【智能路由器】ndpi深度报文分析之协议分析器

【智能路由器】系列文章连接
http://blog.csdn.net/u012819339/article/category/5803489

---

本篇博客讲述ndpi已经实现的QQ协议分析器的实现 以及 编写一个微信协议分析器的框架要点。
关于ndpi源码的叙述在上篇博客已有提及，点击这里（或者查看链接http://blog.csdn.net/u012819339/article/details/52443705）

ndpi协议分析器结构框架

每个协议分析器都必须定义一个独有的id，例如：

/*ndpi_protocol_ids.h*/
#define NDPI_PROTOCOL_QQ                    48

协议分析器格式

#include "ndpi_utils.h"
#ifdef NDPI_PROTOCOL_QQ
.....
#endif

协议分析器的初始化函数，初始化函数中调用ndpi_set_bitmask_protocol_detection来输入自己的注册信息，以及处理数据包的类型

void init_qq_dissector(struct ndpi_detection_module_struct *ndpi_struct, u_int32_t *id, NDPI_PROTOCOL_BITMASK *detection_bitmask)
{
  ndpi_set_bitmask_protocol_detection("QQ", ndpi_struct, detection_bitmask, *id,
                      NDPI_PROTOCOL_QQ,
                      ndpi_search_qq,
                      NDPI_SELECTION_BITMASK_PROTOCOL_V4_V6_UDP_WITH_PAYLOAD,
                      SAVE_DETECTION_BITMASK_AS_UNKNOWN,
                      ADD_TO_DETECTION_BITMASK);

  *id += 1;
}

一个回调函数，用来处理数据包，这个函数就是初始化中注册的函数，这个函数将细分情况，然后交给实际的分析函数，这里分别是ndpi_search_qq_udp和ndpi_search_qq_tcp。虽然也可以直接在这里分析处理，但这样会显得层次结构不清晰

void ndpi_search_qq(struct ndpi_detection_module_struct *ndpi_struct, struct ndpi_flow_struct *flow)
{
  struct ndpi_packet_struct *packet = &flow->packet;

  if (packet->udp != NULL && flow->detected_protocol_stack[0] != NDPI_PROTOCOL_QQ)
    ndpi_search_qq_udp(ndpi_struct, flow);

  if (packet->tcp != NULL && flow->detected_protocol_stack[0] != NDPI_PROTOCOL_QQ)
    ndpi_search_qq_tcp(ndpi_struct, flow);
}

协议分析器的框架就到此为止，然后将初始化函数添加到ndpi_main.c文件的ndpi_init_protocol_defaults函数中，像这样：

static void ndpi_init_protocol_defaults(struct ndpi_detection_module_struct *ndpi_mod) {

  int i;
  ndpi_port_range ports_a[MAX_DEFAULT_PORTS], ports_b[MAX_DEFAULT_PORTS];
  u_int16_t no_master[2] = { NDPI_PROTOCOL_NO_MASTER_PROTO, NDPI_PROTOCOL_NO_MASTER_PROTO },
    custom_master[2];

    /* Reset all settings */
    memset(ndpi_mod->proto_defaults, 0, sizeof(ndpi_mod->proto_defaults));

//......
    ndpi_set_proto_defaults(ndpi_mod, NDPI_PROTOCOL_FUN, NDPI_PROTOCOL_QQ,
                    no_master,
                    no_master, "QQ",
                    ndpi_build_default_ports(ports_a, 0, 0, 0, 0, 0) /* TCP */,
                    ndpi_build_default_ports(ports_b, 0, 0, 0, 0, 0) /* UDP */);


//......
}

宏定义NDPI_PROTOCOL_LONG_STRING和NDPI_PROTOCOL_SHORT_STRING中添加协议名称字符串。
协议分析器的结构框架大概如此了

ndpi协议分析中QQ协议分析器的代码概要

根据packet是tcp还是udp将探测过程被分为两中情况：

void ndpi_search_qq(struct ndpi_detection_module_struct *ndpi_struct, struct ndpi_flow_struct *flow)
{
  struct ndpi_packet_struct *packet = &flow->packet;

  if (packet->udp != NULL && flow->detected_protocol_stack[0] != NDPI_PROTOCOL_QQ)
    ndpi_search_qq_udp(ndpi_struct, flow);

  if (packet->tcp != NULL && flow->detected_protocol_stack[0] != NDPI_PROTOCOL_QQ)
    ndpi_search_qq_tcp(ndpi_struct, flow);
}

看一下qq的udp报文是怎么探测的，在下面这个函数中

static void ndpi_search_qq_udp(struct ndpi_detection_module_struct *ndpi_struct, struct ndpi_flow_struct *flow)
{
  struct ndpi_packet_struct *packet = &flow->packet;

  static const u_int16_t p8000_patt_02[13] =    // maybe version numbers
    { 0x1549, 0x1801, 0x180d, 0x0961, 0x01501, 0x0e35, 0x113f, 0x0b37, 0x1131, 0x163a, 0x1e0d, 0x3619,};
  u_int16_t no_of_patterns = 12, index = 0;


  NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG, "search qq udp.\n");


  if (flow->qq_stage <= 3) {
    if ((packet->payload_packet_len == 27 && ntohs(get_u_int16_t(packet->payload, 0)) == 0x0300
     && packet->payload[2] == 0x01)
    || (packet->payload_packet_len == 84 && ((ntohs(get_u_int16_t(packet->payload, 0)) == 0x000e
                          && packet->payload[2] == 0x35)
                         || (ntohs(get_u_int16_t(packet->payload, 0)) == 0x0015
                             && packet->payload[2] == 0x01)
                         || (ntohs(get_u_int16_t(packet->payload, 0)) == 0x000b
                             && packet->payload[2] == 0x37)
                         || (ntohs(get_u_int16_t(packet->payload, 0)) == 0x0015
                             && packet->payload[2] == 0x49)))
    || (packet->payload_packet_len > 10
        && ((get_u_int16_t(packet->payload, 0) == htons(0x000b) && packet->payload[2] == 0x37)
        || (get_u_int32_t(packet->payload, 0) == htonl(0x04163a00)
            && packet->payload[packet->payload_packet_len - 1] == 0x03
            && packet->payload[4] == packet->payload_packet_len)))) {
      /*
    if (flow->qq_stage == 3 && flow->detected_protocol == NDPI_PROTOCOL_QQ) {
    if (flow->packet_direction_counter[0] > 0 && flow->packet_direction_counter[1] > 0) {
    flow->protocol_subtype = NDPI_PROTOCOL_QQ_SUBTYPE_AUDIO;
    return;
    } else if (flow->packet_counter < 10) {
    return;
    }
    } */
      flow->qq_stage++;
      if (flow->qq_stage == 3) {
    NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG,
         "found qq udp pattern 030001 or 000e35 four times.\n");
    ndpi_int_qq_add_connection(ndpi_struct, flow);
    return;
      }
      return;
    }
    if (packet->payload_packet_len > 2 && (packet->payload[0] == 0x02 || packet->payload[0] == 0x04)) {
      u_int16_t pat = ntohs(get_u_int16_t(packet->payload, 1));
      for (index = 0; index < no_of_patterns; index++) {
    if (pat == p8000_patt_02[index] && packet->payload[packet->payload_packet_len - 1] == 0x03) {
      flow->qq_stage++;
      // maybe we can test here packet->payload[4] == packet->payload_packet_len
      if (flow->qq_stage == 3) {
        NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG,
             "found qq udp pattern 02 ... 03 four times.\n");
        /*
          if (packet->payload[0] == 0x04) {
          ndpi_int_qq_add_connection(ndpi_struct, flow, NDPI_REAL_PROTOCOL);
          return;
          } */
        ndpi_int_qq_add_connection(ndpi_struct, flow);
        return;
      }
      return;
    }
      }
    }
    if (packet->payload_packet_len == 84 && (packet->payload[0] == 0 || packet->payload[0] == 0x03)) {
      u_int16_t pat = ntohs(get_u_int16_t(packet->payload, 1));
      for (index = 0; index < no_of_patterns; index++) {
    if (pat == p8000_patt_02[index]) {
      flow->qq_stage++;
      /*
        if (flow->qq_stage == 3 && flow->packet_direction_counter[0] > 0 &&
        flow->packet_direction_counter[1] > 0) {
        NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG, "found qq udp pattern four times.\n");
        ndpi_int_qq_add_connection(ndpi_struct, flow, NDPI_REAL_PROTOCOL);
        return;
        } else */ if (flow->qq_stage == 3) {
        NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG, "found qq udp pattern four times.\n");
        ndpi_int_qq_add_connection(ndpi_struct, flow);
        return;
      }
      return;
    }
      }
    }
    if (packet->payload_packet_len > 2 && packet->payload[0] == 0x04
    && ((ntohs(get_u_int16_t(packet->payload, 1)) == 0x1549
         || ntohs(get_u_int16_t(packet->payload, 1)) == 0x1801 || ntohs(get_u_int16_t(packet->payload, 1)) == 0x0961)
        ||
        (packet->payload_packet_len > 16
         && (ntohs(get_u_int16_t(packet->payload, 1)) == 0x180d || ntohs(get_u_int16_t(packet->payload, 1)) == 0x096d)
         && ntohl(get_u_int32_t(packet->payload, 12)) == 0x28000000
         && ntohs(get_u_int16_t(packet->payload, 3)) == packet->payload_packet_len))
    && packet->payload[packet->payload_packet_len - 1] == 0x03) {
      flow->qq_stage++;
      if (flow->qq_stage == 3) {
    NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG,
         "found qq udp pattern 04 1159 ... 03 four times.\n");
    ndpi_int_qq_add_connection(ndpi_struct, flow);
    return;
      }
      return;
    }
    if (packet->payload_packet_len > 2 && (packet->payload[0] == 0x06 || packet->payload[0] == 0x02)
    && ntohs(get_u_int16_t(packet->payload, 1)) == 0x0100
    && (packet->payload[packet->payload_packet_len - 1] == 0x00
        || packet->payload[packet->payload_packet_len - 1] == 0x03)) {
      flow->qq_stage++;
      if (flow->qq_stage == 3) {
    NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG,
         "found qq udp pattern 02/06 0100 ... 03/00 four times.\n");
    ndpi_int_qq_add_connection(ndpi_struct, flow);
    return;
      }
      return;
    }

    if (packet->payload_packet_len > 2 && (packet->payload[0] == 0x02)
    && ntohs(get_u_int16_t(packet->payload, 1)) == 0x1131 && packet->payload[packet->payload_packet_len - 1] == 0x03) {
      flow->qq_stage++;
      if (flow->qq_stage == 3) {
    NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG,
         "found qq udp pattern 02 1131 ... 03 four times.\n");
    ndpi_int_qq_add_connection(ndpi_struct, flow);
    return;
      }
      return;
    }

    if (packet->payload_packet_len > 5 && get_u_int16_t(packet->payload, 0) == htons(0x0203) &&
    ntohs(get_u_int16_t(packet->payload, 2)) == packet->payload_packet_len &&
    get_u_int16_t(packet->payload, 4) == htons(0x0b0b)) {
      flow->qq_stage++;
      if (flow->qq_stage == 3) {
    NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG,
         "found qq udp pattern 0203[packet_length_0b0b] three times.\n");
    ndpi_int_qq_add_connection(ndpi_struct, flow);
    return;
      }
      return;
    }

    if (packet->udp->dest == htons(9000) || packet->udp->source == htons(9000)) {
      if (packet->payload_packet_len > 3
      && ntohs(get_u_int16_t(packet->payload, 0)) == 0x0202
      && ntohs(get_u_int16_t(packet->payload, 2)) == packet->payload_packet_len) {
    flow->qq_stage++;
    if (flow->qq_stage == 3) {
      NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG,
           "found qq udp pattern 02 02  four times.\n");
      ndpi_int_qq_add_connection(ndpi_struct, flow);
      return;
    }
    return;
      }

    }
  }

  if (ndpi_is_valid_qq_packet(packet)) {
    flow->qq_stage++;
    if (flow->qq_stage == 3) {
      NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG, "found qq over udp.\n");
      ndpi_int_qq_add_connection(ndpi_struct, flow);
      return;
    }
    NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG, "found qq packet stage %d\n", flow->qq_stage);
    return;
  }

  if (ndpi_is_valid_qq_ft_packet(packet)) {
    flow->qq_stage++;
    if (flow->qq_stage == 3) {
      NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG, "found qq ft over udp.\n");
      ndpi_int_qq_add_connection(ndpi_struct, flow);
      return;
    }
    return;
  }

  if (flow->qq_stage && flow->packet_counter <= 5) {
    return;
  }

  NDPI_LOG(NDPI_PROTOCOL_QQ, ndpi_struct, NDPI_LOG_DEBUG, "QQ excluded\n");
  NDPI_ADD_PROTOCOL_TO_BITMASK(flow->excluded_protocol_bitmask, NDPI_PROTOCOL_QQ);
}

该函数有点长，不过基本都是属于探测OICQ协议的报头，oicq协议是qq用于即时通信的协议，数据流经8000端口，具体oicq协议头的解释可参看arvik的另一篇博客http://blog.csdn.net/u012819339/article/details/50374252

得到payload头指针后起始可以直接强制指针类型转换成 oicqhdr *，就比较方便啦

typedef struct _oicqhdr {
    uint8_t flag;
    uint16_t ver;
    uint16_t command;
    uint16_t sequence;
    uint32_t qq;
} __attribute__((packed)) oicqhdr;

tcp检测就不叙述了，总之都是先抓包分析QQ报文头的结构特征，然后拿到协议分析器中做匹配。

ndpi协议分析中编写一个微信协议分析器的方法指导

协议分析器框架见上文所述，可以参照QQ协议分析器依葫芦画瓢，
这里主要是弄清微信协议的格式，这需要自己抓包分析，arvik已经做好了一部分微信抓包分析的过程，见我的博客微信数据包（链接地址http://blog.csdn.net/u012819339/article/details/52588288），微信客户端有很多模块，比如发红包，漂流瓶，摇一摇，传图片，发语音，发短视频等等，每个模块都可能会有一种协议格式，arvik并没有完全对这些模块进行分析，需要开发者自行抓包。

---

好了，本文到此结束，作者arvik。
【智能路由器】系列文章连接
http://blog.csdn.net/u012819339/article/category/5803489