Network IPC:Socket

• Introduction
• Socket Descriptors
• Addressing
• Connection Establishment
• Data Transfer
• Socket Option

• classical IPC’s 讓跑在同一台電腦上的 processes 可以互相通訊
• network IPC 則可讓跑在不同電腦的 processes 互相通訊
• network IPC 可用在 inter-machine communication 和 intra-machine communication
• 本章節主要討論 TCP/IP Sockets

socket descriptor

• A socket is an abstraction of a communication endpoint
• socket descriptor 在 UNIX 系統中是以 file descriptor 方式實作的
• 對於 file descriptor 操作的那些 function 如 read、write 也同樣能對 socket descriptor 操作

#include <sys/types.h>
#include <sys/socket.h>
 
int socket(int domain, int type, int protocol);
/*return file (socket) descriptor if success, or -1 on error*/

• 建立一個 socket
• domain:
  • AF_INET: IPv4 Internet protocols
  • AF_INET6: IPv6 Internet protocols
  • AF_UNIX or AF_LOCAL: Local communication
• type:
  • SOCK_STREAM: stream TCP
  • SOCK_DGRAM: datagram UDP
• protocol: 0(default), IPPROTO_TCP, IPPROTO_UDP

Socket Descriptors and File I/O Functions

• close
• dup,dup2
• fcntl
• read
• write

Release a socket descriptor

#include <sys/socket.h>
 
int shutdown(int socket, int how);
/*returns 0 if success, or -1 on error*/

• how:
  • SHUT_RD
  • SHUT_WR
  • SHUT_RDWR
• socket communication 是雙向的，相較於 close，利用 shutdown 可以僅關閉單一方向的通訊
• close 會關閉呼叫它的 process 的 file descriptor 參照，file descriptor 參照數量 -1，但其他 process 仍可對 file descriptor 繼續操作，直到 file descriptor 參照數量為 0 時 file descriptor 才會被關閉
• 如果多個 processes 參照同個 file descriptor，而其中一個 process 呼叫 shutdown，則 file descriptor 會立即被關閉，其他 processes 也無法繼續操作 file descriptor

identify socket

• 使用 addressing schemes 來識別 socket
  • AF_UNIX : a pathname
  • AF_INET + SOCK_STREAM + IP_PROTO_TCP : IPv4 address and TCP port number
  • AF_INET + SOCKET_DGRAM + IP_PROTO_UDP : IPv4 address and UDP port number
  • AF_INET6 + SOCK_STREAM + IP_PROTO_TCP : IPv6 address and TCP port number
  • AF_INET6 + SOCKET_DGRAM + IP_PROTO_UDP : IPv6 address and UDP port number

byte ordering

• 電腦世界中有兩種 byte order
  • Big-Endian(正序)
  • Little-Endian(倒序)
• Network Byte Order 是 Big-Endian
• Host Byte Order 則由電腦架構而定，x86 為 Little-Endian
• 我們在寫網路程式時，需要確認 port number 跟 IP address 都是 Network Byte Order，而 C 語言提供了一些函式，讓我們直接假設 Host Byte Order 皆不是 Network Byte Order，而可使用這些函式進行轉換

#include <arpa/inet.h>
 
uint32_t htonl(uint32_t hostlong);
uint16_t htons(uint16_t hostshort);
uint32_t ntohl(uint32_t netlong);
uint16_t ntohs(uint16_t netshort);

• 這些函式可使我們將 unsigned integer,unsigned short integer 在 Host Byte Order 和 Network Byte Order 間轉換
  • h means host
  • n means network
  • l means long
  • s means short

address formats

• Generic:

/*On Linux*/
struct sockaddr
{
    sa_family_t sa_family;    /* address family */
    char sa_data[14];         /* variable-length address */
};
 
/*On some other system*/
struct sockaddr
{
    unsigned char sa_len;     /* total length */
    sa_family_t sa_family;    /* address family */
    char sa_data[14];         /* variable-length address */
};

• IPv4 on Linux

typedef uint16_t in_port_t;
typedef uint32_t in_addr_t;
 
struct in_addr
{
    in_addr_t s_addr;           /* IPv4 address */
};
 
struct sockaddr_in
{
    sa_family_t sin_family;     /* address family */
    in_port_t sin_port;         /* port number */
    struct in_addr sin_addr;    /* IPv4 address */
    unsigned char sin_zero[8];
};

• IPv6 on Linux

typedef uint16_t in_port_t;
struct in6_addr
{
    union
    {                                /* IPv4 address */
        uint8_t __u6_addr8[16];
        uint16_t __u6_addr16[8];
        uint32_t __u6_addr32[4];
    } __in6_u;
};
 
struct sockaddr_in6
{
    sa_family_t sin_family;                /* address family */
    in_port_t sin6_port;                   /* port number */
    uint32_t sin6_flowinfo;                /* IPv6 flow info */
    struct in6_addr sin6_addr;             /* IPv6 address */
    uint32_t sin6_scope_id;                /* IPv6 scope id */
};

conversion of address

• 為了將 address 已可讀的形式印出來，我們須將 numeric address 轉換為 text

#include <arpa/inet.h>
 
const char *inet_ntop(int af, const void *src, char *dst, socklen_t size);
/*Returns: pointer to address string on success, or NULL on error*/

• 將 text formatted address 轉回 numeric address

#include <arpa/inet.h>
 
int inet_pton(int af, const char *src, void *dst);
/*Returns: 1 on success, 0 if the format is invalid, or -1 on error and errno=EAFNOSUPPORT*/

if(inet_pton(AF_INET, argv[1], &addr4) == 1)
{
    printf("IPv4: 0x%08x\n", htonl(addr4.s_addr));
}
 
if(inet_pton(AF_INET6, argv[1], &addr6) == 1)
{
    printf("IPv6: 0x%08x%08x%08x%08x\n",
        htonl(addr6.__in6_u.__u6_addr32[0]),
        htonl(addr6.__in6_u.__u6_addr32[1]),
        htonl(addr6.__in6_u.__u6_addr32[2]),
        htonl(addr6.__in6_u.__u6_addr32[3]));
}

address lookup

• 本機上我們可以由三項條件做位址查詢
  • known hosts
  • known protocols
  • known services
• 另外可通過 DNS 查詢 hostname

• known hosts
  • 在系統中可以查看 /etc/hosts 檔案

struct hostent
{
    char *h_name;          /*official name of host*/
    char **h_aliases;      /* alias list */
    int h_addrtype;        /* host address type */
    int h_length;          /* length of address */
    char **h_addr_list;    /* list of addresses */
}

/* for AF_INET */
#include <sys/socket.h>
 
struct hostent *gethostent(void);
/*Returns: valid pointer if success, or NULL on error*/
 
void sethostent(int stayopen);
void endhostent(void);

• gethostent() is not thread-safe
• sethostent()，若 database 未開啟則 open 已開啟則 rewind
• endhostent()，關閉 host database
• 這兩個 function 在 DNS lookup 時會有不同意義

int main()
{
    int i;
    char buf[64];
    struct hostent *h;
 
    while((h = gethostent()) != NULL)
    {
        if(h->h_addrtype != AF_INET)
            continue;
        printf(“name=%s, addr={ “, h->h_name);
 
        for(i = 0; h->h_addr_list[i] != NULL; i++)
        {
            printf("%s ", inet_ntop(AF_INET, h->h_addr_list[i], buf, sizeof(buf)));
        }
 
        printf("}\n");
    }
 
    return(0);
}

name=cshome.cs.nctu.edu.tw, addr={ 140.113.235.101 }
name=csduty.cs.nctu.edu.tw, addr={ 140.113.235.102 }

• known protocols
  • 在系統中可以查看 /etc/protocols 檔案

truct protoent
{
    char *p_name;        /*officialprotocolname*/
    char **p_aliases;    /* alias list */
    int p_proto;         /* protocol number */
}

#include <netdb.h>
 
struct protoent *getprotoent(void);
/*Returns: valid pointer if success, or NULL on error*/
 
struct protoent *getprotobyname(const char *name);
/*returns a protoent structure for the entry from the database that matches the protocol name,NULL on error*/
 
struct protoent *getprotobynumber(int proto);
/*returns a protoent structure for the entry from the database that matches the protocol number,NULL on error*/
 
void setprotoent(int stayopen);
void endprotoent(void);

• getprotobyname 和 getprotobynumber 都不是 thread-safe

int main()
{
    int i;
    struct protoent *p;
 
    while((p = getprotoent()) != NULL)
    {
        printf("name=%s (%d), ", p->p_name, p->p_proto);
        printf("alias={ ");
 
        for(i = 0; p->p_aliases[i] != NULL; i++)
        {
            printf("%s ", p->p_aliases[i]); printf("}\n");
        }
    }
 
    return(0);
}

name=ip (0), alias={ IP }
name=icmp (1), alias={ ICMP }
name=igmp (2), alias={ IGMP }
...
name=tcp (6), alias={ TCP }
...
name=udp (17), alias={ UDP }
...

• known services
  • 在系統中可以查看 /etc/services 檔案

struct servent
{
    char *s_name;      /*officialservicename*/
    char **S_aliases;  /* alias list */
    int s_port;        /* port number */
    char *s_proto;     /*protocoltouse*/
}

#include <netdb.h>
 
struct servent *getservent(void);
/*Returns: valid pointer if success, or NULL on error*/
 
struct servent *getservbyname(const char *name, const char *proto);
/*returns a servent structure for the entry from the database that matches the service name using protocol proto, or NULL on error*/
 
struct servent *getservbyport(int port, const char *proto);
/*returns a servent structure for the entry from the database that matches the service port(network byte order) using protocol proto, or NULL on error*/
 
void setservent(int stayopen);
 
void endservent(void);

• getservbyport 的 port 須為 network byte order
• getservent, getservbyname, getservbyport 都不是 thread-safe

int main()
{
    int i;
    struct servent *s;
 
    while((s = getservent()) != NULL)
    {
        printf("name=%s (%s/%d), ", s->s_name, s->s_proto, ntohs(s->s_port));
        printf("alias={ ");
 
        for(i = 0; s->s_aliases[i] != NULL; i++)
        {
            printf("%s ", s->s_aliases[i]); printf("}\n");
        }
    }
    return(0);
}

...
name=telnet (tcp/23), alias={ }
name=ftp (tcp/21), alias={ }
name=pop3 (tcp/110), alias={ pop-3 }
name=www (tcp/80), alias={ http }
...

• hostname via DNS

#include <netdb.h>
extern int h_errno;
 
struct hostent *gethostbyname(const char *name);
/*Returns: valid pointer if success, or NULL on error*/
 
#include <sys/socket.h>
 
struct hostent *gethostbyaddr(const void *addr, socklen_t len, int type);
/*Returns: valid pointer if success, or NULL on error*/

• gethostbyaddr 的參數 type 可為 AF_INET 或 AF_INET6
• 此兩個 function 預設是使用 UDP protocol,可使用 sethostent() 變更，如果 stayopen 為 true，會使用 TCP socket 連線，且 query 成功時，連線會保持，若 stayopen 為 false，則採用預設 UDP datagrams
• 此兩個 function 也是 not thread-safe

thread-safe query

• 前面提了許多 function 都是 not thread-safe 的，接下來介紹幾個 thread-safe 的 function

• query address and port

#include <sys/types.h>
#include <sys/socket.h>
#include <netdb.h>
 
int getaddrinfo(const char *node, const char *service, const struct addrinfo *hints, struct addrinfo **res);
/*Returns: zero if success, or nonzero error code on error*/

• node: 要查詢的節點(name or address)
• service: service 名稱
• hints: 查詢標準
  • flags (see the next slide)
  • address family (AF_INET/AF_INET6)
  • socktype (SOCK_DGRAM/SOCK_STREAM, can be 0)
  • protocol (can be 0)
  • Other fields must be zero
• res: return the queried result

struct addrinfo
{
    int ai_flags;
    int ai_family;
    int ai_socktype;
    int ai_protocol;
    size_t ai_addrlen;
    struct sockaddr *ai_addr;
    char *ai_canonname;
    struct addrinfo *ai_next;
};

• 當 getaddrinfo 有 error 時無法從 perror 或 stderr 取得錯誤訊息，需要呼叫 gai_strerror function 將 error code 轉為可讀的文字訊息

#include <sys/types.h>
#include <sys/socket.h>
#include <netdb.h>
 
const char *gai_strerror(int error);

int main(int argc, char *argv[])
{
    int s;
    struct addrinfo hints, *result, *rp;
 
    if (argc < 3)
    {
        fprintf(stderr, "usage: %s host port\n", argv[0]);
        exit(-1);
    }
 
    bzero(&hints, sizeof(struct addrinfo));
    hints.ai_family = AF_INET;                 /* allow IPv4 or IPv6 */
    hints.ai_socktype = SOCK_STREAM;           /* stream socket */
    hints.ai_flags = 0;
    hints.ai_protocol = 0;                     /* any protocol */
 
    if((s=getaddrinfo(argv[1], argv[2], &hints, &result))!=0)
    {
        fprintf(stderr, "getaddrinfo: %s\n", gai_strerror(s));
        exit(-1);
    }
 
    for(rp = result; rp != NULL; rp = rp->ai_next)
    {
        struct sockaddr_in *p = (struct sockaddr_in*) rp->ai_addr;
        printf( "%s:%d\n", inet_ntoa(p->sin_addr), ntohs(p->sin_port));
    }
 
    return(0);
}

$ ./getaddrinfo google.com www
74.125.45.100:80
209.85.171.100:8
74.125.67.100:80

• query of name and service

int getnameinfo(const struct sockaddr *sa, socklen_t salen, char *host, size_t hostlen, char *serv, size_t servlen, int flags);
/*Returns:zero if success,or nonzero on error*/

int main(int argc, char *argv[])
{
    struct sockaddr_in sin;
    char host[64], serv[64];
    int s;
 
    if (argc < 3)
    {
        fprintf(stderr, "usage: %s ip port\n", argv[0]);
        exit(-1);
    }
 
    bzero(&sin, sizeof(sin));
    sin.sin_family = AF_INET;
    sin.sin_addr.s_addr = inet_addr(argv[1]);
    sin.sin_port = htons(atoi(argv[2]));
 
    if((s = getnameinfo((struct sockaddr*) &sin, sizeof(sin), host, sizeof(host), serv, sizeof(serv), 0)) != 0)
    {
        fprintf(stderr, "getnameinfo: %s\n", gai_strerror(s));
        exit(-1);
    }
 
    printf("%s:%s\n", host, serv);
    return(0);
}

server connect step

• 建立 socket file descriptor
• 準備一個 sockaddr_in data structure,將 socket address bind 到 socket descriptor
  • port number 不能小於 1024
• listen for incoming connection
• accept incoming connection

client connect step

• 建立 socket file descriptor
• 準備一個 sockaddr_in data structure
• connect to server

socket function

#include <sys/socket.h>
 
int bind(int socket, const struct sockaddr *address, socklen_t address_len);
/*Returns: zero if success, or -1 on error*/

#include <sys/socket.h>
 
int connect(int socket, const struct sockaddr *address, socklen_t address_len);
/*Returns: zero if success, or -1 on error*/

#include <sys/socket.h>
 
int listen(int socket, int backlog);
/*Returns: zero if success, or -1 on error*/

• backlog is the number of outstanding connect requests in a queue
• backlog 在 Linux 上最大是 128 (SOMAXCONN constant)
• queue 一旦滿了，系統會拒絕其他新的連線

#include <sys/socket.h>
 
int accept(int sockfd, struct sockaddr *restrict address, socklen_t *restrict address_len);
/*Returns: file (socket) descriptor if success, or -1 on error*/

• The returned descriptor is the socket connected to the client
• 新的 socket descriptor 和舊的 sockfd 會有相同 socket type 和 address family
• address 保存了 client address 和 port，若我們不需要這些資訊可將此設為 NULL
• 如果沒有等到 connect request，connect 會 block 直到有連線請求到達

#include <sys/socket.h>
 
int getsockname(int socket, struct sockaddr *restrict address, socklen_t *restrict address_len);
int getpeername(int socket, struct sockaddr *restrict address, socklen_t *restrict address_len);

• getsockname get the local address bound to a socket
• getpeername get the remote address bound to a socket
• address 和 address_len 要先宣告，在呼叫這兩個函式前，address_len 要設成 address 的 data structure(sockaddr) 的長度

send data

• 可以使用 write function 對 file descriptor 操作
• 也可使用較彈性的 send/sendto function

#include <sys/socket.h>
 
ssize_t send(int socket, const void *buffer, size_t length, int flags);
ssize_t sendto(int socket, const void *message, size_t length, int flags, const struct sockaddr *dest_addr, socklen_t dest_len);
/*Returns: number of bytes sent if success, or -1 on error*/

• send function 適用於連線導向的 socket
• sendto function 可用於連線或非連線導向的 socket
• 如果 send function 的 flag 設為 0，基本上相等於 write function
• 在非連線導向中，我們要明確指定 dest_addr 參數

receive data

• 可以使用 read function 對 file descriptor 操作
• 也可使用較彈性的 recv/recvfrom function

#include <sys/socket.h>
 
ssize_t recv(int socket, void *buffer, size_t length, int flags);
ssize_t recvfrom(int socket, void *restrict buffer, size_t length, int flags, struct sockaddr *restrict address, socklen_t *restrict address_len);
/*Returns: number of bytes received if success, 0 if no messages are available and peer has done an orderly shutdown, or -1 on error*/

• recv function 適用於連線導向的 socket
• recvfrom function 可用於連線或非連線導向的 socket
• 如果 recv function 的 flag 設為 0，基本上相等於 read function
• 在非連線導向中，recvfrom 的 address 參數會記錄 data sender 的 address 和 port

#include <sys/socket.h>
 
int setsockopt(int socket, int level, int option_name, const void *option_value, socklen_t option_len);
int getsockopt(int socket, int level, int option_name, void *restrict option_value, socklen_t *restrict option_len);
/*Returns: zero if success, or -1 on error*/

• The level argument identify the protocol (by a protocol number) to apply.For example, IPPROTO_IP, IPPROTO_TCP…

• Beej's Guide to Network Programming 正體中文版