学习Linux-4.12内核网路协议栈（1.6）——协议栈的初始化(inet

xiaoxiao2021-02-27 240

这篇文章主要分析inet_init（）函数的实现过程：

1796 static int __init inet_init(void) 1797 { 1798 struct inet_protosw *q; 1799 struct list_head *r; 1800 int rc = -EINVAL; 1801 1802 sock_skb_cb_check_size(sizeof(struct inet_skb_parm)); 1803 1804 rc = proto_register(&tcp_prot, 1); 1805 if (rc) 1806 goto out; 1807 1808 rc = proto_register(&udp_prot, 1); 1809 if (rc) 1810 goto out_unregister_tcp_proto; 1811 1812 rc = proto_register(&raw_prot, 1); 1813 if (rc) 1814 goto out_unregister_udp_proto; 1815 1816 rc = proto_register(&ping_prot, 1); 1817 if (rc) 1818 goto out_unregister_raw_proto; 1820 /* 1821 * Tell SOCKET that we are alive... 1822 */ 1823 1824 (void)sock_register(&inet_family_ops); 1825 1826 #ifdef CONFIG_SYSCTL 1827 ip_static_sysctl_init(); 1828 #endif 1829 1830 /* 1831 * Add all the base protocols. 1832 */ 1833 1834 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0) 1835 pr_crit("%s: Cannot add ICMP protocol\n", __func__); 1836 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0) 1837 pr_crit("%s: Cannot add UDP protocol\n", __func__); 1838 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0) 1839 pr_crit("%s: Cannot add TCP protocol\n", __func__); 1840 #ifdef CONFIG_IP_MULTICAST 1841 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0) 1842 pr_crit("%s: Cannot add IGMP protocol\n", __func__); 1843 #endif 1844 1845 /* Register the socket-side information for inet_create. */ 1846 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r) 1847 INIT_LIST_HEAD(r); 1848 1849 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q) 1850 inet_register_protosw(q); 1852 /* 1853 * Set the ARP module up 1854 */ 1855 1856 arp_init(); 1857 1858 /* 1859 * Set the IP module up 1860 */ 1861 1862 ip_init(); 1863 1864 /* Setup TCP slab cache for open requests. */ 1865 tcp_init(); 1866 1867 /* Setup UDP memory threshold */ 1868 udp_init(); 1869 1870 /* Add UDP-Lite (RFC 3828) */ 1871 udplite4_register(); 1872 1873 ping_init(); 1874 1875 /* 1876 * Set the ICMP layer up 1877 */ 1878 1879 if (icmp_init() < 0) 1880 panic("Failed to create the ICMP control socket.\n"); 1881 1882 /* 1883 * Initialise the multicast router 1884 */ 1885 #if defined(CONFIG_IP_MROUTE) 1886 if (ip_mr_init()) 1887 pr_crit("%s: Cannot init ipv4 mroute\n", __func__); 1888 #endif 1889 1890 if (init_inet_pernet_ops()) 1891 pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__); 1892 /* 1893 * Initialise per-cpu ipv4 mibs 1894 */ 1895 1896 if (init_ipv4_mibs()) 1897 pr_crit("%s: Cannot init ipv4 mibs\n", __func__); 1898 1899 ipv4_proc_init(); 1900 1901 ipfrag_init(); 1902 1903 dev_add_pack(&ip_packet_type); 1904 1905 ip_tunnel_core_init(); 1906 1907 rc = 0; 1908 out: 1909 return rc; 1910 out_unregister_raw_proto: 1911 proto_unregister(&raw_prot); 1912 out_unregister_udp_proto: 1913 proto_unregister(&udp_prot); 1914 out_unregister_tcp_proto: 1915 proto_unregister(&tcp_prot); 1916 goto out; 1917 } 1918 1919 fs_initcall(inet_init);

其他函数在前面的都介绍过了，这篇文章我们主要介绍下面几个函数：

1856 arp_init(); 1857 1858 /* 1859 * Set the IP module up 1860 */ 1861 1862 ip_init(); 1863 1864 /* Setup TCP slab cache for open requests. */ 1865 tcp_init(); 1866 1867 /* Setup UDP memory threshold */ 1868 udp_init(); 1869 1870 /* Add UDP-Lite (RFC 3828) */ 1871 udplite4_register(); 1872 1873 ping_init(); 1874 1875 /* 1876 * Set the ICMP layer up 1877 */ 1878 1879 if (icmp_init() < 0) 1880 panic("Failed to create the ICMP control socket.\n"); 1881 1882 /* 1883 * Initialise the multicast router 1884 */ 1885 #if defined(CONFIG_IP_MROUTE) 1886 if (ip_mr_init()) 1887 pr_crit("%s: Cannot init ipv4 mroute\n", __func__); 1888 #endif 1889 1890 if (init_inet_pernet_ops()) 1891 pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__); 1892 /* 1893 * Initialise per-cpu ipv4 mibs 1894 */ 1895 1896 if (init_ipv4_mibs()) 1897 pr_crit("%s: Cannot init ipv4 mibs\n", __func__); 1898 1899 ipv4_proc_init(); 1900 1901 ipfrag_init(); 1902 1903 dev_add_pack(&ip_packet_type); 1904 1905 ip_tunnel_core_init();

1. arp_init()

1281 void __init arp_init(void) 1282 { 1283 neigh_table_init(NEIGH_ARP_TABLE, &arp_tbl); 1284 1285 dev_add_pack(&arp_packet_type); 1286 arp_proc_init(); 1287 #ifdef CONFIG_SYSCTL 1288 neigh_sysctl_register(NULL, &arp_tbl.parms, NULL); 1289 #endif 1290 register_netdevice_notifier(&arp_netdev_notifier); 1291 } 这个函数比较简单，就是将arp的协议类型添加到ptype_base[]里面，当IP层收到arp包的时候调用arp_rcv()函数进行处理。

同时也完成/proc/net/arp文件的初始化。

2. ip_init()

//ip子系统初始化 //调用路径inet_init->ip_init //包括路由子系统的初始化，inet_peer缓存的初始化 void __init ip_init(void) { ip_rt_init(); //路由子系统初始化 inet_initpeers();//inet_peer缓存 #if defined(CONFIG_IP_MULTICAST) igmp_mc_init(); #endif } //调用路径ip_init->inet_initpeers //inet_peer缓存初始化 // 1.每一个与linux通过inet协议交换过数据的主机，都会被认为是一个ip peer，linux为每个peer分配一个inet_peer结构。 // 2.inet_peer的主要目的，使不同ip使用不同的ip id生成器。 // 3.系统中所有的inet_peer实例组织成一课avl树，方便查找。 1.2 void __init inet_initpeers(void) { struct sysinfo si; //获取系统内存信息 si_meminfo(&si); //系统中inet_peer数量的阀值 if (si.totalram <= (32768*1024)/PAGE_SIZE) inet_peer_threshold >>= 1; /* max pool size about 1MB on IA32 */ if (si.totalram <= (16384*1024)/PAGE_SIZE) inet_peer_threshold >>= 1; /* about 512KB */ if (si.totalram <= (8192*1024)/PAGE_SIZE) inet_peer_threshold >>= 2; /* about 128KB */ //inet_peer SLAB cache peer_cachep = kmem_cache_create("inet_peer_cache", sizeof(struct inet_peer), 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); //垃圾回收 peer_periodic_timer.expires = jiffies + net_random() % inet_peer_gc_maxtime + inet_peer_gc_maxtime; add_timer(&peer_periodic_timer); }

3. tcp_init()

tcp的初始化内容比较多，后面单独讲

4. udp_init（）

2653 void __init udp_init(void) 2654 { 2655 unsigned long limit; 2656 unsigned int i; 2657 2658 udp_table_init(&udp_table, "UDP"); 2659 limit = nr_free_buffer_pages() / 8; 2660 limit = max(limit, 128UL); 2661 sysctl_udp_mem[0] = limit / 4 * 3; 2662 sysctl_udp_mem[1] = limit; 2663 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; 2664 2665 sysctl_udp_rmem_min = SK_MEM_QUANTUM; 2666 sysctl_udp_wmem_min = SK_MEM_QUANTUM; 2667 2668 /* 16 spinlocks per cpu */ 2669 udp_busylocks_log = ilog2(nr_cpu_ids) + 4; 2670 udp_busylocks = kmalloc(sizeof(spinlock_t) << udp_busylocks_log, 2671 GFP_KERNEL); 2672 if (!udp_busylocks) 2673 panic("UDP: failed to alloc udp_busylocks\n"); 2674 for (i = 0; i < (1U << udp_busylocks_log); i++) 2675 spin_lock_init(udp_busylocks + i); 2676 } 这里主要完成udp_table的初始化：

67 /** 68 * struct udp_table - UDP table 69 * 70 * @hash: hash table, sockets are hashed on (local port) 71 * @hash2: hash table, sockets are hashed on (local port, local address) 72 * @mask: number of slots in hash tables, minus 1 73 * @log: log2(number of slots in hash table) 74 */ 75 struct udp_table { 76 struct udp_hslot *hash; 77 struct udp_hslot *hash2; 78 unsigned int mask; 79 unsigned int log; 80 }; 这个函数主要创建了两个hash表，用于串联本地套接字对应的端口号

5. icmp_init（）

1252 int __init icmp_init(void) 1253 { 1254 return register_pernet_subsys(&icmp_sk_ops); 1255 } 该函数的主要作用是将一个网络协议模块添加到每一个网络命令空间中，然后再执行其ops->init程序进行初始化，一般其ops->init会在其对应的proc目录下，生成一个网络协议模块对应的proc文件或proc目录，并执行一些协议初始化相关的函数。

1247 static struct pernet_operations __net_initdata icmp_sk_ops = { 1248 .init = icmp_sk_init, 1249 .exit = icmp_sk_exit, 1250 }; 1185 static int __net_init icmp_sk_init(struct net *net) 1186 { 1187 int i, err; 1188 1189 net->ipv4.icmp_sk = alloc_percpu(struct sock *); 1190 if (!net->ipv4.icmp_sk) 1191 return -ENOMEM; 1192 1193 for_each_possible_cpu(i) { 1194 struct sock *sk; 1195 1196 err = inet_ctl_sock_create(&sk, PF_INET, 1197 SOCK_RAW, IPPROTO_ICMP, net); 1198 if (err < 0) 1199 goto fail; 1200 1201 *per_cpu_ptr(net->ipv4.icmp_sk, i) = sk; 1202 1203 /* Enough space for 2 64K ICMP packets, including 1204 * sk_buff/skb_shared_info struct overhead. 1205 */ 1206 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); 1207 1208 /* 1209 * Speedup sock_wfree() 1210 */ 1211 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); 1212 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT; 1213 } 1214 1215 /* Control parameters for ECHO replies. */ 1216 net->ipv4.sysctl_icmp_echo_ignore_all = 0; 1217 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1; 1218 1219 /* Control parameter - ignore bogus broadcast responses? */ 1220 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1; 1221 1222 /* 1223 * Configurable global rate limit. 1224 * 1225 * ratelimit defines tokens/packet consumed for dst->rate_token 1226 * bucket ratemask defines which icmp types are ratelimited by 1227 * setting it's bit position. 1228 * 1229 * default: 1230 * dest unreachable (3), source quench (4), 1231 * time exceeded (11), parameter problem (12) 1232 */ 1233 1234 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ; 1235 net->ipv4.sysctl_icmp_ratemask = 0x1818; 1236 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0; 1237 1238 return 0; 1239 1240 fail: 1241 for_each_possible_cpu(i) 1242 inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.icmp_sk, i)); 1243 free_percpu(net->ipv4.icmp_sk); 1244 return err; 1245 } http://blog.csdn.net/lickylin/article/details/24208463

6.init_inet_pernet_ops()

它最后调用inet_init_net(struct net *net)完成以下网路命名空间的初始化，每一个网络设备都会有一个网络命名空间，比如eth1， ath0等，可以查看这里了解。

7. init_ipv4_mibs()

前面说了，每一个网络接口都对应着一个网络命名空间，这些网路空间串联在一个由”struct net”对象组成的结构体中，该函数完成这些空间数据统计的一些信息初始化。

static __net_init int ipv4_mib_init_net(struct net *net) { int i; net->mib.tcp_statistics = alloc_percpu(struct tcp_mib); if (!net->mib.tcp_statistics) goto err_tcp_mib; net->mib.ip_statistics = alloc_percpu(struct ipstats_mib); if (!net->mib.ip_statistics) goto err_ip_mib; for_each_possible_cpu(i) { struct ipstats_mib *af_inet_stats; af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i); u64_stats_init(&af_inet_stats->syncp); } net->mib.net_statistics = alloc_percpu(struct linux_mib); if (!net->mib.net_statistics) goto err_net_mib; net->mib.udp_statistics = alloc_percpu(struct udp_mib); if (!net->mib.udp_statistics) goto err_udp_mib; net->mib.udplite_statistics = alloc_percpu(struct udp_mib); if (!net->mib.udplite_statistics) goto err_udplite_mib; net->mib.icmp_statistics = alloc_percpu(struct icmp_mib); if (!net->mib.icmp_statistics) goto err_icmp_mib; net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib), GFP_KERNEL); if (!net->mib.icmpmsg_statistics) goto err_icmpmsg_mib; tcp_mib_init(net);

8. ipfrag_init（）

IP分片的初始化，其中指定了对应动作所调用的函数

void __init ipfrag_init(void) { ip4_frags_ctl_register(); register_pernet_subsys(&ip4_frags_ops); ip4_frags.hashfn = ip4_hashfn; ip4_frags.constructor = ip4_frag_init; ip4_frags.destructor = ip4_frag_free; ip4_frags.qsize = sizeof(struct ipq); ip4_frags.match = ip4_frag_match; ip4_frags.frag_expire = ip_expire; ip4_frags.frags_cache_name = ip_frag_cache_name; if (inet_frags_init(&ip4_frags)) panic("IP: failed to allocate ip4_frags cache\n"); }

9. ipv4_proc_init()

在proc的ipv4目录下，会根据不同的协议创建不同的文件

static int __init ipv4_proc_init(void) { int rc = 0; if (raw_proc_init()) goto out_raw; if (tcp4_proc_init()) goto out_tcp; if (udp4_proc_init()) goto out_udp; if (ping_proc_init()) goto out_ping; if (ip_misc_proc_init()) goto out_misc;

root:/proc/213/net# cat tcp sl local_address rem_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode 0: 00000000:8240 00000000:0000 0A 00000000:00000000 00:00000000 00000000 0 0 18057 1 d9aa0000 100 0 0 10 0 1: FA01A8C0:C000 00000000:0000 0A 00000000:00000000 00:00000000 00000000 0 0 10027 1 d9828000 100 0 0 10 0 2: 00000000:4E25 00000000:0000 0A 00000000:00000000 00:00000000 00000000 0 0 18624 1 d86b8000 100 0 0 10 0 3: 00000000:0D05 00000000:0000 0A 00000000:00000000 00:00000000 00000000 0 0 3566 1 de690500 100 0 0 10 0 4: 0100007F:1E6B 00000000:0000 0A 00000000:00000000 00:00000000 00000000 0 0 16883 1 d9828500 100 0 0 10 0 5: 00000000:0050 00000000:0000 0A 00000000:00000000 00:00000000 00000000 0 0 3548 1 de690000 100 0 0 10 0 6: 00000000:15B3 00000000:0000 0A 00000000:00000000 00:00000000 00000000 0 0 10551287 1 d86bc600 100 0 0 10 0 7: 00000000:0277 00000000:0000 0A 00000000:00000000 00:00000000 00000000 0 0 18632 1 d86b8500 100 0 0 10 0 8: 0100007F:DAEA 0100007F:0050 06 00000000:00000000 03:00000640 00000000 0 0 0 3 d86c7d80 9: FA01A8C0:A668 0A01A8C0:0B35 06 00000000:00000000 03:000011B2 00000000 0 0 0 3 de52c000 10: FA01A8C0:A664 0A01A8C0:0B35 06 00000000:00000000 03:0000107F 00000000 0 0 0 3 de52ca80 11: FA01A8C0:A669 0A01A8C0:0B35 06 00000000:00000000 03:000011B1 00000000 0 0 0 3 de52c180 12: FA01A8C0:A66A 0A01A8C0:0B35 06 00000000:00000000 03:000011B1 00000000 0 0 0 3 de52c480 13: FA01A8C0:A670 0A01A8C0:0B35 06 00000000:00000000 03:000014B9 00000000 0 0 0 3 d9fb3cc0 14: FA01A8C0:A64B 0A01A8C0:0B35 06 00000000:00000000 03:00000216 00000000 0 0 0 3 d9fb3e40

也许你会问路径里面的213是什么，应该是不同的name space 吧

转载请注明原文地址: https://www.6miu.com/read-7151.html

技术

最新回复(0)