字符驱动设备之poll机制

概念

poll机制引入前,我们来先来说下I/O复用,所谓I/O复用就是解决能够同时操作多个设备的方法,及时处理多个设备的数据,I/O复用的方法大概有如下,应用程序使用poll,select系统调用,应用程序使用多线程技术,比如处理按键一个线程,处理鼠标一个线程,由于线程是并行运行,所有可以同时处理按键和鼠标,工作机制如下,poll和select的机制道理是一样的,监听文件描述符集,当没有文件可以读写则挂入等待队列睡眠,当有一个或多个fd发生变化的,则从阻塞状态返回,转而处理该文件描述符IO操作,poll函数比select函数更强大的功能,更细粒的等待时间

函数介绍

poll()接受一个指向结构"struct pollfd"列表的指针,其中包括了你想测试的文件描述符和事件,事件由一个在结构中事件域的比特掩码确定
struct pollfd
{
int fd; //文件描述符
short event;//等待的需要测试事件
short revents //实际发生了的事件
};
events和revents的值见表

内核源码分析

所有的系统调用,基于都可以在它的名字前加上“sys_”前缀,这就是它在内核中对应的函数.比如系统调用open,rea,write,poll,与之对应的内核函数为:sys_open,sys_read,sys_write,sys_poll.
1.
sys_poll函数位于fs/select.c文件中,代码如下:

asmlinkage long sys_poll(struct pollfd __user *ufds, unsigned int nfds,
			long timeout_msecs)
{
	s64 timeout_jiffies;

	if (timeout_msecs > 0) {
#if HZ > 1000
		/* We can only overflow if HZ > 1000 */
		if (timeout_msecs / 1000 > (s64)0x7fffffffffffffffULL / (s64)HZ)
			timeout_jiffies = -1;
		else
#endif
			timeout_jiffies = msecs_to_jiffies(timeout_msecs);
	} else {
		/* Infinite (< 0) or no (0) timeout */
		timeout_jiffies = timeout_msecs;
	}

	return do_sys_poll(ufds, nfds, &timeout_jiffies);
}

上述的代码对超时参数稍作处理后，直接调用do_sys_poll
2.
do_sys_poll函数也位于位于fs/select.c文件中,我们取出主要部分的代码：

int do_sys_poll(struct pollfd __user *ufds, unsigned int nfds, s64 *timeout)
{
     ……
     struct poll_wqueues table;
     poll_initwait(&table);
     ……
	 fdcount = do_poll(nfds, head, &table, timeout);
     ……
}

poll_initwait函数非常简单，它初始化一个poll_wqueues变量table,table->pt->qproc = __pollwait,从__pollwait的代码可知,它只是把当前进程挂入我们驱动程序里定义的一个队列里而已,__pollwait将在驱动的poll函数里用到,进而调用do_poll函数。
3.do_poll函数位于fs/select.c文件中,代码如下：

static int do_poll(unsigned int nfds,  struct poll_list *list,
		   struct poll_wqueues *wait, s64 *timeout)
{
	int count = 0;
	poll_table* pt = &wait->pt;

	/* Optimise the no-wait case */
	if (!(*timeout))
		pt = NULL;
 
	for (;;) {
		struct poll_list *walk;
		long __timeout;

		set_current_state(TASK_INTERRUPTIBLE);
		for (walk = list; walk != NULL; walk = walk->next) {
			struct pollfd * pfd, * pfd_end;

			pfd = walk->entries;
			pfd_end = pfd + walk->len;
			for (; pfd != pfd_end; pfd++) {
				/*
				 * Fish for events. If we found one, record it
				 * and kill the poll_table, so we don't
				 * needlessly register any other waiters after
				 * this. They'll get immediately deregistered
				 * when we break out and return.
				 */
				if (do_pollfd(pfd, pt)) {
					count++;
					pt = NULL;
				}
			}
		}
		/*
		 * All waiters have already been registered, so don't provide
		 * a poll_table to them on the next loop iteration.
		 */
		pt = NULL;
		if (count || !*timeout || signal_pending(current))
			break;
		count = wait->error;
		if (count)
			break;

		if (*timeout < 0) {
			/* Wait indefinitely */
			__timeout = MAX_SCHEDULE_TIMEOUT;
		} else if (unlikely(*timeout >= (s64)MAX_SCHEDULE_TIMEOUT-1)) {
			/*
			 * Wait for longer than MAX_SCHEDULE_TIMEOUT. Do it in
			 * a loop
			 */
			__timeout = MAX_SCHEDULE_TIMEOUT - 1;
			*timeout -= __timeout;
		} else {
			__timeout = *timeout;
			*timeout = 0;
		}

		__timeout = schedule_timeout(__timeout);
		if (*timeout >= 0)
			*timeout += __timeout;
	}
	__set_current_state(TASK_RUNNING);
	return count;
}

*if (count || !timeout || signal_pending(current))可以看出当count非0,超时,有信号等待处理结束循环,count非0表示do_pollfd至少有一个成功,__timeout = schedule_timeout(__timeout)使进程进入休眠状态一段时间,除了休眠到指定时间被系统唤醒外,还可以被驱动程序唤醒,这就是为什么驱动的poll里要调用poll_wait的原因,后面分析。
4.do_pollfd函数位于fs/select.c文件中,代码如下：

static inline unsigned int do_pollfd(struct pollfd *pollfd, poll_table *pwait)
{
……
			if (file->f_op && file->f_op->poll)
				mask = file->f_op->poll(file, pwait);
……
}

可见,它就是调用我们的驱动程序里注册的poll函数。

驱动程序设计

驱动程序里与poll相关的地方有两处:1.是构造file_operation结构时，要定义自己的poll函数2.是通过poll_wait来调用上面说到的__pollwait函数,pollwait的代码如下：

static inline void poll_wait(struct file * filp, wait_queue_head_t * wait_address, poll_table *p)
{
	if (p && wait_address)
		p->qproc(filp, wait_address, p);
}

p->qproc就是__pollwait函数，从它的代码可知，它只是把当前进程挂入我们驱动程序里定义的一个队列里而已。它的代码如下：

static void __pollwait(struct file *filp, wait_queue_head_t *wait_address,
				poll_table *p)
{
	struct poll_table_entry *entry = poll_get_entry(p);
	if (!entry)
		return;
	get_file(filp);
	entry->filp = filp;
	entry->wait_address = wait_address;
	init_waitqueue_entry(&entry->wait, current);
	add_wait_queue(wait_address, &entry->wait);
}

内核程序设计时我们在fops写我们的poll函数时,我们会调用poll_wait函数(这里调用的是回调函数__pollwait,刚函数在do_sys_poll阶段被注册,可见,poll_wait的作用,只是为了让驱动程序能找到要唤醒的进程)把进程放入等待队列,进程并没有休眠,我们的驱动程序里实现的poll函数是不会引起休眠的.让进程进入休眠,是前面分析的do_sys_poll函数的“__timeout = schedule_timeout(__timeout)”,进入睡眠后如果我们的驱动程序发现情况就绪,可以把这个队列上挂着的进程唤醒
现在来总结一下poll机制：

1. poll > sys_poll > do_sys_poll > poll_initwait,poll_initwait函数注册一下回调函数__pollwait,它就是我们的驱动程序执行poll_wait时,真正被调用的函数。

2. 接下来执行file->f_op->poll,即我们驱动程序里自己实现的poll函数 它会调用poll_wait把自己挂入某个队列,这个队列也是我们的驱动自己定义的;它还判断一下设备是否就绪。

3. 如果设备未就绪,do_sys_poll里会让进程休眠一定时间

4. 进程被唤醒的条件有2：一是上面说的“一定时间”到了，二是被驱动程序唤醒.驱动程序发现条件就绪时,就把“某个队列”上挂着的进程唤醒，这个队列,就是前面通过poll_wait把本进程挂过去的队列。

5. 如果驱动程序没有去唤醒进程,那么shedule_timeout(__timeou)超时后,会重复2,3动作,直到应用程序的poll调用传入的时间到达。

流程图:


driver.c

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/fs.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <asm/uaccess.h>
#include <asm-arm/irq.h>
#include <linux/irq.h>
#include <asm/io.h>
#include <asm-arm/arch-s3c2410/gpio.h>
#include <asm/arch/regs-gpio.h>
#include <asm/hardware.h>
#include <linux/cdev.h>
#include <linux/device.h>
#include <linux/interrupt.h>
#include <linux/poll.h>

int chr_major;
struct cdev  cgs_dev;
static struct class *led_class;
static struct class_device *led_class_dev;
volatile unsigned long *gpfcon = NULL;
volatile unsigned long *gpfdat = NULL;
DECLARE_WAIT_QUEUE_HEAD(button_queue);
static volatile int ev_press = 0;
static unsigned char value;
struct pin_desc
{
   char *name;
   unsigned int pin;  
   unsigned int irq;
   unsigned int key_val;
};
static struct pin_desc  pd[4] = 
{
   {"S1",S3C2410_GPF0,0,0x01},
   {"S2",S3C2410_GPF2,0,0x02},
   {"S3",S3C2410_GPG3,0,0x03},
   {"S4",S3C2410_GPG11,0,0x04},
};
static irqreturn_t buttons_handler(int irq,void *dev_id)
{
    struct pin_desc *desc =  (struct pin_desc*)dev_id;
	if(strcmp("S1",desc->name)==0)
	{
        value = desc->key_val;
		*gpfdat &= ~(1<<6);
		ev_press = 1;                  /* 表示中断发生了 */
        wake_up_interruptible(&button_queue);   /* 唤醒休眠的进程 */
	}
    if(strcmp("S2",desc->name)==0)
	{
        value = desc->key_val;
		*gpfdat &= ~(1<<5);
		ev_press = 1;                  /* 表示中断发生了 */
        wake_up_interruptible(&button_queue);   /* 唤醒休眠的进程 */
	}
	if(strcmp("S3",desc->name)==0)
	{
        value = desc->key_val;
		*gpfdat &= ~(1<<4);
		ev_press = 1;                  /* 表示中断发生了 */
        wake_up_interruptible(&button_queue);   /* 唤醒休眠的进程 */
	}
	if(strcmp("S4",desc->name)==0)
	{
        value =  0x88;
		*gpfdat &= ~((1<<4) | (1<<5) | (1<<6));
		ev_press = 1;                  /* 表示中断发生了 */
        wake_up_interruptible(&button_queue);   /* 唤醒休眠的进程 */
	}
	return IRQ_RETVAL(IRQ_HANDLED);
}
static int cgs_led_open (struct inode *inode, struct file *file)
{   
    int i;
    *gpfcon &= ~((0x3<<(4*2)) | (0x3<<(5*2)) | (0x3<<(6*2)));
	*gpfcon |= ((0x1<<(4*2)) | (0x1<<(5*2)) | (0x1<<(6*2)));
	for(i=0;i<4;i++)
	{
      pd[i].irq = gpio_to_irq(pd[i].pin);
      request_irq(pd[i].irq,buttons_handler, IRQT_BOTHEDGE,pd[i].name,&pd[i]);
	}
	return 0;
}
ssize_t cgs_led_read (struct file *file, char __user *buf, size_t size, loff_t *ppos)
{
   if (size != 1)
		return -EINVAL;
   wait_event_interruptible(button_queue, ev_press);
   copy_to_user(buf,&value,1);
   ev_press = 0;
   return 1;
}
static unsigned cgs_led_poll(struct file *file, poll_table *wait)
{
   unsigned int mask = 0;
   poll_wait(file,&button_queue,wait);
   if(ev_press)
   	   mask |= POLLIN|POLLRDNORM;
   return mask;
}

struct file_operations cgs_fops ={
   .owner   = THIS_MODULE,
   .open   =  cgs_led_open,
   .read    = cgs_led_read,
   .poll    = cgs_led_poll,
};
static int __init cgs_lcd_init(void)
{

    dev_t dev;
	int result;
	cgs_dev.owner = THIS_MODULE;
	result = alloc_chrdev_region(&dev,0, 1,"cgs_led");
	if(result<0)
	{
       return result;
	}
	chr_major = MAJOR(dev);
	cdev_init(&cgs_dev,&cgs_fops);
	cdev_add(&cgs_dev,MKDEV(chr_major,0),1);
	led_class = class_create(THIS_MODULE,"cgs_led");
	led_class_dev = class_device_create(led_class,NULL,MKDEV(chr_major,0),NULL,"cgs_led");
	gpfcon = (volatile unsigned long*)ioremap(0x56000050,16);
	gpfdat = gpfcon + 1;
	return 0;
}
static void __exit cgs_lcd_exit(void)
{   
    int j;
    unregister_chrdev_region(MKDEV(chr_major,0),1);
	cdev_del(&cgs_dev);
    device_destroy(led_class,MKDEV(chr_major,0));
	class_destroy(led_class);
	iounmap(gpfcon);
	for(j=0;j<4;j++)
    {
       free_irq(pd[j].irq, &pd[j]);
	}
}
module_init(cgs_lcd_init);
module_exit(cgs_lcd_exit);
MODULE_LICENSE("GPL");

test.c

#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <stdio.h>
int main(int argc,char *argv[])
{
  int fd;
  int ret;
  unsigned char buf;
  fd = open("/dev/cgs_led",O_RDWR);
  struct pollfd fds[1];
  fds[0].fd = fd;
  fds[0].events = POLLIN;
  if(fd<0)
  {
    printf("can't open!\n");
  }
  while(1)
  {
   ret = poll(fds,1,4000);
   if (ret == 0)
		{
			printf("time out\n");
		}
		else
		{
			read(fd, &key_val, 1);
			printf("key_val = 0x%x\n", key_val);
		}
  }
  return 0;
}