blob: d96c9d9cca96560143ca5ba6a31106c8115bb4b4 [file] [log] [blame]
rjw1f884582022-01-06 17:20:42 +08001/* 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
14#include <net/flow.h>
15#include <net/netns/core.h>
16#include <net/netns/mib.h>
17#include <net/netns/unix.h>
18#include <net/netns/packet.h>
19#include <net/netns/ipv4.h>
20#include <net/netns/ipv6.h>
21#include <net/netns/ieee802154_6lowpan.h>
22#include <net/netns/sctp.h>
23#include <net/netns/dccp.h>
24#include <net/netns/netfilter.h>
25#include <net/netns/x_tables.h>
26#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
27#include <net/netns/conntrack.h>
28#endif
29#include <net/netns/nftables.h>
30#include <net/netns/xfrm.h>
31#include <net/netns/mpls.h>
32#include <net/netns/can.h>
33#include <linux/ns_common.h>
34#include <linux/idr.h>
35#include <linux/skbuff.h>
36
37struct user_namespace;
38struct proc_dir_entry;
39struct net_device;
40struct sock;
41struct ctl_table_header;
42struct net_generic;
43struct sock;
44struct netns_ipvs;
45
46
47#define NETDEV_HASHBITS 8
48#define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
49
50struct net {
51 refcount_t passive; /* To decided when the network
52 * namespace should be freed.
53 */
54 atomic_t count; /* To decided when the network
55 * namespace should be shut down.
56 */
57 spinlock_t rules_mod_lock;
58
59 u32 hash_mix;
60 atomic64_t cookie_gen;
61
62 struct list_head list; /* list of network namespaces */
63 struct list_head cleanup_list; /* namespaces on death row */
64 struct list_head exit_list; /* Use only net_mutex */
65
66 struct user_namespace *user_ns; /* Owning user namespace */
67 struct ucounts *ucounts;
68 spinlock_t nsid_lock;
69 struct idr netns_ids;
70
71 struct ns_common ns;
72
73 struct proc_dir_entry *proc_net;
74 struct proc_dir_entry *proc_net_stat;
75
76#ifdef CONFIG_SYSCTL
77 struct ctl_table_set sysctls;
78#endif
79
80 struct sock *rtnl; /* rtnetlink socket */
81 struct sock *genl_sock;
82
83 struct list_head dev_base_head;
84 struct hlist_head *dev_name_head;
85 struct hlist_head *dev_index_head;
86 unsigned int dev_base_seq; /* protected by rtnl_mutex */
87 int ifindex;
88 unsigned int dev_unreg_count;
89
90 /* core fib_rules */
91 struct list_head rules_ops;
92
93 struct list_head fib_notifier_ops; /* protected by net_mutex */
94
95 struct net_device *loopback_dev; /* The loopback */
96 struct netns_core core;
97 struct netns_mib mib;
98 struct netns_packet packet;
99 struct netns_unix unx;
100 struct netns_ipv4 ipv4;
101#if IS_ENABLED(CONFIG_IPV6)
102 struct netns_ipv6 ipv6;
103#endif
104#if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
105 struct netns_ieee802154_lowpan ieee802154_lowpan;
106#endif
107#if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
108 struct netns_sctp sctp;
109#endif
110#if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
111 struct netns_dccp dccp;
112#endif
113#ifdef CONFIG_NETFILTER
114 struct netns_nf nf;
115 struct netns_xt xt;
116#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
117 struct netns_ct ct;
118#endif
119#if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
120 struct netns_nftables nft;
121#endif
122#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
123 struct netns_nf_frag nf_frag;
124 struct ctl_table_header *nf_frag_frags_hdr;
125#endif
126 struct sock *nfnl;
127 struct sock *nfnl_stash;
128#if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
129 struct list_head nfnl_acct_list;
130#endif
131#if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
132 struct list_head nfct_timeout_list;
133#endif
134#endif
135#ifdef CONFIG_WEXT_CORE
136 struct sk_buff_head wext_nlevents;
137#endif
138 struct net_generic __rcu *gen;
139
140 /* Note : following structs are cache line aligned */
141#ifdef CONFIG_XFRM
142 struct netns_xfrm xfrm;
143#endif
144#if IS_ENABLED(CONFIG_IP_VS)
145 struct netns_ipvs *ipvs;
146#endif
147#if IS_ENABLED(CONFIG_MPLS)
148 struct netns_mpls mpls;
149#endif
150#if IS_ENABLED(CONFIG_CAN)
151 struct netns_can can;
152#endif
153 struct sock *diag_nlsk;
154 atomic_t fnhe_genid;
155} __randomize_layout;
156
157#include <linux/seq_file_net.h>
158
159/* Init's network namespace */
160extern struct net init_net;
161
162#ifdef CONFIG_NET_NS
163struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
164 struct net *old_net);
165
166void net_ns_barrier(void);
167#else /* CONFIG_NET_NS */
168#include <linux/sched.h>
169#include <linux/nsproxy.h>
170static inline struct net *copy_net_ns(unsigned long flags,
171 struct user_namespace *user_ns, struct net *old_net)
172{
173 if (flags & CLONE_NEWNET)
174 return ERR_PTR(-EINVAL);
175 return old_net;
176}
177
178static inline void net_ns_barrier(void) {}
179#endif /* CONFIG_NET_NS */
180
181
182extern struct list_head net_namespace_list;
183
184struct net *get_net_ns_by_pid(pid_t pid);
185struct net *get_net_ns_by_fd(int fd);
186
187#ifdef CONFIG_SYSCTL
188void ipx_register_sysctl(void);
189void ipx_unregister_sysctl(void);
190#else
191#define ipx_register_sysctl()
192#define ipx_unregister_sysctl()
193#endif
194
195#ifdef CONFIG_NET_NS
196void __put_net(struct net *net);
197
198static inline struct net *get_net(struct net *net)
199{
200 atomic_inc(&net->count);
201 return net;
202}
203
204static inline struct net *maybe_get_net(struct net *net)
205{
206 /* Used when we know struct net exists but we
207 * aren't guaranteed a previous reference count
208 * exists. If the reference count is zero this
209 * function fails and returns NULL.
210 */
211 if (!atomic_inc_not_zero(&net->count))
212 net = NULL;
213 return net;
214}
215
216static inline void put_net(struct net *net)
217{
218 if (atomic_dec_and_test(&net->count))
219 __put_net(net);
220}
221
222static inline
223int net_eq(const struct net *net1, const struct net *net2)
224{
225 return net1 == net2;
226}
227
228static inline int check_net(const struct net *net)
229{
230 return atomic_read(&net->count) != 0;
231}
232
233void net_drop_ns(void *);
234
235#else
236
237static inline struct net *get_net(struct net *net)
238{
239 return net;
240}
241
242static inline void put_net(struct net *net)
243{
244}
245
246static inline struct net *maybe_get_net(struct net *net)
247{
248 return net;
249}
250
251static inline
252int net_eq(const struct net *net1, const struct net *net2)
253{
254 return 1;
255}
256
257static inline int check_net(const struct net *net)
258{
259 return 1;
260}
261
262#define net_drop_ns NULL
263#endif
264
265
266typedef struct {
267#ifdef CONFIG_NET_NS
268 struct net *net;
269#endif
270} possible_net_t;
271
272static inline void write_pnet(possible_net_t *pnet, struct net *net)
273{
274#ifdef CONFIG_NET_NS
275 pnet->net = net;
276#endif
277}
278
279static inline struct net *read_pnet(const possible_net_t *pnet)
280{
281#ifdef CONFIG_NET_NS
282 return pnet->net;
283#else
284 return &init_net;
285#endif
286}
287
288#define for_each_net(VAR) \
289 list_for_each_entry(VAR, &net_namespace_list, list)
290
291#define for_each_net_rcu(VAR) \
292 list_for_each_entry_rcu(VAR, &net_namespace_list, list)
293
294#ifdef CONFIG_NET_NS
295#define __net_init
296#define __net_exit
297#define __net_initdata
298#define __net_initconst
299#else
300#define __net_init __init
301#define __net_exit __ref
302#define __net_initdata __initdata
303#define __net_initconst __initconst
304#endif
305
306int peernet2id_alloc(struct net *net, struct net *peer);
307int peernet2id(struct net *net, struct net *peer);
308bool peernet_has_id(struct net *net, struct net *peer);
309struct net *get_net_ns_by_id(struct net *net, int id);
310
311struct pernet_operations {
312 struct list_head list;
313 int (*init)(struct net *net);
314 void (*exit)(struct net *net);
315 void (*exit_batch)(struct list_head *net_exit_list);
316 unsigned int *id;
317 size_t size;
318};
319
320/*
321 * Use these carefully. If you implement a network device and it
322 * needs per network namespace operations use device pernet operations,
323 * otherwise use pernet subsys operations.
324 *
325 * Network interfaces need to be removed from a dying netns _before_
326 * subsys notifiers can be called, as most of the network code cleanup
327 * (which is done from subsys notifiers) runs with the assumption that
328 * dev_remove_pack has been called so no new packets will arrive during
329 * and after the cleanup functions have been called. dev_remove_pack
330 * is not per namespace so instead the guarantee of no more packets
331 * arriving in a network namespace is provided by ensuring that all
332 * network devices and all sockets have left the network namespace
333 * before the cleanup methods are called.
334 *
335 * For the longest time the ipv4 icmp code was registered as a pernet
336 * device which caused kernel oops, and panics during network
337 * namespace cleanup. So please don't get this wrong.
338 */
339int register_pernet_subsys(struct pernet_operations *);
340void unregister_pernet_subsys(struct pernet_operations *);
341int register_pernet_device(struct pernet_operations *);
342void unregister_pernet_device(struct pernet_operations *);
343
344struct ctl_table;
345struct ctl_table_header;
346
347#ifdef CONFIG_SYSCTL
348int net_sysctl_init(void);
349struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
350 struct ctl_table *table);
351void unregister_net_sysctl_table(struct ctl_table_header *header);
352#else
353static inline int net_sysctl_init(void) { return 0; }
354static inline struct ctl_table_header *register_net_sysctl(struct net *net,
355 const char *path, struct ctl_table *table)
356{
357 return NULL;
358}
359static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
360{
361}
362#endif
363
364static inline int rt_genid_ipv4(struct net *net)
365{
366 return atomic_read(&net->ipv4.rt_genid);
367}
368
369static inline void rt_genid_bump_ipv4(struct net *net)
370{
371 atomic_inc(&net->ipv4.rt_genid);
372}
373
374extern void (*__fib6_flush_trees)(struct net *net);
375static inline void rt_genid_bump_ipv6(struct net *net)
376{
377 if (__fib6_flush_trees)
378 __fib6_flush_trees(net);
379}
380
381#if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
382static inline struct netns_ieee802154_lowpan *
383net_ieee802154_lowpan(struct net *net)
384{
385 return &net->ieee802154_lowpan;
386}
387#endif
388
389/* For callers who don't really care about whether it's IPv4 or IPv6 */
390static inline void rt_genid_bump_all(struct net *net)
391{
392 rt_genid_bump_ipv4(net);
393 rt_genid_bump_ipv6(net);
394}
395
396static inline int fnhe_genid(struct net *net)
397{
398 return atomic_read(&net->fnhe_genid);
399}
400
401static inline void fnhe_genid_bump(struct net *net)
402{
403 atomic_inc(&net->fnhe_genid);
404}
405
406#endif /* __NET_NET_NAMESPACE_H */