2010年03月01日 星期一 19时57分02秒
在simpLB的基础上已加入了简单的轮询调度算法，因没有加入连接跟踪机制，目前只能用ping命令测试。
1、simpLB.c
/*简单的轮询调度算法，因没有连接跟踪，目前只能用ping测试*/
#include <linux/kernel.h>
#include <linux/tcp.h> /* for tcphdr */
#include <net/ip.h>
#include <net/tcp.h> /* for csum_tcpudp_magic */
#include <net/udp.h>
#include <net/icmp.h> /* for icmp_send */
#include <net/route.h> /* for ip_route_output */
#include <net/ipv6.h>
#include <net/ip6_route.h>
#include <linux/icmpv6.h>
#include <linux/netfilter.h>
#include <linux/netfilter_ipv4.h>
#include "tools.h"
#define VPORT_TCP 80
#define DPORT_TCP 80
#define VPORT_UDP 4950
#define DPORT_UDP 4950
MODULE_LICENSE("GPL");
/* This is the structure we shall use to register our function */
/* IP address we want to Nat*/
static unsigned char *vmLB_ip = "/xc0/xa8/x7a/x01"; /* 192.168.122.1*/
static unsigned char *vm01_ip = "/xc0/xa8/x63/x65"; /* 192.168.99.101 */
static unsigned char *vm02_ip = "/xc0/xa8/x63/x66"; /* 192.168.99.102 */
static unsigned char *srv_list[2];
static int sahu_id = 0;
/* This is the hook function itself */
unsigned int sahu_pre_routing(unsigned int hooknum,
struct sk_buff *skb,
const struct net_device *in,
const struct net_device *out,
int (*okfn)(struct sk_buff *))
{
unsigned char *srv_addr;
char daddr_str[16];
struct sk_buff *sb = skb;
struct iphdr *iph;
if(!sb) return NF_ACCEPT;
iph = ip_hdr(sb);
if(!iph) return NF_ACCEPT;
if (iph->daddr == *(unsigned int *)vmLB_ip){
srv_addr = srv_list[(sahu_id++)%2];
if(iph->protocol == IPPROTO_TCP){
tcp_dnat_base(skb,*(__be32 *)vmLB_ip,*(__be32 *)srv_addr,htons(VPORT_TCP),htons(DPORT_TCP));
}else if(iph->protocol == IPPROTO_UDP){
udp_dnat_base(skb,*(__be32 *)vmLB_ip,*(__be32 *)srv_addr,htons(VPORT_UDP),htons(DPORT_UDP));
}else{
}
iph->daddr= *(unsigned int *)srv_addr;
ip_send_check(iph);
// skb->local_df = 1;
printk("DNat: %d.%d.%d.%d To:%d.%d.%d.%d/n",
*vmLB_ip, *(vmLB_ip + 1), *(vmLB_ip + 2),*(vmLB_ip + 3),
*srv_addr, *(srv_addr + 1), *(srv_addr + 2),*(srv_addr + 3));
return NF_ACCEPT;
}else{
inet_i2str(iph->daddr,daddr_str);
printk("No DNat for %s/n",daddr_str);
return NF_ACCEPT;
}
}
unsigned int sahu_post_routing(unsigned int hooknum,
struct sk_buff *skb,
const struct net_device *in,
const struct net_device *out,
int (*okfn)(struct sk_buff *))
{
unsigned char *srv_addr;
char saddr_str[16];
struct sk_buff *sb = skb;
struct iphdr *iph;
if(!sb) return NF_ACCEPT;
iph = ip_hdr(sb);
if(!iph) return NF_ACCEPT;
srv_addr = srv_list[(sahu_id+1)%2];
if (iph->saddr == *(unsigned int *)srv_addr){
if(iph->protocol == IPPROTO_TCP){
tcp_snat_base(skb,*(__be32 *)vmLB_ip,*(__be32 *)srv_addr,htons(VPORT_TCP),htons(DPORT_TCP));
}else if(iph->protocol == IPPROTO_UDP){
udp_snat_base(skb,*(__be32 *)vmLB_ip,*(__be32 *)srv_addr,htons(VPORT_UDP),htons(DPORT_UDP));
}else{
}
iph->saddr= *(unsigned int *)vmLB_ip;
ip_send_check(iph);
// skb->local_df = 1;
printk("SNat: %d.%d.%d.%d To:%d.%d.%d.%d/n",
*srv_addr, *(srv_addr + 1), *(srv_addr + 2),*(srv_addr + 3),
*vmLB_ip, *(vmLB_ip + 1), *(vmLB_ip + 2),*(vmLB_ip + 3));
return NF_ACCEPT;
}else{
inet_i2str(iph->saddr,saddr_str);
printk("No SNat for %s/n",saddr_str);
return NF_ACCEPT;
}
}
/* netfilter hooks in this kernel module*/
static struct nf_hook_ops sahu_ops[] __read_mostly = {
{
.hook = sahu_pre_routing,
.owner = THIS_MODULE,
.pf = PF_INET,
.hooknum = NF_INET_PRE_ROUTING,
.priority = 100,
},
{
.hook = sahu_post_routing,
.owner = THIS_MODULE,
.pf = PF_INET,
.hooknum = NF_INET_POST_ROUTING,
.priority = 100,
}
};
/* Initialisation routine */
int init_module()
{
int ret;
ret = nf_register_hooks(sahu_ops,ARRAY_SIZE(sahu_ops));
if(ret<0){
pr_info("can't install simpLB into kernel!/n");
}else{
pr_info("simpLB install into kernel!/n");
}
srv_list[0] = vm01_ip;
srv_list[1] = vm02_ip;
return 0;
}
/* Cleanup routine */
void cleanup_module()
{
nf_unregister_hooks(sahu_ops,ARRAY_SIZE(sahu_ops));
pr_info("simpLB removed from kernel!/n");
}
2、tool.h
//extern static char * inet_i2str(__be32 addr);
int inet_i2str(unsigned int addr,char *addr_str){
unsigned char *p;
int i;
p=(unsigned char *)(&addr);
for(i=0;i<4;i++){
addr_str[i*4+0]=*(p+i)/100+'0';
addr_str[i*4+1]=*(p+i)/10-(*(p+i)/100)*10+'0';
addr_str[i*4+2]=*(p+i)%10+'0';
addr_str[i*4+3]='.';
}
addr_str[15]='/0';
return 0;
}
static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
{
__be32 diff[2] = { ~old, new };
return csum_partial(diff, sizeof(diff), oldsum);
}
static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
{
__be16 diff[2] = { ~old, new };
return csum_partial(diff, sizeof(diff), oldsum);
}
static int tcp_snat_base(struct sk_buff *skb,__be32 vaddr,__be32 daddr,__be16 vport,__be16 dport){
struct tcphdr *tcph;
unsigned int tcphoff;
int oldlen;
tcphoff = ip_hdrlen(skb);
oldlen = skb->len - tcphoff;
tcph = (void *)skb_network_header(skb) + tcphoff;
tcph->source = vport;
tcph->check=
csum_fold(ip_vs_check_diff4(daddr,vaddr,
ip_vs_check_diff2(dport,vport,
~csum_unfold(tcph->check))));
if(skb->ip_summed==CHECKSUM_COMPLETE)
skb->ip_summed=CHECKSUM_NONE;
return 0;
}
static int tcp_dnat_base(struct sk_buff *skb,__be32 vaddr,__be32 daddr,__be16 vport,__be16 dport){
struct tcphdr *tcph;
unsigned int tcphoff;
int oldlen;
tcphoff = ip_hdrlen(skb);
oldlen = skb->len - tcphoff;
tcph = (void *)skb_network_header(skb) + tcphoff;
tcph->dest = dport;
tcph->check=
csum_fold(ip_vs_check_diff4(vaddr,daddr,
ip_vs_check_diff2(vport,dport,
~csum_unfold(tcph->check))));
if(skb->ip_summed==CHECKSUM_COMPLETE)
skb->ip_summed=CHECKSUM_NONE;
return 0;
}
static int udp_snat_base(struct sk_buff *skb,__be32 vaddr,__be32 daddr,__be16 vport,__be16 dport){
struct udphdr *udph;
unsigned int udphoff;
int oldlen;
udphoff = ip_hdrlen(skb);
oldlen = skb->len - udphoff;
udph = (void *)skb_network_header(skb) + udphoff;
udph->source = vport;
udph->check=
csum_fold(ip_vs_check_diff4(daddr,vaddr,
ip_vs_check_diff2(dport,vport,
~csum_unfold(udph->check))));
if(!udph->check){
udph->check = CSUM_MANGLED_0;
}
if(skb->ip_summed==CHECKSUM_COMPLETE)
skb->ip_summed=CHECKSUM_NONE;
return 0;
}
static int udp_dnat_base(struct sk_buff *skb,__be32 vaddr,__be32 daddr,__be16 vport,__be16 dport){
struct udphdr *udph;
unsigned int udphoff;
int oldlen;
udphoff = ip_hdrlen(skb);
oldlen = skb->len - udphoff;
udph = (void *)skb_network_header(skb) + udphoff;
udph->dest = dport;
udph->check=
csum_fold(ip_vs_check_diff4(vaddr,daddr,
ip_vs_check_diff2(vport,dport,
~csum_unfold(udph->check))));
if(!udph->check){
udph->check = CSUM_MANGLED_0;
}
if(skb->ip_summed==CHECKSUM_COMPLETE)
skb->ip_summed=CHECKSUM_NONE;
return 0;
}
3、Makefile
obj-m +=simpLB.o
all:
make -C /lib/modules/`uname -r`/build M=`pwd`
clean:
make -C /lib/modules/`uname -r`/build M=`pwd` clean
install:
/sbin/insmod simpLB.ko
remove:
/sbin/rmmod simpLB
4、测试
测试环境和前一个版本的simpLB模块相同，只是多加了一个Server_2(192.168.99.102)。
4.1、于LoadBalaner上：
编译（make）、安装（make install）simpLB模块。
4.2、于Server_1和Server_2上：
安装netMonitor模块。
4.3、于Client上：
运行：ping 192.168.122.1 -c 4
得到如下字样：
PING vmLB (192.168.122.1) 56(84) bytes of data.
64 bytes from vmLB (192.168.122.1): icmp_seq=1 ttl=63 time=0.238 ms
64 bytes from vmLB (192.168.122.1): icmp_seq=2 ttl=63 time=0.229 ms
64 bytes from vmLB (192.168.122.1): icmp_seq=3 ttl=63 time=0.238 ms
64 bytes from vmLB (192.168.122.1): icmp_seq=4 ttl=63 time=0.283 ms
再分别在LoadBalancer、Server_1和Server_2上 运行dmesg | tail。可以发现，LoadBalaner将4个ping包轮询转发个Server_1和Server_2。