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