Home  >  Article  >  System Tutorial  >  In-depth discussion of the implementation principles and related technologies of Linux's universal two-way circular linked list

In-depth discussion of the implementation principles and related technologies of Linux's universal two-way circular linked list

WBOY
WBOYforward
2024-02-13 23:09:13888browse

In embedded Linux, a two-way circular linked list is a very important data structure. They are widely used in various scenarios, such as kernel modules, drivers, network protocol stacks, etc. In this article, we will delve into the implementation principles and related technologies of Linux's common two-way circular linked list.

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

struct list_head {
    struct list_head *next, *prev;
};

This is the element structure of the linked list. Because it is a circular linked list, both the header and the nodes in the list have this structure. There are two pointers, prev and next, which point to the previous node and the next node in the linked list respectively.

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
/*
 * Simple doubly linked list implementation.
 *
 * Some of the internal functions ("__xxx") are useful when
 * manipulating whole lists rather than single entries, as
 * sometimes we already know the next/prev entries and we can
 * generate better code by using them directly rather than
 * using the generic single-entry routines.
 */

#define LIST_HEAD_INIT(name) { &(name), &(name) }

#define LIST_HEAD(name) \
    struct list_head name = LIST_HEAD_INIT(name)

static inline void INIT_LIST_HEAD(struct list_head *list)
{
    list->next = list;
    list->prev = list;
}
In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

During initialization, the prev and next of the linked list header point to itself.

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
/*
 * Insert a new entry between two known consecutive entries.
 *
 * This is only for internal list manipulation where we know
 * the prev/next entries already!
 */
#ifndef CONFIG_DEBUG_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;
}
#else
extern void __list_add(struct list_head *new,
                  struct list_head *prev,
                  struct list_head *next);
#endif

/**
 * list_add - add a new entry
 * @new: new entry to be added
 * @head: list head to add it after
 *
 * Insert a new entry after the specified head.
 * This is good for implementing stacks.
 */
static inline void list_add(struct list_head *new, struct list_head *head)
{
    __list_add(new, head, head->next);
}


/**
 * list_add_tail - add a new entry
 * @new: new entry to be added
 * @head: list head to add it before
 *
 * Insert a new entry before the specified head.
 * This is useful for implementing queues.
 */
static inline void list_add_tail(struct list_head *new, struct list_head *head)
{
    __list_add(new, head->prev, head);
}
In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

The implementation of a two-way circular linked list has few exceptions and can basically be handled in a public way. Whether adding the first node or other nodes, the method used here is the same.
In addition, the implementation of the linked list API is roughly divided into two layers: an external layer, such as list_add, list_add_tail, used to eliminate some exceptions and call the internal implementation; an internal layer, with double underscores added before the function name, such as _ _list_add is often a common part of several operations, or an implementation after excluding exceptions.

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
/*
 * Delete a list entry by making the prev/next entries
 * point to each other.
 *
 * This is only for internal list manipulation where we know
 * the prev/next entries already!
 */
static inline void __list_del(struct list_head * prev, struct list_head * next)
{
    next->prev = prev;
    prev->next = next;
}

/**
 * list_del - deletes entry from list.
 * @entry: the element to delete from the list.
 * Note: list_empty() on entry does not return true after this, the entry is
 * in an undefined state.
 */
#ifndef CONFIG_DEBUG_LIST
static inline void list_del(struct list_head *entry)
{
    __list_del(entry->prev, entry->next);
    entry->next = LIST_POISON1;
    entry->prev = LIST_POISON2;
}
#else
extern void list_del(struct list_head *entry);
#endif

/**
 * list_del_init - deletes entry from list and reinitialize it.
 * @entry: the element to delete from the list.
 */
static inline void list_del_init(struct list_head *entry)
{
    __list_del(entry->prev, entry->next);
    INIT_LIST_HEAD(entry);
}
In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

list_del is the deletion of nodes in the linked list. The reason why after calling __list_del is to point the next and prev of the deleted element to the special LIST_POSITION1 and LIST_POSITION2 is to debug undefined pointers.
list_del_init is to initialize the pointer in the node again after deleting the node. This deletion method is more practical.

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
/**
 * list_replace - replace old entry by new one
 * @old : the element to be replaced
 * @new : the new element to insert
 *
 * If @old was empty, it will be overwritten.
 */
static inline void list_replace(struct list_head *old,
                struct list_head *new)
{
    new->next = old->next;
    new->next->prev = new;
    new->prev = old->prev;
    new->prev->next = new;
}

static inline void list_replace_init(struct list_head *old,
                    struct list_head *new)
{
    list_replace(old, new);
    INIT_LIST_HEAD(old);
}
In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

list_replace is to replace a node old in the linked list with another node new. From an implementation point of view, even if the linked list where old is located has only one node old, new can successfully replace it. This is the terrible universality of two-way circular linked lists.
list_replace_init will replace old and then initialize it again.

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
/**
 * list_move - delete from one list and add as another's head
 * @list: the entry to move
 * @head: the head that will precede our entry
 */
static inline void list_move(struct list_head *list, struct list_head *head)
{
    __list_del(list->prev, list->next);
    list_add(list, head);
}

/**
 * list_move_tail - delete from one list and add as another's tail
 * @list: the entry to move
 * @head: the head that will follow our entry
 */
static inline void list_move_tail(struct list_head *list,
                  struct list_head *head)
{
    __list_del(list->prev, list->next);
    list_add_tail(list, head);
}
In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

The function of list_move is to remove the list node from the original linked list and add it to the new linked list head.
list_move_tail is only different from list_move when adding a new linked list. list_move is added to the head of the linked list after head, while list_move_tail is added to the tail of the linked list before head.

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
/**
 * list_is_last - tests whether @list is the last entry in list @head
 * @list: the entry to test
 * @head: the head of the list
 */
static inline int list_is_last(const struct list_head *list,
                const struct list_head *head)
{
    return list->next == head;
}
In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

list_is_last determines whether list is at the end of the head linked list.

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
/**
 * list_empty - tests whether a list is empty
 * @head: the list to test.
 */
static inline int list_empty(const struct list_head *head)
{
    return head->next == head;
}

/**
 * list_empty_careful - tests whether a list is empty and not being modified
 * @head: the list to test
 *
 * Description:
 * tests whether a list is empty _and_ checks that no other CPU might be
 * in the process of modifying either member (next or prev)
 *
 * NOTE: using list_empty_careful() without synchronization
 * can only be safe if the only activity that can happen
 * to the list entry is list_del_init(). Eg. it cannot be used
 * if another CPU could re-list_add() it.
 */
static inline int list_empty_careful(const struct list_head *head)
{
    struct list_head *next = head->next;
    return (next == head) && (next == head->prev);
}
In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

list_empty determines whether the head linked list is empty. Empty means that there is only one linked list head, head.
list_empty_careful also determines whether the head linked list is empty, but the check is more strict.

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
/**
 * list_is_singular - tests whether a list has just one entry.
 * @head: the list to test.
 */
static inline int list_is_singular(const struct list_head *head)
{
    return !list_empty(head) && (head->next == head->prev);
}
In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

list_is_singular 判断head中是否只有一个节点,即除链表头head外只有一个节点。

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
static inline void __list_cut_position(struct list_head *list,
        struct list_head *head, struct list_head *entry)
{
    struct list_head *new_first = entry->next;
    list->next = head->next;
    list->next->prev = list;
    list->prev = entry;
    entry->next = list;
    head->next = new_first;
    new_first->prev = head;
}

/**
 * list_cut_position - cut a list into two
 * @list: a new list to add all removed entries
 * @head: a list with entries
 * @entry: an entry within head, could be the head itself
 *    and if so we won't cut the list
 *
 * This helper moves the initial part of @head, up to and
 * including @entry, from @head to @list. You should
 * pass on @entry an element you know is on @head. @list
 * should be an empty list or a list you do not care about
 * losing its data.
 *
 */
static inline void list_cut_position(struct list_head *list,
        struct list_head *head, struct list_head *entry)
{
    if (list_empty(head))
        return;
    if (list_is_singular(head) &&
        (head->next != entry && head != entry))
        return;
    if (entry == head)
        INIT_LIST_HEAD(list);
    else
        __list_cut_position(list, head, entry);
}
In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

list_cut_position 用于把head链表分为两个部分。从head->next一直到entry被从head链表中删除,加入新的链表list。新链表list应该是空的,或者原来的节点都可以被忽略掉。可以看到,list_cut_position中排除了一些意外情况,保证调用__list_cut_position时至少有一个元素会被加入新链表。

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
static inline void __list_splice(const struct list_head *list,
                 struct list_head *prev,
                 struct list_head *next)
{
    struct list_head *first = list->next;
    struct list_head *last = list->prev;

    first->prev = prev;
    prev->next = first;

    last->next = next;
    next->prev = last;
}

/**
 * list_splice - join two lists, this is designed for stacks
 * @list: the new list to add.
 * @head: the place to add it in the first list.
 */
static inline void list_splice(const struct list_head *list,
                struct list_head *head)
{
    if (!list_empty(list))
        __list_splice(list, head, head->next);
}

/**
 * list_splice_tail - join two lists, each list being a queue
 * @list: the new list to add.
 * @head: the place to add it in the first list.
 */
static inline void list_splice_tail(struct list_head *list,
                struct list_head *head)
{
    if (!list_empty(list))
        __list_splice(list, head->prev, head);
}
In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list

list_splice的功能和list_cut_position正相反,它合并两个链表。list_splice把list链表中的节点加入head链表中。在实际操作之前,要先判断list链表是否为空。它保证调用__list_splice时list链表中至少有一个节点可以被合并到head链表中。
list_splice_tail只是在合并链表时插入的位置不同。list_splice是把原来list链表中的节点全加到head链表的头部,而list_splice_tail则是把原来list链表中的节点全加到head链表的尾部。

In-depth discussion of the implementation principles and related technologies of Linuxs universal two-way circular linked list
/**
 * list_splice_init - join two lists and reinitialise the emptied list.
 * @list: the new list to add.
 * @head: the place to add it in the first list.
 *
 * The list at @list is reinitialised
 */
static inline void list_splice_init(struct list_head *list,
                    struct list_head *head)
{
    if (!list_empty(list)) {
        __list_splice(list, head, head->next);
        INIT_LIST_HEAD(list);
    }
}

/**
 * list_splice_tail_init - join two lists and reinitialise the emptied list
 * @list: the new list to add.
 * @head: the place to add it in the first list.
 *
 * Each of the lists is a queue.
 * The list at @list is reinitialised
 */
static inline void list_splice_tail_init(struct list_head *list,
                     struct list_head *head)
{
    if (!list_empty(list)) {
        __list_splice(list, head->prev, head);
        INIT_LIST_HEAD(list);
    }
}

list_splice_init 除了完成list_splice的功能,还把变空了的list链表头重新初始化。
list_splice_tail_init 除了完成list_splice_tail的功能,还吧变空了得list链表头重新初始化。
list操作的API大致如以上所列,包括链表节点添加与删除、节点从一个链表转移到另一个链表、链表中一个节点被替换为另一个节点、链表的合并与拆分、查看链表当前是否为空或者只有一个节点。
接下来,是操作链表遍历时的一些宏,我们也简单介绍一下。

/**
 * list_entry - get the struct for this entry
 * @ptr:    the &struct list_head pointer.
 * @type:    the type of the struct this is embedded in.
 * @member:    the name of the list_struct within the struct.
 */
#define list_entry(ptr, type, member) \
    container_of(ptr, type, member)

list_entry主要用于从list节点查找其内嵌在的结构。比如定义一个结构struct A{ struct list_head list; }; 如果知道结构中链表的地址ptrList,就可以从ptrList进而获取整个结构的地址(即整个结构的指针) struct A *ptrA = list_entry(ptrList, struct A, list);
这种地址翻译的技巧是linux的拿手好戏,container_of随处可见,只是链表节点多被封装在更复杂的结构中,使用专门的list_entry定义也是为了使用方便

/**
 * list_first_entry - get the first element from a list
 * @ptr:    the list head to take the element from.
 * @type:    the type of the struct this is embedded in.
 * @member:    the name of the list_struct within the struct.
 *
 * Note, that list is expected to be not empty.
 */
#define list_first_entry(ptr, type, member) \
    list_entry((ptr)->next, type, member)

list_first_entry是将ptr看完一个链表的链表头,取出其中第一个节点对应的结构地址。使用list_first_entry是应保证链表中至少有一个节点。

/**
 * list_for_each    -    iterate over a list
 * @pos:    the &struct list_head to use as a loop cursor.
 * @head:    the head for your list.
 */
#define list_for_each(pos, head) \
    for (pos = (head)->next; prefetch(pos->next), pos != (head); \
            pos = pos->next)

list_for_each循环遍历链表中的每个节点,从链表头部的第一个节点,一直到链表尾部。中间的prefetch是为了利用平台特性加速链表遍历,在某些平台下定义为空,可以忽略。

/**
 * __list_for_each    -    iterate over a list
 * @pos:    the &struct list_head to use as a loop cursor.
 * @head:    the head for your list.
 *
 * This variant differs from list_for_each() in that it's the
 * simplest possible list iteration code, no prefetching is done.
 * Use this for code that knows the list to be very short (empty
 * or 1 entry) most of the time.
 */
#define __list_for_each(pos, head) \
    for (pos = (head)->next; pos != (head); pos = pos->next)

__list_for_each与list_for_each没什么不同,只是少了prefetch的内容,实现上更为简单易懂。

/**
 * list_for_each_prev    -    iterate over a list backwards
 * @pos:    the &struct list_head to use as a loop cursor.
 * @head:    the head for your list.
 */
#define list_for_each_prev(pos, head) \
    for (pos = (head)->prev; prefetch(pos->prev), pos != (head); \
            pos = pos->prev)

list_for_each_prev与list_for_each的遍历顺序相反,从链表尾逆向遍历到链表头。

/**
 * list_for_each_safe - iterate over a list safe against removal of list entry
 * @pos:    the &struct list_head to use as a loop cursor.
 * @n:        another &struct list_head to use as temporary storage
 * @head:    the head for your list.
 */
#define list_for_each_safe(pos, n, head) \
    for (pos = (head)->next, n = pos->next; pos != (head); \
        pos = n, n = pos->next)

list_for_each_safe 也是链表顺序遍历,只是更加安全。即使在遍历过程中,当前节点从链表中删除,也不会影响链表的遍历。参数上需要加一个暂存的链表节点指针n。

/**
 * list_for_each_prev_safe - iterate over a list backwards safe against removal of list entry
 * @pos:    the &struct list_head to use as a loop cursor.
 * @n:        another &struct list_head to use as temporary storage
 * @head:    the head for your list.
 */
#define list_for_each_prev_safe(pos, n, head) \
    for (pos = (head)->prev, n = pos->prev; \
         prefetch(pos->prev), pos != (head); \
         pos = n, n = pos->prev)

list_for_each_prev_safe 与list_for_each_prev同样是链表逆序遍历,只是加了链表节点删除保护。

/**
 * list_for_each_entry    -    iterate over list of given type
 * @pos:    the type * to use as a loop cursor.
 * @head:    the head for your list.
 * @member:    the name of the list_struct within the struct.
 */
#define list_for_each_entry(pos, head, member)                \
    for (pos = list_entry((head)->next, typeof(*pos), member);    \
         prefetch(pos->member.next), &pos->member != (head);     \
         pos = list_entry(pos->member.next, typeof(*pos), member))

list_for_each_entry不是遍历链表节点,而是遍历链表节点所嵌套进的结构。这个实现上较为复杂,但可以等价于list_for_each加上list_entry的组合。

/**
 * list_for_each_entry_reverse - iterate backwards over list of given type.
 * @pos:    the type * to use as a loop cursor.
 * @head:    the head for your list.
 * @member:    the name of the list_struct within the struct.
 */
#define list_for_each_entry_reverse(pos, head, member)            \
    for (pos = list_entry((head)->prev, typeof(*pos), member);    \
         prefetch(pos->member.prev), &pos->member != (head);     \
         pos = list_entry(pos->member.prev, typeof(*pos), member))

list_for_each_entry_reverse 是逆序遍历链表节点所嵌套进的结构,等价于list_for_each_prev加上list_etnry的组合。

/**
 * list_for_each_entry_continue - continue iteration over list of given type
 * @pos:    the type * to use as a loop cursor.
 * @head:    the head for your list.
 * @member:    the name of the list_struct within the struct.
 *
 * Continue to iterate over list of given type, continuing after
 * the current position.
 */
#define list_for_each_entry_continue(pos, head, member)         \
    for (pos = list_entry(pos->member.next, typeof(*pos), member);    \
         prefetch(pos->member.next), &pos->member != (head);    \
         pos = list_entry(pos->member.next, typeof(*pos), member))

list_for_each_entry_continue也是遍历链表上的节点嵌套的结构。只是并非从链表头开始,而是从结构指针的下一个结构开始,一直到链表尾部。

/**
 * list_for_each_entry_continue_reverse - iterate backwards from the given point
 * @pos:    the type * to use as a loop cursor.
 * @head:    the head for your list.
 * @member:    the name of the list_struct within the struct.
 *
 * Start to iterate over list of given type backwards, continuing after
 * the current position.
 */
#define list_for_each_entry_continue_reverse(pos, head, member)        \
    for (pos = list_entry(pos->member.prev, typeof(*pos), member);    \
         prefetch(pos->member.prev), &pos->member != (head);    \
         pos = list_entry(pos->member.prev, typeof(*pos), member))

list_for_each_entry_continue_reverse 是逆序遍历链表上的节点嵌套的结构。只是并非从链表尾开始,而是从结构指针的前一个结构开始,一直到链表头部。

/**
 * list_for_each_entry_from - iterate over list of given type from the current point
 * @pos:    the type * to use as a loop cursor.
 * @head:    the head for your list.
 * @member:    the name of the list_struct within the struct.
 *
 * Iterate over list of given type, continuing from current position.
 */
#define list_for_each_entry_from(pos, head, member)             \
    for (; prefetch(pos->member.next), &pos->member != (head);    \
         pos = list_entry(pos->member.next, typeof(*pos), member))

list_for_each_entry_from 是从当前结构指针pos开始,顺序遍历链表上的结构指针。

/**
 * list_for_each_entry_safe - iterate over list of given type safe against removal of list entry
 * @pos:    the type * to use as a loop cursor.
 * @n:        another type * to use as temporary storage
 * @head:    the head for your list.
 * @member:    the name of the list_struct within the struct.
 */
#define list_for_each_entry_safe(pos, n, head, member)            \
    for (pos = list_entry((head)->next, typeof(*pos), member),    \
        n = list_entry(pos->member.next, typeof(*pos), member);    \
         &pos->member != (head);                     \
         pos = n, n = list_entry(n->member.next, typeof(*n), member))

list_for_each_entry_safe 也是顺序遍历链表上节点嵌套的结构。只是加了删除节点的保护。

/**
 * list_for_each_entry_safe_continue - continue list iteration safe against removal
 * @pos:    the type * to use as a loop cursor.
 * @n:        another type * to use as temporary storage
 * @head:    the head for your list.
 * @member:    the name of the list_struct within the struct.
 *
 * Iterate over list of given type, continuing after current point,
 * safe against removal of list entry.
 */
#define list_for_each_entry_safe_continue(pos, n, head, member)         \
    for (pos = list_entry(pos->member.next, typeof(*pos), member),         \
        n = list_entry(pos->member.next, typeof(*pos), member);        \
         &pos->member != (head);                        \
         pos = n, n = list_entry(n->member.next, typeof(*n), member))

list_for_each_entry_safe_continue 是从pos的下一个结构指针开始,顺序遍历链表上的结构指针,同时加了节点删除保护。

/**
 * list_for_each_entry_safe_from - iterate over list from current point safe against removal
 * @pos:    the type * to use as a loop cursor.
 * @n:        another type * to use as temporary storage
 * @head:    the head for your list.
 * @member:    the name of the list_struct within the struct.
 *
 * Iterate over list of given type from current point, safe against
 * removal of list entry.
 */
#define list_for_each_entry_safe_from(pos, n, head, member)             \
    for (n = list_entry(pos->member.next, typeof(*pos), member);        \
         &pos->member != (head);                        \
         pos = n, n = list_entry(n->member.next, typeof(*n), member))

list_for_each_entry_safe_from 是从pos开始,顺序遍历链表上的结构指针,同时加了节点删除保护。

/**
 * list_for_each_entry_safe_reverse - iterate backwards over list safe against removal
 * @pos:    the type * to use as a loop cursor.
 * @n:        another type * to use as temporary storage
 * @head:    the head for your list.
 * @member:    the name of the list_struct within the struct.
 *
 * Iterate backwards over list of given type, safe against removal
 * of list entry.
 */
#define list_for_each_entry_safe_reverse(pos, n, head, member)        \
    for (pos = list_entry((head)->prev, typeof(*pos), member),    \
        n = list_entry(pos->member.prev, typeof(*pos), member);    \
         &pos->member != (head);                     \
         pos = n, n = list_entry(n->member.prev, typeof(*n), member))

list_for_each_entry_safe_reverse 是从pos的前一个结构指针开始,逆序遍历链表上的结构指针,同时加了节点删除保护。
至此为止,我们介绍了linux中双向循环链表的结构、所有的操作函数和遍历宏定义。相信以后在linux代码中遇到链表的使用,不会再陌生。

总之,双向循环链表是嵌入式Linux中不可或缺的一部分。它们被广泛应用于各种场景,如内核模块、驱动程序、网络协议栈等。希望本文能够帮助读者更好地理解Linux通用的双向循环链表的实现原理和相关技术。

The above is the detailed content of In-depth discussion of the implementation principles and related technologies of Linux's universal two-way circular linked list. For more information, please follow other related articles on the PHP Chinese website!

Statement:
This article is reproduced at:lxlinux.net. If there is any infringement, please contact admin@php.cn delete