linux内核链表list_head的原理与使用
一般的链表
一般的链表如下,它是将数据结构塞入链表,这样的话,每一种数据结构都需要重新定义。
struct student_node {
char name[20];
int age;
struct student_node *next;
struct student_node *prev;
}
Linux内核中的链表
内核版本:4.1.15, 在目录include/linux/types.h中定义了list_head结构:
struct list_head {
struct list_head *next, *prev;
};
可以看到,很简单,Linux不是将数据结构塞入链表,而是将链表塞入数据结构。
list_head主要的操作
在目录include/linux/types.h中定义了list_head的操作函数:
static inline void INIT_LIST_HEAD(struct list_head *list) //初始化一个链表头,
{ //将其中的next和prev都指向自己
list->next = list;
list->prev = list;
}
static inline void __list_add(struct list_head *_new,
struct list_head *prev,
struct list_head *next)
{
next->prev = _new;
_new->next = next;
_new->prev = prev;
prev->next = _new;
}
static inline void list_add(struct list_head *new, struct list_head *head)//在头部添加一个节点
{
__list_add(new, head, head->next);
}
static inline void list_add_tail(struct list_head *_new, struct list_head *head)//在尾部添加一个节点
{
__list_add(_new, head->prev, head);
}
#define container_of(ptr, type, member) ({ \
const typeof( ((type *)0)->member ) *__mptr = (ptr); \
(type *)( (char *)__mptr - offsetof(type,member) );})
#define list_entry(ptr, type, member) \
container_of(ptr, type, member)
#define list_for_each_entry(pos, head, member) \ //遍历head
for (pos = list_entry((head)->next, typeof(*pos), member); \
&pos->member != (head); \
pos = list_entry(pos->member.next, typeof(*pos), member))
内核链表的使用
struct student {
char name[20];
int age;
struct list_head list;
};
int main()
{
//struct list_head stu_head = {&stu_head, &stu_head};
struct list_head stu_head;
INIT_LIST_HEAD(&stu_head);
struct student *stu1;
stu1 = malloc(sizeof(*stu1));
strcpy(stu1->name, "jack");
stu1->age = 20;
struct student *stu2;
stu2 = malloc(sizeof(*stu2));
strcpy(stu2->name, "bob");
stu2->age = 30;
struct student *stu3;
stu3 = malloc(sizeof(*stu3));
strcpy(stu3->name, "bobc");
stu3->age = 31;
list_add(&stu1->list, &stu_head);
list_add(&stu2->list, &stu_head);
list_add(&stu3->list, &stu_head);
struct student *stu;
list_for_each_entry(stu, &stu_head, list) {
printf("stu->name=%s\n",stu->name);
}
return 0;
}
输出如下:
stu->name=bobc
stu->name=bob
stu->name=jack
可以看到,linux内核中链表操作的都是list_head结构,那么它是怎样通过list_head来访问对应的数据结构里的成员的呢?
Liunx内核链表的原理
从上面的代码可以知道,遍历是通过list_for_each_entry宏来实现的,该宏是一个for循环,可以使用gcc的预编译命令来查看list_for_each_entry展开后的代码:
gcc -E -o main.i main.c
查看main.i中宏的位置,展开后如下:
struct student *stu;
for (stu = ({ const typeof( ((typeof(*stu) *)0)->list ) *__mptr = ((&stu_head)->next); (typeof(*stu) *)( (char *)__mptr - ((size_t) &((typeof(*stu) *)0)->list) );}); &stu->list != (&stu_head); stu = ({ const typeof( ((typeof(*stu) *)0)->list ) *__mptr = (stu->list.next); (typeof(*stu) *)( (char *)__mptr - ((size_t) &((typeof(*stu) *)0)->list) );})) {
printf("stu->name=%s\n",stu->name);
}
展开后分为for循环中的三个部分:
stu = ({ const typeof( ((typeof(*stu) *)0)->list ) *__mptr = ((&stu_head)->next); (typeof(*stu) *)( (char *)__mptr - ((size_t) &((typeof(*stu) *)0)->list) );})
&stu->list != (&stu_head);
stu = ({ const typeof( ((typeof(*stu) *)0)->list ) *__mptr = (stu->list.next); (typeof(*stu) *)( (char *)__mptr - ((size_t) &((typeof(*stu) *)0)->list) );})
分别对应for语句中的三份代码,可以看到,遍历的终止条件是&stu->list != (&stu_head),如果&stu->list = (&stu_head),说明遍历到头节点了,循环结束,下面简化一下第一份代码:
const list_head *__mptr = ((&stu_head)->next);
(struct student *)( (char *)__mptr - ((size_t) &((struct student *)0)->list);
可以看到, 最终stu的值是:
(struct student *)( (char *)__mptr - ((size_t) &((struct student *)0)->list);
将第一步细化实现如下:
/*手动实现从list_head访问student*/
struct list_head *mptr = stu_head.next;
unsigned long long offset = (unsigned long long)(&(((struct student *)0)->list));
mptr = (unsigned long long)mptr - offset;
stu = (struct student *)mptr;
printf("stu = %s %p\n", stu->name, stu);
printf("stu1 = %p\n", stu1);
printf("stu2 = %p\n", stu2);
printf("stu3 = %p\n", stu3);
输出如下:
stu = bobc 0x55d8023972c0
stu1 = 0x55d802397260
stu2 = 0x55d802397290
stu3 = 0x55d8023972c0
可以看到,stu最终等于stu3,因为stu3最后一个插入。画了一份草图便于理解:
可以看到,linux能够实现从list_head访问到student主要是使用了offset,offset是student结构体的起始地址到list_head的大小,再使用list_head的地址减去这段大小即可得到该结构体的起始地址,从而访问结构体成员。