基于linuxthreads2.0.1线程源码如何分析信号量

基于linux threads2.0.1线程源码如何分析信号量,很多新手对此不是很清楚,为了帮助大家解决这个难题,下面小编将为大家详细讲解,有这方面需求的人可以来学习下,希望你能有所收获。

成都创新互联成立与2013年,是专业互联网技术服务公司,拥有项目网站设计制作、网站设计网站策划,项目实施与项目整合能力。我们以让每一个梦想脱颖而出为使命,1280元溧水做网站,已为上家服务,为溧水各地企业和个人服务,联系电话:028-86922220

   
     
 
    
   

/* Semaphores a la POSIX 1003.1b */

#include "pthread.h"
#include "semaphore.h"
#include "internals.h"
#include "restart.h"


#ifndef HAS_COMPARE_AND_SWAP
/* If we have no atomic compare and swap, fake it using an extra spinlock.  */

#include "spinlock.h"
// 等于旧值则更新sem_status为新值,如果一样则不更新,更新则返回1
static inline int compare_and_swap(sem_t *sem, long oldval, long newval)
{
 int ret;
 acquire(&sem->sem_spinlock);
 if ((ret = sem->sem_status == oldval) != 0)
   sem->sem_status = newval;
 release(&sem->sem_spinlock);
 return ret;
}

#else
/* But if we do have an atomic compare and swap, use it!  */

#define compare_and_swap(sem,old,new) \
 __compare_and_swap(&(sem)->sem_status, (old), (new))

#endif


/* The state of a semaphore is represented by a long int encoding
  either the semaphore count if >= 0 and no thread is waiting on it,
  or the head of the list of threads waiting for the semaphore.
  To distinguish the two cases, we encode the semaphore count N
  as 2N+1, so that it has the lowest bit set.

  A sequence of sem_wait operations on a semaphore initialized to N
  result in the following successive states:
    2N+1, 2N-1, ..., 3, 1, &first_waiting_thread, &second_waiting_thread, ...
*/

int sem_init(sem_t *sem, int pshared, unsigned int value)
{
 if (value > SEM_VALUE_MAX) {
   errno = EINVAL;
   return -1;
 }
 // 还没实现
 if (pshared) {
   errno = ENOSYS;
   return -1;
 }
 // 记录资源数
 sem->sem_status = ((long)value << 1) + 1;
 return 0;
}

int sem_wait(sem_t * sem)
{
 long oldstatus, newstatus;
 volatile pthread_t self = thread_self();
 pthread_t * th;

 while (1) {
   do {
     // oldstatus可能是线程或者资源数
     oldstatus = sem->sem_status;
     // 大于1说明有资源,等于1说着0说明没有资源或没有资源并且有线程阻塞
     if ((oldstatus & 1) && (oldstatus != 1))
       newstatus = oldstatus - 2;
     else {
       // 没有可用资源,需要阻塞
       newstatus = (long) self;
       // 保存这时候的资源数或者上一个被阻塞的线程
       self->p_nextwaiting = (pthread_t) oldstatus;
     }
   }
   // sem_status可能指向资源数或者被阻塞的线程链表。赋值成功则返回1,否则返回0
   while (! compare_and_swap(sem, oldstatus, newstatus));
   // self是按偶数地址对齐的,低位为1说明是还有可用资源
   if (newstatus & 1)
     /* We got the semaphore. */
     return 0;
   /* Wait for sem_post or cancellation */
   // 等待被restart或者cancel信号唤醒
   suspend_with_cancellation(self);
   /* This is a cancellation point */
   // 判断是否被取消了,即是被cancel信号唤醒的,不是的话重新判断是否有资源,即回到上面的while(1)
   if (self->p_canceled && self->p_cancelstate == PTHREAD_CANCEL_ENABLE) {
     /* Remove ourselves from the waiting list if we're still on it */
     /* First check if we're at the head of the list. */
     do {
       // 得到被阻塞的第一个线程
       oldstatus = sem->sem_status;
       // 相等说明当前线程是最后一个被阻塞的线程
       if (oldstatus != (long) self) break;
       // 得到该线程被阻塞时的资源数或下一个被阻塞的线程
       newstatus = (long) self->p_nextwaiting;
     }
     // sem_status指向资源数或者下一个被阻塞的线程
     while (! compare_and_swap(sem, oldstatus, newstatus));
     /* Now, check if we're somewhere in the list.
        There's a race condition with sem_post here, but it does not matter:
        the net result is that at the time pthread_exit is called,
        self is no longer reachable from sem->sem_status. */
     // 可能是break或者while为true,不是当前线程并且不是资源数,即oldstatus指向一个其他线程
     if (oldstatus != (long) self && (oldstatus & 1) == 0) {
       th = &(((pthread_t) oldstatus)->p_nextwaiting);
       // 不是资源数,即是线程,从等待的线程链表中删除self线程,因为他即将退出
       while (*th != (pthread_t) 1 && *th != NULL) {
         if (*th == self) {
           *th = self->p_nextwaiting;
           break;
         }
         th = &((*th)->p_nextwaiting);
       }
     }
     // 当前线程退出
     pthread_exit(PTHREAD_CANCELED);
   }
 }
}
// 非阻塞获取信号量
int sem_trywait(sem_t * sem)
{
 long oldstatus, newstatus;

 do {
   oldstatus = sem->sem_status;
   // oldstatus & 1等于0说明是线程,即有线程在等待,或者等于1,都说明没有可用资源,直接返回
   if ((oldstatus & 1) == 0 || (oldstatus == 1)) {
     errno = EAGAIN;
     return -1;
   }
   // 更新资源数
   newstatus = oldstatus - 2;
 }
 // 更新资源数,如果失败说明被其他线程拿到锁了,则重新执行do里面的逻辑,因为数据可能被修改了
 while (! compare_and_swap(sem, oldstatus, newstatus));
 return 0;
}

int sem_post(sem_t * sem)
{
 long oldstatus, newstatus;
 pthread_t th, next_th;

 do {
   oldstatus = sem->sem_status;
   // 说明原来的资源数是0,并且有线程在等待,则更新为1,2n+1即3
   if ((oldstatus & 1) == 0)
     newstatus = 3;
   else {
     if (oldstatus >= SEM_VALUE_MAX) {
       /* Overflow */
       errno = ERANGE;
       return -1;
     }
     // 否则加2,即资源数加一
     newstatus = oldstatus + 2;
   }
 }
 // 更新资源数
 while (! compare_and_swap(sem, oldstatus, newstatus));
 // 如果之前有线程阻塞,则唤醒所有线程,再次竞争获得信号量
 if ((oldstatus & 1) == 0) {
   th = (pthread_t) oldstatus;
   do {
     next_th = th->p_nextwaiting;
     th->p_nextwaiting = NULL;
     restart(th);
     th = next_th;
   } while(th != (pthread_t) 1);
 }
 return 0;
}
// 获取资源数
int sem_getvalue(sem_t * sem, int * sval)
{
 long status = sem->sem_status;
 // 有资源
 if (status & 1)
   // 除以2
   *sval = (int)((unsigned long) status >> 1);
 else
   *sval = 0;
 return 0;
}

int sem_destroy(sem_t * sem)
{
 // 有线程在等待
 if ((sem->sem_status & 1) == 0) {
   errno = EBUSY;
   return -1;
 }
 return 0;
}

             

基于linux threads2.0.1线程源码如何分析信号量

阻塞时的视图。

看完上述内容是否对您有帮助呢?如果还想对相关知识有进一步的了解或阅读更多相关文章,请关注创新互联行业资讯频道,感谢您对创新互联的支持。


当前标题:基于linuxthreads2.0.1线程源码如何分析信号量
标题来源:http://hbruida.cn/article/jiiego.html