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/nexthop.h>
  23 #include <net/netns/ieee802154_6lowpan.h>
  24 #include <net/netns/sctp.h>
  25 #include <net/netns/dccp.h>
  26 #include <net/netns/netfilter.h>
  27 #include <net/netns/x_tables.h>
  28 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  29 #include <net/netns/conntrack.h>
  30 #endif
  31 #include <net/netns/nftables.h>
  32 #include <net/netns/xfrm.h>
  33 #include <net/netns/mpls.h>
  34 #include <net/netns/can.h>
  35 #include <net/netns/xdp.h>
  36 #include <linux/ns_common.h>
  37 #include <linux/idr.h>
  38 #include <linux/skbuff.h>
  39 #include <linux/notifier.h>
  40
  41 struct user_namespace;
  42 struct proc_dir_entry;
  43 struct net_device;
  44 struct sock;
  45 struct ctl_table_header;
  46 struct net_generic;
  47 struct uevent_sock;
  48 struct netns_ipvs;
  49 struct bpf_prog;
  50
  51
  52 #define NETDEV_HASHBITS    8
  53 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
  54
  55 struct net {
  56         /* First cache line can be often dirtied.
  57          * Do not place here read-mostly fields.
  58          */
  59         refcount_t              passive;        /* To decide when the network
  60                                                  * namespace should be freed.
  61                                                  */
  62         refcount_t              count;          /* To decided when the network
  63                                                  *  namespace should be shut down.
  64                                                  */
  65         spinlock_t              rules_mod_lock;
  66
  67         unsigned int            dev_unreg_count;
  68
  69         unsigned int            dev_base_seq;   /* protected by rtnl_mutex */
  70         int                     ifindex;
  71
  72         spinlock_t              nsid_lock;
  73         atomic_t                fnhe_genid;
  74
  75         struct list_head        list;           /* list of network namespaces */
  76         struct list_head        exit_list;      /* To linked to call pernet exit
  77                                                  * methods on dead net (
  78                                                  * pernet_ops_rwsem read locked),
  79                                                  * or to unregister pernet ops
  80                                                  * (pernet_ops_rwsem write locked).
  81                                                  */
  82         struct llist_node       cleanup_list;   /* namespaces on death row */
  83
  84 #ifdef CONFIG_KEYS
  85         struct key_tag          *key_domain;    /* Key domain of operation tag */
  86 #endif
  87         struct user_namespace   *user_ns;       /* Owning user namespace */
  88         struct ucounts          *ucounts;
  89         struct idr              netns_ids;
  90
  91         struct ns_common        ns;
  92
  93         struct list_head        dev_base_head;
  94         struct proc_dir_entry   *proc_net;
  95         struct proc_dir_entry   *proc_net_stat;
  96
  97 #ifdef CONFIG_SYSCTL
  98         struct ctl_table_set    sysctls;
  99 #endif
 100
 101         struct sock             *rtnl;                  /* rtnetlink socket */
 102         struct sock             *genl_sock;
 103
 104         struct uevent_sock      *uevent_sock;           /* uevent socket */
 105
 106         struct hlist_head       *dev_name_head;
 107         struct hlist_head       *dev_index_head;
 108         struct raw_notifier_head        netdev_chain;
 109
 110         /* Note that @hash_mix can be read millions times per second,
 111          * it is critical that it is on a read_mostly cache line.
 112          */
 113         u32                     hash_mix;
 114
 115         struct net_device       *loopback_dev;          /* The loopback */
 116
 117         /* core fib_rules */
 118         struct list_head        rules_ops;
 119
 120         struct netns_core       core;
 121         struct netns_mib        mib;
 122         struct netns_packet     packet;
 123         struct netns_unix       unx;
 124         struct netns_nexthop    nexthop;
 125         struct netns_ipv4       ipv4;
 126 #if IS_ENABLED(CONFIG_IPV6)
 127         struct netns_ipv6       ipv6;
 128 #endif
 129 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
 130         struct netns_ieee802154_lowpan  ieee802154_lowpan;
 131 #endif
 132 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
 133         struct netns_sctp       sctp;
 134 #endif
 135 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
 136         struct netns_dccp       dccp;
 137 #endif
 138 #ifdef CONFIG_NETFILTER
 139         struct netns_nf         nf;
 140         struct netns_xt         xt;
 141 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 142         struct netns_ct         ct;
 143 #endif
 144 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
 145         struct netns_nftables   nft;
 146 #endif
 147 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
 148         struct netns_nf_frag    nf_frag;
 149         struct ctl_table_header *nf_frag_frags_hdr;
 150 #endif
 151         struct sock             *nfnl;
 152         struct sock             *nfnl_stash;
 153 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
 154         struct list_head        nfnl_acct_list;
 155 #endif
 156 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
 157         struct list_head        nfct_timeout_list;
 158 #endif
 159 #endif
 160 #ifdef CONFIG_WEXT_CORE
 161         struct sk_buff_head     wext_nlevents;
 162 #endif
 163         struct net_generic __rcu        *gen;
 164
 165         struct bpf_prog __rcu   *flow_dissector_prog;
 166
 167         /* Note : following structs are cache line aligned */
 168 #ifdef CONFIG_XFRM
 169         struct netns_xfrm       xfrm;
 170 #endif
 171 #if IS_ENABLED(CONFIG_IP_VS)
 172         struct netns_ipvs       *ipvs;
 173 #endif
 174 #if IS_ENABLED(CONFIG_MPLS)
 175         struct netns_mpls       mpls;
 176 #endif
 177 #if IS_ENABLED(CONFIG_CAN)
 178         struct netns_can        can;
 179 #endif
 180 #ifdef CONFIG_XDP_SOCKETS
 181         struct netns_xdp        xdp;
 182 #endif
 183 #if IS_ENABLED(CONFIG_CRYPTO_USER)
 184         struct sock             *crypto_nlsk;
 185 #endif
 186         struct sock             *diag_nlsk;
 187 } __randomize_layout;
 188
 189 #include <linux/seq_file_net.h>
 190
 191 /* Init's network namespace */
 192 extern struct net init_net;
 193
 194 #ifdef CONFIG_NET_NS
 195 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
 196                         struct net *old_net);
 197
 198 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
 199
 200 void net_ns_barrier(void);
 201 #else /* CONFIG_NET_NS */
 202 #include <linux/sched.h>
 203 #include <linux/nsproxy.h>
 204 static inline struct net *copy_net_ns(unsigned long flags,
 205         struct user_namespace *user_ns, struct net *old_net)
 206 {
 207         if (flags & CLONE_NEWNET)
 208                 return ERR_PTR(-EINVAL);
 209         return old_net;
 210 }
 211
 212 static inline void net_ns_get_ownership(const struct net *net,
 213                                         kuid_t *uid, kgid_t *gid)
 214 {
 215         *uid = GLOBAL_ROOT_UID;
 216         *gid = GLOBAL_ROOT_GID;
 217 }
 218
 219 static inline void net_ns_barrier(void) {}
 220 #endif /* CONFIG_NET_NS */
 221
 222
 223 extern struct list_head net_namespace_list;
 224
 225 struct net *get_net_ns_by_pid(pid_t pid);
 226 struct net *get_net_ns_by_fd(int fd);
 227
 228 #ifdef CONFIG_SYSCTL
 229 void ipx_register_sysctl(void);
 230 void ipx_unregister_sysctl(void);
 231 #else
 232 #define ipx_register_sysctl()
 233 #define ipx_unregister_sysctl()
 234 #endif
 235
 236 #ifdef CONFIG_NET_NS
 237 void __put_net(struct net *net);
 238
 239 static inline struct net *get_net(struct net *net)
 240 {
 241         refcount_inc(&net->count);
 242         return net;
 243 }
 244
 245 static inline struct net *maybe_get_net(struct net *net)
 246 {
 247         /* Used when we know struct net exists but we
 248          * aren't guaranteed a previous reference count
 249          * exists.  If the reference count is zero this
 250          * function fails and returns NULL.
 251          */
 252         if (!refcount_inc_not_zero(&net->count))
 253                 net = NULL;
 254         return net;
 255 }
 256
 257 static inline void put_net(struct net *net)
 258 {
 259         if (refcount_dec_and_test(&net->count))
 260                 __put_net(net);
 261 }
 262
 263 static inline
 264 int net_eq(const struct net *net1, const struct net *net2)
 265 {
 266         return net1 == net2;
 267 }
 268
 269 static inline int check_net(const struct net *net)
 270 {
 271         return refcount_read(&net->count) != 0;
 272 }
 273
 274 void net_drop_ns(void *);
 275
 276 #else
 277
 278 static inline struct net *get_net(struct net *net)
 279 {
 280         return net;
 281 }
 282
 283 static inline void put_net(struct net *net)
 284 {
 285 }
 286
 287 static inline struct net *maybe_get_net(struct net *net)
 288 {
 289         return net;
 290 }
 291
 292 static inline
 293 int net_eq(const struct net *net1, const struct net *net2)
 294 {
 295         return 1;
 296 }
 297
 298 static inline int check_net(const struct net *net)
 299 {
 300         return 1;
 301 }
 302
 303 #define net_drop_ns NULL
 304 #endif
 305
 306
 307 typedef struct {
 308 #ifdef CONFIG_NET_NS
 309         struct net *net;
 310 #endif
 311 } possible_net_t;
 312
 313 static inline void write_pnet(possible_net_t *pnet, struct net *net)
 314 {
 315 #ifdef CONFIG_NET_NS
 316         pnet->net = net;
 317 #endif
 318 }
 319
 320 static inline struct net *read_pnet(const possible_net_t *pnet)
 321 {
 322 #ifdef CONFIG_NET_NS
 323         return pnet->net;
 324 #else
 325         return &init_net;
 326 #endif
 327 }
 328
 329 /* Protected by net_rwsem */
 330 #define for_each_net(VAR)                               \
 331         list_for_each_entry(VAR, &net_namespace_list, list)
 332 #define for_each_net_continue_reverse(VAR)              \
 333         list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
 334 #define for_each_net_rcu(VAR)                           \
 335         list_for_each_entry_rcu(VAR, &net_namespace_list, list)
 336
 337 #ifdef CONFIG_NET_NS
 338 #define __net_init
 339 #define __net_exit
 340 #define __net_initdata
 341 #define __net_initconst
 342 #else
 343 #define __net_init      __init
 344 #define __net_exit      __ref
 345 #define __net_initdata  __initdata
 346 #define __net_initconst __initconst
 347 #endif
 348
 349 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
 350 int peernet2id(const struct net *net, struct net *peer);
 351 bool peernet_has_id(const struct net *net, struct net *peer);
 352 struct net *get_net_ns_by_id(const struct net *net, int id);
 353
 354 struct pernet_operations {
 355         struct list_head list;
 356         /*
 357          * Below methods are called without any exclusive locks.
 358          * More than one net may be constructed and destructed
 359          * in parallel on several cpus. Every pernet_operations
 360          * have to keep in mind all other pernet_operations and
 361          * to introduce a locking, if they share common resources.
 362          *
 363          * The only time they are called with exclusive lock is
 364          * from register_pernet_subsys(), unregister_pernet_subsys()
 365          * register_pernet_device() and unregister_pernet_device().
 366          *
 367          * Exit methods using blocking RCU primitives, such as
 368          * synchronize_rcu(), should be implemented via exit_batch.
 369          * Then, destruction of a group of net requires single
 370          * synchronize_rcu() related to these pernet_operations,
 371          * instead of separate synchronize_rcu() for every net.
 372          * Please, avoid synchronize_rcu() at all, where it's possible.
 373          *
 374          * Note that a combination of pre_exit() and exit() can
 375          * be used, since a synchronize_rcu() is guaranteed between
 376          * the calls.
 377          */
 378         int (*init)(struct net *net);
 379         void (*pre_exit)(struct net *net);
 380         void (*exit)(struct net *net);
 381         void (*exit_batch)(struct list_head *net_exit_list);
 382         unsigned int *id;
 383         size_t size;
 384 };
 385
 386 /*
 387  * Use these carefully.  If you implement a network device and it
 388  * needs per network namespace operations use device pernet operations,
 389  * otherwise use pernet subsys operations.
 390  *
 391  * Network interfaces need to be removed from a dying netns _before_
 392  * subsys notifiers can be called, as most of the network code cleanup
 393  * (which is done from subsys notifiers) runs with the assumption that
 394  * dev_remove_pack has been called so no new packets will arrive during
 395  * and after the cleanup functions have been called.  dev_remove_pack
 396  * is not per namespace so instead the guarantee of no more packets
 397  * arriving in a network namespace is provided by ensuring that all
 398  * network devices and all sockets have left the network namespace
 399  * before the cleanup methods are called.
 400  *
 401  * For the longest time the ipv4 icmp code was registered as a pernet
 402  * device which caused kernel oops, and panics during network
 403  * namespace cleanup.   So please don't get this wrong.
 404  */
 405 int register_pernet_subsys(struct pernet_operations *);
 406 void unregister_pernet_subsys(struct pernet_operations *);
 407 int register_pernet_device(struct pernet_operations *);
 408 void unregister_pernet_device(struct pernet_operations *);
 409
 410 struct ctl_table;
 411 struct ctl_table_header;
 412
 413 #ifdef CONFIG_SYSCTL
 414 int net_sysctl_init(void);
 415 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
 416                                              struct ctl_table *table);
 417 void unregister_net_sysctl_table(struct ctl_table_header *header);
 418 #else
 419 static inline int net_sysctl_init(void) { return 0; }
 420 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
 421         const char *path, struct ctl_table *table)
 422 {
 423         return NULL;
 424 }
 425 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
 426 {
 427 }
 428 #endif
 429
 430 static inline int rt_genid_ipv4(const struct net *net)
 431 {
 432         return atomic_read(&net->ipv4.rt_genid);
 433 }
 434
 435 static inline void rt_genid_bump_ipv4(struct net *net)
 436 {
 437         atomic_inc(&net->ipv4.rt_genid);
 438 }
 439
 440 extern void (*__fib6_flush_trees)(struct net *net);
 441 static inline void rt_genid_bump_ipv6(struct net *net)
 442 {
 443         if (__fib6_flush_trees)
 444                 __fib6_flush_trees(net);
 445 }
 446
 447 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
 448 static inline struct netns_ieee802154_lowpan *
 449 net_ieee802154_lowpan(struct net *net)
 450 {
 451         return &net->ieee802154_lowpan;
 452 }
 453 #endif
 454
 455 /* For callers who don't really care about whether it's IPv4 or IPv6 */
 456 static inline void rt_genid_bump_all(struct net *net)
 457 {
 458         rt_genid_bump_ipv4(net);
 459         rt_genid_bump_ipv6(net);
 460 }
 461
 462 static inline int fnhe_genid(const struct net *net)
 463 {
 464         return atomic_read(&net->fnhe_genid);
 465 }
 466
 467 static inline void fnhe_genid_bump(struct net *net)
 468 {
 469         atomic_inc(&net->fnhe_genid);
 470 }
 471
 472 #endif /* __NET_NET_NAMESPACE_H */