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