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 <net/netns/bpf.h>
  37 #include <linux/ns_common.h>
  38 #include <linux/idr.h>
  39 #include <linux/skbuff.h>
  40 #include <linux/notifier.h>
  41
  42 struct user_namespace;
  43 struct proc_dir_entry;
  44 struct net_device;
  45 struct sock;
  46 struct ctl_table_header;
  47 struct net_generic;
  48 struct uevent_sock;
  49 struct netns_ipvs;
  50 struct bpf_prog;
  51
  52
  53 #define NETDEV_HASHBITS    8
  54 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
  55
  56 struct net {
  57         /* First cache line can be often dirtied.
  58          * Do not place here read-mostly fields.
  59          */
  60         refcount_t              passive;        /* To decide when the network
  61                                                  * namespace should be freed.
  62                                                  */
  63         refcount_t              count;          /* To decided when the network
  64                                                  *  namespace should be shut down.
  65                                                  */
  66         spinlock_t              rules_mod_lock;
  67
  68         unsigned int            dev_unreg_count;
  69
  70         unsigned int            dev_base_seq;   /* protected by rtnl_mutex */
  71         int                     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
  94         struct list_head        dev_base_head;
  95         struct proc_dir_entry   *proc_net;
  96         struct proc_dir_entry   *proc_net_stat;
  97
  98 #ifdef CONFIG_SYSCTL
  99         struct ctl_table_set    sysctls;
 100 #endif
 101
 102         struct sock             *rtnl;                  /* rtnetlink socket */
 103         struct sock             *genl_sock;
 104
 105         struct uevent_sock      *uevent_sock;           /* uevent socket */
 106
 107         struct hlist_head       *dev_name_head;
 108         struct hlist_head       *dev_index_head;
 109         struct raw_notifier_head        netdev_chain;
 110
 111         /* Note that @hash_mix can be read millions times per second,
 112          * it is critical that it is on a read_mostly cache line.
 113          */
 114         u32                     hash_mix;
 115
 116         struct net_device       *loopback_dev;          /* The loopback */
 117
 118         /* core fib_rules */
 119         struct list_head        rules_ops;
 120
 121         struct netns_core       core;
 122         struct netns_mib        mib;
 123         struct netns_packet     packet;
 124         struct netns_unix       unx;
 125         struct netns_nexthop    nexthop;
 126         struct netns_ipv4       ipv4;
 127 #if IS_ENABLED(CONFIG_IPV6)
 128         struct netns_ipv6       ipv6;
 129 #endif
 130 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
 131         struct netns_ieee802154_lowpan  ieee802154_lowpan;
 132 #endif
 133 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
 134         struct netns_sctp       sctp;
 135 #endif
 136 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
 137         struct netns_dccp       dccp;
 138 #endif
 139 #ifdef CONFIG_NETFILTER
 140         struct netns_nf         nf;
 141         struct netns_xt         xt;
 142 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 143         struct netns_ct         ct;
 144 #endif
 145 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
 146         struct netns_nftables   nft;
 147 #endif
 148 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
 149         struct netns_nf_frag    nf_frag;
 150         struct ctl_table_header *nf_frag_frags_hdr;
 151 #endif
 152         struct sock             *nfnl;
 153         struct sock             *nfnl_stash;
 154 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
 155         struct list_head        nfnl_acct_list;
 156 #endif
 157 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
 158         struct list_head        nfct_timeout_list;
 159 #endif
 160 #endif
 161 #ifdef CONFIG_WEXT_CORE
 162         struct sk_buff_head     wext_nlevents;
 163 #endif
 164         struct net_generic __rcu        *gen;
 165
 166         /* Used to store attached BPF programs */
 167         struct netns_bpf        bpf;
 168
 169         /* Note : following structs are cache line aligned */
 170 #ifdef CONFIG_XFRM
 171         struct netns_xfrm       xfrm;
 172 #endif
 173
 174         atomic64_t              net_cookie; /* written once */
 175
 176 #if IS_ENABLED(CONFIG_IP_VS)
 177         struct netns_ipvs       *ipvs;
 178 #endif
 179 #if IS_ENABLED(CONFIG_MPLS)
 180         struct netns_mpls       mpls;
 181 #endif
 182 #if IS_ENABLED(CONFIG_CAN)
 183         struct netns_can        can;
 184 #endif
 185 #ifdef CONFIG_XDP_SOCKETS
 186         struct netns_xdp        xdp;
 187 #endif
 188 #if IS_ENABLED(CONFIG_CRYPTO_USER)
 189         struct sock             *crypto_nlsk;
 190 #endif
 191         struct sock             *diag_nlsk;
 192 } __randomize_layout;
 193
 194 #include <linux/seq_file_net.h>
 195
 196 /* Init's network namespace */
 197 extern struct net init_net;
 198
 199 #ifdef CONFIG_NET_NS
 200 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
 201                         struct net *old_net);
 202
 203 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
 204
 205 void net_ns_barrier(void);
 206 #else /* CONFIG_NET_NS */
 207 #include <linux/sched.h>
 208 #include <linux/nsproxy.h>
 209 static inline struct net *copy_net_ns(unsigned long flags,
 210         struct user_namespace *user_ns, struct net *old_net)
 211 {
 212         if (flags & CLONE_NEWNET)
 213                 return ERR_PTR(-EINVAL);
 214         return old_net;
 215 }
 216
 217 static inline void net_ns_get_ownership(const struct net *net,
 218                                         kuid_t *uid, kgid_t *gid)
 219 {
 220         *uid = GLOBAL_ROOT_UID;
 221         *gid = GLOBAL_ROOT_GID;
 222 }
 223
 224 static inline void net_ns_barrier(void) {}
 225 #endif /* CONFIG_NET_NS */
 226
 227
 228 extern struct list_head net_namespace_list;
 229
 230 struct net *get_net_ns_by_pid(pid_t pid);
 231 struct net *get_net_ns_by_fd(int fd);
 232
 233 u64 net_gen_cookie(struct net *net);
 234
 235 #ifdef CONFIG_SYSCTL
 236 void ipx_register_sysctl(void);
 237 void ipx_unregister_sysctl(void);
 238 #else
 239 #define ipx_register_sysctl()
 240 #define ipx_unregister_sysctl()
 241 #endif
 242
 243 #ifdef CONFIG_NET_NS
 244 void __put_net(struct net *net);
 245
 246 static inline struct net *get_net(struct net *net)
 247 {
 248         refcount_inc(&net->count);
 249         return net;
 250 }
 251
 252 static inline struct net *maybe_get_net(struct net *net)
 253 {
 254         /* Used when we know struct net exists but we
 255          * aren't guaranteed a previous reference count
 256          * exists.  If the reference count is zero this
 257          * function fails and returns NULL.
 258          */
 259         if (!refcount_inc_not_zero(&net->count))
 260                 net = NULL;
 261         return net;
 262 }
 263
 264 static inline void put_net(struct net *net)
 265 {
 266         if (refcount_dec_and_test(&net->count))
 267                 __put_net(net);
 268 }
 269
 270 static inline
 271 int net_eq(const struct net *net1, const struct net *net2)
 272 {
 273         return net1 == net2;
 274 }
 275
 276 static inline int check_net(const struct net *net)
 277 {
 278         return refcount_read(&net->count) != 0;
 279 }
 280
 281 void net_drop_ns(void *);
 282
 283 #else
 284
 285 static inline struct net *get_net(struct net *net)
 286 {
 287         return net;
 288 }
 289
 290 static inline void put_net(struct net *net)
 291 {
 292 }
 293
 294 static inline struct net *maybe_get_net(struct net *net)
 295 {
 296         return net;
 297 }
 298
 299 static inline
 300 int net_eq(const struct net *net1, const struct net *net2)
 301 {
 302         return 1;
 303 }
 304
 305 static inline int check_net(const struct net *net)
 306 {
 307         return 1;
 308 }
 309
 310 #define net_drop_ns NULL
 311 #endif
 312
 313
 314 typedef struct {
 315 #ifdef CONFIG_NET_NS
 316         struct net *net;
 317 #endif
 318 } possible_net_t;
 319
 320 static inline void write_pnet(possible_net_t *pnet, struct net *net)
 321 {
 322 #ifdef CONFIG_NET_NS
 323         pnet->net = net;
 324 #endif
 325 }
 326
 327 static inline struct net *read_pnet(const possible_net_t *pnet)
 328 {
 329 #ifdef CONFIG_NET_NS
 330         return pnet->net;
 331 #else
 332         return &init_net;
 333 #endif
 334 }
 335
 336 /* Protected by net_rwsem */
 337 #define for_each_net(VAR)                               \
 338         list_for_each_entry(VAR, &net_namespace_list, list)
 339 #define for_each_net_continue_reverse(VAR)              \
 340         list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
 341 #define for_each_net_rcu(VAR)                           \
 342         list_for_each_entry_rcu(VAR, &net_namespace_list, list)
 343
 344 #ifdef CONFIG_NET_NS
 345 #define __net_init
 346 #define __net_exit
 347 #define __net_initdata
 348 #define __net_initconst
 349 #else
 350 #define __net_init      __init
 351 #define __net_exit      __ref
 352 #define __net_initdata  __initdata
 353 #define __net_initconst __initconst
 354 #endif
 355
 356 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
 357 int peernet2id(const struct net *net, struct net *peer);
 358 bool peernet_has_id(const struct net *net, struct net *peer);
 359 struct net *get_net_ns_by_id(const struct net *net, int id);
 360
 361 struct pernet_operations {
 362         struct list_head list;
 363         /*
 364          * Below methods are called without any exclusive locks.
 365          * More than one net may be constructed and destructed
 366          * in parallel on several cpus. Every pernet_operations
 367          * have to keep in mind all other pernet_operations and
 368          * to introduce a locking, if they share common resources.
 369          *
 370          * The only time they are called with exclusive lock is
 371          * from register_pernet_subsys(), unregister_pernet_subsys()
 372          * register_pernet_device() and unregister_pernet_device().
 373          *
 374          * Exit methods using blocking RCU primitives, such as
 375          * synchronize_rcu(), should be implemented via exit_batch.
 376          * Then, destruction of a group of net requires single
 377          * synchronize_rcu() related to these pernet_operations,
 378          * instead of separate synchronize_rcu() for every net.
 379          * Please, avoid synchronize_rcu() at all, where it's possible.
 380          *
 381          * Note that a combination of pre_exit() and exit() can
 382          * be used, since a synchronize_rcu() is guaranteed between
 383          * the calls.
 384          */
 385         int (*init)(struct net *net);
 386         void (*pre_exit)(struct net *net);
 387         void (*exit)(struct net *net);
 388         void (*exit_batch)(struct list_head *net_exit_list);
 389         unsigned int *id;
 390         size_t size;
 391 };
 392
 393 /*
 394  * Use these carefully.  If you implement a network device and it
 395  * needs per network namespace operations use device pernet operations,
 396  * otherwise use pernet subsys operations.
 397  *
 398  * Network interfaces need to be removed from a dying netns _before_
 399  * subsys notifiers can be called, as most of the network code cleanup
 400  * (which is done from subsys notifiers) runs with the assumption that
 401  * dev_remove_pack has been called so no new packets will arrive during
 402  * and after the cleanup functions have been called.  dev_remove_pack
 403  * is not per namespace so instead the guarantee of no more packets
 404  * arriving in a network namespace is provided by ensuring that all
 405  * network devices and all sockets have left the network namespace
 406  * before the cleanup methods are called.
 407  *
 408  * For the longest time the ipv4 icmp code was registered as a pernet
 409  * device which caused kernel oops, and panics during network
 410  * namespace cleanup.   So please don't get this wrong.
 411  */
 412 int register_pernet_subsys(struct pernet_operations *);
 413 void unregister_pernet_subsys(struct pernet_operations *);
 414 int register_pernet_device(struct pernet_operations *);
 415 void unregister_pernet_device(struct pernet_operations *);
 416
 417 struct ctl_table;
 418 struct ctl_table_header;
 419
 420 #ifdef CONFIG_SYSCTL
 421 int net_sysctl_init(void);
 422 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
 423                                              struct ctl_table *table);
 424 void unregister_net_sysctl_table(struct ctl_table_header *header);
 425 #else
 426 static inline int net_sysctl_init(void) { return 0; }
 427 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
 428         const char *path, struct ctl_table *table)
 429 {
 430         return NULL;
 431 }
 432 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
 433 {
 434 }
 435 #endif
 436
 437 static inline int rt_genid_ipv4(const struct net *net)
 438 {
 439         return atomic_read(&net->ipv4.rt_genid);
 440 }
 441
 442 #if IS_ENABLED(CONFIG_IPV6)
 443 static inline int rt_genid_ipv6(const struct net *net)
 444 {
 445         return atomic_read(&net->ipv6.fib6_sernum);
 446 }
 447 #endif
 448
 449 static inline void rt_genid_bump_ipv4(struct net *net)
 450 {
 451         atomic_inc(&net->ipv4.rt_genid);
 452 }
 453
 454 extern void (*__fib6_flush_trees)(struct net *net);
 455 static inline void rt_genid_bump_ipv6(struct net *net)
 456 {
 457         if (__fib6_flush_trees)
 458                 __fib6_flush_trees(net);
 459 }
 460
 461 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
 462 static inline struct netns_ieee802154_lowpan *
 463 net_ieee802154_lowpan(struct net *net)
 464 {
 465         return &net->ieee802154_lowpan;
 466 }
 467 #endif
 468
 469 /* For callers who don't really care about whether it's IPv4 or IPv6 */
 470 static inline void rt_genid_bump_all(struct net *net)
 471 {
 472         rt_genid_bump_ipv4(net);
 473         rt_genid_bump_ipv6(net);
 474 }
 475
 476 static inline int fnhe_genid(const struct net *net)
 477 {
 478         return atomic_read(&net->fnhe_genid);
 479 }
 480
 481 static inline void fnhe_genid_bump(struct net *net)
 482 {
 483         atomic_inc(&net->fnhe_genid);
 484 }
 485
 486 #endif /* __NET_NET_NAMESPACE_H */