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
207 struct ns_common *get_net_ns(struct ns_common *ns);
208 #else /* CONFIG_NET_NS */
209 #include <linux/sched.h>
210 #include <linux/nsproxy.h>
copy_net_ns(unsigned long flags,struct user_namespace * user_ns,struct net * old_net)211 static inline struct net *copy_net_ns(unsigned long flags,
212 struct user_namespace *user_ns, struct net *old_net)
213 {
214 if (flags & CLONE_NEWNET)
215 return ERR_PTR(-EINVAL);
216 return old_net;
217 }
218
net_ns_get_ownership(const struct net * net,kuid_t * uid,kgid_t * gid)219 static inline void net_ns_get_ownership(const struct net *net,
220 kuid_t *uid, kgid_t *gid)
221 {
222 *uid = GLOBAL_ROOT_UID;
223 *gid = GLOBAL_ROOT_GID;
224 }
225
net_ns_barrier(void)226 static inline void net_ns_barrier(void) {}
227
get_net_ns(struct ns_common * ns)228 static inline struct ns_common *get_net_ns(struct ns_common *ns)
229 {
230 return ERR_PTR(-EINVAL);
231 }
232 #endif /* CONFIG_NET_NS */
233
234
235 extern struct list_head net_namespace_list;
236
237 struct net *get_net_ns_by_pid(pid_t pid);
238 struct net *get_net_ns_by_fd(int fd);
239
240 u64 __net_gen_cookie(struct net *net);
241
242 #ifdef CONFIG_SYSCTL
243 void ipx_register_sysctl(void);
244 void ipx_unregister_sysctl(void);
245 #else
246 #define ipx_register_sysctl()
247 #define ipx_unregister_sysctl()
248 #endif
249
250 #ifdef CONFIG_NET_NS
251 void __put_net(struct net *net);
252
get_net(struct net * net)253 static inline struct net *get_net(struct net *net)
254 {
255 refcount_inc(&net->count);
256 return net;
257 }
258
maybe_get_net(struct net * net)259 static inline struct net *maybe_get_net(struct net *net)
260 {
261 /* Used when we know struct net exists but we
262 * aren't guaranteed a previous reference count
263 * exists. If the reference count is zero this
264 * function fails and returns NULL.
265 */
266 if (!refcount_inc_not_zero(&net->count))
267 net = NULL;
268 return net;
269 }
270
put_net(struct net * net)271 static inline void put_net(struct net *net)
272 {
273 if (refcount_dec_and_test(&net->count))
274 __put_net(net);
275 }
276
277 static inline
net_eq(const struct net * net1,const struct net * net2)278 int net_eq(const struct net *net1, const struct net *net2)
279 {
280 return net1 == net2;
281 }
282
check_net(const struct net * net)283 static inline int check_net(const struct net *net)
284 {
285 return refcount_read(&net->count) != 0;
286 }
287
288 void net_drop_ns(void *);
289
290 #else
291
get_net(struct net * net)292 static inline struct net *get_net(struct net *net)
293 {
294 return net;
295 }
296
put_net(struct net * net)297 static inline void put_net(struct net *net)
298 {
299 }
300
maybe_get_net(struct net * net)301 static inline struct net *maybe_get_net(struct net *net)
302 {
303 return net;
304 }
305
306 static inline
net_eq(const struct net * net1,const struct net * net2)307 int net_eq(const struct net *net1, const struct net *net2)
308 {
309 return 1;
310 }
311
check_net(const struct net * net)312 static inline int check_net(const struct net *net)
313 {
314 return 1;
315 }
316
317 #define net_drop_ns NULL
318 #endif
319
320
321 typedef struct {
322 #ifdef CONFIG_NET_NS
323 struct net *net;
324 #endif
325 } possible_net_t;
326
write_pnet(possible_net_t * pnet,struct net * net)327 static inline void write_pnet(possible_net_t *pnet, struct net *net)
328 {
329 #ifdef CONFIG_NET_NS
330 pnet->net = net;
331 #endif
332 }
333
read_pnet(const possible_net_t * pnet)334 static inline struct net *read_pnet(const possible_net_t *pnet)
335 {
336 #ifdef CONFIG_NET_NS
337 return pnet->net;
338 #else
339 return &init_net;
340 #endif
341 }
342
343 /* Protected by net_rwsem */
344 #define for_each_net(VAR) \
345 list_for_each_entry(VAR, &net_namespace_list, list)
346 #define for_each_net_continue_reverse(VAR) \
347 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
348 #define for_each_net_rcu(VAR) \
349 list_for_each_entry_rcu(VAR, &net_namespace_list, list)
350
351 #ifdef CONFIG_NET_NS
352 #define __net_init
353 #define __net_exit
354 #define __net_initdata
355 #define __net_initconst
356 #else
357 #define __net_init __init
358 #define __net_exit __ref
359 #define __net_initdata __initdata
360 #define __net_initconst __initconst
361 #endif
362
363 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
364 int peernet2id(const struct net *net, struct net *peer);
365 bool peernet_has_id(const struct net *net, struct net *peer);
366 struct net *get_net_ns_by_id(const struct net *net, int id);
367
368 struct pernet_operations {
369 struct list_head list;
370 /*
371 * Below methods are called without any exclusive locks.
372 * More than one net may be constructed and destructed
373 * in parallel on several cpus. Every pernet_operations
374 * have to keep in mind all other pernet_operations and
375 * to introduce a locking, if they share common resources.
376 *
377 * The only time they are called with exclusive lock is
378 * from register_pernet_subsys(), unregister_pernet_subsys()
379 * register_pernet_device() and unregister_pernet_device().
380 *
381 * Exit methods using blocking RCU primitives, such as
382 * synchronize_rcu(), should be implemented via exit_batch.
383 * Then, destruction of a group of net requires single
384 * synchronize_rcu() related to these pernet_operations,
385 * instead of separate synchronize_rcu() for every net.
386 * Please, avoid synchronize_rcu() at all, where it's possible.
387 *
388 * Note that a combination of pre_exit() and exit() can
389 * be used, since a synchronize_rcu() is guaranteed between
390 * the calls.
391 */
392 int (*init)(struct net *net);
393 void (*pre_exit)(struct net *net);
394 void (*exit)(struct net *net);
395 void (*exit_batch)(struct list_head *net_exit_list);
396 unsigned int *id;
397 size_t size;
398 };
399
400 /*
401 * Use these carefully. If you implement a network device and it
402 * needs per network namespace operations use device pernet operations,
403 * otherwise use pernet subsys operations.
404 *
405 * Network interfaces need to be removed from a dying netns _before_
406 * subsys notifiers can be called, as most of the network code cleanup
407 * (which is done from subsys notifiers) runs with the assumption that
408 * dev_remove_pack has been called so no new packets will arrive during
409 * and after the cleanup functions have been called. dev_remove_pack
410 * is not per namespace so instead the guarantee of no more packets
411 * arriving in a network namespace is provided by ensuring that all
412 * network devices and all sockets have left the network namespace
413 * before the cleanup methods are called.
414 *
415 * For the longest time the ipv4 icmp code was registered as a pernet
416 * device which caused kernel oops, and panics during network
417 * namespace cleanup. So please don't get this wrong.
418 */
419 int register_pernet_subsys(struct pernet_operations *);
420 void unregister_pernet_subsys(struct pernet_operations *);
421 int register_pernet_device(struct pernet_operations *);
422 void unregister_pernet_device(struct pernet_operations *);
423
424 struct ctl_table;
425 struct ctl_table_header;
426
427 #ifdef CONFIG_SYSCTL
428 int net_sysctl_init(void);
429 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
430 struct ctl_table *table);
431 void unregister_net_sysctl_table(struct ctl_table_header *header);
432 #else
net_sysctl_init(void)433 static inline int net_sysctl_init(void) { return 0; }
register_net_sysctl(struct net * net,const char * path,struct ctl_table * table)434 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
435 const char *path, struct ctl_table *table)
436 {
437 return NULL;
438 }
unregister_net_sysctl_table(struct ctl_table_header * header)439 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
440 {
441 }
442 #endif
443
rt_genid_ipv4(const struct net * net)444 static inline int rt_genid_ipv4(const struct net *net)
445 {
446 return atomic_read(&net->ipv4.rt_genid);
447 }
448
449 #if IS_ENABLED(CONFIG_IPV6)
rt_genid_ipv6(const struct net * net)450 static inline int rt_genid_ipv6(const struct net *net)
451 {
452 return atomic_read(&net->ipv6.fib6_sernum);
453 }
454 #endif
455
rt_genid_bump_ipv4(struct net * net)456 static inline void rt_genid_bump_ipv4(struct net *net)
457 {
458 atomic_inc(&net->ipv4.rt_genid);
459 }
460
461 extern void (*__fib6_flush_trees)(struct net *net);
rt_genid_bump_ipv6(struct net * net)462 static inline void rt_genid_bump_ipv6(struct net *net)
463 {
464 if (__fib6_flush_trees)
465 __fib6_flush_trees(net);
466 }
467
468 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
469 static inline struct netns_ieee802154_lowpan *
net_ieee802154_lowpan(struct net * net)470 net_ieee802154_lowpan(struct net *net)
471 {
472 return &net->ieee802154_lowpan;
473 }
474 #endif
475
476 /* For callers who don't really care about whether it's IPv4 or IPv6 */
rt_genid_bump_all(struct net * net)477 static inline void rt_genid_bump_all(struct net *net)
478 {
479 rt_genid_bump_ipv4(net);
480 rt_genid_bump_ipv6(net);
481 }
482
fnhe_genid(const struct net * net)483 static inline int fnhe_genid(const struct net *net)
484 {
485 return atomic_read(&net->fnhe_genid);
486 }
487
fnhe_genid_bump(struct net * net)488 static inline void fnhe_genid_bump(struct net *net)
489 {
490 atomic_inc(&net->fnhe_genid);
491 }
492
493 #endif /* __NET_NET_NAMESPACE_H */
494