]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/net/net_namespace.h
netfilter: ipv6: nf_defrag: reduce struct net memory waste
[mirror_ubuntu-bionic-kernel.git] / include / net / net_namespace.h
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/ieee802154_6lowpan.h>
23 #include <net/netns/sctp.h>
24 #include <net/netns/dccp.h>
25 #include <net/netns/netfilter.h>
26 #include <net/netns/x_tables.h>
27 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
28 #include <net/netns/conntrack.h>
29 #endif
30 #include <net/netns/nftables.h>
31 #include <net/netns/xfrm.h>
32 #include <net/netns/mpls.h>
33 #include <net/netns/can.h>
34 #include <linux/ns_common.h>
35 #include <linux/idr.h>
36 #include <linux/skbuff.h>
37
38 struct user_namespace;
39 struct proc_dir_entry;
40 struct net_device;
41 struct sock;
42 struct ctl_table_header;
43 struct net_generic;
44 struct sock;
45 struct netns_ipvs;
46
47
48 #define NETDEV_HASHBITS 8
49 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
50
51 struct net {
52 refcount_t passive; /* To decided when the network
53 * namespace should be freed.
54 */
55 atomic_t count; /* To decided when the network
56 * namespace should be shut down.
57 */
58 spinlock_t rules_mod_lock;
59
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 */
160 extern struct net init_net;
161
162 #ifdef CONFIG_NET_NS
163 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
164 struct net *old_net);
165
166 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
167
168 void net_ns_barrier(void);
169 #else /* CONFIG_NET_NS */
170 #include <linux/sched.h>
171 #include <linux/nsproxy.h>
172 static inline struct net *copy_net_ns(unsigned long flags,
173 struct user_namespace *user_ns, struct net *old_net)
174 {
175 if (flags & CLONE_NEWNET)
176 return ERR_PTR(-EINVAL);
177 return old_net;
178 }
179
180 static inline void net_ns_get_ownership(const struct net *net,
181 kuid_t *uid, kgid_t *gid)
182 {
183 *uid = GLOBAL_ROOT_UID;
184 *gid = GLOBAL_ROOT_GID;
185 }
186
187 static inline void net_ns_barrier(void) {}
188 #endif /* CONFIG_NET_NS */
189
190
191 extern struct list_head net_namespace_list;
192
193 struct net *get_net_ns_by_pid(pid_t pid);
194 struct net *get_net_ns_by_fd(int fd);
195
196 #ifdef CONFIG_SYSCTL
197 void ipx_register_sysctl(void);
198 void ipx_unregister_sysctl(void);
199 #else
200 #define ipx_register_sysctl()
201 #define ipx_unregister_sysctl()
202 #endif
203
204 #ifdef CONFIG_NET_NS
205 void __put_net(struct net *net);
206
207 static inline struct net *get_net(struct net *net)
208 {
209 atomic_inc(&net->count);
210 return net;
211 }
212
213 static inline struct net *maybe_get_net(struct net *net)
214 {
215 /* Used when we know struct net exists but we
216 * aren't guaranteed a previous reference count
217 * exists. If the reference count is zero this
218 * function fails and returns NULL.
219 */
220 if (!atomic_inc_not_zero(&net->count))
221 net = NULL;
222 return net;
223 }
224
225 static inline void put_net(struct net *net)
226 {
227 if (atomic_dec_and_test(&net->count))
228 __put_net(net);
229 }
230
231 static inline
232 int net_eq(const struct net *net1, const struct net *net2)
233 {
234 return net1 == net2;
235 }
236
237 static inline int check_net(const struct net *net)
238 {
239 return atomic_read(&net->count) != 0;
240 }
241
242 void net_drop_ns(void *);
243
244 #else
245
246 static inline struct net *get_net(struct net *net)
247 {
248 return net;
249 }
250
251 static inline void put_net(struct net *net)
252 {
253 }
254
255 static inline struct net *maybe_get_net(struct net *net)
256 {
257 return net;
258 }
259
260 static inline
261 int net_eq(const struct net *net1, const struct net *net2)
262 {
263 return 1;
264 }
265
266 static inline int check_net(const struct net *net)
267 {
268 return 1;
269 }
270
271 #define net_drop_ns NULL
272 #endif
273
274
275 typedef struct {
276 #ifdef CONFIG_NET_NS
277 struct net *net;
278 #endif
279 } possible_net_t;
280
281 static inline void write_pnet(possible_net_t *pnet, struct net *net)
282 {
283 #ifdef CONFIG_NET_NS
284 pnet->net = net;
285 #endif
286 }
287
288 static inline struct net *read_pnet(const possible_net_t *pnet)
289 {
290 #ifdef CONFIG_NET_NS
291 return pnet->net;
292 #else
293 return &init_net;
294 #endif
295 }
296
297 #define for_each_net(VAR) \
298 list_for_each_entry(VAR, &net_namespace_list, list)
299
300 #define for_each_net_rcu(VAR) \
301 list_for_each_entry_rcu(VAR, &net_namespace_list, list)
302
303 #ifdef CONFIG_NET_NS
304 #define __net_init
305 #define __net_exit
306 #define __net_initdata
307 #define __net_initconst
308 #else
309 #define __net_init __init
310 #define __net_exit __ref
311 #define __net_initdata __initdata
312 #define __net_initconst __initconst
313 #endif
314
315 int peernet2id_alloc(struct net *net, struct net *peer);
316 int peernet2id(struct net *net, struct net *peer);
317 bool peernet_has_id(struct net *net, struct net *peer);
318 struct net *get_net_ns_by_id(struct net *net, int id);
319
320 struct pernet_operations {
321 struct list_head list;
322 int (*init)(struct net *net);
323 void (*exit)(struct net *net);
324 void (*exit_batch)(struct list_head *net_exit_list);
325 unsigned int *id;
326 size_t size;
327 };
328
329 /*
330 * Use these carefully. If you implement a network device and it
331 * needs per network namespace operations use device pernet operations,
332 * otherwise use pernet subsys operations.
333 *
334 * Network interfaces need to be removed from a dying netns _before_
335 * subsys notifiers can be called, as most of the network code cleanup
336 * (which is done from subsys notifiers) runs with the assumption that
337 * dev_remove_pack has been called so no new packets will arrive during
338 * and after the cleanup functions have been called. dev_remove_pack
339 * is not per namespace so instead the guarantee of no more packets
340 * arriving in a network namespace is provided by ensuring that all
341 * network devices and all sockets have left the network namespace
342 * before the cleanup methods are called.
343 *
344 * For the longest time the ipv4 icmp code was registered as a pernet
345 * device which caused kernel oops, and panics during network
346 * namespace cleanup. So please don't get this wrong.
347 */
348 int register_pernet_subsys(struct pernet_operations *);
349 void unregister_pernet_subsys(struct pernet_operations *);
350 int register_pernet_device(struct pernet_operations *);
351 void unregister_pernet_device(struct pernet_operations *);
352
353 struct ctl_table;
354 struct ctl_table_header;
355
356 #ifdef CONFIG_SYSCTL
357 int net_sysctl_init(void);
358 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
359 struct ctl_table *table);
360 void unregister_net_sysctl_table(struct ctl_table_header *header);
361 #else
362 static inline int net_sysctl_init(void) { return 0; }
363 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
364 const char *path, struct ctl_table *table)
365 {
366 return NULL;
367 }
368 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
369 {
370 }
371 #endif
372
373 static inline int rt_genid_ipv4(struct net *net)
374 {
375 return atomic_read(&net->ipv4.rt_genid);
376 }
377
378 static inline void rt_genid_bump_ipv4(struct net *net)
379 {
380 atomic_inc(&net->ipv4.rt_genid);
381 }
382
383 extern void (*__fib6_flush_trees)(struct net *net);
384 static inline void rt_genid_bump_ipv6(struct net *net)
385 {
386 if (__fib6_flush_trees)
387 __fib6_flush_trees(net);
388 }
389
390 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
391 static inline struct netns_ieee802154_lowpan *
392 net_ieee802154_lowpan(struct net *net)
393 {
394 return &net->ieee802154_lowpan;
395 }
396 #endif
397
398 /* For callers who don't really care about whether it's IPv4 or IPv6 */
399 static inline void rt_genid_bump_all(struct net *net)
400 {
401 rt_genid_bump_ipv4(net);
402 rt_genid_bump_ipv6(net);
403 }
404
405 static inline int fnhe_genid(struct net *net)
406 {
407 return atomic_read(&net->fnhe_genid);
408 }
409
410 static inline void fnhe_genid_bump(struct net *net)
411 {
412 atomic_inc(&net->fnhe_genid);
413 }
414
415 #endif /* __NET_NET_NAMESPACE_H */