• linux Network Address Translation NAT 转载 还需要整理


    转贴自:http://alexanderlaw.blog.hexun.com/9791596_d.html

    Network Address Translation

     

          地址转换用来改变源/目的地址/端口,是netfilter的一部分,也是通过hook点上注册相应的结构来工作

          Nat注册的hook点和conntrack相同,只是优先级不同,数据包进入netfilter之后先经过conntrack,再经过nat。而在数据包离开netfilter之前先经过nat,再经过conntrack

    1  nat模块的初始化

    1.1       数据结构    ip_nat_standalone.c

    在ip_conntrack结构中有为nat定义的一个nat结构,为什么把这个结构放在ip_conntrack里呢。简单的说,对于非初始化连接的数据包,即后续的数据包,一旦确定它属于某个连接,则可以直接利用连接状态里的nat信息来进行地址转换;而对于初始数据包,必须在nat表里查找相应的规则,确定了地址转换的内容后,将这些信息放到连接跟踪结构的nat参量里面,供后续的数据包使用。

    #ifdef CONFIG_IP_NF_NAT_NEEDED
    
          struct {
    
               struct ip_nat_info info;
    
               union ip_conntrack_nat_help help;
    
    #if defined(CONFIG_IP_NF_TARGET_MASQUERADE) || \
    
          defined(CONFIG_IP_NF_TARGET_MASQUERADE_MODULE)
    
               int masq_index;
    
    #endif
    
    #if defined(CONFIG_IP_NF_RTSP) || defined(CONFIG_IP_NF_RTSP_MODULE)
    
                    struct ip_nat_rtsp_info rtsp_info;
    
    #endif
    
          } nat;
    
    #endif /* CONFIG_IP_NF_NAT_NEEDED */
    
    
    #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
    
          unsigned long mark;
    
    #endif
    

    它包括两个参数,struct ip_nat_info和union ip_conntrack_nat_help,后一个暂时没什么用,只看前一个

    struct ip_nat_info
    
    {
          /* 用来检测该连接是否已经进行过某类nat初始化了,在新的内核中该参数被去掉了,当然,有其它方法来实现它的作用。 */
    
          int initialized;
    
          unsigned int num_manips;
    
          /* 这个就是用来存储关于如何进行地址转换的相关信息的数据结构,其中IP_NAT_MAX_MANIPS代表某个连接的数据包在经过netfilter一次的过程中最多能进行的地址转换的次数,这里是(2*3)=6 。意思大概是说对于某个连接,如果nat表的每条链上都有一条规则:
    
    NF_IP_PRE_ROUTING==>NF_IP_POST_ROUTING
    如果在NF_IP_PRE_ROUTING上做目的转换,要在NF_IP_POST_ROUTING上做反方向上的源转换
    NF_IP_POST_ROUTING==>NF_IP_PRE_ROUTING
    如果在NF_IP_POST_ROUTING上做源转换,要在NF_IP_PRE_ROUTING上做反方向上的目的转换
    NF_IP_LOCAL_OUT==>NF_IP_LOCAL_IN
    如果在NF_IP_LOCAL_OUT做源转换,要在NF_IP_LOCAL_IN上做反方向上的目的转换  
    
    算下来就是最多进行6次地址转换 */
    
          struct ip_nat_info_manip manips[IP_NAT_MAX_MANIPS];
     
          /* 两个全局hash表,用来将所有需要进行地址转换的连接组织起来 */
    
          struct ip_nat_hash bysource, byipsproto;
    
          /* 做特殊用途,通常是NULL */
    
          struct ip_nat_helper *helper;
    
          struct ip_nat_seq seq[IP_CT_DIR_MAX];
    
    };

    ip_nat_info_manip结构定义如下:

    struct ip_nat_info_manip
    
    {
    
          /* 方向,初始或应答 */
    
          u_int8_t direction;
    
     
    
          /* 转换发生的hook点 */
    
          u_int8_t hooknum;
    
     
    
          /* 转换的类型,源还是目的 */
    
          u_int8_t maniptype;
    
     
    
          /* Manipulations to occur at each conntrack in this dirn. */
    
          struct ip_conntrack_manip manip;
    
    };
    
    struct ip_conntrack_manip
    
    {
    
          u_int32_t ip;
    
          union ip_conntrack_manip_proto u;
    
    };
    
     

    ip_nat_hash结构   ip_nat.h

    struct ip_nat_hash
    
    {
    
          struct list_head list;
    
          struct ip_conntrack *conntrack;
    
    };
    
     
    
     

    1.2       init()函数    ip_nat_standalone.c

    static int __init init(void)
    
    {
    
          return init_or_cleanup(1);
    
    }
    
    init()函数直接调用init_or_cleanup()
    
     
    
    static int init_or_cleanup(int init)
    
    {
    
          int ret = 0;
    
    /* nat依赖于conntrack,这个函数是空的 */
    
          need_ip_conntrack();
    
     
    
          if (!init) goto cleanup;
    
    /* 初始化nat规则 */
    
          ret = ip_nat_rule_init();
    
          if (ret < 0) {
    
               printk("ip_nat_init: can't setup rules.\n");
    
               goto cleanup_nothing;
    
          }
    
    /* 初始化nat */
    
          ret = ip_nat_init();
    
          if (ret < 0) {
    
               printk("ip_nat_init: can't setup rules.\n");
    
               goto cleanup_rule_init;
    
          }
    
    
    /* 注册hook,共在四个hook点上注册了函数,分别是:
    
    NF_IP_PRE_ROUTING   ip_nat_fn
    
    NF_IP_POST_ROUTING  ip_nat_out
    
    NF_IP_LOCAL_OUT   ip_nat_local_fn
    
    NF_IP_LOCAL_IN     ip_nat_fn
    
    NF_IP_LOCAL_OUT和NF_IP_LOCAL_IN需要定义CONFIG_IP_NF_NAT_LOCAL
    
    其中在ip_nat_out和ip_nat_local_fn中都会调用ip_nat_fn
    
    */
    
          ret = nf_register_hook(&ip_nat_in_ops);
    
          if (ret < 0) {
    
               printk("ip_nat_init: can't register in hook.\n");
    
               goto cleanup_nat;
    
          }
    
          ret = nf_register_hook(&ip_nat_out_ops);
    
          if (ret < 0) {
    
               printk("ip_nat_init: can't register out hook.\n");
    
               goto cleanup_inops;
    
          }
    
    #ifdef CONFIG_IP_NF_NAT_LOCAL
    
          ret = nf_register_hook(&ip_nat_local_out_ops);
    
          if (ret < 0) {
    
               printk("ip_nat_init: can't register local out hook.\n");
    
               goto cleanup_outops;
    
          }
    
          ret = nf_register_hook(&ip_nat_local_in_ops);
    
          if (ret < 0) {
    
               printk("ip_nat_init: can't register local in hook.\n");
    
               goto cleanup_localoutops;
    
          }
    
    #endif
    
          return ret;
    
     
    
     cleanup:
    
    #ifdef CONFIG_IP_NF_NAT_LOCAL
    
          nf_unregister_hook(&ip_nat_local_in_ops);
    
     cleanup_localoutops:
    
          nf_unregister_hook(&ip_nat_local_out_ops);
    
     cleanup_outops:
    
    #endif
    
          nf_unregister_hook(&ip_nat_out_ops);
    
     cleanup_inops:
    
          nf_unregister_hook(&ip_nat_in_ops);
    
     cleanup_nat:
    
          ip_nat_cleanup();
    
     cleanup_rule_init:
    
          ip_nat_rule_cleanup();
    
     cleanup_nothing:
    
          MUST_BE_READ_WRITE_UNLOCKED(&ip_nat_lock);
    
          return ret;
    
    }

    1.3  ip_nat_rule_init()函数  ip_nat_rule.c

    int __init ip_nat_rule_init(void)

    {

          int ret;

    /* 注册nat表 */

          ret = ipt_register_table(&nat_table);

          if (ret != 0)

               return ret;

    /* 注册了两个target,一个是snat一个是dnat  */

          ret = ipt_register_target(&ipt_snat_reg);

          if (ret != 0)

               goto unregister_table;

          ret = ipt_register_target(&ipt_dnat_reg);

          if (ret != 0)

               goto unregister_snat;

          return ret;

     unregister_snat:

          ipt_unregister_target(&ipt_snat_reg);

     unregister_table:

          ipt_unregister_table(&nat_table);

          return ret;

    }

    看一下nat表的初始化:

    static struct ipt_table nat_table = {

          .name        = "nat",

          .table         = &nat_initial_table.repl,

          .valid_hooks     = NAT_VALID_HOOKS,

          .lock          = RW_LOCK_UNLOCKED,

          .me      = THIS_MODULE,

    };

    和filter表的初始化类似,一开始规则都是空的

    两个target的初始化:

    static struct ipt_target ipt_snat_reg = {

          .name        = "SNAT",

          .target       = ipt_snat_target,

          .checkentry = ipt_snat_checkentry,

    };

    static struct ipt_target ipt_dnat_reg = {

          .name        = "DNAT",

          .target       = ipt_dnat_target,

          .checkentry = ipt_dnat_checkentry,

    };

    两个target函数分别是ipt_snat_target和ipt_dnat_target

    1.4  ip_nat_init()函数  ipt_nat_core.c

    int __init ip_nat_init(void)

    {

          size_t i;

          /* nat的hash表大小和conntrack的hash表相同 */

          ip_nat_htable_size = ip_conntrack_htable_size;

          /* 初始化了一个叫bysource的全局链表指针 */

          bysource = vmalloc(sizeof(struct list_head) * ip_nat_htable_size*2);

          if (!bysource) {

               return -ENOMEM;

          }

          /* 全局链表指针byipsproto,在bysource之后。bysource和byipsproto实际上也是两个hash表,每个节点是一个ip_nat_hash结构,包含一个list_head和一个ip_conntrack。有点特别的就是nat用两个hash表来组织地址转换的数据结构,其本质是一样的,只是所使用的hash算法不同,bysource一般用于SNAT的处理,计算bysource的hash值的函数是hash_by_src();byipsproto用于DNAT的处理,计算byipsproto的hash值的函数是hash_by_ipsproto()。*/

          byipsproto = bysource + ip_nat_htable_size;

          /* 注册一些内建的协议,&protos是用来维护nat模块中用到的协议结构ip_nat_protocol的全局链表 */

          WRITE_LOCK(&ip_nat_lock);

          list_append(&protos, &ip_nat_protocol_tcp);

          list_append(&protos, &ip_nat_protocol_udp);

          list_append(&protos, &ip_nat_protocol_icmp);

          WRITE_UNLOCK(&ip_nat_lock);

          for (i = 0; i < ip_nat_htable_size; i++) {

    /* 初始化bysource和byipsproto中的所有链表,两个数组的大小都是ip_nat_htables_size,数组的每个节点是一个链表头 */

               INIT_LIST_HEAD(&bysource[i]);

               INIT_LIST_HEAD(&byipsproto[i]);

          }

          IP_NF_ASSERT(ip_conntrack_destroyed == NULL);

    /* 初始化一个ip_conntrack_destroyed函数,ip_nat_cleanup_conntrack(struct ip_conntrack *conn) 的作用是在bysource和byipproto链表中删除conn对应的节点 */

          ip_conntrack_destroyed = &ip_nat_cleanup_conntrack;

         

          /* Initialize fake conntrack so that NAT will skip it */

          ip_conntrack_untracked.nat.info.initialized |=

               (1 << IP_NAT_MANIP_SRC) | (1 << IP_NAT_MANIP_DST);

          return 0;

    }

    2  地址转换的过程

    2.1  ip_nat_fn函数  ip_nat_standalone.c

    ip_nat_fn()是nat中的主要函数,nat在netfilter中注册了四个hook,最终都会调用该函数

    static unsigned int
    
    ip_nat_fn(unsigned int hooknum,
    
            struct sk_buff **pskb,
    
            const struct net_device *in,
    
            const struct net_device *out,
    
            int (*okfn)(struct sk_buff *))
    
    {
    
          struct ip_conntrack *ct;
    
          enum ip_conntrack_info ctinfo;
    
          struct ip_nat_info *info;
    
            /* 根据所在的hook点判断转换类型是源地址转换还是目的地址转换,为0(IP_NAT_MANIP_SRC)表示源地址转换,为1(IP_NAT_MANIP_DST)表示目的地址转换 */
    
          enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum);
    
     
    
          /* 前面函数中已经处理过分片的情况,这里应该不会再出现分片包了. */
    
          IP_NF_ASSERT(!((*pskb)->nh.iph->frag_off
    
                      & htons(IP_MF|IP_OFFSET)));
    
     
    
          /*因为地址转换会修改数据包,所以这里先初始化将其设置为“未修改”标志,后面进行数据包修改时再来重置这个标志*/
    
          (*pskb)->nfcache |= NFC_UNKNOWN;
    
     
    
          /* 校验和 */
    
          if ((*pskb)->ip_summed == CHECKSUM_HW)
    
               if (skb_checksum_help(pskb, (out == NULL)))
    
                     return NF_DROP;
    
    /*取得数据包的连接状态*/
    
          ct = ip_conntrack_get(*pskb, &ctinfo);
    
          /* 如果找不到对应连接,则应该直接放行它,而不再对其进行转换处理,特别地,ICMP重定向报文将会被丢弃*/
    
          if (!ct) {
    
               /* Exception: ICMP redirect to new connection (not in
    
                       hash table yet).  We must not let this through, in
    
                       case we're doing NAT to the same network. */
    
               if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
    
                     struct icmphdr hdr;
    
     
    
                     if (skb_copy_bits(*pskb, (*pskb)->nh.iph->ihl*4,
    
                                  &hdr, sizeof(hdr)) == 0
    
                         && hdr.type == ICMP_REDIRECT)
    
                          return NF_DROP;
    
               }
    
               return NF_ACCEPT;
    
          }
    
    /* 判断连接状态,调用相应的处理函数*/
    
          switch (ctinfo) {
    
          case IP_CT_RELATED:
    
          case IP_CT_RELATED+IP_CT_IS_REPLY:
    
               if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
    
                     if (!icmp_reply_translation(pskb, ct, hooknum,
    
                                          CTINFO2DIR(ctinfo)))
    
                          return NF_DROP;
    
                     else
    
                           return NF_ACCEPT;
    
               }
    
               /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
    
    /* 如果是一个初始连接的数据包 */
    
          case IP_CT_NEW:
    
               info = &ct->nat.info;
    
     
    
               WRITE_LOCK(&ip_nat_lock);
    
    /* 观察这个连接中的nat部分是否已经被初始化过了,如果有则跳过下面的部分,直接进行地址转换,如果没有,进一步判断 */    
    
          if (!(info->initialized & (1 << maniptype))
    
    #ifndef CONFIG_IP_NF_NAT_LOCAL
    
                   && !(ct->status & IPS_CONFIRMED)
    
    #endif
    
                   ) {
    
                     unsigned int ret;
    
          /* 如果该连接是由expect创建的,并且有expect函数,则在这里调用 */
    
                     if (ct->master
    
                         && master_ct(ct)->nat.info.helper
    
                         && master_ct(ct)->nat.info.helper->expect) {
    
                          ret = call_expect(master_ct(ct), pskb,
    
                                        hooknum, ct, info);
    
                     } else {
    
    #ifdef CONFIG_IP_NF_NAT_LOCAL
    
                          /* LOCAL_IN hook doesn't have a chain!  */
    
                          if (hooknum == NF_IP_LOCAL_IN)
    
                                ret = alloc_null_binding(ct, info,
    
                                                  hooknum);
    
                          else
    
    #endif
    
          /* 既没有被nat修改过,也不是由expect创建,这是一个初始的数据包,开始在nat表中查找规则 */
    
                     ret = ip_nat_rule_find(pskb, hooknum, in, out, ct, info);
    
                     }
    
     
    
                     if (ret != NF_ACCEPT) {
    
                          WRITE_UNLOCK(&ip_nat_lock);
    
                          return ret;
    
                     }
    
               } else
    
    /* 如果该连接的nat部分已经被初始化了,打印调试信息 */
    
                     DEBUGP("Already setup manip %s for ct %p\n",
    
                            maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
    
                            ct);
    
               WRITE_UNLOCK(&ip_nat_lock);
    
               break;
    
     
    
          default:
    
               /* ESTABLISHED */
    
               IP_NF_ASSERT(ctinfo == IP_CT_ESTABLISHED
    
                          || ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
    
               info = &ct->nat.info;
    
          }
    
     
    
          IP_NF_ASSERT(info);
    
          /* 前面已经修改了连接跟踪表,这里正式修改了数据包里的地址 */
    
          return do_bindings(ct, ctinfo, info, hooknum, pskb);
    
    }
    
     

    2.2           ip_nat_rule_find函数  ip_nat_rule.c

    int ip_nat_rule_find(struct sk_buff **pskb,

                    unsigned int hooknum,

                    const struct net_device *in,

                    const struct net_device *out,

                    struct ip_conntrack *ct,

                    struct ip_nat_info *info)

    {

          int ret;

    /* 调用ipt_do_tables函数,第五个参数是&nat_table  */

          ret = ipt_do_table(pskb, hooknum, in, out, &nat_table, NULL);

          if (ret == NF_ACCEPT) {

               if (!(info->initialized & (1 << HOOK2MANIP(hooknum))))

                     /* NUL mapping */

                     ret = alloc_null_binding(ct, info, hooknum);

          }

          return ret;

    }

    nat表和filter表一样,都是通过调用ipt_do_table函数来工作的

    ipt_do_table查找表中的所有entry,如果match全都匹配,则调用target函数

    此时的target函数就是在nat初始化时注册的ipt_snat_target和ipt_dnat_target

    2.3           ipt_s(d)nat_target函数  ip_nat_rule.c

    static unsigned int ipt_snat_target(struct sk_buff **pskb,

                              const struct net_device *in,

                              const struct net_device *out,

                              unsigned int hooknum,

                              const void *targinfo,

                              void *userinfo)

    {

          struct ip_conntrack *ct;

          enum ip_conntrack_info ctinfo;

          IP_NF_ASSERT(hooknum == NF_IP_POST_ROUTING);

    /* 取得数据包的连接状态 */

          ct = ip_conntrack_get(*pskb, &ctinfo);

          /* Connection must be valid and new. */

          IP_NF_ASSERT(ct && (ctinfo == IP_CT_NEW || ctinfo == IP_CT_RELATED));

          IP_NF_ASSERT(out);

          return ip_nat_setup_info(ct, targinfo, hooknum);

    }

    ipt_dnat_target和ipt_snat_target差不多,都是调用ip_nat_setup_info完成地址转换,这里的targinfo参数来自ipt_entry_target结构的unsigned char data[0]参数,一个长度为0的数组,指向target的末尾

    static unsigned int ipt_dnat_target(struct sk_buff **pskb,

                              const struct net_device *in,

                              const struct net_device *out,

                              unsigned int hooknum,

                              const void *targinfo,

                              void *userinfo)

    {

          struct ip_conntrack *ct;

          enum ip_conntrack_info ctinfo;

    #ifdef CONFIG_IP_NF_NAT_LOCAL

          IP_NF_ASSERT(hooknum == NF_IP_PRE_ROUTING

                    || hooknum == NF_IP_LOCAL_OUT);

    #else

          IP_NF_ASSERT(hooknum == NF_IP_PRE_ROUTING);

    #endif

          ct = ip_conntrack_get(*pskb, &ctinfo);

          /* Connection must be valid and new. */

          IP_NF_ASSERT(ct && (ctinfo == IP_CT_NEW || ctinfo == IP_CT_RELATED));

          return ip_nat_setup_info(ct, targinfo, hooknum);

    }

    2.4           ip_nat_setup_info()函数  ip_nat_rule.c

    unsigned int

    ip_nat_setup_info(struct ip_conntrack *conntrack,   /* 数据包的连接状态 */

                 const struct ip_nat_multi_range *mr,       /* 转换后的地址池 */

                 unsigned int hooknum)                    /* hook点 */

    {

          struct ip_conntrack_tuple new_tuple, inv_tuple, reply;

          struct ip_conntrack_tuple orig_tp;

          struct ip_nat_info *info = &conntrack->nat.info;

          int in_hashes = info->initialized;

          MUST_BE_WRITE_LOCKED(&ip_nat_lock);

          IP_NF_ASSERT(hooknum == NF_IP_PRE_ROUTING

                    || hooknum == NF_IP_POST_ROUTING

                    || hooknum == NF_IP_LOCAL_IN

                    || hooknum == NF_IP_LOCAL_OUT);

          IP_NF_ASSERT(info->num_manips < IP_NAT_MAX_MANIPS);

          IP_NF_ASSERT(!(info->initialized & (1 << HOOK2MANIP(hooknum))));

          /* 对当前状态的应答方向的tuple调用invert_tuplepr取反,得到一个orig_tp,如果之前没有进行过地址或端口转换,通常这里得到的orig_tp就等于初始方向的tuple */

          invert_tuplepr(&orig_tp, conntrack->tuplehash[IP_CT_DIR_REPLY].tuple);

          do {

          /* 进行地址转换,new_tuple为转换后的地址的tuple */

            if (!get_unique_tuple(&new_tuple,&orig_tp,mr,conntrack,hooknum))

            {

                     DEBUGP("ip_nat_setup_info: Can't get unique for %p.\n",

                            conntrack);

                     return NF_DROP;

               }

          /* 对new_tuple取反,得到经过转换后的应答方向的tuple  */

               invert_tuplepr(&reply, &new_tuple);

          /* 修改conntrack中的应答方向的reply tuple,在这之前还要检查如果该reply tuple已经在hash表里存在了,即被其它连接占用(存在初始方向tuple不同,应答方向tuple相同的连接),则还要回头继续修改 */

          } while (!ip_conntrack_alter_reply(conntrack, &reply));

          /* 对orig_tp取反,实际上又得到了原conntrack的reply_tuple…… */

          invert_tuplepr(&inv_tuple, &orig_tp);

          /* 将所作转换的相关信息保存到连接状态conntrack里,这样该连接的后续数据包就可以直接利用这些信息进行地址转换,不用重新查找nat表了 */

          /* 如果是源地址改变(SNAT) */

          if (!ip_ct_tuple_src_equal(&new_tuple, &orig_tp)) {

               /* In this direction, a source manip. */

               info->manips[info->num_manips++] =

                     ((struct ip_nat_info_manip)

                      { IP_CT_DIR_ORIGINAL, hooknum,

                        IP_NAT_MANIP_SRC, new_tuple.src });

               IP_NF_ASSERT(info->num_manips < IP_NAT_MAX_MANIPS);

               /* 在相对的hook点上必然有对应的目的地址改变(DNAT) */

               info->manips[info->num_manips++] =

                     ((struct ip_nat_info_manip)

                          /* opposite_hook即是求当前hook点的对应hook点 */

                      { IP_CT_DIR_REPLY, opposite_hook[hooknum],

                        IP_NAT_MANIP_DST, orig_tp.src });

               IP_NF_ASSERT(info->num_manips <= IP_NAT_MAX_MANIPS);

          }

          /* 如果是目的地址改变(DNAT) */

          if (!ip_ct_tuple_dst_equal(&new_tuple, &orig_tp)) {

               /* In this direction, a destination manip */

               info->manips[info->num_manips++] =

                     ((struct ip_nat_info_manip)

                      { IP_CT_DIR_ORIGINAL, hooknum,

                        IP_NAT_MANIP_DST, reply.src });

               IP_NF_ASSERT(info->num_manips < IP_NAT_MAX_MANIPS);

               /* In the reverse direction, a source manip. */

               info->manips[info->num_manips++] =

                     ((struct ip_nat_info_manip)

                      { IP_CT_DIR_REPLY, opposite_hook[hooknum],

                        IP_NAT_MANIP_SRC, inv_tuple.src });

               IP_NF_ASSERT(info->num_manips <= IP_NAT_MAX_MANIPS);

          }

          /* 如果这个连接不是某个连接的预期的连接(子连接),则在全局链表helpers查找对应的ip_nat_helper结构 */

          if (!conntrack->master)

               info->helper = LIST_FIND(&helpers, helper_cmp, struct ip_nat_helper *, &reply);

          /* 转换完了,标记一下 */

          info->initialized |= (1 << HOOK2MANIP(hooknum));

          /* 将所做的地址转换的数据结构加入到全局hash表bysource和byipsproto中,如果该地址转换是某地址转换基础上的再次转换,则用replace_in_hashes替换,反之则用place_in_hashes */

          if (in_hashes) {

               IP_NF_ASSERT(info->bysource.conntrack);

               replace_in_hashes(conntrack, info);

          } else {

               place_in_hashes(conntrack, info);

          }

          return NF_ACCEPT;

    }

    2.5           get_unique_tuple ()函数  ip_nat_core.c

    get_unique_tuple,获得一个唯一的tuple,就是说除了要做地址/段口的转换,还要保证转换得到的tuple是唯一的。

    很复杂的一个函数。。。

    第三个参数是用来替换的地址或端口的范围

    static int

    get_unique_tuple(struct ip_conntrack_tuple *tuple,

                const struct ip_conntrack_tuple *orig_tuple,

                const struct ip_nat_multi_range *mrr,

                struct ip_conntrack *conntrack,

                unsigned int hooknum)

    {

          struct ip_nat_protocol *proto

               = find_nat_proto(orig_tuple->dst.protonum);

          struct ip_nat_range *rptr;

          unsigned int i;

          int ret;

          struct ip_nat_multi_range *mr = (void *)mrr;

          /* 下面这一段比较晕,和p2p,udp打洞等技术有关。 */

          if (hooknum == NF_IP_POST_ROUTING) {

          /* ip_conntrack_manip结构包含一个ip地址和一个协议端口 */

               struct ip_conntrack_manip *manip;

          /* find_appropriate_src函数先调用hash_by_src函数计算orig_tuple的hash值,然后去bysource表里查找,如果能找到源地址和端口都匹配的连接,并且如果该连接的地址/端口本身就满足目标地址/端口范围的话,就直接返回查到的这个连接的源ip */

               manip = find_appropriate_src(orig_tuple, mr);

               if (manip) {

                     /* Apply same source manipulation. */

                     *tuple = ((struct ip_conntrack_tuple)

                            { *manip, orig_tuple->dst });

                     DEBUGP("get_unique_tuple: Found current src map\n");

                     /* 还要保证连接跟踪表里没有这个连接 */

                     if (!ip_nat_used_tuple(tuple, conntrack))

                          return 1;

               }

          }

          /* orig_tuple是转换之前的,tuple是转换之后的 */

          *tuple = *orig_tuple;

          /* 循环 ,尝试mr参数所指定的地址/端口范围,直到能满足其tuple是唯一的 */

          while ((rptr = find_best_ips_proto_fast(tuple, mr, conntrack, hooknum))

                 != NULL) {

               DEBUGP("Found best for "); DUMP_TUPLE(tuple);

          /*  IP_NAT_MANIP_SRC, 进行SNAT

    IP_NAT_MANIP_DST 进行DNAT

    IP_NAT_RANGE_MAP_IPS 在range里指定了IP地址

    IP_NAT_RANGE_PROTO_SPECIFIED 在range里指定了port

    如果没有指定协议端口范围,或者满足了所指定的范围 */

               if ((!(rptr->flags & IP_NAT_RANGE_PROTO_SPECIFIED)

                    || proto->in_range(tuple, HOOK2MANIP(hooknum),

                                &rptr->min, &rptr->max))

                   && !ip_nat_used_tuple(tuple, conntrack)) {

                     ret = 1;

                     goto clear_fulls;

               } else {

                     if (proto->unique_tuple(tuple, rptr,

                                      HOOK2MANIP(hooknum),

                                      conntrack)) {

                          /* Must be unique. */

                          IP_NF_ASSERT(!ip_nat_used_tuple(tuple,

                                                 conntrack));

                          ret = 1;

                          goto clear_fulls;

                     } else if (HOOK2MANIP(hooknum) == IP_NAT_MANIP_DST) {

                          /* Try implicit source NAT; protocol

                                       may be able to play with ports to

                                       make it unique. */

                          struct ip_nat_range r

                                = { IP_NAT_RANGE_MAP_IPS,

                                    tuple->src.ip, tuple->src.ip,

                                    { 0 }, { 0 } };

                          DEBUGP("Trying implicit mapping\n");

                          if (proto->unique_tuple(tuple, &r,

                                           IP_NAT_MANIP_SRC,

                                           conntrack)) {

                                /* Must be unique. */

                                IP_NF_ASSERT(!ip_nat_used_tuple

                                           (tuple, conntrack));

                                ret = 1;

                                goto clear_fulls;

                          }

                     }

                     DEBUGP("Protocol can't get unique tuple %u.\n",

                            hooknum);

               }

               /* Eliminate that from range, and try again. */

               rptr->flags |= IP_NAT_RANGE_FULL;

               *tuple = *orig_tuple;

          }

          ret = 0;

     clear_fulls:

          /* Clear full flags. */

          IP_NF_ASSERT(mr->rangesize >= 1);

          for (i = 0; i < mr->rangesize; i++)

               mr->range[i].flags &= ~IP_NAT_RANGE_FULL;

          return ret;

    }

     
    http代理服务器(3-4-7层代理)-网络事件库公共组件、内核kernel驱动 摄像头驱动 tcpip网络协议栈、netfilter、bridge 好像看过!!!! 但行好事 莫问前程 --身高体重180的胖子
  • 相关阅读:
    文本域光标操作(选、添、删、取)的jQuery扩展
    jQuery插件,将内容插入到光标处
    onmouseout,mouseover经过子元素也触发的问题解决方案
    【M4】非必要不提供default 构造方法
    【M3】绝对不要以多态方式处理数组
    100亿个数字找出最大的10个
    【M2】最好使用C++转型操作符
    【M26】限制某个class所能产生的对象数量
    理解extern
    变量的属性
  • 原文地址:https://www.cnblogs.com/codestack/p/10878918.html
Copyright © 2020-2023  润新知