linux - Linux 内核中的 DECLARE_COMPLETION_ONSTACK 功能

Question

谁能帮助我理解“DECLARE_COMPLETION_ONSTACK”和静态(DECLARE_COMPLETION(comp))或动态初始化之间的区别？

我找到了一些引用资料，如下所示(http://lxr.free-electrons.com/source/drivers/spi/spidev.c#L110):

110 spidev_sync(struct spidev_data *spidev, struct spi_message *message)
111 {
112         DECLARE_COMPLETION_ONSTACK(done);
113         int status;
114 
115         message->complete = spidev_complete;
116         message->context = &done;
117 
118         spin_lock_irq(&spidev->spi_lock);
119         if (spidev->spi == NULL)
120                 status = -ESHUTDOWN;
121         else
122                 status = spi_async(spidev->spi, message);
123         spin_unlock_irq(&spidev->spi_lock);
124 
125         if (status == 0) {
126                 wait_for_completion(&done);
127                 status = message->status;
128                 if (status == 0)
129                         status = message->actual_length;
130         }
131         return status;
132 }

看起来这里是在等待 spi_sync() 函数完成，我在理解上述函数时有几个疑问

2:这是否意味着这个函数不是真正的并发？(因为这个完成变量强加了一种序列化)

BR, & 斯里达尔

Answer 1

不开启LOCK_DEP没有区别。

它仅对锁定调试功能有用。

查看下面的内核代码，

47/*
48 * Lockdep needs to run a non-constant initializer for on-stack
49 * completions - so we use the _ONSTACK() variant for those that
50 * are on the kernel stack:
51 */

59#ifdef CONFIG_LOCKDEP
60# define DECLARE_COMPLETION_ONSTACK(work) \
61  struct completion work = COMPLETION_INITIALIZER_ONSTACK(work)
62#else
63# define DECLARE_COMPLETION_ONSTACK(work) DECLARE_COMPLETION(work)
64#endif