]> git.ipfire.org Git - thirdparty/linux.git/blame - include/net/net_namespace.h
Merge tag 'io_uring-5.7-2020-05-22' of git://git.kernel.dk/linux-block
[thirdparty/linux.git] / include / net / net_namespace.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
5f256bec
EB
2/*
3 * Operations on the network namespace
4 */
5#ifndef __NET_NET_NAMESPACE_H
6#define __NET_NET_NAMESPACE_H
7
60063497 8#include <linux/atomic.h>
c122e14d 9#include <linux/refcount.h>
5f256bec
EB
10#include <linux/workqueue.h>
11#include <linux/list.h>
bee95250 12#include <linux/sysctl.h>
fbdeaed4 13#include <linux/uidgid.h>
5f256bec 14
6a662719 15#include <net/flow.h>
8efa6e93 16#include <net/netns/core.h>
852566f5 17#include <net/netns/mib.h>
a0a53c8b 18#include <net/netns/unix.h>
2aaef4e4 19#include <net/netns/packet.h>
8afd351c 20#include <net/netns/ipv4.h>
b0f159db 21#include <net/netns/ipv6.h>
ab84be7e 22#include <net/netns/nexthop.h>
633fc86f 23#include <net/netns/ieee802154_6lowpan.h>
4db67e80 24#include <net/netns/sctp.h>
67019cc9 25#include <net/netns/dccp.h>
f3c1a44a 26#include <net/netns/netfilter.h>
8d870052 27#include <net/netns/x_tables.h>
dfdb8d79
AD
28#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
29#include <net/netns/conntrack.h>
30#endif
99633ab2 31#include <net/netns/nftables.h>
d62ddc21 32#include <net/netns/xfrm.h>
0189197f 33#include <net/netns/mpls.h>
8e8cda6d 34#include <net/netns/can.h>
1d0dc069 35#include <net/netns/xdp.h>
435d5f4b 36#include <linux/ns_common.h>
04c52dec
PNA
37#include <linux/idr.h>
38#include <linux/skbuff.h>
a30c7b42 39#include <linux/notifier.h>
a0a53c8b 40
038e7332 41struct user_namespace;
457c4cbc 42struct proc_dir_entry;
2774c7ab 43struct net_device;
97c53cac 44struct sock;
1597fbc0 45struct ctl_table_header;
dec827d1 46struct net_generic;
94e5e308 47struct uevent_sock;
2553d064 48struct netns_ipvs;
d58e468b 49struct bpf_prog;
1597fbc0 50
7c28bd0b
ED
51
52#define NETDEV_HASHBITS 8
53#define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
54
5f256bec 55struct net {
2a06b898
ED
56 /* First cache line can be often dirtied.
57 * Do not place here read-mostly fields.
58 */
aad12c23 59 refcount_t passive; /* To decide when the network
a685e089
AV
60 * namespace should be freed.
61 */
273c28bc 62 refcount_t count; /* To decided when the network
a685e089 63 * namespace should be shut down.
5f256bec 64 */
8e602ce2
ED
65 spinlock_t rules_mod_lock;
66
2a06b898
ED
67 unsigned int dev_unreg_count;
68
69 unsigned int dev_base_seq; /* protected by rtnl_mutex */
70 int ifindex;
71
72 spinlock_t nsid_lock;
73 atomic_t fnhe_genid;
33cf7c90 74
5f256bec 75 struct list_head list; /* list of network namespaces */
19efbd93 76 struct list_head exit_list; /* To linked to call pernet exit
4420bf21
KT
77 * methods on dead net (
78 * pernet_ops_rwsem read locked),
79 * or to unregister pernet ops
80 * (pernet_ops_rwsem write locked).
19efbd93 81 */
65b7b5b9
KT
82 struct llist_node cleanup_list; /* namespaces on death row */
83
9b242610
DH
84#ifdef CONFIG_KEYS
85 struct key_tag *key_domain; /* Key domain of operation tag */
86#endif
038e7332 87 struct user_namespace *user_ns; /* Owning user namespace */
70328660 88 struct ucounts *ucounts;
0c7aecd4 89 struct idr netns_ids;
038e7332 90
435d5f4b 91 struct ns_common ns;
98f842e6 92
2a06b898 93 struct list_head dev_base_head;
457c4cbc
EB
94 struct proc_dir_entry *proc_net;
95 struct proc_dir_entry *proc_net_stat;
881d966b 96
73455092
AV
97#ifdef CONFIG_SYSCTL
98 struct ctl_table_set sysctls;
99#endif
95bdfccb 100
8e602ce2
ED
101 struct sock *rtnl; /* rtnetlink socket */
102 struct sock *genl_sock;
2774c7ab 103
94e5e308
CB
104 struct uevent_sock *uevent_sock; /* uevent socket */
105
881d966b
EB
106 struct hlist_head *dev_name_head;
107 struct hlist_head *dev_index_head;
a30c7b42
JP
108 struct raw_notifier_head netdev_chain;
109
2a06b898
ED
110 /* Note that @hash_mix can be read millions times per second,
111 * it is critical that it is on a read_mostly cache line.
112 */
113 u32 hash_mix;
114
115 struct net_device *loopback_dev; /* The loopback */
97c53cac 116
5fd30ee7
DL
117 /* core fib_rules */
118 struct list_head rules_ops;
5fd30ee7 119
8efa6e93 120 struct netns_core core;
852566f5 121 struct netns_mib mib;
2aaef4e4 122 struct netns_packet packet;
a0a53c8b 123 struct netns_unix unx;
ab84be7e 124 struct netns_nexthop nexthop;
8afd351c 125 struct netns_ipv4 ipv4;
dfd56b8b 126#if IS_ENABLED(CONFIG_IPV6)
b0f159db
DL
127 struct netns_ipv6 ipv6;
128#endif
633fc86f
AA
129#if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
130 struct netns_ieee802154_lowpan ieee802154_lowpan;
131#endif
4db67e80
EB
132#if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
133 struct netns_sctp sctp;
134#endif
67019cc9
PE
135#if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
136 struct netns_dccp dccp;
137#endif
8d870052 138#ifdef CONFIG_NETFILTER
f3c1a44a 139 struct netns_nf nf;
8d870052 140 struct netns_xt xt;
dfdb8d79
AD
141#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
142 struct netns_ct ct;
c038a767 143#endif
99633ab2
PNA
144#if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
145 struct netns_nftables nft;
146#endif
c038a767
AW
147#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
148 struct netns_nf_frag nf_frag;
9ce7bc03 149 struct ctl_table_header *nf_frag_frags_hdr;
dfdb8d79 150#endif
cd8c20b6
AD
151 struct sock *nfnl;
152 struct sock *nfnl_stash;
3499abb2
AS
153#if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
154 struct list_head nfnl_acct_list;
155#endif
19576c94
PN
156#if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
157 struct list_head nfct_timeout_list;
158#endif
d62ddc21 159#endif
3d23e349 160#ifdef CONFIG_WEXT_CORE
b333b3d2 161 struct sk_buff_head wext_nlevents;
8d870052 162#endif
1c87733d 163 struct net_generic __rcu *gen;
8e602ce2 164
d58e468b
PP
165 struct bpf_prog __rcu *flow_dissector_prog;
166
8e602ce2
ED
167 /* Note : following structs are cache line aligned */
168#ifdef CONFIG_XFRM
169 struct netns_xfrm xfrm;
170#endif
f318903c
DB
171
172 atomic64_t net_cookie; /* written once */
173
8b4d14d8 174#if IS_ENABLED(CONFIG_IP_VS)
61b1ab45 175 struct netns_ipvs *ipvs;
0189197f
EB
176#endif
177#if IS_ENABLED(CONFIG_MPLS)
178 struct netns_mpls mpls;
8e8cda6d
MK
179#endif
180#if IS_ENABLED(CONFIG_CAN)
181 struct netns_can can;
1d0dc069
BT
182#endif
183#ifdef CONFIG_XDP_SOCKETS
184 struct netns_xdp xdp;
91b05a7e
OM
185#endif
186#if IS_ENABLED(CONFIG_CRYPTO_USER)
187 struct sock *crypto_nlsk;
8b4d14d8 188#endif
51d7cccf 189 struct sock *diag_nlsk;
3859a271 190} __randomize_layout;
5f256bec 191
c0f39322
DL
192#include <linux/seq_file_net.h>
193
4fabcd71 194/* Init's network namespace */
5f256bec 195extern struct net init_net;
a4aa834a 196
d727abcb 197#ifdef CONFIG_NET_NS
e67e16ea
JP
198struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
199 struct net *old_net);
225c0a01 200
fbdeaed4
TH
201void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
202
7866cc57 203void net_ns_barrier(void);
d727abcb
EB
204#else /* CONFIG_NET_NS */
205#include <linux/sched.h>
206#include <linux/nsproxy.h>
038e7332
EB
207static inline struct net *copy_net_ns(unsigned long flags,
208 struct user_namespace *user_ns, struct net *old_net)
9dd776b6 209{
d727abcb
EB
210 if (flags & CLONE_NEWNET)
211 return ERR_PTR(-EINVAL);
212 return old_net;
9dd776b6 213}
7866cc57 214
fbdeaed4
TH
215static inline void net_ns_get_ownership(const struct net *net,
216 kuid_t *uid, kgid_t *gid)
217{
218 *uid = GLOBAL_ROOT_UID;
219 *gid = GLOBAL_ROOT_GID;
220}
221
7866cc57 222static inline void net_ns_barrier(void) {}
d727abcb 223#endif /* CONFIG_NET_NS */
225c0a01
DL
224
225
226extern struct list_head net_namespace_list;
9dd776b6 227
e67e16ea 228struct net *get_net_ns_by_pid(pid_t pid);
0f5258cd 229struct net *get_net_ns_by_fd(int fd);
30ffee84 230
5a95cbb8
DB
231u64 net_gen_cookie(struct net *net);
232
535d3ae9
RK
233#ifdef CONFIG_SYSCTL
234void ipx_register_sysctl(void);
235void ipx_unregister_sysctl(void);
236#else
237#define ipx_register_sysctl()
238#define ipx_unregister_sysctl()
239#endif
240
d4655795 241#ifdef CONFIG_NET_NS
e67e16ea 242void __put_net(struct net *net);
5f256bec
EB
243
244static inline struct net *get_net(struct net *net)
245{
273c28bc 246 refcount_inc(&net->count);
5f256bec
EB
247 return net;
248}
249
077130c0
EB
250static inline struct net *maybe_get_net(struct net *net)
251{
252 /* Used when we know struct net exists but we
253 * aren't guaranteed a previous reference count
254 * exists. If the reference count is zero this
255 * function fails and returns NULL.
256 */
273c28bc 257 if (!refcount_inc_not_zero(&net->count))
077130c0
EB
258 net = NULL;
259 return net;
260}
261
5f256bec
EB
262static inline void put_net(struct net *net)
263{
273c28bc 264 if (refcount_dec_and_test(&net->count))
5f256bec
EB
265 __put_net(net);
266}
267
878628fb
YH
268static inline
269int net_eq(const struct net *net1, const struct net *net2)
270{
271 return net1 == net2;
272}
a685e089 273
4ee806d5
DS
274static inline int check_net(const struct net *net)
275{
3e3ab9cc 276 return refcount_read(&net->count) != 0;
4ee806d5
DS
277}
278
e67e16ea 279void net_drop_ns(void *);
a685e089 280
d4655795 281#else
b9f75f45 282
d4655795
PE
283static inline struct net *get_net(struct net *net)
284{
285 return net;
286}
287
288static inline void put_net(struct net *net)
289{
290}
291
5d1e4468
DL
292static inline struct net *maybe_get_net(struct net *net)
293{
294 return net;
295}
296
297static inline
298int net_eq(const struct net *net1, const struct net *net2)
299{
300 return 1;
301}
a685e089 302
4ee806d5
DS
303static inline int check_net(const struct net *net)
304{
305 return 1;
306}
307
a685e089 308#define net_drop_ns NULL
5d1e4468
DL
309#endif
310
311
0c5c9fb5 312typedef struct {
8f424b5f 313#ifdef CONFIG_NET_NS
0c5c9fb5
EB
314 struct net *net;
315#endif
316} possible_net_t;
8f424b5f 317
0c5c9fb5 318static inline void write_pnet(possible_net_t *pnet, struct net *net)
8f424b5f 319{
0c5c9fb5
EB
320#ifdef CONFIG_NET_NS
321 pnet->net = net;
322#endif
8f424b5f
ED
323}
324
0c5c9fb5 325static inline struct net *read_pnet(const possible_net_t *pnet)
8f424b5f 326{
0c5c9fb5
EB
327#ifdef CONFIG_NET_NS
328 return pnet->net;
8f424b5f 329#else
0c5c9fb5 330 return &init_net;
8f424b5f 331#endif
0c5c9fb5 332}
5d1e4468 333
f0b07bb1 334/* Protected by net_rwsem */
5f256bec
EB
335#define for_each_net(VAR) \
336 list_for_each_entry(VAR, &net_namespace_list, list)
afa0df59
JP
337#define for_each_net_continue_reverse(VAR) \
338 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
11a28d37
JB
339#define for_each_net_rcu(VAR) \
340 list_for_each_entry_rcu(VAR, &net_namespace_list, list)
341
4665079c
PE
342#ifdef CONFIG_NET_NS
343#define __net_init
344#define __net_exit
022cbae6 345#define __net_initdata
04a6f82c 346#define __net_initconst
4665079c
PE
347#else
348#define __net_init __init
bd721ea7 349#define __net_exit __ref
022cbae6 350#define __net_initdata __initdata
04a6f82c 351#define __net_initconst __initconst
4665079c 352#endif
5f256bec 353
d4e4fdf9 354int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
56f200c7
GN
355int peernet2id(const struct net *net, struct net *peer);
356bool peernet_has_id(const struct net *net, struct net *peer);
357struct net *get_net_ns_by_id(const struct net *net, int id);
0c7aecd4 358
5f256bec
EB
359struct pernet_operations {
360 struct list_head list;
6056415d
KT
361 /*
362 * Below methods are called without any exclusive locks.
363 * More than one net may be constructed and destructed
364 * in parallel on several cpus. Every pernet_operations
365 * have to keep in mind all other pernet_operations and
366 * to introduce a locking, if they share common resources.
367 *
8518e9bb
KT
368 * The only time they are called with exclusive lock is
369 * from register_pernet_subsys(), unregister_pernet_subsys()
370 * register_pernet_device() and unregister_pernet_device().
371 *
6056415d
KT
372 * Exit methods using blocking RCU primitives, such as
373 * synchronize_rcu(), should be implemented via exit_batch.
374 * Then, destruction of a group of net requires single
375 * synchronize_rcu() related to these pernet_operations,
376 * instead of separate synchronize_rcu() for every net.
377 * Please, avoid synchronize_rcu() at all, where it's possible.
d7d99872
ED
378 *
379 * Note that a combination of pre_exit() and exit() can
380 * be used, since a synchronize_rcu() is guaranteed between
381 * the calls.
6056415d 382 */
5f256bec 383 int (*init)(struct net *net);
d7d99872 384 void (*pre_exit)(struct net *net);
5f256bec 385 void (*exit)(struct net *net);
72ad937a 386 void (*exit_batch)(struct list_head *net_exit_list);
c7d03a00 387 unsigned int *id;
f875bae0 388 size_t size;
5f256bec
EB
389};
390
17edde52
EB
391/*
392 * Use these carefully. If you implement a network device and it
393 * needs per network namespace operations use device pernet operations,
394 * otherwise use pernet subsys operations.
395 *
4edf547b
JB
396 * Network interfaces need to be removed from a dying netns _before_
397 * subsys notifiers can be called, as most of the network code cleanup
398 * (which is done from subsys notifiers) runs with the assumption that
399 * dev_remove_pack has been called so no new packets will arrive during
400 * and after the cleanup functions have been called. dev_remove_pack
401 * is not per namespace so instead the guarantee of no more packets
402 * arriving in a network namespace is provided by ensuring that all
403 * network devices and all sockets have left the network namespace
404 * before the cleanup methods are called.
17edde52
EB
405 *
406 * For the longest time the ipv4 icmp code was registered as a pernet
407 * device which caused kernel oops, and panics during network
408 * namespace cleanup. So please don't get this wrong.
409 */
e67e16ea
JP
410int register_pernet_subsys(struct pernet_operations *);
411void unregister_pernet_subsys(struct pernet_operations *);
412int register_pernet_device(struct pernet_operations *);
413void unregister_pernet_device(struct pernet_operations *);
f875bae0 414
95bdfccb
EB
415struct ctl_table;
416struct ctl_table_header;
d62c612e 417
2ca794e5 418#ifdef CONFIG_SYSCTL
e67e16ea
JP
419int net_sysctl_init(void);
420struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
421 struct ctl_table *table);
422void unregister_net_sysctl_table(struct ctl_table_header *header);
48c74958
EB
423#else
424static inline int net_sysctl_init(void) { return 0; }
425static inline struct ctl_table_header *register_net_sysctl(struct net *net,
426 const char *path, struct ctl_table *table)
427{
428 return NULL;
429}
430static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
431{
432}
433#endif
434
56f200c7 435static inline int rt_genid_ipv4(const struct net *net)
b42664f8 436{
ca4c3fc2 437 return atomic_read(&net->ipv4.rt_genid);
b42664f8
ND
438}
439
8f34e53b
DA
440#if IS_ENABLED(CONFIG_IPV6)
441static inline int rt_genid_ipv6(const struct net *net)
442{
443 return atomic_read(&net->ipv6.fib6_sernum);
444}
445#endif
446
ca4c3fc2 447static inline void rt_genid_bump_ipv4(struct net *net)
b42664f8 448{
ca4c3fc2 449 atomic_inc(&net->ipv4.rt_genid);
450}
451
705f1c86 452extern void (*__fib6_flush_trees)(struct net *net);
ca4c3fc2 453static inline void rt_genid_bump_ipv6(struct net *net)
454{
705f1c86
HFS
455 if (__fib6_flush_trees)
456 __fib6_flush_trees(net);
ca4c3fc2 457}
ca4c3fc2 458
599018a7
LR
459#if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
460static inline struct netns_ieee802154_lowpan *
461net_ieee802154_lowpan(struct net *net)
462{
463 return &net->ieee802154_lowpan;
464}
599018a7
LR
465#endif
466
ca4c3fc2 467/* For callers who don't really care about whether it's IPv4 or IPv6 */
468static inline void rt_genid_bump_all(struct net *net)
469{
470 rt_genid_bump_ipv4(net);
471 rt_genid_bump_ipv6(net);
b42664f8 472}
95bdfccb 473
56f200c7 474static inline int fnhe_genid(const struct net *net)
5aad1de5
TT
475{
476 return atomic_read(&net->fnhe_genid);
477}
478
479static inline void fnhe_genid_bump(struct net *net)
480{
481 atomic_inc(&net->fnhe_genid);
482}
483
5f256bec 484#endif /* __NET_NET_NAMESPACE_H */