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         struct ctl_table_header *nf_frag_frags_hdr;
 132 #endif
 133         struct sock             *nfnl;
 134         struct sock             *nfnl_stash;
 135 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
 136         struct list_head        nfnl_acct_list;
 137 #endif
 138 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
 139         struct list_head        nfct_timeout_list;
 140 #endif
 141 #endif
 142 #ifdef CONFIG_WEXT_CORE
 143         struct sk_buff_head     wext_nlevents;
 144 #endif
 145         struct net_generic __rcu        *gen;
 146
 147         /* Note : following structs are cache line aligned */
 148 #ifdef CONFIG_XFRM
 149         struct netns_xfrm       xfrm;
 150 #endif
 151 #if IS_ENABLED(CONFIG_IP_VS)
 152         struct netns_ipvs       *ipvs;
 153 #endif
 154 #if IS_ENABLED(CONFIG_MPLS)
 155         struct netns_mpls       mpls;
 156 #endif
 157 #if IS_ENABLED(CONFIG_CAN)
 158         struct netns_can        can;
 159 #endif
 160         struct sock             *diag_nlsk;
 161         atomic_t                fnhe_genid;
 162 } __randomize_layout;
 163
 164 #include <linux/seq_file_net.h>
 165
 166 /* Init's network namespace */
 167 extern struct net init_net;
 168
 169 #ifdef CONFIG_NET_NS
 170 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
 171                         struct net *old_net);
 172
 173 void net_ns_barrier(void);
 174 #else /* CONFIG_NET_NS */
 175 #include <linux/sched.h>
 176 #include <linux/nsproxy.h>
 177 static inline struct net *copy_net_ns(unsigned long flags,
 178         struct user_namespace *user_ns, struct net *old_net)
 179 {
 180         if (flags & CLONE_NEWNET)
 181                 return ERR_PTR(-EINVAL);
 182         return old_net;
 183 }
 184
 185 static inline void net_ns_barrier(void) {}
 186 #endif /* CONFIG_NET_NS */
 187
 188
 189 extern struct list_head net_namespace_list;
 190
 191 struct net *get_net_ns_by_pid(pid_t pid);
 192 struct net *get_net_ns_by_fd(int fd);
 193
 194 #ifdef CONFIG_SYSCTL
 195 void ipx_register_sysctl(void);
 196 void ipx_unregister_sysctl(void);
 197 #else
 198 #define ipx_register_sysctl()
 199 #define ipx_unregister_sysctl()
 200 #endif
 201
 202 #ifdef CONFIG_NET_NS
 203 void __put_net(struct net *net);
 204
 205 static inline struct net *get_net(struct net *net)
 206 {
 207         refcount_inc(&net->count);
 208         return net;
 209 }
 210
 211 static inline struct net *maybe_get_net(struct net *net)
 212 {
 213         /* Used when we know struct net exists but we
 214          * aren't guaranteed a previous reference count
 215          * exists.  If the reference count is zero this
 216          * function fails and returns NULL.
 217          */
 218         if (!refcount_inc_not_zero(&net->count))
 219                 net = NULL;
 220         return net;
 221 }
 222
 223 static inline void put_net(struct net *net)
 224 {
 225         if (refcount_dec_and_test(&net->count))
 226                 __put_net(net);
 227 }
 228
 229 static inline
 230 int net_eq(const struct net *net1, const struct net *net2)
 231 {
 232         return net1 == net2;
 233 }
 234
 235 static inline int check_net(const struct net *net)
 236 {
 237         return refcount_read(&net->count) != 0;
 238 }
 239
 240 void net_drop_ns(void *);
 241
 242 #else
 243
 244 static inline struct net *get_net(struct net *net)
 245 {
 246         return net;
 247 }
 248
 249 static inline void put_net(struct net *net)
 250 {
 251 }
 252
 253 static inline struct net *maybe_get_net(struct net *net)
 254 {
 255         return net;
 256 }
 257
 258 static inline
 259 int net_eq(const struct net *net1, const struct net *net2)
 260 {
 261         return 1;
 262 }
 263
 264 static inline int check_net(const struct net *net)
 265 {
 266         return 1;
 267 }
 268
 269 #define net_drop_ns NULL
 270 #endif
 271
 272
 273 typedef struct {
 274 #ifdef CONFIG_NET_NS
 275         struct net *net;
 276 #endif
 277 } possible_net_t;
 278
 279 static inline void write_pnet(possible_net_t *pnet, struct net *net)
 280 {
 281 #ifdef CONFIG_NET_NS
 282         pnet->net = net;
 283 #endif
 284 }
 285
 286 static inline struct net *read_pnet(const possible_net_t *pnet)
 287 {
 288 #ifdef CONFIG_NET_NS
 289         return pnet->net;
 290 #else
 291         return &init_net;
 292 #endif
 293 }
 294
 295 /* Protected by net_rwsem */
 296 #define for_each_net(VAR)                               \
 297         list_for_each_entry(VAR, &net_namespace_list, list)
 298
 299 #define for_each_net_rcu(VAR)                           \
 300         list_for_each_entry_rcu(VAR, &net_namespace_list, list)
 301
 302 #ifdef CONFIG_NET_NS
 303 #define __net_init
 304 #define __net_exit
 305 #define __net_initdata
 306 #define __net_initconst
 307 #else
 308 #define __net_init      __init
 309 #define __net_exit      __ref
 310 #define __net_initdata  __initdata
 311 #define __net_initconst __initconst
 312 #endif
 313
 314 int peernet2id_alloc(struct net *net, struct net *peer);
 315 int peernet2id(struct net *net, struct net *peer);
 316 bool peernet_has_id(struct net *net, struct net *peer);
 317 struct net *get_net_ns_by_id(struct net *net, int id);
 318
 319 struct pernet_operations {
 320         struct list_head list;
 321         /*
 322          * Below methods are called without any exclusive locks.
 323          * More than one net may be constructed and destructed
 324          * in parallel on several cpus. Every pernet_operations
 325          * have to keep in mind all other pernet_operations and
 326          * to introduce a locking, if they share common resources.
 327          *
 328          * The only time they are called with exclusive lock is
 329          * from register_pernet_subsys(), unregister_pernet_subsys()
 330          * register_pernet_device() and unregister_pernet_device().
 331          *
 332          * Exit methods using blocking RCU primitives, such as
 333          * synchronize_rcu(), should be implemented via exit_batch.
 334          * Then, destruction of a group of net requires single
 335          * synchronize_rcu() related to these pernet_operations,
 336          * instead of separate synchronize_rcu() for every net.
 337          * Please, avoid synchronize_rcu() at all, where it's possible.
 338          */
 339         int (*init)(struct net *net);
 340         void (*exit)(struct net *net);
 341         void (*exit_batch)(struct list_head *net_exit_list);
 342         unsigned int *id;
 343         size_t size;
 344 };
 345
 346 /*
 347  * Use these carefully.  If you implement a network device and it
 348  * needs per network namespace operations use device pernet operations,
 349  * otherwise use pernet subsys operations.
 350  *
 351  * Network interfaces need to be removed from a dying netns _before_
 352  * subsys notifiers can be called, as most of the network code cleanup
 353  * (which is done from subsys notifiers) runs with the assumption that
 354  * dev_remove_pack has been called so no new packets will arrive during
 355  * and after the cleanup functions have been called.  dev_remove_pack
 356  * is not per namespace so instead the guarantee of no more packets
 357  * arriving in a network namespace is provided by ensuring that all
 358  * network devices and all sockets have left the network namespace
 359  * before the cleanup methods are called.
 360  *
 361  * For the longest time the ipv4 icmp code was registered as a pernet
 362  * device which caused kernel oops, and panics during network
 363  * namespace cleanup.   So please don't get this wrong.
 364  */
 365 int register_pernet_subsys(struct pernet_operations *);
 366 void unregister_pernet_subsys(struct pernet_operations *);
 367 int register_pernet_device(struct pernet_operations *);
 368 void unregister_pernet_device(struct pernet_operations *);
 369
 370 struct ctl_table;
 371 struct ctl_table_header;
 372
 373 #ifdef CONFIG_SYSCTL
 374 int net_sysctl_init(void);
 375 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
 376                                              struct ctl_table *table);
 377 void unregister_net_sysctl_table(struct ctl_table_header *header);
 378 #else
 379 static inline int net_sysctl_init(void) { return 0; }
 380 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
 381         const char *path, struct ctl_table *table)
 382 {
 383         return NULL;
 384 }
 385 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
 386 {
 387 }
 388 #endif
 389
 390 static inline int rt_genid_ipv4(struct net *net)
 391 {
 392         return atomic_read(&net->ipv4.rt_genid);
 393 }
 394
 395 static inline void rt_genid_bump_ipv4(struct net *net)
 396 {
 397         atomic_inc(&net->ipv4.rt_genid);
 398 }
 399
 400 extern void (*__fib6_flush_trees)(struct net *net);
 401 static inline void rt_genid_bump_ipv6(struct net *net)
 402 {
 403         if (__fib6_flush_trees)
 404                 __fib6_flush_trees(net);
 405 }
 406
 407 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
 408 static inline struct netns_ieee802154_lowpan *
 409 net_ieee802154_lowpan(struct net *net)
 410 {
 411         return &net->ieee802154_lowpan;
 412 }
 413 #endif
 414
 415 /* For callers who don't really care about whether it's IPv4 or IPv6 */
 416 static inline void rt_genid_bump_all(struct net *net)
 417 {
 418         rt_genid_bump_ipv4(net);
 419         rt_genid_bump_ipv6(net);
 420 }
 421
 422 static inline int fnhe_genid(struct net *net)
 423 {
 424         return atomic_read(&net->fnhe_genid);
 425 }
 426
 427 static inline void fnhe_genid_bump(struct net *net)
 428 {
 429         atomic_inc(&net->fnhe_genid);
 430 }
 431
 432 #endif /* __NET_NET_NAMESPACE_H */