实验三：OpenFlow协议分析实践

一、实验目的

能够运用 wireshark 对 OpenFlow 协议数据交互过程进行抓包；
能够借助包解析工具，分析与解释 OpenFlow协议的数据包交互过程与机制。

二、实验环境

下载虚拟机软件Oracle VisualBox；
在虚拟机中安装Ubuntu 20.04 Desktop amd64，并完整安装Mininet；

三、实验要求

（一）基本要求

搭建下图所示拓扑，完成相关 IP 配置，并实现主机与主机之间的 IP 通信。用抓包软件获取控制器与交换机之间的通信数据包。

1. 配置网段

2. 配置ip地址

3. 保存拓扑为python文件

4. 运行sudo wireshark命令，并选择any模式进行抓包，开启另一个终端，命令行运行031902241.py文件，运行pingall

5. 查看抓包结果，分析OpenFlow协议中交换机与控制器的消息交互过程（截图以其中一个交换机为例）

• OFPT_HELLO 源端口6633 -> 目的端口34068，从控制器到交换机
• 也有源端口34068 -> 目的端口6633的，即交换机到控制器的另一个包，此处协议为openflow1.5控制器与交换机建立连接，并使用OpenFlow 1.0
• OFPT_FEATURES_REQUEST 源端口6633 -> 目的端口34068，从控制器到交换机控制器请求交换器的特征信息
• OFPT_SET_CONFIG 源端口6633 -> 目的端口34068，从控制器到交换机控制器要求交换机按照所给出的信息进行配置
• OFPT_PORT_STATUS 源端口35828 -> 目的端口6633，从交换机到控制器当交换机端口发生变化时，交换机告知控制器相应的端口状态
• OFPT_FEATURES_REPLY 源端口34068 -> 目的端口6633，从交换机到控制器交换机告知控制器它的特征信息
• OFPT_PACKET_IN 源端口34068 -> 目的端口6633，从交换机到控制器交换机告知控制器有数据包进来，请求控制器指示
• OFPT_PACKET_OUT 源端口6633 -> 目的端口35844，从控制器到交换机控制器要求交换机按照所给出的action进行处理
• OFPT_FLOW_MOD 源端口6633 -> 目的端口35844，从控制器到交换机控制器对交换机进行流表的添加、删除、变更等操作
• 上述OFPT_PACKET_IN、OFPT_PACKET_OUT、OFPT_FLOW_MOD三种消息报文的交互会频繁多次出现在交换机和控制器之间。

6.画出相关交互图或流程图：

7.回答问题：交换机与控制器建立通信时是使用TCP协议还是UDP协议？

如图所示为（Transmission Control Protocol）TCP协议。

（二）进阶要求

将抓包结果对照OpenFlow源码，了解OpenFlow主要消息类型对应的数据结构定义。相关数据结构可在openflow安装目录openflow/include/openflow当中的openflow.h头文件中查询到
1. HELLO

struct ofp_header {

    uint8_t version;    /* OFP_VERSION. */

    uint8_t type;       /* One of the OFPT_ constants. */

    uint16_t length;    /* Length including this ofp_header. */

    uint32_t xid;       /* Transaction id associated with this packet.

                           Replies use the same id as was in the request

                           to facilitate pairing. */

};

struct ofp_hello {

    struct ofp_header header;

};

可以看到对应了HELLO报文的四个参数

2. FEATURES_REQUEST

源码参数格式与HELLO相同，与上述ofp_header结构体中数据相同

3.SET_CONFIG

控制器下发的交换机配置数据结构体

/* Switch configuration. */

struct ofp_switch_config {

    struct ofp_header header;

    uint16_t flags;             /* OFPC_* flags. */

    uint16_t miss_send_len;     /* Max bytes of new flow that datapath should

                                   send to the controller. */

};

4. PORT_STATUS

/* A physical port has changed in the datapath */

struct ofp_port_status {

    struct ofp_header header;

    uint8_t reason;          /* One of OFPPR_*. */

    uint8_t pad[7];          /* Align to 64-bits. */

    struct ofp_phy_port desc;

};

5. FEATURES_REPLY

struct ofp_switch_features {

    struct ofp_header header;

    uint64_t datapath_id;   /* Datapath unique ID.  The lower 48-bits are for

                               a MAC address, while the upper 16-bits are

                               implementer-defined. */

    uint32_t n_buffers;     /* Max packets buffered at once. */

    uint8_t n_tables;       /* Number of tables supported by datapath. */

    uint8_t pad[3];         /* Align to 64-bits. */

    /* Features. */

    uint32_t capabilities;  /* Bitmap of support "ofp_capabilities". */

    uint32_t actions;       /* Bitmap of supported "ofp_action_type"s. */

    /* Port info.*/

    struct ofp_phy_port ports[0];  /* Port definitions.  The number of ports

                                      is inferred from the length field in

                                      the header. */

};

/* Description of a physical port */

struct ofp_phy_port {

    uint16_t port_no;

    uint8_t hw_addr[OFP_ETH_ALEN];

    char name[OFP_MAX_PORT_NAME_LEN]; /* Null-terminated */

    uint32_t config;        /* Bitmap of OFPPC_* flags. */

    uint32_t state;         /* Bitmap of OFPPS_* flags. */

    /* Bitmaps of OFPPF_* that describe features.  All bits zeroed if

     * unsupported or unavailable. */

    uint32_t curr;          /* Current features. */

    uint32_t advertised;    /* Features being advertised by the port. */

    uint32_t supported;     /* Features supported by the port. */

    uint32_t peer;          /* Features advertised by peer. */

};

可以看到与图中信息一一对应，包括交换机物理端口的信息

6. PACKET_IN

PACKET_IN有两种情况：

1. 交换机查找流表，发现没有匹配条目,但是这种包没有抓到过

enum ofp_packet_in_reason {

    OFPR_NO_MATCH,          /* No matching flow. */

    OFPR_ACTION             /* Action explicitly output to controller. */

};

1. 有匹配条目,对应的action是OUTPUT=CONTROLLER,固定收到向控制器发送包

struct ofp_packet_in {

    struct ofp_header header;

    uint32_t buffer_id;     /* ID assigned by datapath. */

    uint16_t total_len;     /* Full length of frame. */

    uint16_t in_port;       /* Port on which frame was received. */

    uint8_t reason;         /* Reason packet is being sent (one of OFPR_*) */

    uint8_t pad;

    uint8_t data[0];        /* Ethernet frame, halfway through 32-bit word,

                               so the IP header is 32-bit aligned.  The

                               amount of data is inferred from the length

                               field in the header.  Because of padding,

                               offsetof(struct ofp_packet_in, data) ==

                               sizeof(struct ofp_packet_in) - 2. */

};

7. PACKET_OUT

struct ofp_packet_out {

    struct ofp_header header;

    uint32_t buffer_id;           /* ID assigned by datapath (-1 if none). */

    uint16_t in_port;             /* Packet's input port (OFPP_NONE if none). */

    uint16_t actions_len;         /* Size of action array in bytes. */

    struct ofp_action_header actions[0]; /* Actions. */

    /* uint8_t data[0]; */        /* Packet data.  The length is inferred

                                     from the length field in the header.

                                     (Only meaningful if buffer_id == -1.) */

};

8. FLOW_MOD

struct ofp_flow_mod {

    struct ofp_header header;

    struct ofp_match match;      /* Fields to match */

    uint64_t cookie;             /* Opaque controller-issued identifier. */

    /* Flow actions. */

    uint16_t command;             /* One of OFPFC_*. */

    uint16_t idle_timeout;        /* Idle time before discarding (seconds). */

    uint16_t hard_timeout;        /* Max time before discarding (seconds). */

    uint16_t priority;            /* Priority level of flow entry. */

    uint32_t buffer_id;           /* Buffered packet to apply to (or -1).

                                     Not meaningful for OFPFC_DELETE*. */

    uint16_t out_port;            /* For OFPFC_DELETE* commands, require

                                     matching entries to include this as an

                                     output port.  A value of OFPP_NONE

                                     indicates no restriction. */

    uint16_t flags;               /* One of OFPFF_*. */

    struct ofp_action_header actions[0]; /* The action length is inferred

                                            from the length field in the

                                            header. */

};

struct ofp_action_header {

    uint16_t type;                  /* One of OFPAT_*. */

    uint16_t len;                   /* Length of action, including this

                                       header.  This is the length of action,

                                       including any padding to make it

                                       64-bit aligned. */

    uint8_t pad[4];

};

（一）个人心得