虚机中访问外网；NAT中的POSTROUTING是怎么搞的？

时间：2018-02-19 19:19:18 阅读：365 评论：0 收藏：0 [点我收藏+]

标签：bsp ret int hook finish 记录 intel ring font

看下docker中是怎么配置的网络

在虚机中访问外网：设定了qemu，在主机上添加路由：sudo iptables -t nat -I POSTROUTING -s 192.168.1.110 -j SNAT --to-source 192.168.0.108

设置了这句话就可以访问外网了。

设置了两个虚拟机：

tap0 (192.168.129.1) --->

tap1 (192,168.130.1) --->

增加nat的NAT的表项设置： sudo iptables -t nat -I POSTROUTING -s 192.168.128.0/20 -j SNAT --to-source 192.168.0.108

同时去访问我的云主机：121.X.X.X，从两个主机中都能ping得通，这说明在NAT记录了这个地址，记录着

两个典型包：

192.168.129.110 --->云主机 ( 192.168.0.108 ---> 云主机)

192.168.130.110 --->云主机 ( 192.168.0.108 ---> 云主机)

NAT内部是怎么记录的这个转换？是记录咋的？从云主机IP中回来了一个包，目的地址是192.168.0.108，怎么分别分流到 192.168.129.110 和 192.168.130.110 两个 IP地址中。

难道是端口的信息在里面？接受数据包的流程

#0  icmp_rcv (skb=0xffff88007c9efc00) at net/ipv4/icmp.c:973
#1  0xffffffff816d97af in ip_local_deliver_finish (
    net=0xffffffff81ed8680 <init_net>, sk=<optimized out>, 
    skb=0xffff88007c9efc00) at net/ipv4/ip_input.c:216
#2  0xffffffff816d9e45 in NF_HOOK_THRESH (thresh=<optimized out>, 
    okfn=<optimized out>, out=<optimized out>, in=<optimized out>, 
    skb=<optimized out>, sk=<optimized out>, net=<optimized out>, 
    hook=<optimized out>, pf=<optimized out>)
    at ./include/linux/netfilter.h:232
#3  NF_HOOK (okfn=<optimized out>, out=<optimized out>, 
    in=<optimized out>, skb=<optimized out>, sk=<optimized out>, 
    net=<optimized out>, hook=<optimized out>, pf=<optimized out>)
    at ./include/linux/netfilter.h:255
#4  ip_local_deliver (skb=0xffff88007c9efc00)
    at net/ipv4/ip_input.c:257
#5  0xffffffff816d9a7b in dst_input (skb=<optimized out>)
    at ./include/net/dst.h:507
#6  ip_rcv_finish (net=0xffffffff81ed8680 <init_net>, 
    sk=<optimized out>, skb=0xffff88007c9efc00)
    at net/ipv4/ip_input.c:396
#7  0xffffffff816da11e in NF_HOOK_THRESH (thresh=<optimized out>, 
    okfn=<optimized out>, out=<optimized out>, in=<optimized out>, 
    skb=<optimized out>, sk=<optimized out>, net=<optimized out>, 
    hook=<optimized out>, pf=<optimized out>)
    at ./include/linux/netfilter.h:232
#8  NF_HOOK (okfn=<optimized out>, out=<optimized out>, 
    in=<optimized out>, skb=<optimized out>, sk=<optimized out>, 
    net=<optimized out>, hook=<optimized out>, pf=<optimized out>)
    at ./include/linux/netfilter.h:255
#9  ip_rcv (skb=0xffff88007c9efc00, dev=0xffff88007c530000, 
    pt=<optimized out>, orig_dev=<optimized out>)
    at net/ipv4/ip_input.c:487
#10 0xffffffff81684eea in __netif_receive_skb_core (
    skb=0xffff88007c9efc00, pfmemalloc=<optimized out>)
    at net/core/dev.c:4211
#11 0xffffffff816878cd in __netif_receive_skb (skb=<optimized out>)
    at net/core/dev.c:4249
#12 0xffffffff8168793d in netif_receive_skb_internal (
    skb=0xffff88007c9efc00) at net/core/dev.c:4277
#13 0xffffffff81688582 in napi_skb_finish (skb=<optimized out>, 
    ret=<optimized out>) at net/core/dev.c:4626
---Type <return> to continue, or q <return> to quit---
#14 napi_gro_receive (napi=0xffff88007c530b70, skb=0xffff88007c9efc00)
    at net/core/dev.c:4658
#15 0xffffffff81532db1 in e1000_receive_skb (skb=<optimized out>, 
    vlan=<optimized out>, status=<optimized out>, 
    adapter=<optimized out>)
    at drivers/net/ethernet/intel/e1000/e1000_main.c:4035
#16 e1000_clean_rx_irq (adapter=0xffff88007c5308c0, 
    rx_ring=<optimized out>, work_done=<optimized out>, 
    work_to_do=<optimized out>)
    at drivers/net/ethernet/intel/e1000/e1000_main.c:4491
#17 0xffffffff81531bb0 in e1000_clean (napi=0xffff88007c530b70, 
    budget=64) at drivers/net/ethernet/intel/e1000/e1000_main.c:3836
#18 0xffffffff8168968a in napi_poll (repoll=<optimized out>, 
    n=<optimized out>) at net/core/dev.c:5158
#19 net_rx_action (h=<optimized out>) at net/core/dev.c:5223
#20 0xffffffff8187c0d9 in __do_softirq () at kernel/softirq.c:284
#21 0xffffffff81058f70 in invoke_softirq () at kernel/softirq.c:364
#22 irq_exit () at kernel/softirq.c:405
#23 0xffffffff8187be94 in exiting_irq ()
    at ./arch/x86/include/asm/apic.h:659
#24 do_IRQ (regs=0xffffc9000006be08) at arch/x86/kernel/irq.c:251
#25 0xffffffff8187a4bf in common_interrupt ()
    at arch/x86/entry/entry_64.S:520
#26 0xffffc9000006be08 in ?? ()
#27 0x0000000000000000 in ?? ()

设置完SNAT后接收icmp包: NAT是

当服务器14.17.88.99回复了一个数据包后(src=14.17.88.99 dst=115.22.112.12)，进入到wan侧接口的PRE_ROUTING链时，
则在调用其nat相关的hook函数后，会调用函数ip_nat_packet获取到 origin tuple 值，然后再根据 origin tuple，计算出反方向的tuple，
即为new_tuple.src = 14.17.88.99 new_tuple.dst = 192.168.1.123,然后就会根据这个新的tuple修改其目的ip地址，
修改后的数据包的目的地址即为192.168.1.123 。然后再查找路由，将数据发送到正常的lan口。这就是nat的De-SNAT

路由地址：

ipt_do_table -->

nf_nat_ipv4_fn

在nf_nat_ipv4_fn函数中，首先上来是：nf_ct_get，ct: conntrack,　其中涉及到的数据结构有：

ip_conntrace_info / nf_conn_nat

下面的链接中有一个，详细解释了当设置ＳＮＡＴ之后，出包和进包的一个流程

http://blog.csdn.net/lickylin/article/details/36740207

当数据到达路由器的wan0口，进入到PRE_ROUTING时，会先建立一个nf_conn结构，和两个nf_conntrack_tuple（origin 与reply）

问题

１）prerouting 在哪里？