进程调度函数schedule()解读

在linux系统中，单处理器也是多线程处理信号、事件等。这就需要一个核心算法来进行进程调度。这个算法就是CFS（Completely Fair Scheduler）。在《LInux Kernel Development》一书中用一句话总结CFS进程调度：

运行rbtree树中最左边叶子节点所代表的那个进程。

在一个自平衡二叉搜索树红黑树rbtree的树节点中，存储了下一个应该运行进程的数据。在这里我们看到了二叉搜索树的完美运用。

而进程调度的主要入口函数就是schedule()。它定义在文件kernel/sched.c中。

我们先看一个在等待队列中进行进程调度的例子：

<!-- lang: cpp -->
DEFINE_WAIT(wait); //申明等待队列

add_wait_queue(q,&wait); //把我们用的q队列加入到wait等待队列中
while(!condition){ //当等待事件没有来临时
     prepare_to_wait(&q,&wait,TASK_INTERRUPTIBLE); //将q 从TASK_RUNNING或者其他状态置为TASK_INTERRUPTIBLE不可运行的休眠状态。同时接受信号&&事件来唤醒它
     if(signal_pending(current))  //如果有来自从处理器的信号
     { processingsignal();}//处理信号
     schedule(); //调用红黑树中的下一个进程
}
finish_wait(&q,&wait); //将进程设置为TASK_RUNNING并移出等待队列.

其实我们可以这么理解这段代码。现在有一个任务要等待事件到来才能运行，怎么实现呢？就是阻塞加查询。但是这样会使得这段代码独占整个操作系统。为了解决这个问题，就在阻塞查询之中加入了队列和进程调度schedule()，从而不耽误其它线程的执行。

再来看一看schedule()函数的结构：

schedule()函数结构

<!-- lang: cpp -->
asmlinkage void __sched schedule(void)  ///定义通过堆栈传值
{
struct task_struct *prev, *next;
unsigned long *switch_count;
struct rq *rq;
int cpu;

/*At the end of this function, it will check if need_resched() return
true, if that indeed happen, then goto here.*/
need_resched:

/*current process won't be preempted after call preemept_disable()*/
preempt_disable(); //不让优先占有当前进程
cpu = smp_processor_id();
rq = cpu_rq(cpu);
/* rcu_sched_qs ? */
rcu_sched_qs(cpu);

/* prev point to current task_struct */
prev = rq->curr;

/* get current task_struct's context switch count */
switch_count = &prev->nivcsw;

/* kernel_flag is "the big kernel lock". 
 * This spinlock is taken and released recursively by lock_kernel()
 * and unlock_kernel(). It is transparently dropped and reacquired
 * over schedule(). It is used to protect legacy code that hasn't
 * been migrated to a proper locking design yet.
 * In task_struct, there is a member lock_depth, which is inited -1,
 * indicates that the current task have no kernel lock.
 * When lock_depth >=0 indicate that it own kernel lock.
 * During context switching, it is not permitted that the task  
 * switched away remain own kernel lock , so in scedule(),it
 * call release_kernel_lock(), release kernel lock.
 */
release_kernel_lock(prev);

need_resched_nonpreemptible:

schedule_debug(prev);

if (sched_feat(HRTICK))
    hrtick_clear(rq);

/* occupy current rq's lock */
raw_spin_lock_irq(&rq->lock); //占有rq自旋锁

/* update rq's clock,this function will call sched_clock_cpu() */
update_rq_clock(rq);

/* clear bit in task_struct's thread_struct's flag TIF_NEED_RESCHED.
 * In case that it will be rescheduled, because it prepare to give
 * up cpu.
 */
clear_tsk_need_resched(prev);



if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
    if (unlikely(signal_pending_state(prev->state, prev)))
        prev->state = TASK_RUNNING;
    else
        deactivate_task(rq, prev, 1);
    switch_count = &prev->nvcsw;
}

/* For none-SMP, pre_schedule is NULL */
pre_schedule(rq, prev);


if (unlikely(!rq->nr_running))
    idle_balance(cpu, rq);

put_prev_task(rq, prev);


next = pick_next_task(rq);


if (likely(prev != next)) {
    sched_info_switch(prev, next);
    perf_event_task_sched_out(prev, next);

    rq->nr_switches++;
    rq->curr = next;
    ++*switch_count;

    context_switch(rq, prev, next); /* unlocks the rq */
    /*
     * the context switch might have flipped the stack from under
     * us, hence refresh the local variables.
     */
    cpu = smp_processor_id();
    rq = cpu_rq(cpu);
} else
  raw_spin_unlock_irq(&rq->lock);//current task still occupy cpu

post_schedule(rq);

if (unlikely(reacquire_kernel_lock(current) < 0)) {
    prev = rq->curr;
    switch_count = &prev->nivcsw;
    goto need_resched_nonpreemptible;
}

preempt_enable_no_resched();
if (need_resched())
    goto need_resched;
}
EXPORT_SYMBOL(schedule);

schedule()函数的目的在于用另一个进程替换当前正在运行的进程。因此，这个函数的主要结果就是设置一个名为next的变量，以便它指向所选中的 代替current的进程的描述符。如果在系统中没有可运行进程的优先级大于current的优先级，那么，结果是next与current一致，没有进程切换发生。

References

[1].UNDERSTANDING THE LINUX KERNEL. Page 276

[2].Linux Kernel Development. Page 52

[3].http://hi.baidu.com/zengzhaonong/item/20d9e8207b04cb8f6e2cc323