clock_nanosleep避免过度睡眠

最新推荐文章于 2025-07-28 22:36:32 发布

nerdX

最新推荐文章于 2025-07-28 22:36:32 发布

阅读量8.3k

点赞数 2

CC 4.0 BY-SA版权

分类专栏： Linux环境编程文章标签： linux环境编程 c

本文链接：https://blog.youkuaiyun.com/nerdx/article/details/20646901

Linux环境编程专栏收录该内容

6 篇文章

订阅专栏

本文探讨了使用信号处理重启休眠导致的进程oversleeping问题。通过使用clock_gettime获取当前时间，并设置绝对时间来调用clock_nanosleep，避免了相对休眠方式带来的时间不准确问题。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

/*
 *   “oversleeping” problem is particularly marked for a process that uses a loop to re start
 *   a sleep that is interrupted by a signal handler. If signals are delivered at a high rate
 *   , then a relative sleep (of the type performed by nanosleep()) can lead to large inaccuracies 
 *   in the time a process spends sleeping. We can avoid the oversleeping problem by making an
 *   initial call to clock_gettime() to retrieve the time, adding the  desired amount to that time, 
 *   and then calling  clock_nanosleep() with the TIMER_ABSTIME  flag (and restarting the system
 *   call if it is interrupted by a signal handler).
 */

struct timespec request;
/* Retrieve current value of CLOCK_REALTIME clock */
if (clock_gettime(CLOCK_REALTIME, &request) == -1) 
    errExit("clock_gettime");
request.tv_sec += 20;               /* Sleep for 20 seconds from now */
s = clock_nanosleep(CLOCK_REALTIME, TIMER_ABSTIME, &request, NULL); 
if (s != 0) {
    if (s == EINTR)
        printf("Interrupted by signal handler\n");
    else
        errExitEN(s, "clock_nanosleep");
}