BUG: scheduling while atomic 分析【转】

本文转载自：https://blog.csdn.net/cfy_phonex/article/details/12090943

遇到一个典型的ｓｃｈｅｄｕｌｅ问题。

1.  
   <3>[26578.636839] C1 [ swapper/1] BUG: scheduling while atomic: swapper/1/0/0x00000002
2.  
   <6>[26578.636869] C0 [ kworker/u:1] CPU1 is up
3.  
   <4>[26578.636900] C1 [ swapper/1] Modules linked in: bcm15500_i2c_ts
4.  
   <4>[26578.636961] C1 [ swapper/1] [<c00146d0>] (unwind_backtrace+0x0/0x11c) from [<c0602684>] (__schedule+0x70/0x6e0)
5.  
   <4>[26578.636991] C1 [ swapper/1] [<c0602684>] (__schedule+0x70/0x6e0) from [<c06030ec>] (schedule_preempt_disabled+0x14/0x20)
6.  
   <4>[26578.637052] C1 [ swapper/1] [<c06030ec>] (schedule_preempt_disabled+0x14/0x20) from [<c000f05c>] (cpu_idle+0xf0/0x104)
7.  
   <4>[26578.637083] C1 [ swapper/1] [<c000f05c>] (cpu_idle+0xf0/0x104) from [<c05e98e0>] (cpu_die+0x2c/0x5c)
8.  
   <3>[26578.637510] C1 [ swapper/1] BUG: scheduling while atomic: swapper/1/0/0x00000002
9.  
   <4>[26578.637510] C1 [ swapper/1] Modules linked in: bcm15500_i2c_ts
10.  
    <4>[26578.637602] C1 [ swapper/1] [<c00146d0>] (unwind_backtrace+0x0/0x11c) from [<c0602684>] (__schedule+0x70/0x6e0)
11.  
    <4>[26578.637663] C1 [ swapper/1] [<c0602684>] (__schedule+0x70/0x6e0) from [<c06030ec>] (schedule_preempt_disabled+0x14/0x20)
12.  
    <4>[26578.637724] C1 [ swapper/1] [<c06030ec>] (schedule_preempt_disabled+0x14/0x20) from [<c000f05c>] (cpu_idle+0xf0/0x104)
13.  
    <4>[26578.637754] C1 [ swapper/1] [<c000f05c>] (cpu_idle+0xf0/0x104) from [<c05e98e0>] (cpu_die+0x2c/0x5c)
14.  
    <3>[26578.648069] C1 [ swapper/1] BUG: scheduling while atomic: swapper/1/0/0x00000002

查看源代码

1.  
   /*
2.  
   * __schedule() is the main scheduler function.
3.  
   */
4.  
   static void __sched __schedule(void)
5.  
   {
6.  
   struct task_struct *prev, *next;
7.  
   unsigned long *switch_count;
8.  
   struct rq *rq;
9.  
   int cpu;
10.  
     
11.  
    need_resched:
12.  
    preempt_disable();
13.  
    cpu = smp_processor_id();
14.  
    rq = cpu_rq(cpu);
15.  
    rcu_note_context_switch(cpu);
16.  
    prev = rq->curr;
17.  
     
18.  
    schedule_debug(prev);
19.  
    　　　　....
20.  
    ｝

1.  
   /*
2.  
   * Print scheduling while atomic bug:
3.  
   */
4.  
   static noinline void __schedule_bug(struct task_struct *prev)
5.  
   {
6.  
   if (oops_in_progress)
7.  
   return;
8.  
    
9.  
   printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
10.  
    prev->comm, prev->pid, preempt_count());
11.  
     
12.  
    debug_show_held_locks(prev);
13.  
    print_modules();
14.  
    if (irqs_disabled())
15.  
    print_irqtrace_events(prev);
16.  
     
17.  
    dump_stack();
18.  
    }
19.  
     
20.  
    /*
21.  
    * Various schedule()-time debugging checks and statistics:
22.  
    */
23.  
    static inline void schedule_debug(struct task_struct *prev)
24.  
    {
25.  
    /*
26.  
    * Test if we are atomic. Since do_exit() needs to call into
27.  
    * schedule() atomically, we ignore that path for now.
28.  
    * Otherwise, whine if we are scheduling when we should not be.
29.  
    */
30.  
    if (unlikely(in_atomic_preempt_off() && !prev->exit_state))
31.  
    __schedule_bug(prev);
32.  
    rcu_sleep_check();
33.  
     
34.  
    profile_hit(SCHED_PROFILING, __builtin_return_address(0));
35.  
     
36.  
    schedstat_inc(this_rq(), sched_count);
37.  
    }

可以看出，　满足如下条件将会打印该出错信息

unlikely(in_atomic_preempt_off() && !prev->exit_state

为0表示TASK_RUNNING状态,当前进程在运行;　并且处于原子状态,,那么就不能切换给其它的进程

1.  
   Linux/include/linux/sched.h
2.  
    
3.  
   /*
4.  
   * Task state bitmask. NOTE! These bits are also
5.  
   * encoded in fs/proc/array.c: get_task_state().
6.  
   *
7.  
   * We have two separate sets of flags: task->state
8.  
   * is about runnability, while task->exit_state are
9.  
   * about the task exiting. Confusing, but this way
10.  
    * modifying one set can't modify the other one by
11.  
    * mistake.
12.  
    */
13.  
    #define TASK_RUNNING 0
14.  
    #define TASK_INTERRUPTIBLE 1
15.  
    #define TASK_UNINTERRUPTIBLE 2
16.  
    #define __TASK_STOPPED 4
17.  
    #define __TASK_TRACED 8
18.  
    /* in tsk->exit_state */
19.  
    #define EXIT_ZOMBIE 16
20.  
    #define EXIT_DEAD 32
21.  
    /* in tsk->state again */
22.  
    #define TASK_DEAD 64
23.  
    #define TASK_WAKEKILL 128
24.  
    #define TASK_WAKING 256
25.  
    #define TASK_STATE_MAX 512

1.  
   kernel/include/linux/hardirq.h
2.  
    
3.  
   #if defined(CONFIG_PREEMPT_COUNT)
4.  
   # define PREEMPT_CHECK_OFFSET 1
5.  
   #else
6.  
   # define PREEMPT_CHECK_OFFSET 0
7.  
   #endif
8.  
    
9.  
   /*
10.  
    * Are we running in atomic context? WARNING: this macro cannot
11.  
    * always detect atomic context; in particular, it cannot know about
12.  
    * held spinlocks in non-preemptible kernels. Thus it should not be
13.  
    * used in the general case to determine whether sleeping is possible.
14.  
    * Do not use in_atomic() in driver code.
15.  
    */
16.  
    #define in_atomic() ((preempt_count() & ~PREEMPT_ACTIVE) != 0)
17.  
     
18.  
    /*
19.  
    * Check whether we were atomic before we did preempt_disable():
20.  
    * (used by the scheduler, *after* releasing the kernel lock)
21.  
    */
22.  
    #define in_atomic_preempt_off() \
23.  
    ((preempt_count() & ~PREEMPT_ACTIVE) != PREEMPT_CHECK_OFFSET)

结论整理

linux内核打印"BUG: scheduling while atomic"和"bad: scheduling from the idle thread"错误的时候，

通常是在中断处理函数中调用了可以休眠的函数，如semaphore,mutex,sleep之类的可休眠的函数，

而linux内核要求在中断处理的时候，不允许系统调度，不允许抢占，要等到中断处理完成才能做其他事情。

因此，要充分考虑中断处理的时间，一定不能太久。

另外一个能产生此问题的是在idle进程里面，做了不该做的事情。现在Linux用于很多手持式设备，为了降低功耗，

通常的作法是在idle进程里面降低CPU或RAM的频率、关闭一些设备等等。要保证这些动作的原子性才能确保

不发生"bad: scheduling from the idle thread"这样的错误！

禁止内核抢占是指内核不会主动的抢占你的process，但是现在是你在自己的程序中主动call schedule()，

kernel并不能阻止你这么作。

Scheduling while atomic" means that a thread has called schedule() during an operation which is supposed to be atomic (ie uninterrupted).

1.  
   190 NOTE: ***** WARNING *****
2.  
   191 NEVER SLEEP IN A COMPLETION HANDLER. These are normally called
3.  
   192 during hardware interrupt processing. If you can, defer substantial
4.  
   193 work to a tasklet (bottom half) to keep system latencies low. You'll
5.  
   194 probably need to use spinlocks to protect data structures you manipulate
6.  
   195 in completion handlers.

1.  
   GFP_ATOMIC is used when
2.  
   (a) you are inside a completion handler, an interrupt, bottom half, tasklet or timer, or
3.  
   (b) you are holding a spinlock or rwlock (does not apply to semaphores), or
4.  
   (c) current->state != TASK_RUNNING, this is the case only after you've changed it.