linux 捕获信号处理中遇到的死锁

tag: 信号 signal  sigchld  死锁 堆栈

我们的程序需要捕获信号自己处理，所以尝试对1-32的信号处理（后面33-64的信号不处理）。
但是在调试代码时，发现一个线程死锁的问题。
程序目的：捕获信号，然后打印堆栈。

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

伪代码如下： 设置捕获信号函数() {  //设置信号处理函数  sigact.sa_sigaction = TsSigHandler;      //  //这里捕获了很多信号，包括SIGCHLD：子进程结束，父进程会收到该信号  sigaction( SIGSEGV, &sigact, NULL );  ....  sigaction( SIGCHLD, &sigact, NULL );     }    信号处理函数：TsSigHandler {  //调用打印堆栈函数  PrintStack(); }    打印堆栈函数PrintStack {  //打印堆栈  backtrace();  backtrace_symbols();      //调用system函数执行一些命令  system("xxxxxx"); }

Thread 12 (Thread 0xf7dd2b90 (LWP 5770)):
以下是一个让我觉得奇怪的堆栈，奇怪之处：
1.死锁了：__lll_lock_wait_private
2.获得了2个信号：<signal handler called>，为什么不是一个一个信号处理

堆栈如下：

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

#0  0xffffe410 in __kernel_vsyscall () #1  0x002a0783 in __lll_lock_wait_private () from /lib/libc.so.6 #2  0x001f8448 in _L_lock_124 () from /lib/libc.so.6 #3  0x001f7f8b in do_system () from /lib/libc.so.6 #4  0x001f8412 in system () from /lib/libc.so.6 #5  0x00317ead in system () from /lib/libpthread.so.0 #6  0x080f95c1 in PrintStack() () #7  0x080f9844 in TsSigHandler(int, siginfo*, void*) () #8  <signal handler called> #9  0xffffe410 in __kernel_vsyscall () #10 0x001eb1a9 in sigprocmask () from /lib/libc.so.6 #11 0x001f8132 in do_system () from /lib/libc.so.6 #12 0x001f8412 in system () from /lib/libc.so.6 #13 0x00317ead in system () from /lib/libpthread.so.0 #14 0x080f95c1 in PrintStack() () #15 0x080f9844 in TsSigHandler(int, siginfo*, void*) () #16 <signal handler called> #17 0x002338ec in memcpy () from /lib/libc.so.6 #18 0x0804fa02 in boom () #19 0x080dbd9c in RunCmd () #20 0x080dbf12 in CmdParse () #21 0x080dc705 in OspTeleDaemon () #22 0x080f8817 in OspTaskTemplateFunc(void*) () #23 0x0030f832 in start_thread () from /lib/libpthread.so.0 #24 0x00293e0e in clone () from /lib/libc.so.6

#18 0x0804fa02 in boom ()
 boom()是我写的一个制造崩溃的函数：
 
 char *pBoom = NULL;
 memcpy( pBoom, "aaaa", 100 );

#16 <signal handler called>
触发信号

#15 0x080f9844 in TsSigHandler(int, siginfo*, void*) ()
 TsSigHandler是信号处理函数。通过以下代码设置：
 struct sigaction sigact;
 sigemptyset( &sigact.sa_mask );
 sigact.sa_flags = SA_ONESHOT | SA_SIGINFO;
 sigact.sa_sigaction = TsSigHandler;
 信号触发后，由TsSigHandler函数处理
 
#14 0x080f95c1 in PrintStack() ()
 TsSigHandler函数中调用PrintStack函数打印堆栈。

#13 0x00317ead in system () from /lib/libpthread.so.0
 PrintStack函数中调用了system函数做一些额外的事情，例如执行gcore（事实证明，这种方法是有点问题的）。

#11 0x001f8132 in do_system () from /lib/libc.so.6
 system调用了do_system
 
#10 0x001eb1a9 in sigprocmask () from /lib/libc.so.6
 do_system调用sigprocmask

#8  <signal handler called>
 关键来了：这是获取到了另外一个信号：SIGCHLD。

#7  0x080f9844 in TsSigHandler(int, siginfo*, void*) ()
 又调用信号处理函数TsSigHandler
#3  0x001f7f8b in do_system () from /lib/libc.so.6
 system调用do_system，调用流程和上面当然是一样的
 
#2  0x001f8448 in _L_lock_124 () from /lib/libc.so.6
#1  0x002a0783 in __lll_lock_wait_private () from /lib/libc.so.6
 nice！锁住了。。

分析:

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123

/* Execute LINE as a shell command, returning its status.  */ static int do_system (const char *line) {   int status, save;   pid_t pid;   struct sigaction sa; #ifndef _LIBC_REENTRANT   struct sigaction intr, quit; #endif   sigset_t omask;     sa.sa_handler = SIG_IGN;   sa.sa_flags = 0;   __sigemptyset (&sa.sa_mask);     DO_LOCK ();   if (ADD_REF () == 0)     {       if (__sigaction (SIGINT, &sa, &intr) < 0)     {       (void) SUB_REF ();       goto out;     }       if (__sigaction (SIGQUIT, &sa, &quit) < 0)     {       save = errno;       (void) SUB_REF ();       goto out_restore_sigint;     }     }   DO_UNLOCK ();     /* We reuse the bitmap in the 'sa' structure.  */   __sigaddset (&sa.sa_mask, SIGCHLD);   save = errno;   if (__sigprocmask (SIG_BLOCK, &sa.sa_mask, &omask) < 0)     { #ifndef _LIBC       if (errno == ENOSYS)     __set_errno (save);       else #endif     {       DO_LOCK ();       if (SUB_REF () == 0)         {           save = errno;           (void) __sigaction (SIGQUIT, &quit, (struct sigaction *) NULL);         out_restore_sigint:           (void) __sigaction (SIGINT, &intr, (struct sigaction *) NULL);           __set_errno (save);         }     out:       DO_UNLOCK ();       return -1;     }     }   #ifdef CLEANUP_HANDLER   CLEANUP_HANDLER; #endif   #ifdef FORK   pid = FORK (); #else   pid = __fork (); #endif   if (pid == (pid_t) 0)     {       /* Child side.  */       const char *new_argv[4];       new_argv[0] = SHELL_NAME;       new_argv[1] = "-c";       new_argv[2] = line;       new_argv[3] = NULL;         /* Restore the signals.  */       (void) __sigaction (SIGINT, &intr, (struct sigaction *) NULL);       (void) __sigaction (SIGQUIT, &quit, (struct sigaction *) NULL);       (void) __sigprocmask (SIG_SETMASK, &omask, (sigset_t *) NULL);       INIT_LOCK ();         /* Exec the shell.  */       (void) __execve (SHELL_PATH, (char *const *) new_argv, __environ);       _exit (127);     }   else if (pid < (pid_t) 0)     /* The fork failed.  */     status = -1;   else     /* Parent side.  */     {       /* Note the system() is a cancellation point.  But since we call      waitpid() which itself is a cancellation point we do not      have to do anything here.  */       if (TEMP_FAILURE_RETRY (__waitpid (pid, &status, 0)) != pid)     status = -1;     }   #ifdef CLEANUP_HANDLER   CLEANUP_RESET; #endif     save = errno;   DO_LOCK ();   if ((SUB_REF () == 0        && (__sigaction (SIGINT, &intr, (struct sigaction *) NULL)        | __sigaction (SIGQUIT, &quit, (struct sigaction *) NULL)) != 0)       || __sigprocmask (SIG_SETMASK, &omask, (sigset_t *) NULL) != 0)     { #ifndef _LIBC       /* glibc cannot be used on systems without waitpid.  */       if (errno == ENOSYS)     __set_errno (save);       else #endif     status = -1;     }   DO_UNLOCK ();     return status; }

system()函数执行的大体过程是：fork()->exec()->waitpid()，
waitpid用于等待子进程执行完毕。
但是在子进程执行完毕时，会产生SIGCHLD信号，
而SIGCHLD信号会唤醒wait中的进程，这就是看到了2个信号的原因，

解决方法：
1.忽略SIGCHLD信号：其实这个信号一般情况下应该被忽略，除非你的程序需要对这种情况做非常特殊的处理
2.不要在这里调用system()

to do： 有空了记得补详细些