linux2.6.30 内核netfilter部分IPV4发包流程分析

转载请注明 作者: Alanx Email:zhangsuozhu@tom.com    QQ:8540426
                                                                     http://hi.baidu.com/alanx/

当内核准备好要发包时，将调用以下函数，其中最重要的结构是struct sk_buff请查相关资料
int ip_output(struct sk_buff *skb)
{
        struct net_device *dev = skb_dst(skb)->dev; 取发包设备

        IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len); 此处应为snmp网网协议相关的操作

        skb->dev = dev; 设发发包的设备
        skb->protocol = htons(ETH_P_IP); 指定IP协议

        return NF_HOOK_COND(PF_INET, NF_INET_POST_ROUTING, skb, NULL, dev,
                            ip_finish_output,
                            !(IPCB(skb)->flags & IPSKB_REROUTED));
}

最后一句是调用勾子，就是netfilter机制最著名的勾子函数。别急，我们像下看。
NF_HOOK_COND()是一个宏，定义如下:

#define NF_HOOK_COND(pf, hook, skb, indev, outdev, okfn, cond)                 /
({int __ret;                                                                   /
if ((__ret=nf_hook_thresh(pf, hook, (skb), indev, outdev, okfn, INT_MIN, cond)) == 1)/
        __ret = (okfn)(skb);                                                   /
__ret;})

再像下看：
static inline int nf_hook_thresh(u_int8_t pf, unsigned int hook,
                                 struct sk_buff *skb,
                                 struct net_device *indev,
                                 struct net_device *outdev,
                                 int (*okfn)(struct sk_buff *), int thresh,
                                 int cond)
{
        if (!cond)
                return 1;
#ifndef CONFIG_NETFILTER_DEBUG
        if (list_empty(&nf_hooks[pf][hook]))
                return 1;
#endif
        return nf_hook_slow(pf, hook, skb, indev, outdev, okfn, thresh);
}
好像明白点什么了,嗯。cond非0时或nf_hooks[pf][hook]链表为空时，int nf_hook_thresh()返回1。接着NF_HOOK_COND()会调用(okfn)(skb)的指针函数。
根据传过来的参数，我们得知，实际上就是调用：ip_finish_output(sk).最后我们再分析ip_finish_output(sk)，先让我们看看cond和nf_hooks[pf][hook]链表吧。
cond:
    根据我们传来的参数得知就是!(IPCB(skb)->flags & IPSKB_REROUTED),具体没细看，猜想可能是和路由重发有关吧。感兴趣的朋友可以了解一下。

nf_hooks[pf][hook]：
    根据上层传过来的参数据应为：nf_hooks[PF_INET][NF_INET_POST_ROUTING] 。 让我们看看它的定义struct list_head nf_hooks[NFPROTO_NUMPROTO][NF_MAX_HOOKS] __read_mostly;
明白了吧。就是的一链表数组。我们这里用的是其中的nf_hooks[PF_INET][NF_INET_POST_ROUTING]。解释一下的话，应 为PF＿INET协议时，路由发包的勾子函数链表。当我们用iptables或tc时。用netlink和内核通信，在这个链表上挂上很多结点。
有什么用呢？我们接着往下看：nf_hook_slow(pf, hook, skb, indev, outdev, okfn, thresh)
我们看一下它的构成：

nt nf_hook_slow(u_int8_t pf, unsigned int hook, struct sk_buff *skb,
                 struct net_device *indev,
                 struct net_device *outdev,
                 int (*okfn)(struct sk_buff *),
                 int hook_thresh)
{
        struct list_head *elem;
        unsigned int verdict;
        int ret = 0;

        /* We may already have this, but read-locks nest anyway */
        rcu_read_lock();

        elem = &nf_hooks[pf][hook];
next_hook:
        verdict = nf_iterate(&nf_hooks[pf][hook], skb, hook, indev,
                             outdev, &elem, okfn, hook_thresh);
        if (verdict == NF_ACCEPT || verdict == NF_STOP) {
                ret = 1;
        } else if (verdict == NF_DROP) {
                kfree_skb(skb);
                ret = -EPERM;   
        } else if ((verdict & NF_VERDICT_MASK) == NF_QUEUE) {
                if (!nf_queue(skb, elem, pf, hook, indev, outdev, okfn,
                              verdict >> NF_VERDICT_BITS))
                        goto next_hook;
        }
        rcu_read_unlock();
        return ret;
}

关键是这句 verdict = nf_iterate(&nf_hooks[pf][hook], skb, hook, indev,
                             outdev, &elem, okfn, hook_thresh);
根据上面返回的值，决定这个包是否通过NF_ACCEPT 或 丢弃 NF_DROP 还是进行流控入排队 NF_QUEUE.注意。这是个循环goto next_hook

接下来，我们看看关键的verdict = nf_iterate(&nf_hooks[pf][hook], skb, hook, indev, outdev, &elem, okfn, hook_thresh);
unsigned int nf_iterate(struct list_head *head,
                        struct sk_buff *skb,
                        unsigned int hook,
                        const struct net_device *indev,
                        const struct net_device *outdev,
                        struct list_head **i,
                        int (*okfn)(struct sk_buff *),
                        int hook_thresh)
{
        unsigned int verdict;

        /*
         * The caller must not block between calls to this
         * function because of risk of continuing from deleted element.
         */
        list_for_each_continue_rcu(*i, head) {
                struct nf_hook_ops *elem = (struct nf_hook_ops *)*i;

                if (hook_thresh > elem->priority)
                        continue;

                /* Optimization: we don't need to hold module
                   reference here, since function can't sleep. --RR */
                verdict = elem->hook(hook, skb, indev, outdev, okfn);
                if (verdict != NF_ACCEPT) {
#ifdef CONFIG_NETFILTER_DEBUG
                        if (unlikely((verdict & NF_VERDICT_MASK)
                                                        > NF_MAX_VERDICT)) {
                                NFDEBUG("Evil return from %p(%u)./n",
                                        elem->hook, hook);
                                continue;
                        }
#endif
                        if (verdict != NF_REPEAT)
                                return verdict;
                        *i = (*i)->prev;
                }
        }
        return NF_ACCEPT;
}
其中
1、ist_for_each_continue_rcu(*i, head) 是循环遍历head 既刚才所说的nf_hooks[PF_INET][NF_INET_POST_ROUTING]。

2、struct nf_hook_ops *elem = (struct nf_hook_ops *)*i; 取出上面链表的结点。

3、if (hook_thresh > elem->priority) 如果优先级小。则下一个结点这里 hook_thresh为INT_MIN
        continue;
         
4、verdict = elem->hook(hook, skb, indev, outdev, okfn);结于调到勾子了。万岁。最激动人心的事。上面说了勾子是我们用iptables或tc通过与内核通讯的netlink，挂到nf_hooks上的。

让我们看看。 nf_hooks上的结点的结构吧。

typedef unsigned int nf_hookfn(unsigned int hooknum,
                               struct sk_buff *skb,
                               const struct net_device *in,
                               const struct net_device *out,
                               int (*okfn)(struct sk_buff *));

struct nf_hook_ops
{                      
        struct list_head list;    链表
               
        /* User fills in from here down. */
        nf_hookfn *hook;    勾子函数，就是上面调用的
        struct module *owner; 模块
        u_int8_t pf;        如:PF_INT
        unsigned int hooknum;    如:NF_INET_POST_ROUTING
        /* Hooks are ordered in ascending priority. */
        int priority;        优先级：如INT＿MIN等  
};

               
等以上的都处理完，最后，要么把包丢弃了。要最一开始的ip_finish_output处理发包啦！先写到此了！