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