Linux下产生死锁问题分析
死锁是指两个或两个以上的进程在执行过程中，因争夺资源而造成的一种互相等待的现象。死锁要产生必须具备四个必要条件：1. 互斥条件 2. 请求和保持条件 3.不可剥夺条件 4. 环路等待条件。由于资源占用是互斥的，当某个进程提出申请资源后，使得有关进程在无外力协助下，永远分配不到必需的资源而无法继续运行，这就产生了一种特殊现象死锁。
下面举一个Linux环境下产生死锁的程序（首先是驱动部分）：
#include <linux/module.h>
#include <linux/init.h>
#include <linux/fs.h>
#include <asm/uaccess.h>
#include <linux/wait.h>
#include <linux/semaphore.h>
#include <linux/device.h>
#include <linux/cdev.h>
#include <linux/sched.h>
MODULE_LICENSE("GPL");
#define init_MUTEX(LOCKNAME) sema_init(LOCKNAME,1)
#define DEVICE_NAME "CDEV_ZHU"
static struct class *cdev_class;
struct cdev dev_c;
dev_t dev;
static ssize_t globalvar_read(struct file *, char *, size_t, loff_t*);
static ssize_t globalvar_write(struct file *, const char *, size_t, loff_t*);
struct file_operations globalvar_fops =
{
read: globalvar_read, write: globalvar_write,
};
static int global_var = 0;
static struct semaphore sem;
static wait_queue_head_t outq;
static int flag = 0;
static int __init globalvar_init(void)
{
int ret,err;
ret = alloc_chrdev_region(&dev,0,1,DEVICE_NAME) ;
if (ret)
{
printk("globalvar register failure");
}
else
{
cdev_init(&dev_c,&globalvar_fops);
err = cdev_add(&dev_c,dev,1);
if(err)
{
printk(KERN_NOTICE "error %d adding FC_dev\n",err);
unregister_chrdev_region(dev, 1);
return err;
}
else
{
printk("device register success! \n");
}
cdev_class = class_create(THIS_MODULE,DEVICE_NAME);
if(IS_ERR(cdev_class))
{
printk("ERR:cannot create a cdev_class\n");
unregister_chrdev_region(dev, 1);
return -1;
}
device_create(cdev_class, NULL, dev, 0, DEVICE_NAME);
init_MUTEX(&sem);
init_waitqueue_head(&outq);
}
return ret;
}
static void __exit globalvar_exit(void)
{
device_destroy(cdev_class,dev);
class_destroy(cdev_class);
unregister_chrdev_region(dev,1);
printk("globalvar exit \n");
}
static ssize_t globalvar_read(struct file *filp, char *buf, size_t len, loff_t *off)
{
/*
//正常情况下这里不应该注释，顺序应该是先wait_event_interruptible，再down_interruptible才不会导致死锁
if (wait_event_interruptible(outq, flag != 0))
{
return - ERESTARTSYS;
}
*/
if (down_interruptible(&sem))
{
return - ERESTARTSYS;
}
/*
交换了 down_interruptible 和 wait_event_interruptible 会造成死锁通过添加打印语句，可以发现会打印 “size semaphore”其它的像“wake up”,
global_write()函数中的”write_down”和”waking up” 都不会打印,说明在up(&sem)之后 global_read()会立刻获取该信号量，然后进入睡眠。
*/
printk(“size semaphore \n”);
if (wait_event_interruptible(outq, flag != 0))
{
return - ERESTARTSYS;
}
printk("wake up !\n");
flag = 0;
if (copy_to_user(buf, &global_var, sizeof(int)))
{
up(&sem);
return - EFAULT;
}
up(&sem);
return sizeof(int);
}
static ssize_t globalvar_write(struct file *filp, const char *buf, size_t len, loff_t *off)
{
if (down_interruptible(&sem))
{
return - ERESTARTSYS;
}
if (copy_from_user(&global_var, buf, sizeof(int)))
{
up(&sem);
return - EFAULT;
}
up(&sem);
flag = 1;
printk("write done!\n");
wake_up_interruptible(&outq);
printk("waking up \n");
return sizeof(int);
}
module_init(globalvar_init);
module_exit(globalvar_exit);
/* /kernel/semaphore.c */
void up(struct semaphore *sem)
{
unsigned long flags;
spin_lock_irqsave(&sem->lock, flags);
if (likely(list_empty(&sem->wait_list)))
sem->count++;
else
__up(sem);
spin_unlock_irqrestore(&sem->lock, flags);
}
static noinline void __sched __up(struct semaphore *sem)
{
struct semaphore_waiter *waiter = list_first_entry(&sem->wait_list,
struct semaphore_waiter, list);
list_del(&waiter->list);
waiter->up = 1;
wake_up_process(waiter->task);
}