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/netfilter.h>
26 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
27 #include <net/netns/conntrack.h>
28 #endif
29 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
30 #include <net/netns/flow_table.h>
31 #endif
32 #include <net/netns/nftables.h>
33 #include <net/netns/xfrm.h>
34 #include <net/netns/mpls.h>
35 #include <net/netns/can.h>
36 #include <net/netns/xdp.h>
37 #include <net/netns/smc.h>
38 #include <net/netns/bpf.h>
39 #include <net/netns/mctp.h>
40 #include <net/netns/vsock.h>
41 #include <net/net_trackers.h>
42 #include <linux/ns_common.h>
43 #include <linux/idr.h>
44 #include <linux/skbuff.h>
45 #include <linux/notifier.h>
46 #include <linux/xarray.h>
47
48 struct user_namespace;
49 struct proc_dir_entry;
50 struct net_device;
51 struct sock;
52 struct ctl_table_header;
53 struct net_generic;
54 struct uevent_sock;
55 struct netns_ipvs;
56 struct bpf_prog;
57
58
59 #define NETDEV_HASHBITS 8
60 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
61
62 struct net {
63 /* First cache line can be often dirtied.
64 * Do not place here read-mostly fields.
65 */
66 refcount_t passive; /* To decide when the network
67 * namespace should be freed.
68 */
69 spinlock_t rules_mod_lock;
70
71 unsigned int dev_base_seq; /* protected by rtnl_mutex */
72 u32 ifindex;
73
74 spinlock_t nsid_lock;
75 atomic_t fnhe_genid;
76
77 struct list_head list; /* list of network namespaces */
78 struct list_head exit_list; /* To linked to call pernet exit
79 * methods on dead net (
80 * pernet_ops_rwsem read locked),
81 * or to unregister pernet ops
82 * (pernet_ops_rwsem write locked).
83 */
84 struct llist_node defer_free_list;
85 struct llist_node cleanup_list; /* namespaces on death row */
86
87 struct list_head ptype_all;
88 struct list_head ptype_specific;
89
90 #ifdef CONFIG_KEYS
91 struct key_tag *key_domain; /* Key domain of operation tag */
92 #endif
93 struct user_namespace *user_ns; /* Owning user namespace */
94 struct ucounts *ucounts;
95 struct idr netns_ids;
96
97 struct ns_common ns;
98 struct ref_tracker_dir refcnt_tracker;
99 struct ref_tracker_dir notrefcnt_tracker; /* tracker for objects not
100 * refcounted against netns
101 */
102 struct list_head dev_base_head;
103 struct proc_dir_entry *proc_net;
104 struct proc_dir_entry *proc_net_stat;
105
106 #ifdef CONFIG_SYSCTL
107 struct ctl_table_set sysctls;
108 #endif
109
110 struct sock *rtnl; /* rtnetlink socket */
111 struct sock *genl_sock;
112
113 struct uevent_sock *uevent_sock; /* uevent socket */
114
115 struct hlist_head *dev_name_head;
116 struct hlist_head *dev_index_head;
117 struct xarray dev_by_index;
118 struct raw_notifier_head netdev_chain;
119
120 /* Note that @hash_mix can be read millions times per second,
121 * it is critical that it is on a read_mostly cache line.
122 */
123 u32 hash_mix;
124 bool is_dying;
125
126 struct net_device *loopback_dev; /* The loopback */
127
128 /* core fib_rules */
129 struct list_head rules_ops;
130
131 struct netns_core core;
132 struct netns_mib mib;
133 struct netns_packet packet;
134 #if IS_ENABLED(CONFIG_UNIX)
135 struct netns_unix unx;
136 #endif
137 struct netns_nexthop nexthop;
138 struct netns_ipv4 ipv4;
139 #if IS_ENABLED(CONFIG_IPV6)
140 struct netns_ipv6 ipv6;
141 #endif
142 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
143 struct netns_ieee802154_lowpan ieee802154_lowpan;
144 #endif
145 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
146 struct netns_sctp sctp;
147 #endif
148 #ifdef CONFIG_NETFILTER
149 struct netns_nf nf;
150 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
151 struct netns_ct ct;
152 #endif
153 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
154 struct netns_nftables nft;
155 #endif
156 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
157 struct netns_ft ft;
158 #endif
159 #endif
160 #ifdef CONFIG_WEXT_CORE
161 struct sk_buff_head wext_nlevents;
162 #endif
163 struct net_generic __rcu *gen;
164
165 /* Used to store attached BPF programs */
166 struct netns_bpf bpf;
167
168 /* Note : following structs are cache line aligned */
169 #ifdef CONFIG_XFRM
170 struct netns_xfrm xfrm;
171 #endif
172
173 u64 net_cookie; /* written once */
174
175 #if IS_ENABLED(CONFIG_IP_VS)
176 struct netns_ipvs *ipvs;
177 #endif
178 #if IS_ENABLED(CONFIG_MPLS)
179 struct netns_mpls mpls;
180 #endif
181 #if IS_ENABLED(CONFIG_CAN)
182 struct netns_can can;
183 #endif
184 #ifdef CONFIG_XDP_SOCKETS
185 struct netns_xdp xdp;
186 #endif
187 #if IS_ENABLED(CONFIG_MCTP)
188 struct netns_mctp mctp;
189 #endif
190 #if IS_ENABLED(CONFIG_CRYPTO_USER)
191 struct sock *crypto_nlsk;
192 #endif
193 struct sock *diag_nlsk;
194 #if IS_ENABLED(CONFIG_SMC)
195 struct netns_smc smc;
196 #endif
197 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL
198 /* Move to a better place when the config guard is removed. */
199 struct mutex rtnl_mutex;
200 #endif
201 #if IS_ENABLED(CONFIG_VSOCKETS)
202 struct netns_vsock vsock;
203 #endif
204 } __randomize_layout;
205
206 #include <linux/seq_file_net.h>
207
208 /* Init's network namespace */
209 extern struct net init_net;
210
211 #ifdef CONFIG_NET_NS
212 struct net *copy_net_ns(u64 flags, struct user_namespace *user_ns,
213 struct net *old_net);
214
215 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
216
217 void net_ns_barrier(void);
218
219 struct ns_common *get_net_ns(struct ns_common *ns);
220 struct net *get_net_ns_by_fd(int fd);
221 extern struct task_struct *cleanup_net_task;
222
223 #else /* CONFIG_NET_NS */
224 #include <linux/sched.h>
225 #include <linux/nsproxy.h>
copy_net_ns(u64 flags,struct user_namespace * user_ns,struct net * old_net)226 static inline struct net *copy_net_ns(u64 flags,
227 struct user_namespace *user_ns, struct net *old_net)
228 {
229 if (flags & CLONE_NEWNET)
230 return ERR_PTR(-EINVAL);
231 return old_net;
232 }
233
net_ns_get_ownership(const struct net * net,kuid_t * uid,kgid_t * gid)234 static inline void net_ns_get_ownership(const struct net *net,
235 kuid_t *uid, kgid_t *gid)
236 {
237 *uid = GLOBAL_ROOT_UID;
238 *gid = GLOBAL_ROOT_GID;
239 }
240
net_ns_barrier(void)241 static inline void net_ns_barrier(void) {}
242
get_net_ns(struct ns_common * ns)243 static inline struct ns_common *get_net_ns(struct ns_common *ns)
244 {
245 return ERR_PTR(-EINVAL);
246 }
247
get_net_ns_by_fd(int fd)248 static inline struct net *get_net_ns_by_fd(int fd)
249 {
250 return ERR_PTR(-EINVAL);
251 }
252 #endif /* CONFIG_NET_NS */
253
254
255 extern struct list_head net_namespace_list;
256
257 struct net *get_net_ns_by_pid(pid_t pid);
258
259 #ifdef CONFIG_SYSCTL
260 void ipx_register_sysctl(void);
261 void ipx_unregister_sysctl(void);
262 #else
263 #define ipx_register_sysctl()
264 #define ipx_unregister_sysctl()
265 #endif
266
267 #ifdef CONFIG_NET_NS
268 void __put_net(struct net *net);
269
to_net_ns(struct ns_common * ns)270 static inline struct net *to_net_ns(struct ns_common *ns)
271 {
272 return container_of(ns, struct net, ns);
273 }
274
275 /* Try using get_net_track() instead */
get_net(struct net * net)276 static inline struct net *get_net(struct net *net)
277 {
278 ns_ref_inc(net);
279 return net;
280 }
281
maybe_get_net(struct net * net)282 static inline struct net *maybe_get_net(struct net *net)
283 {
284 /* Used when we know struct net exists but we
285 * aren't guaranteed a previous reference count
286 * exists. If the reference count is zero this
287 * function fails and returns NULL.
288 */
289 if (!ns_ref_get(net))
290 net = NULL;
291 return net;
292 }
293
294 /* Try using put_net_track() instead */
put_net(struct net * net)295 static inline void put_net(struct net *net)
296 {
297 if (ns_ref_put(net))
298 __put_net(net);
299 }
300
301 static inline
net_eq(const struct net * net1,const struct net * net2)302 int net_eq(const struct net *net1, const struct net *net2)
303 {
304 return net1 == net2;
305 }
306
check_net(const struct net * net)307 static inline int check_net(const struct net *net)
308 {
309 return ns_ref_read(net) != 0;
310 }
311
312 void net_drop_ns(void *);
313 void net_passive_dec(struct net *net);
314
315 #else
316
get_net(struct net * net)317 static inline struct net *get_net(struct net *net)
318 {
319 return net;
320 }
321
put_net(struct net * net)322 static inline void put_net(struct net *net)
323 {
324 }
325
maybe_get_net(struct net * net)326 static inline struct net *maybe_get_net(struct net *net)
327 {
328 return net;
329 }
330
331 static inline
net_eq(const struct net * net1,const struct net * net2)332 int net_eq(const struct net *net1, const struct net *net2)
333 {
334 return 1;
335 }
336
check_net(const struct net * net)337 static inline int check_net(const struct net *net)
338 {
339 return 1;
340 }
341
342 #define net_drop_ns NULL
343
net_passive_dec(struct net * net)344 static inline void net_passive_dec(struct net *net)
345 {
346 refcount_dec(&net->passive);
347 }
348 #endif
349
net_passive_inc(struct net * net)350 static inline void net_passive_inc(struct net *net)
351 {
352 refcount_inc(&net->passive);
353 }
354
355 /* Returns true if the netns initialization is completed successfully */
net_initialized(const struct net * net)356 static inline bool net_initialized(const struct net *net)
357 {
358 return READ_ONCE(net->list.next);
359 }
360
__netns_tracker_alloc(struct net * net,netns_tracker * tracker,bool refcounted,gfp_t gfp)361 static inline void __netns_tracker_alloc(struct net *net,
362 netns_tracker *tracker,
363 bool refcounted,
364 gfp_t gfp)
365 {
366 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
367 ref_tracker_alloc(refcounted ? &net->refcnt_tracker :
368 &net->notrefcnt_tracker,
369 tracker, gfp);
370 #endif
371 }
372
netns_tracker_alloc(struct net * net,netns_tracker * tracker,gfp_t gfp)373 static inline void netns_tracker_alloc(struct net *net, netns_tracker *tracker,
374 gfp_t gfp)
375 {
376 __netns_tracker_alloc(net, tracker, true, gfp);
377 }
378
__netns_tracker_free(struct net * net,netns_tracker * tracker,bool refcounted)379 static inline void __netns_tracker_free(struct net *net,
380 netns_tracker *tracker,
381 bool refcounted)
382 {
383 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
384 ref_tracker_free(refcounted ? &net->refcnt_tracker :
385 &net->notrefcnt_tracker, tracker);
386 #endif
387 }
388
get_net_track(struct net * net,netns_tracker * tracker,gfp_t gfp)389 static inline struct net *get_net_track(struct net *net,
390 netns_tracker *tracker, gfp_t gfp)
391 {
392 get_net(net);
393 netns_tracker_alloc(net, tracker, gfp);
394 return net;
395 }
396
put_net_track(struct net * net,netns_tracker * tracker)397 static inline void put_net_track(struct net *net, netns_tracker *tracker)
398 {
399 __netns_tracker_free(net, tracker, true);
400 put_net(net);
401 }
402
403 typedef struct {
404 #ifdef CONFIG_NET_NS
405 struct net __rcu *net;
406 #endif
407 } possible_net_t;
408
write_pnet(possible_net_t * pnet,struct net * net)409 static inline void write_pnet(possible_net_t *pnet, struct net *net)
410 {
411 #ifdef CONFIG_NET_NS
412 rcu_assign_pointer(pnet->net, net);
413 #endif
414 }
415
read_pnet(const possible_net_t * pnet)416 static inline struct net *read_pnet(const possible_net_t *pnet)
417 {
418 #ifdef CONFIG_NET_NS
419 return rcu_dereference_protected(pnet->net, true);
420 #else
421 return &init_net;
422 #endif
423 }
424
read_pnet_rcu(const possible_net_t * pnet)425 static inline struct net *read_pnet_rcu(const possible_net_t *pnet)
426 {
427 #ifdef CONFIG_NET_NS
428 return rcu_dereference(pnet->net);
429 #else
430 return &init_net;
431 #endif
432 }
433
434 /* Protected by net_rwsem */
435 #define for_each_net(VAR) \
436 list_for_each_entry(VAR, &net_namespace_list, list)
437 #define for_each_net_continue_reverse(VAR) \
438 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
439 #define for_each_net_rcu(VAR) \
440 list_for_each_entry_rcu(VAR, &net_namespace_list, list)
441
442 #ifdef CONFIG_NET_NS
443 #define __net_init
444 #define __net_exit
445 #define __net_initdata
446 #define __net_initconst
447 #else
448 #define __net_init __init
449 #define __net_exit __ref
450 #define __net_initdata __initdata
451 #define __net_initconst __initconst
452 #endif
453
454 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
455 int peernet2id(const struct net *net, struct net *peer);
456 bool peernet_has_id(const struct net *net, struct net *peer);
457 struct net *get_net_ns_by_id(const struct net *net, int id);
458
459 struct pernet_operations {
460 struct list_head list;
461 /*
462 * Below methods are called without any exclusive locks.
463 * More than one net may be constructed and destructed
464 * in parallel on several cpus. Every pernet_operations
465 * have to keep in mind all other pernet_operations and
466 * to introduce a locking, if they share common resources.
467 *
468 * The only time they are called with exclusive lock is
469 * from register_pernet_subsys(), unregister_pernet_subsys()
470 * register_pernet_device() and unregister_pernet_device().
471 *
472 * Exit methods using blocking RCU primitives, such as
473 * synchronize_rcu(), should be implemented via exit_batch.
474 * Then, destruction of a group of net requires single
475 * synchronize_rcu() related to these pernet_operations,
476 * instead of separate synchronize_rcu() for every net.
477 * Please, avoid synchronize_rcu() at all, where it's possible.
478 *
479 * Note that a combination of pre_exit() and exit() can
480 * be used, since a synchronize_rcu() is guaranteed between
481 * the calls.
482 */
483 int (*init)(struct net *net);
484 void (*pre_exit)(struct net *net);
485 void (*exit)(struct net *net);
486 void (*exit_batch)(struct list_head *net_exit_list);
487 /* Following method is called with RTNL held. */
488 void (*exit_rtnl)(struct net *net,
489 struct list_head *dev_kill_list);
490 unsigned int * const id;
491 const size_t size;
492 };
493
494 /*
495 * Use these carefully. If you implement a network device and it
496 * needs per network namespace operations use device pernet operations,
497 * otherwise use pernet subsys operations.
498 *
499 * Network interfaces need to be removed from a dying netns _before_
500 * subsys notifiers can be called, as most of the network code cleanup
501 * (which is done from subsys notifiers) runs with the assumption that
502 * dev_remove_pack has been called so no new packets will arrive during
503 * and after the cleanup functions have been called. dev_remove_pack
504 * is not per namespace so instead the guarantee of no more packets
505 * arriving in a network namespace is provided by ensuring that all
506 * network devices and all sockets have left the network namespace
507 * before the cleanup methods are called.
508 *
509 * For the longest time the ipv4 icmp code was registered as a pernet
510 * device which caused kernel oops, and panics during network
511 * namespace cleanup. So please don't get this wrong.
512 */
513 int register_pernet_subsys(struct pernet_operations *);
514 void unregister_pernet_subsys(struct pernet_operations *);
515 int register_pernet_device(struct pernet_operations *);
516 void unregister_pernet_device(struct pernet_operations *);
517
518 struct ctl_table;
519
520 #define register_net_sysctl(net, path, table) \
521 register_net_sysctl_sz(net, path, table, ARRAY_SIZE(table))
522 #ifdef CONFIG_SYSCTL
523 int net_sysctl_init(void);
524 struct ctl_table_header *register_net_sysctl_sz(struct net *net, const char *path,
525 struct ctl_table *table, size_t table_size);
526 void unregister_net_sysctl_table(struct ctl_table_header *header);
527 #else
net_sysctl_init(void)528 static inline int net_sysctl_init(void) { return 0; }
register_net_sysctl_sz(struct net * net,const char * path,struct ctl_table * table,size_t table_size)529 static inline struct ctl_table_header *register_net_sysctl_sz(struct net *net,
530 const char *path, struct ctl_table *table, size_t table_size)
531 {
532 return NULL;
533 }
unregister_net_sysctl_table(struct ctl_table_header * header)534 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
535 {
536 }
537 #endif
538
rt_genid_ipv4(const struct net * net)539 static inline int rt_genid_ipv4(const struct net *net)
540 {
541 return atomic_read(&net->ipv4.rt_genid);
542 }
543
544 #if IS_ENABLED(CONFIG_IPV6)
rt_genid_ipv6(const struct net * net)545 static inline int rt_genid_ipv6(const struct net *net)
546 {
547 return atomic_read(&net->ipv6.fib6_sernum);
548 }
549 #endif
550
rt_genid_bump_ipv4(struct net * net)551 static inline void rt_genid_bump_ipv4(struct net *net)
552 {
553 atomic_inc(&net->ipv4.rt_genid);
554 }
555
556 extern void (*__fib6_flush_trees)(struct net *net);
rt_genid_bump_ipv6(struct net * net)557 static inline void rt_genid_bump_ipv6(struct net *net)
558 {
559 if (__fib6_flush_trees)
560 __fib6_flush_trees(net);
561 }
562
563 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
564 static inline struct netns_ieee802154_lowpan *
net_ieee802154_lowpan(struct net * net)565 net_ieee802154_lowpan(struct net *net)
566 {
567 return &net->ieee802154_lowpan;
568 }
569 #endif
570
571 /* For callers who don't really care about whether it's IPv4 or IPv6 */
rt_genid_bump_all(struct net * net)572 static inline void rt_genid_bump_all(struct net *net)
573 {
574 rt_genid_bump_ipv4(net);
575 rt_genid_bump_ipv6(net);
576 }
577
fnhe_genid(const struct net * net)578 static inline int fnhe_genid(const struct net *net)
579 {
580 return atomic_read(&net->fnhe_genid);
581 }
582
fnhe_genid_bump(struct net * net)583 static inline void fnhe_genid_bump(struct net *net)
584 {
585 atomic_inc(&net->fnhe_genid);
586 }
587
588 #ifdef CONFIG_NET
589 void net_ns_init(void);
590 #else
net_ns_init(void)591 static inline void net_ns_init(void) {}
592 #endif
593
594 #endif /* __NET_NET_NAMESPACE_H */
595