Linux內核--網絡協議棧深入分析(五)--套接字的綁定、監聽、連接和斷開

本文分析基於Linux Kernel 3.2.1

更多請查看 Linux內核--網絡內核實現分析

1、套接字的綁定

創建完套接字服務器端會在應用層使用bind函數驚醒套接字的綁定，這時會產生系統調用，sys_bind內核函數進行套接字。

系統調用函數的具體實現

1. SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
2. {
3. struct socket *sock;
4. struct sockaddr_storage address;
5. int err, fput_needed;
7. sock = sockfd_lookup_light(fd, &err, &fput_needed);
8. if (sock) {
9. err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
10. if (err >= 0) {
11. err = security_socket_bind(sock,
12. (struct sockaddr *)&address,
13. addrlen);
14. if (!err)
15. err = sock->ops->bind(sock,
16. (struct sockaddr *)
17. &address, addrlen);
18. }
19. fput_light(sock->file, fput_needed);
20. }
21. return err;
22. }

首先調用函數sockfd_lookup_light()函數通過文件描述符來查找對應的套接字sock。

1. static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
2. {
3. struct file *file;
4. struct socket *sock;
6. *err = -EBADF;
7. file = fget_light(fd, fput_needed);
8. if (file) {
9. sock = sock_from_file(file, err);
10. if (sock)
11. return sock;
12. fput_light(file, *fput_needed);
13. }
14. return NULL;
15. }

上面函數中先調用fget_light函數通過文件描述符返回對應的文件結構，然後調用函數sock_from_file函數返回該文件對應的套接字結構體地址，它存儲在file->private_data屬性中。

再回到sys_bind函數，在返回了對應的套接字結構之後，調用move_addr_to_kernel將用戶地址空間的socket拷貝到內核空間。

然後調用INET協議族的操作集中bind函數inet_bind函數將socket地址（內核空間）和socket綁定。

1. int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
2. {
3. struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
4. struct sock *sk = sock->sk;
5. struct inet_sock *inet = inet_sk(sk);
6. unsigned short snum;
7. int chk_addr_ret;
8. int err;
10. //RAW類型套接字若有自己的bind函數，則使用之
11. if (sk->sk_prot->bind) {
12. err = sk->sk_prot->bind(sk, uaddr, addr_len);
13. goto out;
14. }
15. err = -EINVAL;
16. .....................
17. //地址合法性檢查
18. chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
20. /* Not specified by any standard per-se, however it breaks too
21. * many applications when removed. It is unfortunate since
22. * allowing applications to make a non-local bind solves
23. * several problems with systems using dynamic addressing.
24. * (ie. your servers still start up even if your ISDN link
25. * is temporarily down)
26. */
27. err = -EADDRNOTAVAIL;
28. if (!sysctl_ip_nonlocal_bind &&
29. !(inet->freebind || inet->transparent) &&
30. addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
31. chk_addr_ret != RTN_LOCAL &&
32. chk_addr_ret != RTN_MULTICAST &&
33. chk_addr_ret != RTN_BROADCAST)
34. goto out;
36. snum = ntohs(addr->sin_port);
37. err = -EACCES;
38. if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
39. goto out;
41. /* We keep a pair of addresses. rcv_saddr is the one
42. * used by hash lookups, and saddr is used for transmit.
43. *
44. * In the BSD API these are the same except where it
45. * would be illegal to use them (multicast/broadcast) in
46. * which case the sending device address is used.
47. */
48. lock_sock(sk);
50. /* Check these errors (active socket, double bind). */
51. err = -EINVAL;
52. if (sk->sk_state != TCP_CLOSE || inet->inet_num)//如果sk的狀態是CLOSE或者本地端口已經被綁定
53. goto out_release_sock;
55. inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;//設置源地址
56. if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
57. inet->inet_saddr = 0; /* Use device */
59. /* Make sure we are allowed to bind here. */
60. if (sk->sk_prot->get_port(sk, snum)) {
61. inet->inet_saddr = inet->inet_rcv_saddr = 0;
62. err = -EADDRINUSE;
63. goto out_release_sock;
64. }
66. if (inet->inet_rcv_saddr)
67. sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
68. if (snum)
69. sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
70. inet->inet_sport = htons(inet->inet_num);//設置源端口號，標明該端口已經被占用
71. inet->inet_daddr = 0;
72. inet->inet_dport = 0;
73. sk_dst_reset(sk);
74. err = 0;
75. out_release_sock:
76. release_sock(sk);
77. out:
78. return err;
79. }

這樣套接字綁定結束。