include/net/net_namespace.h

   1 /*
   2  * Operations on the network namespace
   3  */
   4 #ifndef __NET_NET_NAMESPACE_H
   5 #define __NET_NET_NAMESPACE_H
   6
   7 #include <linux/atomic.h>
   8 #include <linux/workqueue.h>
   9 #include <linux/list.h>
  10 #include <linux/sysctl.h>
  11
  12 #include <net/netns/core.h>
  13 #include <net/netns/mib.h>
  14 #include <net/netns/unix.h>
  15 #include <net/netns/packet.h>
  16 #include <net/netns/ipv4.h>
  17 #include <net/netns/ipv6.h>
  18 #include <net/netns/sctp.h>
  19 #include <net/netns/dccp.h>
  20 #include <net/netns/netfilter.h>
  21 #include <net/netns/x_tables.h>
  22 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  23 #include <net/netns/conntrack.h>
  24 #endif
  25 #include <net/netns/xfrm.h>
  26
  27 struct user_namespace;
  28 struct proc_dir_entry;
  29 struct net_device;
  30 struct sock;
  31 struct ctl_table_header;
  32 struct net_generic;
  33 struct sock;
  34 struct netns_ipvs;
  35
  36
  37 #define NETDEV_HASHBITS    8
  38 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
  39
  40 struct net {
  41         atomic_t                passive;        /* To decided when the network
  42                                                  * namespace should be freed.
  43                                                  */
  44         atomic_t                count;          /* To decided when the network
  45                                                  *  namespace should be shut down.
  46                                                  */
  47 #ifdef NETNS_REFCNT_DEBUG
  48         atomic_t                use_count;      /* To track references we
  49                                                  * destroy on demand
  50                                                  */
  51 #endif
  52         spinlock_t              rules_mod_lock;
  53
  54         struct list_head        list;           /* list of network namespaces */
  55         struct list_head        cleanup_list;   /* namespaces on death row */
  56         struct list_head        exit_list;      /* Use only net_mutex */
  57
  58         struct user_namespace   *user_ns;       /* Owning user namespace */
  59
  60         unsigned int            proc_inum;
  61
  62         struct proc_dir_entry   *proc_net;
  63         struct proc_dir_entry   *proc_net_stat;
  64
  65 #ifdef CONFIG_SYSCTL
  66         struct ctl_table_set    sysctls;
  67 #endif
  68
  69         struct sock             *rtnl;                  /* rtnetlink socket */
  70         struct sock             *genl_sock;
  71
  72         struct list_head        dev_base_head;
  73         struct hlist_head       *dev_name_head;
  74         struct hlist_head       *dev_index_head;
  75         unsigned int            dev_base_seq;   /* protected by rtnl_mutex */
  76         int                     ifindex;
  77
  78         /* core fib_rules */
  79         struct list_head        rules_ops;
  80
  81
  82         struct net_device       *loopback_dev;          /* The loopback */
  83         struct netns_core       core;
  84         struct netns_mib        mib;
  85         struct netns_packet     packet;
  86         struct netns_unix       unx;
  87         struct netns_ipv4       ipv4;
  88 #if IS_ENABLED(CONFIG_IPV6)
  89         struct netns_ipv6       ipv6;
  90 #endif
  91 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
  92         struct netns_sctp       sctp;
  93 #endif
  94 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
  95         struct netns_dccp       dccp;
  96 #endif
  97 #ifdef CONFIG_NETFILTER
  98         struct netns_nf         nf;
  99         struct netns_xt         xt;
 100 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 101         struct netns_ct         ct;
 102 #endif
 103 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
 104         struct netns_nf_frag    nf_frag;
 105 #endif
 106         struct sock             *nfnl;
 107         struct sock             *nfnl_stash;
 108 #endif
 109 #ifdef CONFIG_WEXT_CORE
 110         struct sk_buff_head     wext_nlevents;
 111 #endif
 112         struct net_generic __rcu        *gen;
 113
 114         /* Note : following structs are cache line aligned */
 115 #ifdef CONFIG_XFRM
 116         struct netns_xfrm       xfrm;
 117 #endif
 118 #if IS_ENABLED(CONFIG_IP_VS)
 119         struct netns_ipvs       *ipvs;
 120 #endif
 121         struct sock             *diag_nlsk;
 122         atomic_t                fnhe_genid;
 123 };
 124
 125 /*
 126  * ifindex generation is per-net namespace, and loopback is
 127  * always the 1st device in ns (see net_dev_init), thus any
 128  * loopback device should get ifindex 1
 129  */
 130
 131 #define LOOPBACK_IFINDEX        1
 132
 133 #include <linux/seq_file_net.h>
 134
 135 /* Init's network namespace */
 136 extern struct net init_net;
 137
 138 #ifdef CONFIG_NET_NS
 139 extern struct net *copy_net_ns(unsigned long flags,
 140         struct user_namespace *user_ns, struct net *old_net);
 141
 142 #else /* CONFIG_NET_NS */
 143 #include <linux/sched.h>
 144 #include <linux/nsproxy.h>
 145 static inline struct net *copy_net_ns(unsigned long flags,
 146         struct user_namespace *user_ns, struct net *old_net)
 147 {
 148         if (flags & CLONE_NEWNET)
 149                 return ERR_PTR(-EINVAL);
 150         return old_net;
 151 }
 152 #endif /* CONFIG_NET_NS */
 153
 154
 155 extern struct list_head net_namespace_list;
 156
 157 extern struct net *get_net_ns_by_pid(pid_t pid);
 158 extern struct net *get_net_ns_by_fd(int pid);
 159
 160 #ifdef CONFIG_NET_NS
 161 extern void __put_net(struct net *net);
 162
 163 static inline struct net *get_net(struct net *net)
 164 {
 165         atomic_inc(&net->count);
 166         return net;
 167 }
 168
 169 static inline struct net *maybe_get_net(struct net *net)
 170 {
 171         /* Used when we know struct net exists but we
 172          * aren't guaranteed a previous reference count
 173          * exists.  If the reference count is zero this
 174          * function fails and returns NULL.
 175          */
 176         if (!atomic_inc_not_zero(&net->count))
 177                 net = NULL;
 178         return net;
 179 }
 180
 181 static inline void put_net(struct net *net)
 182 {
 183         if (atomic_dec_and_test(&net->count))
 184                 __put_net(net);
 185 }
 186
 187 static inline
 188 int net_eq(const struct net *net1, const struct net *net2)
 189 {
 190         return net1 == net2;
 191 }
 192
 193 extern void net_drop_ns(void *);
 194
 195 #else
 196
 197 static inline struct net *get_net(struct net *net)
 198 {
 199         return net;
 200 }
 201
 202 static inline void put_net(struct net *net)
 203 {
 204 }
 205
 206 static inline struct net *maybe_get_net(struct net *net)
 207 {
 208         return net;
 209 }
 210
 211 static inline
 212 int net_eq(const struct net *net1, const struct net *net2)
 213 {
 214         return 1;
 215 }
 216
 217 #define net_drop_ns NULL
 218 #endif
 219
 220
 221 #ifdef NETNS_REFCNT_DEBUG
 222 static inline struct net *hold_net(struct net *net)
 223 {
 224         if (net)
 225                 atomic_inc(&net->use_count);
 226         return net;
 227 }
 228
 229 static inline void release_net(struct net *net)
 230 {
 231         if (net)
 232                 atomic_dec(&net->use_count);
 233 }
 234 #else
 235 static inline struct net *hold_net(struct net *net)
 236 {
 237         return net;
 238 }
 239
 240 static inline void release_net(struct net *net)
 241 {
 242 }
 243 #endif
 244
 245 #ifdef CONFIG_NET_NS
 246
 247 static inline void write_pnet(struct net **pnet, struct net *net)
 248 {
 249         *pnet = net;
 250 }
 251
 252 static inline struct net *read_pnet(struct net * const *pnet)
 253 {
 254         return *pnet;
 255 }
 256
 257 #else
 258
 259 #define write_pnet(pnet, net)   do { (void)(net);} while (0)
 260 #define read_pnet(pnet)         (&init_net)
 261
 262 #endif
 263
 264 #define for_each_net(VAR)                               \
 265         list_for_each_entry(VAR, &net_namespace_list, list)
 266
 267 #define for_each_net_rcu(VAR)                           \
 268         list_for_each_entry_rcu(VAR, &net_namespace_list, list)
 269
 270 #ifdef CONFIG_NET_NS
 271 #define __net_init
 272 #define __net_exit
 273 #define __net_initdata
 274 #define __net_initconst
 275 #else
 276 #define __net_init      __init
 277 #define __net_exit      __exit_refok
 278 #define __net_initdata  __initdata
 279 #define __net_initconst __initconst
 280 #endif
 281
 282 struct pernet_operations {
 283         struct list_head list;
 284         int (*init)(struct net *net);
 285         void (*exit)(struct net *net);
 286         void (*exit_batch)(struct list_head *net_exit_list);
 287         int *id;
 288         size_t size;
 289 };
 290
 291 /*
 292  * Use these carefully.  If you implement a network device and it
 293  * needs per network namespace operations use device pernet operations,
 294  * otherwise use pernet subsys operations.
 295  *
 296  * Network interfaces need to be removed from a dying netns _before_
 297  * subsys notifiers can be called, as most of the network code cleanup
 298  * (which is done from subsys notifiers) runs with the assumption that
 299  * dev_remove_pack has been called so no new packets will arrive during
 300  * and after the cleanup functions have been called.  dev_remove_pack
 301  * is not per namespace so instead the guarantee of no more packets
 302  * arriving in a network namespace is provided by ensuring that all
 303  * network devices and all sockets have left the network namespace
 304  * before the cleanup methods are called.
 305  *
 306  * For the longest time the ipv4 icmp code was registered as a pernet
 307  * device which caused kernel oops, and panics during network
 308  * namespace cleanup.   So please don't get this wrong.
 309  */
 310 extern int register_pernet_subsys(struct pernet_operations *);
 311 extern void unregister_pernet_subsys(struct pernet_operations *);
 312 extern int register_pernet_device(struct pernet_operations *);
 313 extern void unregister_pernet_device(struct pernet_operations *);
 314
 315 struct ctl_table;
 316 struct ctl_table_header;
 317
 318 #ifdef CONFIG_SYSCTL
 319 extern int net_sysctl_init(void);
 320 extern struct ctl_table_header *register_net_sysctl(struct net *net,
 321         const char *path, struct ctl_table *table);
 322 extern void unregister_net_sysctl_table(struct ctl_table_header *header);
 323 #else
 324 static inline int net_sysctl_init(void) { return 0; }
 325 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
 326         const char *path, struct ctl_table *table)
 327 {
 328         return NULL;
 329 }
 330 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
 331 {
 332 }
 333 #endif
 334
 335 static inline int rt_genid_ipv4(struct net *net)
 336 {
 337         return atomic_read(&net->ipv4.rt_genid);
 338 }
 339
 340 static inline void rt_genid_bump_ipv4(struct net *net)
 341 {
 342         atomic_inc(&net->ipv4.rt_genid);
 343 }
 344
 345 #if IS_ENABLED(CONFIG_IPV6)
 346 static inline int rt_genid_ipv6(struct net *net)
 347 {
 348         return atomic_read(&net->ipv6.rt_genid);
 349 }
 350
 351 static inline void rt_genid_bump_ipv6(struct net *net)
 352 {
 353         atomic_inc(&net->ipv6.rt_genid);
 354 }
 355 #else
 356 static inline int rt_genid_ipv6(struct net *net)
 357 {
 358         return 0;
 359 }
 360
 361 static inline void rt_genid_bump_ipv6(struct net *net)
 362 {
 363 }
 364 #endif
 365
 366 /* For callers who don't really care about whether it's IPv4 or IPv6 */
 367 static inline void rt_genid_bump_all(struct net *net)
 368 {
 369         rt_genid_bump_ipv4(net);
 370         rt_genid_bump_ipv6(net);
 371 }
 372
 373 static inline int fnhe_genid(struct net *net)
 374 {
 375         return atomic_read(&net->fnhe_genid);
 376 }
 377
 378 static inline void fnhe_genid_bump(struct net *net)
 379 {
 380         atomic_inc(&net->fnhe_genid);
 381 }
 382
 383 #endif /* __NET_NET_NAMESPACE_H */