Linux内核如何接收不同类型以太帧

本文档的Copyleft归yfydz所有，使用GPL发布，可以自由拷贝，转载，转载时请保持文档的完整性，严禁用于任何商业用途。
msn: [e-mail]
来源：http://yfydz.cublog.cn

1. 前言
以太头中除了6字节目的MAC地址、6字节源MAC地址外，还有两字节的以太帧类型值，如IPv4为0x0800，ARP为0x0806等，网卡驱动收到以太帧后通过接口函数netif_receive_skb()(netif_rx实际最后也是调用netif_receive_skb)交到上层，而这个接口函数就完成对以太帧类型的区分，交到不同的协议处理程序。如果想自己编写某一以太类型帧的处理程序，需要自己添加相应的
以下Linux内核代码为2.6.17.11。
 
2. 数据结构

每种协议都要定义一个packet_type结构，引导进入相关的协议数据处理函数，所有节点组成一个链表(HASH链表)。
/* include/linux/netdevice.h */
struct packet_type {
 __be16   type; /* This is really htons(ether_type). */
 struct net_device *dev; /* NULL is wildcarded here      */
 int   (*func) (struct sk_buff *,
      struct net_device *,
      struct packet_type *,
      struct net_device *);
 void   *af_packet_priv;
 struct list_head list;
};
参数说明：
type：以太帧类型，16位
dev：所附着的网卡设备，如果为NULL则匹配全部网卡
func：协议入口接收处理函数
af_packet_priv：协议私有数据
list：链表扣
一般各协议的packet_type结构都是静态存在，初始化时只提供type和func两个参数就可以了，每个协议在初始化时都要将此结构加入到系统类型链表中。

3. 处理函数

3.1 添加节点
/* net/core/dev.c */
/**
 * dev_add_pack - add packet handler
 * @pt: packet type declaration
 *
 * Add a protocol handler to the networking stack. The passed &packet_type
 * is linked into kernel lists and may not be freed until it has been
 * removed from the kernel lists.
 *
 * This call does not sleep therefore it can not
 * guarantee all CPU's that are in middle of receiving packets
 * will see the new packet type (until the next received packet).
 */
void dev_add_pack(struct packet_type *pt)
{
 int hash;
 spin_lock_bh(&ptype_lock);
// 如果类型是全部以太类型，则节点链接到ptype_all链
 if (pt->type == htons(ETH_P_ALL)) {
  netdev_nit++;
  list_add_rcu(&pt->list, &ptype_all);
 } else {
// 根据协议类型取个HASH，共15个HASH链表
  hash = ntohs(pt->type) & 15;
// 将节点链接到HASH链表中，list_add_rcu是加了smp_wmb()保护的list_add链表操作
  list_add_rcu(&pt->list, &ptype_base[hash]);
 }
 spin_unlock_bh(&ptype_lock);
}

3.2  删除节点

/**
 * __dev_remove_pack  - remove packet handler
 * @pt: packet type declaration
 *
 * Remove a protocol handler that was previously added to the kernel
 * protocol handlers by dev_add_pack(). The passed &packet_type is removed
 * from the kernel lists and can be freed or reused once this function
 * returns.
 *
 *      The packet type might still be in use by receivers
 * and must not be freed until after all the CPU's have gone
 * through a quiescent state.
 */
void __dev_remove_pack(struct packet_type *pt)
{
 struct list_head *head;
 struct packet_type *pt1;
 spin_lock_bh(&ptype_lock);
// 根据协议类型找是在ptype_all表还是某一HASH链表中
 if (pt->type == htons(ETH_P_ALL)) {
  netdev_nit--;
  head = &ptype_all;
 } else
  head = &ptype_base[ntohs(pt->type) & 15];
// 直接用地址比对进行查找，而不是类型，因为同一个类型也可能有多个节点
 list_for_each_entry(pt1, head, list) {
  if (pt == pt1) {
   list_del_rcu(&pt->list);
   goto out;
  }
 }
 printk(KERN_WARNING "dev_remove_pack: %p not found.\n", pt);
out:
 spin_unlock_bh(&ptype_lock);
}
/**
 * dev_remove_pack  - remove packet handler
 * @pt: packet type declaration
 *
 * Remove a protocol handler that was previously added to the kernel
 * protocol handlers by dev_add_pack(). The passed &packet_type is removed
 * from the kernel lists and can be freed or reused once this function
 * returns.
 *
 * This call sleeps to guarantee that no CPU is looking at the packet
 * type after return.
 */
// 只是__dev_remove_pack()的包裹函数
void dev_remove_pack(struct packet_type *pt)
{
 __dev_remove_pack(pt);
 
 synchronize_net();
}

4. 实例

4.1 IP
/* net/ipv4/af_inet.c */
static struct packet_type ip_packet_type = {
 .type = __constant_htons(ETH_P_IP),
 .func = ip_rcv, // IP接收数据的入口点
};
static int __init inet_init(void)
{
......
 dev_add_pack(&ip_packet_type);
......
由于IP协议部分不能作为内核模块，所以是没有卸载函数的，没必要调用dev_remove_pack()函数。

4.2 8021q vlan
/* net/8021q/vlan.c */
static struct packet_type vlan_packet_type = {
 .type = __constant_htons(ETH_P_8021Q),
 .func = vlan_skb_recv, /* VLAN receive method */
};
......
static int __init vlan_proto_init(void)
{
......
 dev_add_pack(&vlan_packet_type);
......

static void __exit vlan_cleanup_module(void)
{
......
 dev_remove_pack(&vlan_packet_type);
......
由于VLAN可为模块方式存在，所以在模块清除函数中要调用dev_remove_pack()。
 
5. 网络接收

网卡驱动收到数据包构造出skb后，通过接口函数netif_receive_skb()传递到上层进行协议处理分配。

/* net/core/dev.c */
int netif_receive_skb(struct sk_buff *skb)
{
......
// 先查处理所有以太类型的链表各节点
 list_for_each_entry_rcu(ptype, &ptype_all, list) {
  if (!ptype->dev || ptype->dev == skb->dev) {
   if (pt_prev)
    ret = deliver_skb(skb, pt_prev, orig_dev);
   pt_prev = ptype;
  }
 }
......
// 再查指定协议的HASH链表
 list_for_each_entry_rcu(ptype, &ptype_base[ntohs(type)&15], list) {
  if (ptype->type == type &&
      (!ptype->dev || ptype->dev == skb->dev)) {
   if (pt_prev)
    ret = deliver_skb(skb, pt_prev, orig_dev);
   pt_prev = ptype;
  }
 }
......

// 该函数就是调用个协议的接收函数处理该skb包，进入第三层网络层处理
static __inline__ int deliver_skb(struct sk_buff *skb,
      struct packet_type *pt_prev,
      struct net_device *orig_dev)
{
 atomic_inc(&skb->users);
 return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}
 
6. 结论

通过链表挂接方式，Linux内核可以很容易的添加各种协议的接收处理函数。
数据流程:
          _____________
         |             |
网卡驱动--->netif_rx()--->netif_receive_skb()->deliver_skb()->packet_type.func