include/net/net_namespace.h

   1 /* SPDX-License-Identifier: GPL-2.0 */
   2 /*
   3  * Operations on the network namespace
   4  */
   5 #ifndef __NET_NET_NAMESPACE_H
   6 #define __NET_NET_NAMESPACE_H
   7
   8 #include <linux/atomic.h>
   9 #include <linux/refcount.h>
  10 #include <linux/workqueue.h>
  11 #include <linux/list.h>
  12 #include <linux/sysctl.h>
  13 #include <linux/uidgid.h>
  14
  15 #include <net/flow.h>
  16 #include <net/netns/core.h>
  17 #include <net/netns/mib.h>
  18 #include <net/netns/unix.h>
  19 #include <net/netns/packet.h>
  20 #include <net/netns/ipv4.h>
  21 #include <net/netns/ipv6.h>
  22 #include <net/netns/ieee802154_6lowpan.h>
  23 #include <net/netns/sctp.h>
  24 #include <net/netns/dccp.h>
  25 #include <net/netns/netfilter.h>
  26 #include <net/netns/x_tables.h>
  27 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  28 #include <net/netns/conntrack.h>
  29 #endif
  30 #include <net/netns/nftables.h>
  31 #include <net/netns/xfrm.h>
  32 #include <net/netns/mpls.h>
  33 #include <net/netns/can.h>
  34 #include <linux/ns_common.h>
  35 #include <linux/idr.h>
  36 #include <linux/skbuff.h>
  37
  38 struct user_namespace;
  39 struct proc_dir_entry;
  40 struct net_device;
  41 struct sock;
  42 struct ctl_table_header;
  43 struct net_generic;
  44 struct uevent_sock;
  45 struct netns_ipvs;
  46
  47
  48 #define NETDEV_HASHBITS    8
  49 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
  50
  51 struct net {
  52         refcount_t              passive;        /* To decided when the network
  53                                                  * namespace should be freed.
  54                                                  */
  55         refcount_t              count;          /* To decided when the network
  56                                                  *  namespace should be shut down.
  57                                                  */
  58         spinlock_t              rules_mod_lock;
  59
  60         atomic64_t              cookie_gen;
  61
  62         struct list_head        list;           /* list of network namespaces */
  63         struct list_head        exit_list;      /* To linked to call pernet exit
  64                                                  * methods on dead net (
  65                                                  * pernet_ops_rwsem read locked),
  66                                                  * or to unregister pernet ops
  67                                                  * (pernet_ops_rwsem write locked).
  68                                                  */
  69         struct llist_node       cleanup_list;   /* namespaces on death row */
  70
  71         struct user_namespace   *user_ns;       /* Owning user namespace */
  72         struct ucounts          *ucounts;
  73         spinlock_t              nsid_lock;
  74         struct idr              netns_ids;
  75
  76         struct ns_common        ns;
  77
  78         struct proc_dir_entry   *proc_net;
  79         struct proc_dir_entry   *proc_net_stat;
  80
  81 #ifdef CONFIG_SYSCTL
  82         struct ctl_table_set    sysctls;
  83 #endif
  84
  85         struct sock             *rtnl;                  /* rtnetlink socket */
  86         struct sock             *genl_sock;
  87
  88         struct uevent_sock      *uevent_sock;           /* uevent socket */
  89
  90         struct list_head        dev_base_head;
  91         struct hlist_head       *dev_name_head;
  92         struct hlist_head       *dev_index_head;
  93         unsigned int            dev_base_seq;   /* protected by rtnl_mutex */
  94         int                     ifindex;
  95         unsigned int            dev_unreg_count;
  96
  97         /* core fib_rules */
  98         struct list_head        rules_ops;
  99
 100         struct list_head        fib_notifier_ops;  /* Populated by
 101                                                     * register_pernet_subsys()
 102                                                     */
 103         struct net_device       *loopback_dev;          /* The loopback */
 104         struct netns_core       core;
 105         struct netns_mib        mib;
 106         struct netns_packet     packet;
 107         struct netns_unix       unx;
 108         struct netns_ipv4       ipv4;
 109 #if IS_ENABLED(CONFIG_IPV6)
 110         struct netns_ipv6       ipv6;
 111 #endif
 112 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
 113         struct netns_ieee802154_lowpan  ieee802154_lowpan;
 114 #endif
 115 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
 116         struct netns_sctp       sctp;
 117 #endif
 118 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
 119         struct netns_dccp       dccp;
 120 #endif
 121 #ifdef CONFIG_NETFILTER
 122         struct netns_nf         nf;
 123         struct netns_xt         xt;
 124 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 125         struct netns_ct         ct;
 126 #endif
 127 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
 128         struct netns_nftables   nft;
 129 #endif
 130 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
 131         struct netns_nf_frag    nf_frag;
 132         struct ctl_table_header *nf_frag_frags_hdr;
 133 #endif
 134         struct sock             *nfnl;
 135         struct sock             *nfnl_stash;
 136 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
 137         struct list_head        nfnl_acct_list;
 138 #endif
 139 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
 140         struct list_head        nfct_timeout_list;
 141 #endif
 142 #endif
 143 #ifdef CONFIG_WEXT_CORE
 144         struct sk_buff_head     wext_nlevents;
 145 #endif
 146         struct net_generic __rcu        *gen;
 147
 148         /* Note : following structs are cache line aligned */
 149 #ifdef CONFIG_XFRM
 150         struct netns_xfrm       xfrm;
 151 #endif
 152 #if IS_ENABLED(CONFIG_IP_VS)
 153         struct netns_ipvs       *ipvs;
 154 #endif
 155 #if IS_ENABLED(CONFIG_MPLS)
 156         struct netns_mpls       mpls;
 157 #endif
 158 #if IS_ENABLED(CONFIG_CAN)
 159         struct netns_can        can;
 160 #endif
 161         struct sock             *diag_nlsk;
 162         atomic_t                fnhe_genid;
 163 } __randomize_layout;
 164
 165 #include <linux/seq_file_net.h>
 166
 167 /* Init's network namespace */
 168 extern struct net init_net;
 169
 170 #ifdef CONFIG_NET_NS
 171 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
 172                         struct net *old_net);
 173
 174 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
 175
 176 void net_ns_barrier(void);
 177 #else /* CONFIG_NET_NS */
 178 #include <linux/sched.h>
 179 #include <linux/nsproxy.h>
 180 static inline struct net *copy_net_ns(unsigned long flags,
 181         struct user_namespace *user_ns, struct net *old_net)
 182 {
 183         if (flags & CLONE_NEWNET)
 184                 return ERR_PTR(-EINVAL);
 185         return old_net;
 186 }
 187
 188 static inline void net_ns_get_ownership(const struct net *net,
 189                                         kuid_t *uid, kgid_t *gid)
 190 {
 191         *uid = GLOBAL_ROOT_UID;
 192         *gid = GLOBAL_ROOT_GID;
 193 }
 194
 195 static inline void net_ns_barrier(void) {}
 196 #endif /* CONFIG_NET_NS */
 197
 198
 199 extern struct list_head net_namespace_list;
 200
 201 struct net *get_net_ns_by_pid(pid_t pid);
 202 struct net *get_net_ns_by_fd(int fd);
 203
 204 #ifdef CONFIG_SYSCTL
 205 void ipx_register_sysctl(void);
 206 void ipx_unregister_sysctl(void);
 207 #else
 208 #define ipx_register_sysctl()
 209 #define ipx_unregister_sysctl()
 210 #endif
 211
 212 #ifdef CONFIG_NET_NS
 213 void __put_net(struct net *net);
 214
 215 static inline struct net *get_net(struct net *net)
 216 {
 217         refcount_inc(&net->count);
 218         return net;
 219 }
 220
 221 static inline struct net *maybe_get_net(struct net *net)
 222 {
 223         /* Used when we know struct net exists but we
 224          * aren't guaranteed a previous reference count
 225          * exists.  If the reference count is zero this
 226          * function fails and returns NULL.
 227          */
 228         if (!refcount_inc_not_zero(&net->count))
 229                 net = NULL;
 230         return net;
 231 }
 232
 233 static inline void put_net(struct net *net)
 234 {
 235         if (refcount_dec_and_test(&net->count))
 236                 __put_net(net);
 237 }
 238
 239 static inline
 240 int net_eq(const struct net *net1, const struct net *net2)
 241 {
 242         return net1 == net2;
 243 }
 244
 245 static inline int check_net(const struct net *net)
 246 {
 247         return refcount_read(&net->count) != 0;
 248 }
 249
 250 void net_drop_ns(void *);
 251
 252 #else
 253
 254 static inline struct net *get_net(struct net *net)
 255 {
 256         return net;
 257 }
 258
 259 static inline void put_net(struct net *net)
 260 {
 261 }
 262
 263 static inline struct net *maybe_get_net(struct net *net)
 264 {
 265         return net;
 266 }
 267
 268 static inline
 269 int net_eq(const struct net *net1, const struct net *net2)
 270 {
 271         return 1;
 272 }
 273
 274 static inline int check_net(const struct net *net)
 275 {
 276         return 1;
 277 }
 278
 279 #define net_drop_ns NULL
 280 #endif
 281
 282
 283 typedef struct {
 284 #ifdef CONFIG_NET_NS
 285         struct net *net;
 286 #endif
 287 } possible_net_t;
 288
 289 static inline void write_pnet(possible_net_t *pnet, struct net *net)
 290 {
 291 #ifdef CONFIG_NET_NS
 292         pnet->net = net;
 293 #endif
 294 }
 295
 296 static inline struct net *read_pnet(const possible_net_t *pnet)
 297 {
 298 #ifdef CONFIG_NET_NS
 299         return pnet->net;
 300 #else
 301         return &init_net;
 302 #endif
 303 }
 304
 305 /* Protected by net_rwsem */
 306 #define for_each_net(VAR)                               \
 307         list_for_each_entry(VAR, &net_namespace_list, list)
 308
 309 #define for_each_net_rcu(VAR)                           \
 310         list_for_each_entry_rcu(VAR, &net_namespace_list, list)
 311
 312 #ifdef CONFIG_NET_NS
 313 #define __net_init
 314 #define __net_exit
 315 #define __net_initdata
 316 #define __net_initconst
 317 #else
 318 #define __net_init      __init
 319 #define __net_exit      __ref
 320 #define __net_initdata  __initdata
 321 #define __net_initconst __initconst
 322 #endif
 323
 324 int peernet2id_alloc(struct net *net, struct net *peer);
 325 int peernet2id(struct net *net, struct net *peer);
 326 bool peernet_has_id(struct net *net, struct net *peer);
 327 struct net *get_net_ns_by_id(struct net *net, int id);
 328
 329 struct pernet_operations {
 330         struct list_head list;
 331         /*
 332          * Below methods are called without any exclusive locks.
 333          * More than one net may be constructed and destructed
 334          * in parallel on several cpus. Every pernet_operations
 335          * have to keep in mind all other pernet_operations and
 336          * to introduce a locking, if they share common resources.
 337          *
 338          * The only time they are called with exclusive lock is
 339          * from register_pernet_subsys(), unregister_pernet_subsys()
 340          * register_pernet_device() and unregister_pernet_device().
 341          *
 342          * Exit methods using blocking RCU primitives, such as
 343          * synchronize_rcu(), should be implemented via exit_batch.
 344          * Then, destruction of a group of net requires single
 345          * synchronize_rcu() related to these pernet_operations,
 346          * instead of separate synchronize_rcu() for every net.
 347          * Please, avoid synchronize_rcu() at all, where it's possible.
 348          */
 349         int (*init)(struct net *net);
 350         void (*exit)(struct net *net);
 351         void (*exit_batch)(struct list_head *net_exit_list);
 352         unsigned int *id;
 353         size_t size;
 354 };
 355
 356 /*
 357  * Use these carefully.  If you implement a network device and it
 358  * needs per network namespace operations use device pernet operations,
 359  * otherwise use pernet subsys operations.
 360  *
 361  * Network interfaces need to be removed from a dying netns _before_
 362  * subsys notifiers can be called, as most of the network code cleanup
 363  * (which is done from subsys notifiers) runs with the assumption that
 364  * dev_remove_pack has been called so no new packets will arrive during
 365  * and after the cleanup functions have been called.  dev_remove_pack
 366  * is not per namespace so instead the guarantee of no more packets
 367  * arriving in a network namespace is provided by ensuring that all
 368  * network devices and all sockets have left the network namespace
 369  * before the cleanup methods are called.
 370  *
 371  * For the longest time the ipv4 icmp code was registered as a pernet
 372  * device which caused kernel oops, and panics during network
 373  * namespace cleanup.   So please don't get this wrong.
 374  */
 375 int register_pernet_subsys(struct pernet_operations *);
 376 void unregister_pernet_subsys(struct pernet_operations *);
 377 int register_pernet_device(struct pernet_operations *);
 378 void unregister_pernet_device(struct pernet_operations *);
 379
 380 struct ctl_table;
 381 struct ctl_table_header;
 382
 383 #ifdef CONFIG_SYSCTL
 384 int net_sysctl_init(void);
 385 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
 386                                              struct ctl_table *table);
 387 void unregister_net_sysctl_table(struct ctl_table_header *header);
 388 #else
 389 static inline int net_sysctl_init(void) { return 0; }
 390 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
 391         const char *path, struct ctl_table *table)
 392 {
 393         return NULL;
 394 }
 395 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
 396 {
 397 }
 398 #endif
 399
 400 static inline int rt_genid_ipv4(struct net *net)
 401 {
 402         return atomic_read(&net->ipv4.rt_genid);
 403 }
 404
 405 static inline void rt_genid_bump_ipv4(struct net *net)
 406 {
 407         atomic_inc(&net->ipv4.rt_genid);
 408 }
 409
 410 extern void (*__fib6_flush_trees)(struct net *net);
 411 static inline void rt_genid_bump_ipv6(struct net *net)
 412 {
 413         if (__fib6_flush_trees)
 414                 __fib6_flush_trees(net);
 415 }
 416
 417 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
 418 static inline struct netns_ieee802154_lowpan *
 419 net_ieee802154_lowpan(struct net *net)
 420 {
 421         return &net->ieee802154_lowpan;
 422 }
 423 #endif
 424
 425 /* For callers who don't really care about whether it's IPv4 or IPv6 */
 426 static inline void rt_genid_bump_all(struct net *net)
 427 {
 428         rt_genid_bump_ipv4(net);
 429         rt_genid_bump_ipv6(net);
 430 }
 431
 432 static inline int fnhe_genid(struct net *net)
 433 {
 434         return atomic_read(&net->fnhe_genid);
 435 }
 436
 437 static inline void fnhe_genid_bump(struct net *net)
 438 {
 439         atomic_inc(&net->fnhe_genid);
 440 }
 441
 442 #endif /* __NET_NET_NAMESPACE_H */