include/net/net_namespace.h

   1 /*
   2  * Operations on the network namespace
   3  */
   4 #ifndef __NET_NET_NAMESPACE_H
   5 #define __NET_NET_NAMESPACE_H
   6
   7 #include <linux/atomic.h>
   8 #include <linux/workqueue.h>
   9 #include <linux/list.h>
  10 #include <linux/sysctl.h>
  11
  12 #include <net/netns/core.h>
  13 #include <net/netns/mib.h>
  14 #include <net/netns/unix.h>
  15 #include <net/netns/packet.h>
  16 #include <net/netns/ipv4.h>
  17 #include <net/netns/ipv6.h>
  18 #include <net/netns/ieee802154_6lowpan.h>
  19 #include <net/netns/sctp.h>
  20 #include <net/netns/dccp.h>
  21 #include <net/netns/netfilter.h>
  22 #include <net/netns/x_tables.h>
  23 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  24 #include <net/netns/conntrack.h>
  25 #endif
  26 #include <net/netns/nftables.h>
  27 #include <net/netns/xfrm.h>
  28
  29 struct user_namespace;
  30 struct proc_dir_entry;
  31 struct net_device;
  32 struct sock;
  33 struct ctl_table_header;
  34 struct net_generic;
  35 struct sock;
  36 struct netns_ipvs;
  37
  38
  39 #define NETDEV_HASHBITS    8
  40 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
  41
  42 struct net {
  43         atomic_t                passive;        /* To decided when the network
  44                                                  * namespace should be freed.
  45                                                  */
  46         atomic_t                count;          /* To decided when the network
  47                                                  *  namespace should be shut down.
  48                                                  */
  49 #ifdef NETNS_REFCNT_DEBUG
  50         atomic_t                use_count;      /* To track references we
  51                                                  * destroy on demand
  52                                                  */
  53 #endif
  54         spinlock_t              rules_mod_lock;
  55
  56         struct list_head        list;           /* list of network namespaces */
  57         struct list_head        cleanup_list;   /* namespaces on death row */
  58         struct list_head        exit_list;      /* Use only net_mutex */
  59
  60         struct user_namespace   *user_ns;       /* Owning user namespace */
  61
  62         unsigned int            proc_inum;
  63
  64         struct proc_dir_entry   *proc_net;
  65         struct proc_dir_entry   *proc_net_stat;
  66
  67 #ifdef CONFIG_SYSCTL
  68         struct ctl_table_set    sysctls;
  69 #endif
  70
  71         struct sock             *rtnl;                  /* rtnetlink socket */
  72         struct sock             *genl_sock;
  73
  74         struct list_head        dev_base_head;
  75         struct hlist_head       *dev_name_head;
  76         struct hlist_head       *dev_index_head;
  77         unsigned int            dev_base_seq;   /* protected by rtnl_mutex */
  78         int                     ifindex;
  79         unsigned int            dev_unreg_count;
  80
  81         /* core fib_rules */
  82         struct list_head        rules_ops;
  83
  84
  85         struct net_device       *loopback_dev;          /* The loopback */
  86         struct netns_core       core;
  87         struct netns_mib        mib;
  88         struct netns_packet     packet;
  89         struct netns_unix       unx;
  90         struct netns_ipv4       ipv4;
  91 #if IS_ENABLED(CONFIG_IPV6)
  92         struct netns_ipv6       ipv6;
  93 #endif
  94 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
  95         struct netns_ieee802154_lowpan  ieee802154_lowpan;
  96 #endif
  97 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
  98         struct netns_sctp       sctp;
  99 #endif
 100 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
 101         struct netns_dccp       dccp;
 102 #endif
 103 #ifdef CONFIG_NETFILTER
 104         struct netns_nf         nf;
 105         struct netns_xt         xt;
 106 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 107         struct netns_ct         ct;
 108 #endif
 109 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
 110         struct netns_nftables   nft;
 111 #endif
 112 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
 113         struct netns_nf_frag    nf_frag;
 114 #endif
 115         struct sock             *nfnl;
 116         struct sock             *nfnl_stash;
 117 #endif
 118 #ifdef CONFIG_WEXT_CORE
 119         struct sk_buff_head     wext_nlevents;
 120 #endif
 121         struct net_generic __rcu        *gen;
 122
 123         /* Note : following structs are cache line aligned */
 124 #ifdef CONFIG_XFRM
 125         struct netns_xfrm       xfrm;
 126 #endif
 127 #if IS_ENABLED(CONFIG_IP_VS)
 128         struct netns_ipvs       *ipvs;
 129 #endif
 130         struct sock             *diag_nlsk;
 131         atomic_t                fnhe_genid;
 132 };
 133
 134 /*
 135  * ifindex generation is per-net namespace, and loopback is
 136  * always the 1st device in ns (see net_dev_init), thus any
 137  * loopback device should get ifindex 1
 138  */
 139
 140 #define LOOPBACK_IFINDEX        1
 141
 142 #include <linux/seq_file_net.h>
 143
 144 /* Init's network namespace */
 145 extern struct net init_net;
 146
 147 #ifdef CONFIG_NET_NS
 148 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
 149                         struct net *old_net);
 150
 151 #else /* CONFIG_NET_NS */
 152 #include <linux/sched.h>
 153 #include <linux/nsproxy.h>
 154 static inline struct net *copy_net_ns(unsigned long flags,
 155         struct user_namespace *user_ns, struct net *old_net)
 156 {
 157         if (flags & CLONE_NEWNET)
 158                 return ERR_PTR(-EINVAL);
 159         return old_net;
 160 }
 161 #endif /* CONFIG_NET_NS */
 162
 163
 164 extern struct list_head net_namespace_list;
 165
 166 struct net *get_net_ns_by_pid(pid_t pid);
 167 struct net *get_net_ns_by_fd(int pid);
 168
 169 #ifdef CONFIG_SYSCTL
 170 void ipx_register_sysctl(void);
 171 void ipx_unregister_sysctl(void);
 172 #else
 173 #define ipx_register_sysctl()
 174 #define ipx_unregister_sysctl()
 175 #endif
 176
 177 #ifdef CONFIG_NET_NS
 178 void __put_net(struct net *net);
 179
 180 static inline struct net *get_net(struct net *net)
 181 {
 182         atomic_inc(&net->count);
 183         return net;
 184 }
 185
 186 static inline struct net *maybe_get_net(struct net *net)
 187 {
 188         /* Used when we know struct net exists but we
 189          * aren't guaranteed a previous reference count
 190          * exists.  If the reference count is zero this
 191          * function fails and returns NULL.
 192          */
 193         if (!atomic_inc_not_zero(&net->count))
 194                 net = NULL;
 195         return net;
 196 }
 197
 198 static inline void put_net(struct net *net)
 199 {
 200         if (atomic_dec_and_test(&net->count))
 201                 __put_net(net);
 202 }
 203
 204 static inline
 205 int net_eq(const struct net *net1, const struct net *net2)
 206 {
 207         return net1 == net2;
 208 }
 209
 210 void net_drop_ns(void *);
 211
 212 #else
 213
 214 static inline struct net *get_net(struct net *net)
 215 {
 216         return net;
 217 }
 218
 219 static inline void put_net(struct net *net)
 220 {
 221 }
 222
 223 static inline struct net *maybe_get_net(struct net *net)
 224 {
 225         return net;
 226 }
 227
 228 static inline
 229 int net_eq(const struct net *net1, const struct net *net2)
 230 {
 231         return 1;
 232 }
 233
 234 #define net_drop_ns NULL
 235 #endif
 236
 237
 238 #ifdef NETNS_REFCNT_DEBUG
 239 static inline struct net *hold_net(struct net *net)
 240 {
 241         if (net)
 242                 atomic_inc(&net->use_count);
 243         return net;
 244 }
 245
 246 static inline void release_net(struct net *net)
 247 {
 248         if (net)
 249                 atomic_dec(&net->use_count);
 250 }
 251 #else
 252 static inline struct net *hold_net(struct net *net)
 253 {
 254         return net;
 255 }
 256
 257 static inline void release_net(struct net *net)
 258 {
 259 }
 260 #endif
 261
 262 #ifdef CONFIG_NET_NS
 263
 264 static inline void write_pnet(struct net **pnet, struct net *net)
 265 {
 266         *pnet = net;
 267 }
 268
 269 static inline struct net *read_pnet(struct net * const *pnet)
 270 {
 271         return *pnet;
 272 }
 273
 274 #else
 275
 276 #define write_pnet(pnet, net)   do { (void)(net);} while (0)
 277 #define read_pnet(pnet)         (&init_net)
 278
 279 #endif
 280
 281 #define for_each_net(VAR)                               \
 282         list_for_each_entry(VAR, &net_namespace_list, list)
 283
 284 #define for_each_net_rcu(VAR)                           \
 285         list_for_each_entry_rcu(VAR, &net_namespace_list, list)
 286
 287 #ifdef CONFIG_NET_NS
 288 #define __net_init
 289 #define __net_exit
 290 #define __net_initdata
 291 #define __net_initconst
 292 #else
 293 #define __net_init      __init
 294 #define __net_exit      __exit_refok
 295 #define __net_initdata  __initdata
 296 #define __net_initconst __initconst
 297 #endif
 298
 299 struct pernet_operations {
 300         struct list_head list;
 301         int (*init)(struct net *net);
 302         void (*exit)(struct net *net);
 303         void (*exit_batch)(struct list_head *net_exit_list);
 304         int *id;
 305         size_t size;
 306 };
 307
 308 /*
 309  * Use these carefully.  If you implement a network device and it
 310  * needs per network namespace operations use device pernet operations,
 311  * otherwise use pernet subsys operations.
 312  *
 313  * Network interfaces need to be removed from a dying netns _before_
 314  * subsys notifiers can be called, as most of the network code cleanup
 315  * (which is done from subsys notifiers) runs with the assumption that
 316  * dev_remove_pack has been called so no new packets will arrive during
 317  * and after the cleanup functions have been called.  dev_remove_pack
 318  * is not per namespace so instead the guarantee of no more packets
 319  * arriving in a network namespace is provided by ensuring that all
 320  * network devices and all sockets have left the network namespace
 321  * before the cleanup methods are called.
 322  *
 323  * For the longest time the ipv4 icmp code was registered as a pernet
 324  * device which caused kernel oops, and panics during network
 325  * namespace cleanup.   So please don't get this wrong.
 326  */
 327 int register_pernet_subsys(struct pernet_operations *);
 328 void unregister_pernet_subsys(struct pernet_operations *);
 329 int register_pernet_device(struct pernet_operations *);
 330 void unregister_pernet_device(struct pernet_operations *);
 331
 332 struct ctl_table;
 333 struct ctl_table_header;
 334
 335 #ifdef CONFIG_SYSCTL
 336 int net_sysctl_init(void);
 337 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
 338                                              struct ctl_table *table);
 339 void unregister_net_sysctl_table(struct ctl_table_header *header);
 340 #else
 341 static inline int net_sysctl_init(void) { return 0; }
 342 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
 343         const char *path, struct ctl_table *table)
 344 {
 345         return NULL;
 346 }
 347 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
 348 {
 349 }
 350 #endif
 351
 352 static inline int rt_genid_ipv4(struct net *net)
 353 {
 354         return atomic_read(&net->ipv4.rt_genid);
 355 }
 356
 357 static inline void rt_genid_bump_ipv4(struct net *net)
 358 {
 359         atomic_inc(&net->ipv4.rt_genid);
 360 }
 361
 362 #if IS_ENABLED(CONFIG_IPV6)
 363 static inline int rt_genid_ipv6(struct net *net)
 364 {
 365         return atomic_read(&net->ipv6.rt_genid);
 366 }
 367
 368 static inline void rt_genid_bump_ipv6(struct net *net)
 369 {
 370         atomic_inc(&net->ipv6.rt_genid);
 371 }
 372 #else
 373 static inline int rt_genid_ipv6(struct net *net)
 374 {
 375         return 0;
 376 }
 377
 378 static inline void rt_genid_bump_ipv6(struct net *net)
 379 {
 380 }
 381 #endif
 382
 383 /* For callers who don't really care about whether it's IPv4 or IPv6 */
 384 static inline void rt_genid_bump_all(struct net *net)
 385 {
 386         rt_genid_bump_ipv4(net);
 387         rt_genid_bump_ipv6(net);
 388 }
 389
 390 static inline int fnhe_genid(struct net *net)
 391 {
 392         return atomic_read(&net->fnhe_genid);
 393 }
 394
 395 static inline void fnhe_genid_bump(struct net *net)
 396 {
 397         atomic_inc(&net->fnhe_genid);
 398 }
 399
 400 #endif /* __NET_NET_NAMESPACE_H */