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/netfilter.h>
  26 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  27 #include <net/netns/conntrack.h>
  28 #endif
  29 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
  30 #include <net/netns/flow_table.h>
  31 #endif
  32 #include <net/netns/nftables.h>
  33 #include <net/netns/xfrm.h>
  34 #include <net/netns/mpls.h>
  35 #include <net/netns/can.h>
  36 #include <net/netns/xdp.h>
  37 #include <net/netns/smc.h>
  38 #include <net/netns/bpf.h>
  39 #include <net/netns/mctp.h>
  40 #include <net/net_trackers.h>
  41 #include <linux/ns_common.h>
  42 #include <linux/idr.h>
  43 #include <linux/skbuff.h>
  44 #include <linux/notifier.h>
  45 #include <linux/xarray.h>
  46
  47 struct user_namespace;
  48 struct proc_dir_entry;
  49 struct net_device;
  50 struct sock;
  51 struct ctl_table_header;
  52 struct net_generic;
  53 struct uevent_sock;
  54 struct netns_ipvs;
  55 struct bpf_prog;
  56
  57
  58 #define NETDEV_HASHBITS    8
  59 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
  60
  61 struct net {
  62         /* First cache line can be often dirtied.
  63          * Do not place here read-mostly fields.
  64          */
  65         refcount_t              passive;        /* To decide when the network
  66                                                  * namespace should be freed.
  67                                                  */
  68         spinlock_t              rules_mod_lock;
  69
  70         unsigned int            dev_base_seq;   /* protected by rtnl_mutex */
  71         u32                     ifindex;
  72
  73         spinlock_t              nsid_lock;
  74         atomic_t                fnhe_genid;
  75
  76         struct list_head        list;           /* list of network namespaces */
  77         struct list_head        exit_list;      /* To linked to call pernet exit
  78                                                  * methods on dead net (
  79                                                  * pernet_ops_rwsem read locked),
  80                                                  * or to unregister pernet ops
  81                                                  * (pernet_ops_rwsem write locked).
  82                                                  */
  83         struct llist_node       cleanup_list;   /* namespaces on death row */
  84
  85 #ifdef CONFIG_KEYS
  86         struct key_tag          *key_domain;    /* Key domain of operation tag */
  87 #endif
  88         struct user_namespace   *user_ns;       /* Owning user namespace */
  89         struct ucounts          *ucounts;
  90         struct idr              netns_ids;
  91
  92         struct ns_common        ns;
  93         struct ref_tracker_dir  refcnt_tracker;
  94         struct ref_tracker_dir  notrefcnt_tracker; /* tracker for objects not
  95                                                     * refcounted against netns
  96                                                     */
  97         struct list_head        dev_base_head;
  98         struct proc_dir_entry   *proc_net;
  99         struct proc_dir_entry   *proc_net_stat;
 100
 101 #ifdef CONFIG_SYSCTL
 102         struct ctl_table_set    sysctls;
 103 #endif
 104
 105         struct sock             *rtnl;                  /* rtnetlink socket */
 106         struct sock             *genl_sock;
 107
 108         struct uevent_sock      *uevent_sock;           /* uevent socket */
 109
 110         struct hlist_head       *dev_name_head;
 111         struct hlist_head       *dev_index_head;
 112         struct xarray           dev_by_index;
 113         struct raw_notifier_head        netdev_chain;
 114
 115         /* Note that @hash_mix can be read millions times per second,
 116          * it is critical that it is on a read_mostly cache line.
 117          */
 118         u32                     hash_mix;
 119
 120         struct net_device       *loopback_dev;          /* The loopback */
 121
 122         /* core fib_rules */
 123         struct list_head        rules_ops;
 124
 125         struct netns_core       core;
 126         struct netns_mib        mib;
 127         struct netns_packet     packet;
 128 #if IS_ENABLED(CONFIG_UNIX)
 129         struct netns_unix       unx;
 130 #endif
 131         struct netns_nexthop    nexthop;
 132         struct netns_ipv4       ipv4;
 133 #if IS_ENABLED(CONFIG_IPV6)
 134         struct netns_ipv6       ipv6;
 135 #endif
 136 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
 137         struct netns_ieee802154_lowpan  ieee802154_lowpan;
 138 #endif
 139 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
 140         struct netns_sctp       sctp;
 141 #endif
 142 #ifdef CONFIG_NETFILTER
 143         struct netns_nf         nf;
 144 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 145         struct netns_ct         ct;
 146 #endif
 147 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
 148         struct netns_nftables   nft;
 149 #endif
 150 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
 151         struct netns_ft ft;
 152 #endif
 153 #endif
 154 #ifdef CONFIG_WEXT_CORE
 155         struct sk_buff_head     wext_nlevents;
 156 #endif
 157         struct net_generic __rcu        *gen;
 158
 159         /* Used to store attached BPF programs */
 160         struct netns_bpf        bpf;
 161
 162         /* Note : following structs are cache line aligned */
 163 #ifdef CONFIG_XFRM
 164         struct netns_xfrm       xfrm;
 165 #endif
 166
 167         u64                     net_cookie; /* written once */
 168
 169 #if IS_ENABLED(CONFIG_IP_VS)
 170         struct netns_ipvs       *ipvs;
 171 #endif
 172 #if IS_ENABLED(CONFIG_MPLS)
 173         struct netns_mpls       mpls;
 174 #endif
 175 #if IS_ENABLED(CONFIG_CAN)
 176         struct netns_can        can;
 177 #endif
 178 #ifdef CONFIG_XDP_SOCKETS
 179         struct netns_xdp        xdp;
 180 #endif
 181 #if IS_ENABLED(CONFIG_MCTP)
 182         struct netns_mctp       mctp;
 183 #endif
 184 #if IS_ENABLED(CONFIG_CRYPTO_USER)
 185         struct sock             *crypto_nlsk;
 186 #endif
 187         struct sock             *diag_nlsk;
 188 #if IS_ENABLED(CONFIG_SMC)
 189         struct netns_smc        smc;
 190 #endif
 191 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL
 192         /* Move to a better place when the config guard is removed. */
 193         struct mutex            rtnl_mutex;
 194 #endif
 195 } __randomize_layout;
 196
 197 #include <linux/seq_file_net.h>
 198
 199 /* Init's network namespace */
 200 extern struct net init_net;
 201
 202 #ifdef CONFIG_NET_NS
 203 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
 204                         struct net *old_net);
 205
 206 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
 207
 208 void net_ns_barrier(void);
 209
 210 struct ns_common *get_net_ns(struct ns_common *ns);
 211 struct net *get_net_ns_by_fd(int fd);
 212 #else /* CONFIG_NET_NS */
 213 #include <linux/sched.h>
 214 #include <linux/nsproxy.h>
 215 static inline struct net *copy_net_ns(unsigned long flags,
 216         struct user_namespace *user_ns, struct net *old_net)
 217 {
 218         if (flags & CLONE_NEWNET)
 219                 return ERR_PTR(-EINVAL);
 220         return old_net;
 221 }
 222
 223 static inline void net_ns_get_ownership(const struct net *net,
 224                                         kuid_t *uid, kgid_t *gid)
 225 {
 226         *uid = GLOBAL_ROOT_UID;
 227         *gid = GLOBAL_ROOT_GID;
 228 }
 229
 230 static inline void net_ns_barrier(void) {}
 231
 232 static inline struct ns_common *get_net_ns(struct ns_common *ns)
 233 {
 234         return ERR_PTR(-EINVAL);
 235 }
 236
 237 static inline struct net *get_net_ns_by_fd(int fd)
 238 {
 239         return ERR_PTR(-EINVAL);
 240 }
 241 #endif /* CONFIG_NET_NS */
 242
 243
 244 extern struct list_head net_namespace_list;
 245
 246 struct net *get_net_ns_by_pid(pid_t pid);
 247
 248 #ifdef CONFIG_SYSCTL
 249 void ipx_register_sysctl(void);
 250 void ipx_unregister_sysctl(void);
 251 #else
 252 #define ipx_register_sysctl()
 253 #define ipx_unregister_sysctl()
 254 #endif
 255
 256 #ifdef CONFIG_NET_NS
 257 void __put_net(struct net *net);
 258
 259 /* Try using get_net_track() instead */
 260 static inline struct net *get_net(struct net *net)
 261 {
 262         refcount_inc(&net->ns.count);
 263         return net;
 264 }
 265
 266 static inline struct net *maybe_get_net(struct net *net)
 267 {
 268         /* Used when we know struct net exists but we
 269          * aren't guaranteed a previous reference count
 270          * exists.  If the reference count is zero this
 271          * function fails and returns NULL.
 272          */
 273         if (!refcount_inc_not_zero(&net->ns.count))
 274                 net = NULL;
 275         return net;
 276 }
 277
 278 /* Try using put_net_track() instead */
 279 static inline void put_net(struct net *net)
 280 {
 281         if (refcount_dec_and_test(&net->ns.count))
 282                 __put_net(net);
 283 }
 284
 285 static inline
 286 int net_eq(const struct net *net1, const struct net *net2)
 287 {
 288         return net1 == net2;
 289 }
 290
 291 static inline int check_net(const struct net *net)
 292 {
 293         return refcount_read(&net->ns.count) != 0;
 294 }
 295
 296 void net_drop_ns(void *);
 297
 298 #else
 299
 300 static inline struct net *get_net(struct net *net)
 301 {
 302         return net;
 303 }
 304
 305 static inline void put_net(struct net *net)
 306 {
 307 }
 308
 309 static inline struct net *maybe_get_net(struct net *net)
 310 {
 311         return net;
 312 }
 313
 314 static inline
 315 int net_eq(const struct net *net1, const struct net *net2)
 316 {
 317         return 1;
 318 }
 319
 320 static inline int check_net(const struct net *net)
 321 {
 322         return 1;
 323 }
 324
 325 #define net_drop_ns NULL
 326 #endif
 327
 328 /* Returns true if the netns initialization is completed successfully */
 329 static inline bool net_initialized(const struct net *net)
 330 {
 331         return READ_ONCE(net->list.next);
 332 }
 333
 334 static inline void __netns_tracker_alloc(struct net *net,
 335                                          netns_tracker *tracker,
 336                                          bool refcounted,
 337                                          gfp_t gfp)
 338 {
 339 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
 340         ref_tracker_alloc(refcounted ? &net->refcnt_tracker :
 341                                        &net->notrefcnt_tracker,
 342                           tracker, gfp);
 343 #endif
 344 }
 345
 346 static inline void netns_tracker_alloc(struct net *net, netns_tracker *tracker,
 347                                        gfp_t gfp)
 348 {
 349         __netns_tracker_alloc(net, tracker, true, gfp);
 350 }
 351
 352 static inline void __netns_tracker_free(struct net *net,
 353                                         netns_tracker *tracker,
 354                                         bool refcounted)
 355 {
 356 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
 357        ref_tracker_free(refcounted ? &net->refcnt_tracker :
 358                                      &net->notrefcnt_tracker, tracker);
 359 #endif
 360 }
 361
 362 static inline struct net *get_net_track(struct net *net,
 363                                         netns_tracker *tracker, gfp_t gfp)
 364 {
 365         get_net(net);
 366         netns_tracker_alloc(net, tracker, gfp);
 367         return net;
 368 }
 369
 370 static inline void put_net_track(struct net *net, netns_tracker *tracker)
 371 {
 372         __netns_tracker_free(net, tracker, true);
 373         put_net(net);
 374 }
 375
 376 typedef struct {
 377 #ifdef CONFIG_NET_NS
 378         struct net __rcu *net;
 379 #endif
 380 } possible_net_t;
 381
 382 static inline void write_pnet(possible_net_t *pnet, struct net *net)
 383 {
 384 #ifdef CONFIG_NET_NS
 385         rcu_assign_pointer(pnet->net, net);
 386 #endif
 387 }
 388
 389 static inline struct net *read_pnet(const possible_net_t *pnet)
 390 {
 391 #ifdef CONFIG_NET_NS
 392         return rcu_dereference_protected(pnet->net, true);
 393 #else
 394         return &init_net;
 395 #endif
 396 }
 397
 398 static inline struct net *read_pnet_rcu(possible_net_t *pnet)
 399 {
 400 #ifdef CONFIG_NET_NS
 401         return rcu_dereference(pnet->net);
 402 #else
 403         return &init_net;
 404 #endif
 405 }
 406
 407 /* Protected by net_rwsem */
 408 #define for_each_net(VAR)                               \
 409         list_for_each_entry(VAR, &net_namespace_list, list)
 410 #define for_each_net_continue_reverse(VAR)              \
 411         list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
 412 #define for_each_net_rcu(VAR)                           \
 413         list_for_each_entry_rcu(VAR, &net_namespace_list, list)
 414
 415 #ifdef CONFIG_NET_NS
 416 #define __net_init
 417 #define __net_exit
 418 #define __net_initdata
 419 #define __net_initconst
 420 #else
 421 #define __net_init      __init
 422 #define __net_exit      __ref
 423 #define __net_initdata  __initdata
 424 #define __net_initconst __initconst
 425 #endif
 426
 427 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
 428 int peernet2id(const struct net *net, struct net *peer);
 429 bool peernet_has_id(const struct net *net, struct net *peer);
 430 struct net *get_net_ns_by_id(const struct net *net, int id);
 431
 432 struct pernet_operations {
 433         struct list_head list;
 434         /*
 435          * Below methods are called without any exclusive locks.
 436          * More than one net may be constructed and destructed
 437          * in parallel on several cpus. Every pernet_operations
 438          * have to keep in mind all other pernet_operations and
 439          * to introduce a locking, if they share common resources.
 440          *
 441          * The only time they are called with exclusive lock is
 442          * from register_pernet_subsys(), unregister_pernet_subsys()
 443          * register_pernet_device() and unregister_pernet_device().
 444          *
 445          * Exit methods using blocking RCU primitives, such as
 446          * synchronize_rcu(), should be implemented via exit_batch.
 447          * Then, destruction of a group of net requires single
 448          * synchronize_rcu() related to these pernet_operations,
 449          * instead of separate synchronize_rcu() for every net.
 450          * Please, avoid synchronize_rcu() at all, where it's possible.
 451          *
 452          * Note that a combination of pre_exit() and exit() can
 453          * be used, since a synchronize_rcu() is guaranteed between
 454          * the calls.
 455          */
 456         int (*init)(struct net *net);
 457         void (*pre_exit)(struct net *net);
 458         void (*exit)(struct net *net);
 459         void (*exit_batch)(struct list_head *net_exit_list);
 460         /* Following method is called with RTNL held. */
 461         void (*exit_batch_rtnl)(struct list_head *net_exit_list,
 462                                 struct list_head *dev_kill_list);
 463         unsigned int * const id;
 464         const size_t size;
 465 };
 466
 467 /*
 468  * Use these carefully.  If you implement a network device and it
 469  * needs per network namespace operations use device pernet operations,
 470  * otherwise use pernet subsys operations.
 471  *
 472  * Network interfaces need to be removed from a dying netns _before_
 473  * subsys notifiers can be called, as most of the network code cleanup
 474  * (which is done from subsys notifiers) runs with the assumption that
 475  * dev_remove_pack has been called so no new packets will arrive during
 476  * and after the cleanup functions have been called.  dev_remove_pack
 477  * is not per namespace so instead the guarantee of no more packets
 478  * arriving in a network namespace is provided by ensuring that all
 479  * network devices and all sockets have left the network namespace
 480  * before the cleanup methods are called.
 481  *
 482  * For the longest time the ipv4 icmp code was registered as a pernet
 483  * device which caused kernel oops, and panics during network
 484  * namespace cleanup.   So please don't get this wrong.
 485  */
 486 int register_pernet_subsys(struct pernet_operations *);
 487 void unregister_pernet_subsys(struct pernet_operations *);
 488 int register_pernet_device(struct pernet_operations *);
 489 void unregister_pernet_device(struct pernet_operations *);
 490
 491 struct ctl_table;
 492
 493 #define register_net_sysctl(net, path, table)   \
 494         register_net_sysctl_sz(net, path, table, ARRAY_SIZE(table))
 495 #ifdef CONFIG_SYSCTL
 496 int net_sysctl_init(void);
 497 struct ctl_table_header *register_net_sysctl_sz(struct net *net, const char *path,
 498                                              struct ctl_table *table, size_t table_size);
 499 void unregister_net_sysctl_table(struct ctl_table_header *header);
 500 #else
 501 static inline int net_sysctl_init(void) { return 0; }
 502 static inline struct ctl_table_header *register_net_sysctl_sz(struct net *net,
 503         const char *path, struct ctl_table *table, size_t table_size)
 504 {
 505         return NULL;
 506 }
 507 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
 508 {
 509 }
 510 #endif
 511
 512 static inline int rt_genid_ipv4(const struct net *net)
 513 {
 514         return atomic_read(&net->ipv4.rt_genid);
 515 }
 516
 517 #if IS_ENABLED(CONFIG_IPV6)
 518 static inline int rt_genid_ipv6(const struct net *net)
 519 {
 520         return atomic_read(&net->ipv6.fib6_sernum);
 521 }
 522 #endif
 523
 524 static inline void rt_genid_bump_ipv4(struct net *net)
 525 {
 526         atomic_inc(&net->ipv4.rt_genid);
 527 }
 528
 529 extern void (*__fib6_flush_trees)(struct net *net);
 530 static inline void rt_genid_bump_ipv6(struct net *net)
 531 {
 532         if (__fib6_flush_trees)
 533                 __fib6_flush_trees(net);
 534 }
 535
 536 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
 537 static inline struct netns_ieee802154_lowpan *
 538 net_ieee802154_lowpan(struct net *net)
 539 {
 540         return &net->ieee802154_lowpan;
 541 }
 542 #endif
 543
 544 /* For callers who don't really care about whether it's IPv4 or IPv6 */
 545 static inline void rt_genid_bump_all(struct net *net)
 546 {
 547         rt_genid_bump_ipv4(net);
 548         rt_genid_bump_ipv6(net);
 549 }
 550
 551 static inline int fnhe_genid(const struct net *net)
 552 {
 553         return atomic_read(&net->fnhe_genid);
 554 }
 555
 556 static inline void fnhe_genid_bump(struct net *net)
 557 {
 558         atomic_inc(&net->fnhe_genid);
 559 }
 560
 561 #ifdef CONFIG_NET
 562 void net_ns_init(void);
 563 #else
 564 static inline void net_ns_init(void) {}
 565 #endif
 566
 567 #endif /* __NET_NET_NAMESPACE_H */