Cisco与Linux的NAT-Linux实现Cisco风格的NAT

既然看到了Cisco的NAT比较灵活，那么Linux能否实现呢？答案是肯定的！因为Linux的Netfilter是超级灵活的，Linux的NAT不灵活是因为iptables程序的不灵活，xtables-addons的RAWNAT已经朝static nat迈出了重要的一步，是iptables限制了Linux的static nat发展！于是我抛开iptables，先基于Netfilter把内核模块实现，然后用procfs作为用户接口，看看怎么实现Cisco风格的static nat。顺带说一句，之所以做这个程序，是因为我们在产品中真的遇到了这个需求，玩过SIP和FTP的都知道，然而因为工期受限，又怕自己做的这个不稳定，效率也没有优化，因此只能放在这里玩玩，不登大雅之堂。
首先，我们看一下基本原理，我们不希望一条NAT绑定任何N元组或者说流，只是一个一对一的地址映射，以源地址转换为例，在从内到外的方向将源地址A转换为B，在从外到内的方向将源目标地址B转换为A！必须记住，任何时候，源地址转换都在POSTROUTING上来做，而目标地址转换都在PREROUTING上来做，按照上述的陈述，以下图说明：

有了上图作为指示，我们就知道该怎么做了：
1.内核中维护一个映射表，仅仅映射两个地址；
2.在PREROUTING和POSTROUTING两个HOOK点上基于上述的映射表执行NAT动作；
3.实现一个用户接口，可以从用户态进行地址映射的配置
以上3点比较容易实现，实际上使用xtables-addons的RAWNAT其实也能实现static nat，然而要想实现两个方向的自动匹配NAT，必然要配置两条甚至多条，最蛋疼的就是明明就是一条映射，非要写成match的形式，所以还是做成Cisco风格的吧。不管怎样，下面的这个代码的实际nat部分还是使用了RAWNAT的代码！

代码如下：

#include <linux/ip.h>
#include <linux/ipv6.h>
#include <linux/module.h>
#include <linux/skbuff.h>
#include <linux/tcp.h>
#include <linux/udp.h>
#include <linux/list.h>
#include <linux/sysfs.h>
#include <linux/fs.h>
#include <linux/proc_fs.h>
#include <linux/version.h>
#include <linux/netfilter.h>
#include <net/ip.h>
#include "compat_xtables.h"

static inline __be32
remask(__be32 addr, __be32 repl, unsigned int shift)
{
	uint32_t mask = (shift == 32) ? 0 : (~(uint32_t)0 >> shift);
	return htonl((ntohl(addr) & mask) | (ntohl(repl) & ~mask));
}

static void rawnat4_update_l4(struct sk_buff *skb, __be32 oldip, __be32 newip)
{
	struct iphdr *iph = ip_hdr(skb);
	void *transport_hdr = (void *)iph + ip_hdrlen(skb);
	struct tcphdr *tcph;
	struct udphdr *udph;
	bool cond;

	switch (iph->protocol) {
	case IPPROTO_TCP:
		tcph = transport_hdr;
		inet_proto_csum_replace4(&tcph->check, skb, oldip, newip, true);
		break;
	case IPPROTO_UDP:
	case IPPROTO_UDPLITE:
		udph = transport_hdr;
		cond = udph->check != 0;
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 19)
		cond |= skb->ip_summed == CHECKSUM_PARTIAL;
#endif
		if (cond) {
			inet_proto_csum_replace4(&udph->check, skb,
				oldip, newip, true);
			if (udph->check == 0)
				udph->check = CSUM_MANGLED_0;
		}
		break;
	}
}

static unsigned int rawnat4_writable_part(const struct iphdr *iph)
{
	unsigned int wlen = sizeof(*iph);

	switch (iph->protocol) {
	case IPPROTO_TCP:
		wlen += sizeof(struct tcphdr);
		break;
	case IPPROTO_UDP:
		wlen += sizeof(struct udphdr);
		break;
	}
	return wlen;
}

//实现源地址转换
static unsigned int
rawsnat(struct sk_buff **pskb, __be32 addr)
{
	struct iphdr *iph;
	__be32 new_addr;

	iph = ip_hdr(*pskb);
	new_addr = remask(iph->saddr, addr, 32);
	if (iph->saddr == new_addr) {
		return NF_ACCEPT;
	}

	if (!skb_make_writable(pskb, rawnat4_writable_part(iph))){
		return NF_DROP;
	}

	iph = ip_hdr(*pskb);
	csum_replace4(&iph->check, iph->saddr, new_addr);
	rawnat4_update_l4(*pskb, iph->saddr, new_addr);
	iph->saddr = new_addr;
	return NF_ACCEPT;
}

//实现目标地址转换
static unsigned int
rawdnat(struct sk_buff **pskb, __be32 addr)
{
	struct iphdr *iph;
	__be32 new_addr;

	iph = ip_hdr(*pskb);
	new_addr = remask(iph->daddr, addr, 32);
	if (iph->daddr == new_addr)
		return NF_ACCEPT;

	if (!skb_make_writable(pskb, rawnat4_writable_part(iph)))
		return NF_DROP;

	iph = ip_hdr(*pskb);
	csum_replace4(&iph->check, iph->daddr, new_addr);
	rawnat4_update_l4(*pskb, iph->daddr, new_addr);
	iph->daddr = new_addr;
	return NF_ACCEPT;
}

//定义数据结构
struct addr_map {
	struct list_head list;
	__be32 addr[2];
	int type;  //0：源地址转换；1：目标地址转换
};
//全局的map list
static LIST_HEAD(map_list);

static unsigned int ipv4_static_nat_pre(unsigned int hooknum,
                     struct sk_buff *skb,
                     const struct net_device *in,
                     const struct net_device *out,
                     int (*okfn)(struct sk_buff *))
{
	__be32 new_daddr = 0x0;
	struct addr_map *map;
	const struct iphdr *iph = ip_hdr(skb);
	if (list_empty(&map_list)) {
		return NF_ACCEPT;
	}
	//查找是否需要做目标地址转换
	list_for_each_entry(map, &map_list, list) {
		if (map->addr[((map->type-1)&0x00000001)] == iph->daddr) {
			new_daddr = map->addr[map->type&0x00000001];
			break;
		}
	}
	if (new_daddr == 0) {
		return NF_ACCEPT;
	}
	return rawdnat(&skb, new_daddr);
}

static unsigned int ipv4_static_nat_post(unsigned int hooknum,
                     struct sk_buff *skb,
                     const struct net_device *in,
                     const struct net_device *out,
                     int (*okfn)(struct sk_buff *))
{
	__be32 new_saddr = 0x0;
	struct addr_map *map;
	const struct iphdr *iph = ip_hdr(skb);
	if (list_empty(&map_list)) {
		return NF_ACCEPT;
	}
	//查找是否做源地址转换
	list_for_each_entry(map, &map_list, list) {
		if (map->addr[map->type&0x00000001] == iph->saddr) {
			new_saddr = map->addr[((map->type-1)&0x00000001)];
			break;
		}
	}
	if (new_saddr == 0) {
		return NF_ACCEPT;
	}
	return rawsnat(&skb, new_saddr);
}

static struct nf_hook_ops ipv4_static_nat[] __read_mostly = {
    {
        .hook        = ipv4_static_nat_pre,
        .owner        = THIS_MODULE,
        .pf        = NFPROTO_IPV4,
        .hooknum    = NF_INET_PRE_ROUTING,
        .priority    = NF_IP_PRI_NAT_SRC+1,
    },
    {
        .hook        = ipv4_static_nat_post,
        .owner        = THIS_MODULE,
        .pf        = NFPROTO_IPV4,
        .hooknum    = NF_INET_PRE_ROUTING,
        .priority    = NF_IP_PRI_RAW+1,
    },
};

//以下是定义用户接口
//如果需要添加一条source转换。则：
//echo +172.16.4.34-128.129.4.34 >/proc/STATIC_Nat/source
struct proc_dir_entry *nat_entry = NULL;

static ssize_t write_snat(struct file *file, const char __user *buf,
				   size_t count, loff_t *ppos)
{
	struct addr_map *am = NULL;
	char addr_temp[20] = {0};
	__be32 addr1 = 0, addr2 = 0;
	int ret = count;
	int i = 1;
	for (; i < 48; i++) {
		if (buf[i] == '-') {
			memcpy(addr_temp, buf+1, i-1);
			break;
		}
	}
	addr1 = in_aton(addr_temp);
	addr2 = in_aton(buf + i + 1);
	
	if (buf[0] == '+') {
		am = kzalloc(sizeof(struct addr_map), GFP_KERNEL);
		INIT_LIST_HEAD(&am->list);
		am->addr[0] = addr1;
		am->addr[1] = addr2;
		am->type = 0;
		list_add(&am->list, &map_list);
	} else if(buf[0] == '-') {
		//Remove TODO
	}
	return ret;
}

static ssize_t write_dnat(struct file *file, const char __user *buf,
				   size_t count, loff_t *ppos)
{
	//TODO
	return 0;
}
static ssize_t read_snat(struct file *file, char __user *buf,
				   size_t count, loff_t *ppos)
{
	//TODO
	return 0;
}
static ssize_t read_dnat(struct file *file, char __user *buf,
				   size_t count, loff_t *ppos)
{
	//TODO
	return 0;
}

static const struct file_operations proc_snat_operations = {
	.read		= read_snat,
	.write		= write_snat,
};
static const struct file_operations proc_dnat_operations = {
	.read		= read_dnat,
	.write		= write_dnat,
};

static int __init static_nat_zy_init(void)
{
	int ret = 0;
	ret = nf_register_hooks(ipv4_static_nat, ARRAY_SIZE(ipv4_static_nat));
	if (ret < 0) {
		printk("ipv4_static_nat: can't register hooks.\n");
	} /* test */else {
		nat_entry = proc_mkdir("STATIC_Nat", NULL);
		proc_create("source", S_IWUSR, nat_entry, &proc_snat_operations);
		proc_create("destination", S_IWUSR, nat_entry, &proc_dnat_operations);
	}  
		
	return ret;
}

static void __exit static_nat_zy_exit(void)
{
	
	remove_proc_entry("source", nat_entry);
	remove_proc_entry("destination", nat_entry);
	remove_proc_entry("STATIC_Nat", NULL);
	nf_unregister_hooks(ipv4_static_nat, ARRAY_SIZE(ipv4_static_nat));
	return;
}

module_init(static_nat_zy_init);
module_exit(static_nat_zy_exit);
//保留原作者
MODULE_AUTHOR("Jan Engelhardt <[e-mail]>");
MODULE_AUTHOR("wangran <[e-mail]>");
MODULE_DESCRIPTION("Static NAT");
MODULE_LICENSE("GPL");

如此就可以实现Cisoc风格的NAT了。代码的优化空间还是有很多的，比如list可以换成hash...