温馨提示×

温馨提示×

您好,登录后才能下订单哦!

密码登录×
登录注册×
其他方式登录
点击 登录注册 即表示同意《亿速云用户服务条款》

基于linuxthreads2.0.1线程源码如何分析线程库的初始化和线程的管理

发布时间:2021-12-09 09:36:21 阅读:128 作者:柒染 栏目:大数据
Linux服务器限时活动,0元免费领,库存有限,领完即止! 点击查看>>

今天就跟大家聊聊有关基于linuxthreads2.0.1线程源码如何分析线程库的初始化和线程的管理,可能很多人都不太了解,为了让大家更加了解,小编给大家总结了以下内容,希望大家根据这篇文章可以有所收获。

初步分析一下线程的初始化和管理。

线程库的初始化代码如下。

   
     
 
    
    // 在main函数之前执行该函数void __pthread_initialize(void) __attribute__((constructor));void __pthread_initialize(void){  struct sigaction sa;  sigset_t mask;  /* We may be called by others.  This may happen if the constructors     are not called in the order we need.  */  if (__pthread_initial_thread_bos != NULL)    return;  /* For the initial stack, reserve at least STACK_SIZE bytes of stack     below the current stack address, and align that on a     STACK_SIZE boundary. */  __pthread_initial_thread_bos =    // 按STACK_SIZE大小对齐    (char *)(((long)CURRENT_STACK_FRAME - 2 * STACK_SIZE) & ~(STACK_SIZE - 1));  /* Update the descriptor for the initial thread. */  // 即main函数代表的主进程id  __pthread_initial_thread.p_pid = getpid();  /* If we have special thread_self processing, initialize that for the     main thread now.  */#ifdef INIT_THREAD_SELF  INIT_THREAD_SELF(&__pthread_initial_thread);#endif  /* Setup signal handlers for the initial thread.     Since signal handlers are shared between threads, these settings     will be inherited by all other threads. */  // 为两个信号注册处理函数  sa.sa_handler = __pthread_sighandler;  sigemptyset(&sa.sa_mask);  sa.sa_flags = SA_RESTART; /* does not matter for regular threads, but                               better for the thread manager */  sigaction(PTHREAD_SIG_RESTART, &sa, NULL);  sa.sa_handler = pthread_handle_sigcancel;  sa.sa_flags = 0;  sigaction(PTHREAD_SIG_CANCEL, &sa, NULL);  /* Initially, block PTHREAD_SIG_RESTART. Will be unblocked on demand. */  // 屏蔽restart信号  sigemptyset(&mask);  sigaddset(&mask, PTHREAD_SIG_RESTART);  sigprocmask(SIG_BLOCK, &mask, NULL);  /* Register an exit function to kill all other threads. */  /* Do it early so that user-registered atexit functions are called     before pthread_exit_process. */  // 注册退出时执行的函数  __on_exit(pthread_exit_process, NULL);}
   
     
 
    
    AI代码助手复制代码

在执行main函数之前会先执行__pthread_initialize函数,该函数做的事情主要有

1 在栈上分配一块内存。

2 保存当前进程,进main函数对应的进程的pid。

3 注册两个信号处理函数。

4 注册退出时执行的函数

接下来我们会调用pthread_create进行线程的创建。我们来看看该函数做了什么。

   
     
 
    
    int pthread_create(pthread_t *thread, const pthread_attr_t *attr,                   void * (*start_routine)(void *), void *arg){  pthread_t self = thread_self();  struct pthread_request request;  // 还没执行过pthread_initialize_manager则执行,用于初始化manager线程  if (__pthread_manager_request < 0) {    if (pthread_initialize_manager() < 0) return EAGAIN;  }  // 给manager发一下请求  request.req_thread = self;  request.req_kind = REQ_CREATE;  request.req_args.create.attr = attr;  request.req_args.create.fn = start_routine;  request.req_args.create.arg = arg;  // 获取当前线程的信号掩码  sigprocmask(SIG_SETMASK, (const sigset_t *) NULL,              &request.req_args.create.mask);  // 通过管道写入,通知manager线程,新建一个线程  __libc_write(__pthread_manager_request, (char *) &request, sizeof(request));  // 挂起,等待manager唤醒  suspend(self);  // 等于0说明创建成功,否则返回失败的错误码,p_retval在pthread_handle_create中设置  if (self->p_retcode == 0) *thread = (pthread_t) self->p_retval;  return self->p_retcode;}
   
     
 
    
    AI代码助手复制代码

我们发现,该函数没有做实际的事情,他通过往管道写了一些数据。这时候就要先看pthread_initialize_manager函数了。

   
     
 
    
    static int pthread_initialize_manager(void){  int manager_pipe[2];  /* Setup stack for thread manager */  // 在堆上分配一块内存用于manager线程的栈  __pthread_manager_thread_bos = malloc(THREAD_MANAGER_STACK_SIZE);  if (__pthread_manager_thread_bos == NULL) return -1;  // limit  __pthread_manager_thread_tos =    __pthread_manager_thread_bos + THREAD_MANAGER_STACK_SIZE;  /* Setup pipe to communicate with thread manager */  if (pipe(manager_pipe) == -1) {    free(__pthread_manager_thread_bos);    return -1;  }  __pthread_manager_request = manager_pipe[1]; /* writing end */  __pthread_manager_reader = manager_pipe[0]; /* reading end */  /* Start the thread manager */  // 新建一个manager线程,manager_pipe是__thread_manager函数的入参  if (__clone(__pthread_manager,  __pthread_manager_thread_tos,  CLONE_VM | CLONE_FS | CLONE_FILES | CLONE_SIGHAND,  (void *)(long)manager_pipe[0]) == -1) {    free(__pthread_manager_thread_bos);    __libc_close(manager_pipe[0]);    __libc_close(manager_pipe[1]);    __pthread_manager_request = -1;    return -1;  }  return 0;}
   
     
 
    
    AI代码助手复制代码

该函数做了几件事情

1 在堆上申请一块内存用作manager线程的栈

2 创建了一个管道,用于manager线程和其他线程通信。

3 然后新建了一个进程,然后执行__pthread_manager函数。(具体可参考http://www.man7.org/linux/man-pages/man2/clone.2.html)

manager线程是linuxthreads线程库比较重要的存在,他是管理其他线程的线程。我们接着看_pthread_manager函数的代码。

   
     
 
    
    /* The server thread managing requests for thread creation and termination */int __pthread_manager(void *arg){  // 管道的读端  int reqfd = (long)arg;  sigset_t mask;  fd_set readfds;  struct timeval timeout;  int n;  struct pthread_request request;  /* If we have special thread_self processing, initialize it.  */#ifdef INIT_THREAD_SELF  INIT_THREAD_SELF(&__pthread_manager_thread);#endif  /* Block all signals except PTHREAD_SIG_RESTART */  // 初始化为全1  sigfillset(&mask);  // 设置某一位为0,这里设置可以处理restart信号  sigdelset(&mask, PTHREAD_SIG_RESTART);  // 设置进程的信号掩码  sigprocmask(SIG_SETMASK, &mask, NULL);  /* Enter server loop */  while(1) {    // 清0    FD_ZERO(&readfds);    // 置某位为1,位数由reqfd算得,这里是管道读端的文件描述符    FD_SET(reqfd, &readfds);    // 阻塞的超时时间    timeout.tv_sec = 2;    timeout.tv_usec = 0;    // 定时阻塞等待管道有数据可读    n = __select(FD_SETSIZE, &readfds, NULL, NULL, &timeout);    /* Check for termination of the main thread */    // 父进程id为1说明主进程(线程)已经退出,子进程被init(pid=1)进程接管了,    if (getppid() == 1) {      // 0说明不需要给主线程发,因为他已经退出了      pthread_kill_all_threads(SIGKILL, 0);      return 0;    }    /* Check for dead children */    if (terminated_children) {      terminated_children = 0;      pthread_reap_children();    }    /* Read and execute request */    // 管道有数据可读    if (n == 1 && FD_ISSET(reqfd, &readfds)) {      // 读出来放到request      n = __libc_read(reqfd, (char *)&request, sizeof(request));      ASSERT(n == sizeof(request));      switch(request.req_kind) {      // 创建线程      case REQ_CREATE:        request.req_thread->p_retcode =          pthread_handle_create((pthread_t *) &request.req_thread->p_retval,                                request.req_args.create.attr,                                request.req_args.create.fn,                                request.req_args.create.arg,                                request.req_args.create.mask,                                request.req_thread->p_pid);        // 唤醒父线程        restart(request.req_thread);        break;      case REQ_FREE:        pthread_handle_free(request.req_args.free.thread);        break;      case REQ_PROCESS_EXIT:        pthread_handle_exit(request.req_thread,                            request.req_args.exit.code);        break;      case REQ_MAIN_THREAD_EXIT:        // 标记主线程退出        main_thread_exiting = 1;        // 其他线程已经退出了,只有主线程了,唤醒主线程,主线程也退出,见pthread_exit,如果还有子线程没退出则主线程不能退出        if (__pthread_main_thread->p_nextlive == __pthread_main_thread) {          restart(__pthread_main_thread);          return 0;        }        break;      }    }  }}
   
     
 
    
    AI代码助手复制代码

该函数是manager线程的主要代码。他类似一个服务器一起。接收其他线程发过来的信息,然后处理。在switch那里可以看到具体的处理。这里我们只看线程创建的逻辑。函数是pthread_handle_create。

   
     
 
    
    // pthread_create发送信号给manager,manager调该函数创建线程static int pthread_handle_create(pthread_t *thread, const pthread_attr_t *attr,                                 void * (*start_routine)(void *), void *arg,                                 sigset_t mask, int father_pid){  int sseg;  int pid;  pthread_t new_thread;  int i;  /* Find a free stack segment for the current stack */  sseg = 0;  while (1) {    while (1) {      if (sseg >= num_stack_segments) {        if (pthread_grow_stack_segments() == -1) return EAGAIN;      }      if (stack_segments[sseg] == 0) break;      sseg++;    }    // 标记已使用    stack_segments[sseg] = 1;    // 存储线程元数据的地方    new_thread = THREAD_SEG(sseg);    /* Allocate space for stack and thread descriptor. */    // 给线程分配栈    if (mmap((caddr_t)((char *)(new_thread+1) - INITIAL_STACK_SIZE), INITIAL_STACK_SIZE, PROT_READ | PROT_WRITE | PROT_EXEC, MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED | MAP_GROWSDOWN, -1, 0)        != (caddr_t) -1) break;    /* It seems part of this segment is already mapped. Leave it marked       as reserved (to speed up future scans) and try the next. */    sseg++;  }  /* Initialize the thread descriptor */  new_thread->p_nextwaiting = NULL;  new_thread->p_spinlock = 0;  new_thread->p_signal = 0;  new_thread->p_signal_jmp = NULL;  new_thread->p_cancel_jmp = NULL;  new_thread->p_terminated = 0;  new_thread->p_detached = attr == NULL ? 0 : attr->detachstate;  new_thread->p_exited = 0;  new_thread->p_retval = NULL;  new_thread->p_joining = NULL;  new_thread->p_cleanup = NULL;  new_thread->p_cancelstate = PTHREAD_CANCEL_ENABLE;  new_thread->p_canceltype = PTHREAD_CANCEL_DEFERRED;  new_thread->p_canceled = 0;  new_thread->p_errno = 0;  new_thread->p_h_errno = 0;  new_thread->p_initial_fn = start_routine;  new_thread->p_initial_fn_arg = arg;  new_thread->p_initial_mask = mask;  for (i = 0; i < PTHREAD_KEYS_MAX; i++) new_thread->p_specific[i] = NULL;  /* Do the cloning */  pid = __clone(pthread_start_thread, new_thread,		(CLONE_VM | CLONE_FS | CLONE_FILES | CLONE_SIGHAND		 | PTHREAD_SIG_RESTART),		new_thread);  /* Check if cloning succeeded */  if (pid == -1) {    /* Free the stack */    munmap((caddr_t)((char *)(new_thread+1) - INITIAL_STACK_SIZE),	   INITIAL_STACK_SIZE);    stack_segments[sseg] = 0;    return EAGAIN;  }  /* Set the priority and policy for the new thread, if available. */  if (attr != NULL && attr->schedpolicy != SCHED_OTHER) {    switch(attr->inheritsched) {    case PTHREAD_EXPLICIT_SCHED:      sched_setscheduler(pid, attr->schedpolicy, &attr->schedparam);      break;    case PTHREAD_INHERIT_SCHED:      { struct sched_param father_param;        int father_policy;        father_policy = sched_getscheduler(father_pid);        sched_getparam(father_pid, &father_param);        sched_setscheduler(pid, father_policy, &father_param);      }      break;    }  }  /* Insert new thread in doubly linked list of active threads */  // 头插法,插入主线程和其他线程之间,  new_thread->p_prevlive = __pthread_main_thread;  new_thread->p_nextlive = __pthread_main_thread->p_nextlive;  __pthread_main_thread->p_nextlive->p_prevlive = new_thread;  __pthread_main_thread->p_nextlive = new_thread;  /* Set pid field of the new thread, in case we get there before the     child starts. */  new_thread->p_pid = pid;  /* We're all set */  *thread = new_thread;  return 0;}
   
     
 
    
    AI代码助手复制代码

该函数分配一个tcb结构体表示新的线程。然后分配一个线程栈,调用clone新建一个进程。最后链接到线程链表中。最后执行pthread_start_thread函数。该函数代码如下。

   
     
 
    
    // 传给clone函数的参数static int pthread_start_thread(void *arg){  // 新建的线程  pthread_t self = (pthread_t) arg;  void * outcome;  /* Initialize special thread_self processing, if any.  */#ifdef INIT_THREAD_SELF  INIT_THREAD_SELF(self);#endif  /* Make sure our pid field is initialized, just in case we get there     before our father has initialized it. */  // 记录线程对应进程的id  self->p_pid = getpid();  /* Initial signal mask is that of the creating thread. (Otherwise,     we'd just inherit the mask of the thread manager.) */  // 设置线程的信号掩码,值继承于父线程  sigprocmask(SIG_SETMASK, &self->p_initial_mask, NULL);  /* Run the thread code */  // 开始执行线程的主函数  outcome = self->p_initial_fn(self->p_initial_fn_arg);  /* Exit with the given return value */  // 执行完退出  pthread_exit(outcome);  return 0;}
   
     
 
    
    AI代码助手复制代码

没有太多逻辑,执行用户传进来的函数。执行完后退出。

看完上述内容,你们对基于linuxthreads2.0.1线程源码如何分析线程库的初始化和线程的管理有进一步的了解吗?如果还想了解更多知识或者相关内容,请关注亿速云行业资讯频道,感谢大家的支持。

亿速云「云服务器」,即开即用、新一代英特尔至强铂金CPU、三副本存储NVMe SSD云盘,价格低至29元/月。点击查看>>

向AI问一下细节

免责声明:本站发布的内容(图片、视频和文字)以原创、转载和分享为主,文章观点不代表本网站立场,如果涉及侵权请联系站长邮箱:is@yisu.com进行举报,并提供相关证据,一经查实,将立刻删除涉嫌侵权内容。

原文链接:https://my.oschina.net/u/4217331/blog/4379545

AI

开发者交流群×