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
47struct user_namespace;
48struct proc_dir_entry;
49struct net_device;
50struct sock;
51struct ctl_table_header;
52struct net_generic;
53struct uevent_sock;
54struct netns_ipvs;
55struct bpf_prog;
56
57
58#define NETDEV_HASHBITS 8
59#define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
60
61struct 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 */
204extern struct net init_net;
205
206#ifdef CONFIG_NET_NS
207struct net *copy_net_ns(u64 flags, struct user_namespace *user_ns,
208 struct net *old_net);
209
210void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
211
212void net_ns_barrier(void);
213
214struct ns_common *get_net_ns(struct ns_common *ns);
215struct net *get_net_ns_by_fd(int fd);
216extern struct task_struct *cleanup_net_task;
217
218#else /* CONFIG_NET_NS */
219#include <linux/sched.h>
220#include <linux/nsproxy.h>
221static inline struct net *copy_net_ns(u64 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
229static 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
236static inline void net_ns_barrier(void) {}
237
238static inline struct ns_common *get_net_ns(struct ns_common *ns)
239{
240 return ERR_PTR(-EINVAL);
241}
242
243static inline struct net *get_net_ns_by_fd(int fd)
244{
245 return ERR_PTR(-EINVAL);
246}
247#endif /* CONFIG_NET_NS */
248
249
250extern struct list_head net_namespace_list;
251
252struct net *get_net_ns_by_pid(pid_t pid);
253
254#ifdef CONFIG_SYSCTL
255void ipx_register_sysctl(void);
256void ipx_unregister_sysctl(void);
257#else
258#define ipx_register_sysctl()
259#define ipx_unregister_sysctl()
260#endif
261
262#ifdef CONFIG_NET_NS
263void __put_net(struct net *net);
264
265static inline struct net *to_net_ns(struct ns_common *ns)
266{
267 return container_of(ns, struct net, ns);
268}
269
270/* Try using get_net_track() instead */
271static inline struct net *get_net(struct net *net)
272{
273 ns_ref_inc(net);
274 return net;
275}
276
277static inline struct net *maybe_get_net(struct net *net)
278{
279 /* Used when we know struct net exists but we
280 * aren't guaranteed a previous reference count
281 * exists. If the reference count is zero this
282 * function fails and returns NULL.
283 */
284 if (!ns_ref_get(net))
285 net = NULL;
286 return net;
287}
288
289/* Try using put_net_track() instead */
290static inline void put_net(struct net *net)
291{
292 if (ns_ref_put(net))
293 __put_net(net);
294}
295
296static inline
297int net_eq(const struct net *net1, const struct net *net2)
298{
299 return net1 == net2;
300}
301
302static inline int check_net(const struct net *net)
303{
304 return ns_ref_read(net) != 0;
305}
306
307void net_drop_ns(void *);
308void net_passive_dec(struct net *net);
309
310#else
311
312static inline struct net *get_net(struct net *net)
313{
314 return net;
315}
316
317static inline void put_net(struct net *net)
318{
319}
320
321static inline struct net *maybe_get_net(struct net *net)
322{
323 return net;
324}
325
326static inline
327int net_eq(const struct net *net1, const struct net *net2)
328{
329 return 1;
330}
331
332static inline int check_net(const struct net *net)
333{
334 return 1;
335}
336
337#define net_drop_ns NULL
338
339static inline void net_passive_dec(struct net *net)
340{
341 refcount_dec(&net->passive);
342}
343#endif
344
345static inline void net_passive_inc(struct net *net)
346{
347 refcount_inc(r: &net->passive);
348}
349
350/* Returns true if the netns initialization is completed successfully */
351static inline bool net_initialized(const struct net *net)
352{
353 return READ_ONCE(net->list.next);
354}
355
356static inline void __netns_tracker_alloc(struct net *net,
357 netns_tracker *tracker,
358 bool refcounted,
359 gfp_t gfp)
360{
361#ifdef CONFIG_NET_NS_REFCNT_TRACKER
362 ref_tracker_alloc(refcounted ? &net->refcnt_tracker :
363 &net->notrefcnt_tracker,
364 tracker, gfp);
365#endif
366}
367
368static inline void netns_tracker_alloc(struct net *net, netns_tracker *tracker,
369 gfp_t gfp)
370{
371 __netns_tracker_alloc(net, tracker, refcounted: true, gfp);
372}
373
374static inline void __netns_tracker_free(struct net *net,
375 netns_tracker *tracker,
376 bool refcounted)
377{
378#ifdef CONFIG_NET_NS_REFCNT_TRACKER
379 ref_tracker_free(refcounted ? &net->refcnt_tracker :
380 &net->notrefcnt_tracker, tracker);
381#endif
382}
383
384static inline struct net *get_net_track(struct net *net,
385 netns_tracker *tracker, gfp_t gfp)
386{
387 get_net(net);
388 netns_tracker_alloc(net, tracker, gfp);
389 return net;
390}
391
392static inline void put_net_track(struct net *net, netns_tracker *tracker)
393{
394 __netns_tracker_free(net, tracker, refcounted: true);
395 put_net(net);
396}
397
398typedef struct {
399#ifdef CONFIG_NET_NS
400 struct net __rcu *net;
401#endif
402} possible_net_t;
403
404static inline void write_pnet(possible_net_t *pnet, struct net *net)
405{
406#ifdef CONFIG_NET_NS
407 rcu_assign_pointer(pnet->net, net);
408#endif
409}
410
411static inline struct net *read_pnet(const possible_net_t *pnet)
412{
413#ifdef CONFIG_NET_NS
414 return rcu_dereference_protected(pnet->net, true);
415#else
416 return &init_net;
417#endif
418}
419
420static inline struct net *read_pnet_rcu(const possible_net_t *pnet)
421{
422#ifdef CONFIG_NET_NS
423 return rcu_dereference(pnet->net);
424#else
425 return &init_net;
426#endif
427}
428
429/* Protected by net_rwsem */
430#define for_each_net(VAR) \
431 list_for_each_entry(VAR, &net_namespace_list, list)
432#define for_each_net_continue_reverse(VAR) \
433 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
434#define for_each_net_rcu(VAR) \
435 list_for_each_entry_rcu(VAR, &net_namespace_list, list)
436
437#ifdef CONFIG_NET_NS
438#define __net_init
439#define __net_exit
440#define __net_initdata
441#define __net_initconst
442#else
443#define __net_init __init
444#define __net_exit __ref
445#define __net_initdata __initdata
446#define __net_initconst __initconst
447#endif
448
449int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
450int peernet2id(const struct net *net, struct net *peer);
451bool peernet_has_id(const struct net *net, struct net *peer);
452struct net *get_net_ns_by_id(const struct net *net, int id);
453
454struct pernet_operations {
455 struct list_head list;
456 /*
457 * Below methods are called without any exclusive locks.
458 * More than one net may be constructed and destructed
459 * in parallel on several cpus. Every pernet_operations
460 * have to keep in mind all other pernet_operations and
461 * to introduce a locking, if they share common resources.
462 *
463 * The only time they are called with exclusive lock is
464 * from register_pernet_subsys(), unregister_pernet_subsys()
465 * register_pernet_device() and unregister_pernet_device().
466 *
467 * Exit methods using blocking RCU primitives, such as
468 * synchronize_rcu(), should be implemented via exit_batch.
469 * Then, destruction of a group of net requires single
470 * synchronize_rcu() related to these pernet_operations,
471 * instead of separate synchronize_rcu() for every net.
472 * Please, avoid synchronize_rcu() at all, where it's possible.
473 *
474 * Note that a combination of pre_exit() and exit() can
475 * be used, since a synchronize_rcu() is guaranteed between
476 * the calls.
477 */
478 int (*init)(struct net *net);
479 void (*pre_exit)(struct net *net);
480 void (*exit)(struct net *net);
481 void (*exit_batch)(struct list_head *net_exit_list);
482 /* Following method is called with RTNL held. */
483 void (*exit_rtnl)(struct net *net,
484 struct list_head *dev_kill_list);
485 unsigned int * const id;
486 const size_t size;
487};
488
489/*
490 * Use these carefully. If you implement a network device and it
491 * needs per network namespace operations use device pernet operations,
492 * otherwise use pernet subsys operations.
493 *
494 * Network interfaces need to be removed from a dying netns _before_
495 * subsys notifiers can be called, as most of the network code cleanup
496 * (which is done from subsys notifiers) runs with the assumption that
497 * dev_remove_pack has been called so no new packets will arrive during
498 * and after the cleanup functions have been called. dev_remove_pack
499 * is not per namespace so instead the guarantee of no more packets
500 * arriving in a network namespace is provided by ensuring that all
501 * network devices and all sockets have left the network namespace
502 * before the cleanup methods are called.
503 *
504 * For the longest time the ipv4 icmp code was registered as a pernet
505 * device which caused kernel oops, and panics during network
506 * namespace cleanup. So please don't get this wrong.
507 */
508int register_pernet_subsys(struct pernet_operations *);
509void unregister_pernet_subsys(struct pernet_operations *);
510int register_pernet_device(struct pernet_operations *);
511void unregister_pernet_device(struct pernet_operations *);
512
513struct ctl_table;
514
515#define register_net_sysctl(net, path, table) \
516 register_net_sysctl_sz(net, path, table, ARRAY_SIZE(table))
517#ifdef CONFIG_SYSCTL
518int net_sysctl_init(void);
519struct ctl_table_header *register_net_sysctl_sz(struct net *net, const char *path,
520 struct ctl_table *table, size_t table_size);
521void unregister_net_sysctl_table(struct ctl_table_header *header);
522#else
523static inline int net_sysctl_init(void) { return 0; }
524static inline struct ctl_table_header *register_net_sysctl_sz(struct net *net,
525 const char *path, struct ctl_table *table, size_t table_size)
526{
527 return NULL;
528}
529static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
530{
531}
532#endif
533
534static inline int rt_genid_ipv4(const struct net *net)
535{
536 return atomic_read(v: &net->ipv4.rt_genid);
537}
538
539#if IS_ENABLED(CONFIG_IPV6)
540static inline int rt_genid_ipv6(const struct net *net)
541{
542 return atomic_read(v: &net->ipv6.fib6_sernum);
543}
544#endif
545
546static inline void rt_genid_bump_ipv4(struct net *net)
547{
548 atomic_inc(v: &net->ipv4.rt_genid);
549}
550
551extern void (*__fib6_flush_trees)(struct net *net);
552static inline void rt_genid_bump_ipv6(struct net *net)
553{
554 if (__fib6_flush_trees)
555 __fib6_flush_trees(net);
556}
557
558#if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
559static inline struct netns_ieee802154_lowpan *
560net_ieee802154_lowpan(struct net *net)
561{
562 return &net->ieee802154_lowpan;
563}
564#endif
565
566/* For callers who don't really care about whether it's IPv4 or IPv6 */
567static inline void rt_genid_bump_all(struct net *net)
568{
569 rt_genid_bump_ipv4(net);
570 rt_genid_bump_ipv6(net);
571}
572
573static inline int fnhe_genid(const struct net *net)
574{
575 return atomic_read(v: &net->fnhe_genid);
576}
577
578static inline void fnhe_genid_bump(struct net *net)
579{
580 atomic_inc(v: &net->fnhe_genid);
581}
582
583#ifdef CONFIG_NET
584void net_ns_init(void);
585#else
586static inline void net_ns_init(void) {}
587#endif
588
589#endif /* __NET_NET_NAMESPACE_H */
590