gettimeofday函数_内核态调测工具：ftrace(六)-ftrace应用：跟踪&统计函数耗时

问题

实际应用中，当系统CPU压力较大，某些固定频率触发的任务突然不稳定起来，或者出现（比较关注性能的）某函数的执行耗时出现波动，这种情况是为什么呢？是函数本身不稳定，还是系统导致？

举例test.c：

#include <stdio.h>
#include <sys/time.h>

int delaytime(int t) {
        int i, j;
        for (i = 0; i < t; i++)
                for (j = 0; j < 5000; j++);
}

void main() {
        struct timeval tv1, tv2;

        gettimeofday(&tv1, NULL);
        delaytime(100);
        gettimeofday(&tv2, NULL);

        printf("diff: %ldusn", 1000000 * (tv2.tv_sec - tv1.tv_sec) + (tv2.tv_usec - tv1.tv_usec));
}

如下执行：

$ stress -c 8 &
$ taskset -c 6 ./test
diff: 1045us
$ taskset -c 6 ./test
diff: 1124us
$ taskset -c 6 ./test
diff: 1444us
$ taskset -c 6 ./test
diff: 2162us
$ taskset -c 6 ./test
diff: 11995us
$ taskset -c 6 ./test
diff: 1067us

如上，一个固定的for循环在CPU加压的情况下，执行耗时不再稳定。

Ftrace跟踪

利用之前讲到的知识

Hello小崔：内核态调测工具：ftrace(三)-ftrace在C程序中应用​zhuanlan.zhihu.com

在test.c中添加跟踪代码：

$ cat test_trace.c
#include <stdio.h>
#include <sys/time.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>

int delaytime(int t) {
        int i, j;
        for (i = 0; i < t; i++)
                for (j = 0; j < 5000; j++);
}

void main() {
        struct timeval tv1, tv2;
        int fd_sched = open("/sys/kernel/debug/tracing/events/sched/enable", O_CREAT | O_RDWR, 0666);
        int fd_irq = open("/sys/kernel/debug/tracing/events/irq/enable", O_CREAT | O_RDWR, 0666);
        int fd_mark = open("/sys/kernel/debug/tracing/trace_marker", O_CREAT | O_RDWR, 0666);
        int fd_trace = open("/sys/kernel/debug/tracing/tracing_on", O_CREAT | O_RDWR, 0666);

        write(fd_trace, "1", 2); // enable ftrace
        write(fd_sched, "1", 2); // enable sched event
        write(fd_irq, "1", 2); // enable irq event

        while (1) {