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