Chinaunix首页 | 论坛 | 博客
  • 博客访问: 959397
  • 博文数量: 145
  • 博客积分: 1302
  • 博客等级: 中尉
  • 技术积分: 1778
  • 用 户 组: 普通用户
  • 注册时间: 2006-03-07 16:00
文章分类

全部博文(145)

文章存档

2018年(1)

2016年(1)

2015年(6)

2014年(4)

2013年(59)

2012年(32)

2011年(36)

2009年(1)

2007年(2)

2006年(3)

分类: LINUX

2015-06-27 10:05:17

原文地址:工作队列(workqueue) 作者:leonwang202

项目需要,在驱动模块里用内核计时器timer_list实现了一个状态机。
郁闷的是,运行时总报错“Scheduling while atomic”,网上搜了一下:

"Scheduling while atomic" indicates that you've tried to sleep somewhere that you shouldn't - like within a spinlock-protected critical section or an interrupt handler.

改进程序,在计时器里使用了workqueue,搞定问题。顺便把workqueue的实现代码总结了一下


一、workqueue简介

workqueue与tasklet类似,都是允许内核代码请求某个函数在将来的时间被调用(抄《ldd3》上的)
每个workqueue就是一个内核进程。

workqueue与tasklet的区别:
   1.tasklet是通过软中断实现的,在软中断上下文中运行,tasklet代码必须是原子的
     workqueue是通过内核进程实现的,就没有上述限制的,最爽的是,工作队列函数可以休眠
        
     PS: 我的驱动模块就是印在计时器中调用了可休眠函数,所以出现了cheduling while atomic告警
         内核计时器也是通过软中断实现的

   2.tasklet始终运行在被初始提交的同一处理器上,workqueue不一定
   3.tasklet不能确定延时时间(即使很短),workqueue可以设定延迟时间


二、workqueue的API

 
workqueue的API自2.6.20后发生了变化

  1. #include <linux/workqueue.h>
  2. struct workqueue_struct;
  3. struct work_struct;

  4. struct workqueue_struct *create_workqueue(const char *name);
  5. void destroy_workqueue(struct workqueue_struct *queue);

  6. INIT_WORK(_work, _func);
  7. INIT_DELAYED_WORK(_work, _func);

  8. int queue_work(struct workqueue_struct *wq, struct work_struct *work);
  9. int queue_delayed_work(struct workqueue_struct *wq,struct delayed_work *dwork, unsigned long delay);

  10. int queue_delayed_work_on(int cpu, struct workqueue_struct *wq,
  11.             struct delayed_work *dwork, unsigned long delay);

  12. int cancel_work_sync(struct work_struct *work);
  13. int cancel_delayed_work_sync(struct delayed_work *dwork);

  14. void flush_workqueue(struct workqueue_struct *wq);
Workqueue编程接口

序号

接口函数

说明

1

create_workqueue

用于创建一个workqueue队列,为系统中的每个CPU都创建一个内核线程。输入参数:

@nameworkqueue的名称

2

create_singlethread_workqueue

用于创建workqueue,只创建一个内核线程。输入参数:

@nameworkqueue名称

3

destroy_workqueue

释放workqueue队列。输入参数:

@ workqueue_struct:需要释放的workqueue队列指针

4

schedule_work

调度执行一个具体的任务,执行的任务将会被挂入Linux系统提供的workqueue——keventd_wq输入参数:

@ work_struct:具体任务对象指针

5

schedule_delayed_work

延迟一定时间去执行一个具体的任务,功能与schedule_work类似,多了一个延迟时间,输入参数:

@work_struct:具体任务对象指针

@delay:延迟时间

6

queue_work

调度执行一个指定workqueue中的任务。输入参数:

@ workqueue_struct:指定的workqueue指针

@work_struct:具体任务对象指针

7

queue_delayed_work

延迟调度执行一个指定workqueue中的任务,功能与queue_work类似,输入参数多了一个delay



下面实例是不指定delay时间的workqueue
(代码基于2.6.24)

  1. struct my_work_stuct{
  2.     int test;
  3.     struct work_stuct save;
  4. };

  5. struct my_work_stuct test_work;
  6. struct workqueue_struct *test_workqueue;

  7. void do_save(struct work_struct *p_work)
  8. {
  9.     struct my_work_struct *p_test_work = container_of(p_work, struct my_work_stuct, save);
  10.     printk("%d\n",p_test_work->test);
  11. }
  12.   
  13. void test_init()
  14. {
  15.     test_workqueue = create_workqueue("test_workqueue");
  16.     if (!test_workqueue)
  17.         panic("Failed to create test_workqueue\n");

  18.     INIT_WORK(&(test_work.save), do_save);

  19.     queue_work(test_workqueue, &(test_work.save));
  20. }

  21. void test_destory(void)
  22. {
  23.     if(test_workqueue)
  24.         destroy_workqueue(test_workqueue);
  25. }


三、workqueue的实现


工作队列workqueue不是通过软中断实现的,它是通过内核进程实现的



首先,创建一个workqueue,实际上就是建立一个内核进程

  1. create_workqueue("tap_workqueue")
  2. --> __create_workqueue(“tap_workqueue”, 0, 0)
  3. --> __create_workqueue_key((name), (singlethread), (freezeable), NULL, NULL){
  4.          wq = kzalloc(sizeof(*wq), GFP_KERNEL);
  5.          wq->cpu_wq = alloc_percpu(struct cpu_workqueue_struct);
  6.          wq->name = name;
  7.          wq->singlethread = singlethread;
  8.          wq->freezeable = freezeable;
  9.          INIT_LIST_HEAD(&wq->list);

  10.          for_each_possible_cpu(cpu) {
  11.              cwq = init_cpu_workqueue(wq, cpu);
  12.              err = create_workqueue_thread(cwq, cpu);
  13.              start_workqueue_thread(cwq, cpu);
  14.          }
  15.     }

create_workqueue_thread 建立了一个内核进程 worker_thread(linux_2_6_24/kernel/workqueue.c)
  1. create_workqueue_thread(struct cpu_workqueue_struct *cwq, int cpu)
  2. {
  3.     struct workqueue_struct *wq = cwq->wq;

  4.     const char *fmt = is_single_threaded(wq) ? "%s" : "%s/%d";
  5.     struct task_struct *p;

  6.     p = kthread_create(worker_thread, cwq, fmt, wq->name, cpu);

  7.     if (IS_ERR(p))
  8.         return PTR_ERR(p);

  9.     cwq->thread = p;

  10.     return 0;
  11. }

内核进程worker_thread做的事情很简单,死循环而已,不停的执行workqueue上的work_list
(linux_2_6_24/kernel/workqueue.c)

  1. int worker_thread (void *__cwq)
  2. {
  3.     struct cpu_workqueue_struct *cwq = __cwq;
  4.     /*下面定义等待队列项*/
  5.     DEFINE_WAIT(wait);

  6.     /*下面freezeable一般为0*/
  7.     if (cwq->wq->freezeable)
  8.         set_freezable();

  9.     /*提高优先级别*/
  10.     set_user_nice(current, -5);

  11.     for (;;) {
  12.         /*在cwq->more_work上等待, 若有人调用queue_work,该函数将调用wake_up(&cwq->more_work) 激活本进程*/
  13.         prepare_to_wait(&cwq->more_work, &wait, TASK_INTERRUPTIBLE);

  14.         /*work队列空则切换出去*/
  15.         if (!freezing(current) && !kthread_should_stop() && list_empty(&cwq->worklist))
  16.             schedule();


  17.         /*切换回来则结束等待 说明有人唤醒cwq->more_work上的等待 有work需要处理*/
  18.         finish_wait(&cwq->more_work, &wait);

  19.         /*下面空,因为没有定义电源管理*/
  20.         try_to_freeze();

  21.         if (kthread_should_stop())
  22.             break;

  23.         /*run_workqueue依次处理工作队列上所有的work*/
  24.         run_workqueue(cwq);
  25.     }
  26.     return 0;
  27. }


  28. /*run_workqueue依次处理工作队列上所有的work*/
  29. static void run_workqueue(struct cpu_workqueue_struct *cwq)
  30. {
  31.     spin_lock_irq(&cwq->lock);
  32.     cwq->run_depth++;
  33.     if (cwq->run_depth > 3) {
  34.         /* morton gets to eat his hat */
  35.         printk("%s: recursion depth exceeded: %d\n",
  36.             __FUNCTION__, cwq->run_depth);
  37.         dump_stack();
  38.     }

  39.     while (!list_empty(&cwq->worklist)) {
  40.         struct work_struct *work = list_entry(cwq->worklist.next,
  41.                         struct work_struct, entry);
  42.         work_func_t f = work->func;
  43. #ifdef CONFIG_LOCKDEP
  44.         /*
  45.          * It is permissible to free the struct work_struct
  46.          * from inside the function that is called from it,
  47.          * this we need to take into account for lockdep too.
  48.          * To avoid bogus "held lock freed" warnings as well
  49.          * as problems when looking into work->lockdep_map,
  50.          * make a copy and use that here.
  51.          */
  52.         struct lockdep_map lockdep_map = work->lockdep_map;
  53. #endif

  54.         cwq->current_work = work;
  55.         list_del_init(cwq->worklist.next);
  56.         spin_unlock_irq(&cwq->lock);

  57.         BUG_ON(get_wq_data(work) != cwq);
  58.         work_clear_pending(work);
  59.         lock_acquire(&cwq->wq->lockdep_map, 0, 0, 0, 2, _THIS_IP_);
  60.         lock_acquire(&lockdep_map, 0, 0, 0, 2, _THIS_IP_);

  61.         f(work); /*执行work项中的func*/
  62.        
  63.         lock_release(&lockdep_map, 1, _THIS_IP_);
  64.         lock_release(&cwq->wq->lockdep_map, 1, _THIS_IP_);

  65.         if (unlikely(in_atomic() || lockdep_depth(current) > 0)) {
  66.             printk(KERN_ERR "BUG: workqueue leaked lock or atomic: "
  67.                     "%s/0x%08x/%d\n",
  68.                     current->comm, preempt_count(),
  69.                         task_pid_nr(current));
  70.             printk(KERN_ERR " last function: ");
  71.             print_symbol("%s\n", (unsigned long)f);
  72.             debug_show_held_locks(current);
  73.             dump_stack();
  74.         }

  75.         spin_lock_irq(&cwq->lock);
  76.         cwq->current_work = NULL;
  77.     }
  78.     cwq->run_depth--;
  79.     spin_unlock_irq(&cwq->lock);
  80. }


将一个work加入到指定workqueue的work_list中(文件linux_2_6_24/kernel/workqueue.c)

  int fastcall queue_work(struct workqueue_struct *wq, struct work_struct *work)
  1. {
  2.     int ret = 0;

  3.     if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) {
  4.         BUG_ON(!list_empty(&work->entry));
  5.         __queue_work(wq_per_cpu(wq, get_cpu()), work);
  6.         put_cpu();
  7.         ret = 1;
  8.     }
  9.     return ret;
  10. }


  11. /* Preempt must be disabled. */
  12. static void __queue_work(struct cpu_workqueue_struct *cwq, struct work_struct *work)
  13. {
  14.     unsigned long flags;

  15.     spin_lock_irqsave(&cwq->lock, flags);
  16.     insert_work(cwq, work, 1);
  17.     spin_unlock_irqrestore(&cwq->lock, flags);
  18. }

  19. static void insert_work(struct cpu_workqueue_struct *cwq,
  20.                 struct work_struct *work, int tail)
  21. {
  22.     set_wq_data(work, cwq);
  23.     /*
  24.      * Ensure that we get the right work->data if we see the
  25.      * result of list_add() below, see try_to_grab_pending().
  26.      */
  27.     smp_wmb();
  28.     if (tail)
  29.         list_add_tail(&work->entry, &cwq->worklist);
  30.     else
  31.         list_add(&work->entry, &cwq->worklist);
  32.     wake_up(&cwq->more_work);
  33. }



四、共享队列

其实内核有自己的一个workqueue,叫keventd_wq,这个工作队列也叫做“共享队列”。
do_basic_setup --> init_workqueues --> create_workqueue("events");

若驱动模块使用的workqueue功能很简单的话,可以使用“共享队列”,不用自己再建一个队列
使用共享队列,有这样一套API

  1. int schedule_work(struct work_struct *work)
  2. {
  3.     queue_work(keventd_wq, work);
  4. }

  5. int schedule_delayed_work(struct delayed_work *dwork,unsigned long delay)
  6. {
  7.     timer_stats_timer_set_start_info(&dwork->timer);
  8.     return queue_delayed_work(keventd_wq, dwork, delay);
  9. }

  10. void flush_scheduled_work(void)
  11. {
  12.     flush_workqueue(keventd_wq);
  13. }


阅读(2039) | 评论(0) | 转发(0) |
给主人留下些什么吧!~~