))
158                 {
159                     fgets(buff2, 1023, fp);
160                     if (buff2[strlen(buff2) - 1] == '\n')
161                         buff2[strlen(buff2) - 1] = 0;
162                     snprintf(buff3, 255, "c++filt %s\n", buff1);//用c++filt 使函数名可视化
163                     FILE* fp2 = popen(buff3, "r");
164                     if (fp2 && fgets(buff3, 255, fp2)) {
165                         fprintf(_pLogFile, ">%s %s", buff2, buff3);
166                         pclose(fp2);
167                     }
168                     else
169                         fprintf(_pLogFile, ">%s %s", buff2, buff1);
170                 }
171                 pclose(fp);
172             }
173         }
174         fprintf(_pLogFile, "-----------------------------------2 end\n");
175         fflush(_pLogFile);
176     }
177 
178     //core dump
179     pid_t pid = fork();
180     if (pid == 0) {
181 #ifdef _FORCE_CORE_DUMP
182 #ifndef _CORE_SIZE
183 #define _CORE_SIZE (256 * 1024 * 1024)
184 #endif /* _CORE_SIZE */
185         struct rlimit limit = {
186             .rlim_cur = _CORE_SIZE,
187             .rlim_max = _CORE_SIZE };
188 
189         setrlimit(RLIMIT_CORE, &limit);
190 #endif /* _FORCE_CORE_DUMP */
191         /* reset the signal handler to default handler,
192         * then raise the corresponding signal. */
193         signal(signo, SIG_DFL);
194         raise(signo);
195         _exit(EXIT_FAILURE);
196     }
197     else
198         _restart();//自动重启
199 }
200 /*    const char* sigstr[] = {
201         "01 SIGHUP 挂起（hangup）",
202         "02 SIGINT 中断，当用户从键盘按^c键或^break键时",
203         "03 SIGQUIT 退出，当用户从键盘按quit键时",
204         "04 SIGILL 非法指令",
205         "05 SIGTRAP 跟踪陷阱（trace trap），启动进程，跟踪代码的执行",
206         "06 SIGIOT IOT指令",
207         "07 SIGEMT EMT指令",
208         "08 SIGFPE 浮点运算溢出",
209         "09 SIGKILL 杀死、终止进程 ",
210         "10 SIGBUS 总线错误",
211         "11 SIGSEGV 段违例（segmentation  violation），进程试图去访问其虚地址空间以外的位置",
212         "12 SIGSYS 系统调用中参数错，如系统调用号非法",
213         "13 SIGPIPE 向某个非读管道中写入数据",
214         "14 SIGALRM 闹钟。当某进程希望在某时间后接收信号时发此信号",
215         "15 SIGTERM 软件终止（software  termination）",
216         "16 SIGUSR1 用户自定义信号1",
217         "17 SIGUSR2 用户自定义信号2",
218         "18 SIGCLD 某个子进程死",
219         "19 SIGPWR 电源故障"
220     };*/
221 
222 
223 //设置线程优先级(设置为普通优先级)
224 //参数为线程句柄,或 std::thread::native_handle()
225 void SetThrPriorityNormal(void* handle)
226 {
227     sched_param sch;
228     int policy, policy2;
229     pthread_getschedparam(handle, &policy, &sch);
230     sch.sched_priority = 0;
231     if (pthread_setschedparam(handle, SCHED_FIFO, &sch))
232         LOG(LOGLV_NOTICE, "Failed to setschedparam normal: %s \n", std::strerror(errno));
233     pthread_getschedparam(handle, &policy2, &sch);
234     LOG(LOGLV_NOTICE, "ThreadPriority  %d -> %d\n", policy, policy2);
235 }
236 //设置线程优先级(设置为高(HIGHEST)优先级)
237 //参数为线程句柄,或 std::thread::native_handle()
238 void SetThrPriorityHighst(void* handle)
239 {
240     sched_param sch;
241     int policy, policy2;
242     pthread_getschedparam(handle, &policy, &sch);
243     sch.sched_priority = -10;
244     if (pthread_setschedparam(handle, SCHED_FIFO, &sch))
245         LOG(LOGLV_NOTICE, "Failed to setschedparam normal: %s \n", std::strerror(errno));
246     pthread_getschedparam(handle, &policy2, &sch);
247     LOG(LOGLV_NOTICE, "ThreadPriority  %d -> %d\n", policy, policy2);
248 }

这些也是网上搜别人的代码，然后自己修改了的，可以直接运。

编译时，要加参数 -g 。或者加 -g3 -ggdb -gstabs。

 

2. win获取崩溃类型和生成dump文件

代码奉上:
  1 //Dump调试
  2 #include <DbgHelp.h>
  3 #include <shellapi.h>
  4 
  5 typedef  BOOL(*MINIDUMPWRITEDUMP)(HANDLE hProcess,
  6     DWORD ProcessId,
  7     HANDLE hFile,
  8     MINIDUMP_TYPE DumpType,
  9     PMINIDUMP_EXCEPTION_INFORMATION ExceptionParam,
 10     PMINIDUMP_USER_STREAM_INFORMATION UserStreamParam,
 11     PMINIDUMP_CALLBACK_INFORMATION CallbackParam);
 12 
 13 MINIDUMPWRITEDUMP  pDumpFunc = NULL; // 声明函数指针
 14 const char* seh_filer(DWORD code)
 15 {
 16     switch (code)
 17     {
 18     case EXCEPTION_ACCESS_VIOLATION:
 19         return ("存储保护异常(解空指针)");
 20     case EXCEPTION_DATATYPE_MISALIGNMENT:
 21         return ("数据类型未对齐异常");
 22     case EXCEPTION_BREAKPOINT:
 23         return ("中断异常");
 24     case EXCEPTION_SINGLE_STEP:
 25         return ("单步中断异常");
 26     case EXCEPTION_ARRAY_BOUNDS_EXCEEDED:
 27         return ("数组越界异常");
 28     case EXCEPTION_FLT_DENORMAL_OPERAND:
 29     case EXCEPTION_FLT_DIVIDE_BY_ZERO:
 30     case EXCEPTION_FLT_INEXACT_RESULT:
 31     case EXCEPTION_FLT_INVALID_OPERATION:
 32     case EXCEPTION_FLT_OVERFLOW:
 33     case EXCEPTION_FLT_STACK_CHECK:
 34     case EXCEPTION_FLT_UNDERFLOW:
 35         return ("浮点数计算异常");
 36     case EXCEPTION_INT_DIVIDE_BY_ZERO:
 37         return ("被0除异常");
 38     case EXCEPTION_INT_OVERFLOW:
 39         return ("数据溢出异常");
 40     case EXCEPTION_IN_PAGE_ERROR:
 41         return ("页错误异常");
 42     case EXCEPTION_ILLEGAL_INSTRUCTION:
 43         return ("非法指令异常");
 44     case EXCEPTION_STACK_OVERFLOW:
 45         return ("堆栈溢出异常");
 46     case EXCEPTION_INVALID_HANDLE:
 47         return ("无效句柄异常");
 48     default:
 49         if (code & (1 << 29))
 50             return ("用户自定义的软件异常");
 51         else
 52             return ("其它异常");
 53     }
 54 }
 55 // 参数lpExceptionInfo包含了异常信息，由系统提供
 56 LONG WINAPI ExceptionFilterFunc(struct _EXCEPTION_POINTERS* lpExceptionInfo)
 57 {
 58     //if(pDumpFunc==NULL) return -1;
 59 
 60     LONG ret = EXCEPTION_CONTINUE_SEARCH;
 61     // 根据当前线程ID,时间创建dmp文件
 62     int irand = rand() % 10000;
 63     char strFile[100];
 64     SYSTEMTIME st;
 65     DWORD ul = lpExceptionInfo->ExceptionRecord->ExceptionCode;
 66     LOG(LOGLV_CRIT, "exception %X,%s\n", ul, seh_filer(ul));
 67     GetLocalTime(&st);
 68     ::CreateDirectory("dumps\\", NULL);
 69     _snprintf_s(strFile, 99, "dumps\\DMP-%d_%04d%02d%02d_%02d%02d%02d.%03d_%04d.dmp", ::GetCurrentThreadId(), st.wYear, st.wMonth, st.wDay, st.wHour, st.wMinute, st.wSecond, st.wMilliseconds, irand);
 70     HANDLE hFile = ::CreateFile(strFile, GENERIC_WRITE, FILE_SHARE_WRITE, NULL, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL);
 71     if (hFile != INVALID_HANDLE_VALUE)
 72     {
 73         MINIDUMP_EXCEPTION_INFORMATION ExInfo;
 74         ExInfo.ThreadId = ::GetCurrentThreadId();
 75         ExInfo.ExceptionPointers = lpExceptionInfo;
 76         ExInfo.ClientPointers = NULL;
 77         MINIDUMP_TYPE mdt = (MINIDUMP_TYPE)(//MiniDumpWithFullMemory |
 78             MiniDumpWithHandleData | MiniDumpFilterModulePaths |
 79             MiniDumpWithProcessThreadData |
 80             MiniDumpWithIndirectlyReferencedMemory |
 81             MiniDumpWithPrivateReadWriteMemory |
 82             MiniDumpWithFullMemoryInfo |
 83             MiniDumpWithUnloadedModules |
 84             //MiniDumpWithDataSegs |
 85             MiniDumpWithThreadInfo );
 86 
 87         //加载DebugDump函数 //MiniDumpWriteDump 函数在头文件(DbgHelp.h)中有定义
 88         HMODULE hDll = LoadLibrary("dbghelp.dll");
 89         if (hDll == NULL)
 90             LOG(LOGLV_ALERT, "Error: load dbghelp.dll fail\n\n");
 91         else
 92         {
 93             //load MiniDumpWriteDump function
 94             pDumpFunc = (MINIDUMPWRITEDUMP)::GetProcAddress(hDll, "MiniDumpWriteDump");
 95             if (pDumpFunc == NULL)
 96                 LOG(LOGLV_ALERT, "Error: load MiniDumpWriteDump function fail\n\n");
 97             else {
 98                 //将崩溃信息记录到dmp文件中
 99                 BOOL bRet = pDumpFunc(GetCurrentProcess(), GetCurrentProcessId(), hFile, mdt, &ExInfo, NULL, NULL);
100                 if (bRet) {
101                     ret = EXCEPTION_EXECUTE_HANDLER;
102                     LOG(LOGLV_WARN, "DumpFile: %s\n", strFile);
103                 }
104                 else
105                     LOG(LOGLV_CRIT, "call MiniDumpWriteDump fail...\n");
106             }
107             //卸载 dbghelp.dll
108             if (hDll != NULL)
109                 FreeLibrary(hDll);
110         }
111         ::CloseHandle(hFile);
112     }
113     else
114         LOG(LOGLV_CRIT, "CreateFile fail... %s\n", strFile);
115     char *fullpth;
116     _get_pgmptr(&fullpth);
117     char* cmdl = strstr(GetCommandLine(), ".exe ");
118     ShellExecute(NULL, "open", fullpth, (cmdl ? (cmdl + 5) : nullptr), Log.CurPath(), SW_SHOW);
119     return ret;
120 }
121 
122 //安装崩溃时//打印当前线程的函数调用堆栈
123 //该流程只对本线程有效，如果是多线程，需要对每个线程都做调用处理。
124 void installTrace() {
125     // 设置异常捕获函数
126     ::SetUnhandledExceptionFilter(ExceptionFilterFunc);
127 }
128 
129 //设置线程优先级(设置为普通优先级)
130 ////如果进程的优先级为最高,则降低2级
131 //参数为线程句柄,或 std::thread::native_handle()
132 void SetThrPriorityNormal(void* handle)
133 {
134     int r0 = GetThreadPriority(handle);
135     int r = 0;
136     //int p = GetPriorityClass(GetCurrentProcess());
137     //if (p > 100)//如果进程的优先级为最高,则降低2级
138     //    r = -2;
139     int r1 = SetThreadPriority(handle, r);
140         r1 = GetThreadPriority(handle);
141     LOG(LOGLV_NOTICE, "ThreadPriority  %d -> %d\n", r0, r1);
142 }
143 //设置线程优先级(设置为高(HIGHEST)优先级)
144 //参数为线程句柄,或 std::thread::native_handle()
145 void SetThrPriorityHighst(void* handle)
146 {
147     int r0 = GetThreadPriority(handle);
148     int r1 = SetThreadPriority(handle, THREAD_PRIORITY_HIGHEST);
149         r1 = GetThreadPriority(handle);
150     LOG(LOGLV_NOTICE, "ThreadPriority  %d -> %d\n", r0, r1);
151 }

网上搜索，说是 win有办法可以获取崩溃的函数和类名，但是比较复杂，就没有搞了。

 

3. 使用方式
在线程开始处调用 installTrace 安装异常捕获即可。
注意：每条线程都安装，如果要捕获未处理异常的话。
posted @ 2018-09-12 10:51  _Ong  阅读(2035)  评论(2)  编辑  收藏  举报