xref: /openbmc/linux/include/net/net_namespace.h (revision 4f3db074)
1 /*
2  * Operations on the network namespace
3  */
4 #ifndef __NET_NET_NAMESPACE_H
5 #define __NET_NET_NAMESPACE_H
6 
7 #include <linux/atomic.h>
8 #include <linux/workqueue.h>
9 #include <linux/list.h>
10 #include <linux/sysctl.h>
11 
12 #include <net/flow.h>
13 #include <net/netns/core.h>
14 #include <net/netns/mib.h>
15 #include <net/netns/unix.h>
16 #include <net/netns/packet.h>
17 #include <net/netns/ipv4.h>
18 #include <net/netns/ipv6.h>
19 #include <net/netns/ieee802154_6lowpan.h>
20 #include <net/netns/sctp.h>
21 #include <net/netns/dccp.h>
22 #include <net/netns/netfilter.h>
23 #include <net/netns/x_tables.h>
24 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
25 #include <net/netns/conntrack.h>
26 #endif
27 #include <net/netns/nftables.h>
28 #include <net/netns/xfrm.h>
29 #include <net/netns/mpls.h>
30 #include <linux/ns_common.h>
31 
32 struct user_namespace;
33 struct proc_dir_entry;
34 struct net_device;
35 struct sock;
36 struct ctl_table_header;
37 struct net_generic;
38 struct sock;
39 struct netns_ipvs;
40 
41 
42 #define NETDEV_HASHBITS    8
43 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
44 
45 struct net {
46 	atomic_t		passive;	/* To decided when the network
47 						 * namespace should be freed.
48 						 */
49 	atomic_t		count;		/* To decided when the network
50 						 *  namespace should be shut down.
51 						 */
52 	spinlock_t		rules_mod_lock;
53 
54 	atomic64_t		cookie_gen;
55 
56 	struct list_head	list;		/* list of network namespaces */
57 	struct list_head	cleanup_list;	/* namespaces on death row */
58 	struct list_head	exit_list;	/* Use only net_mutex */
59 
60 	struct user_namespace   *user_ns;	/* Owning user namespace */
61 	struct idr		netns_ids;
62 
63 	struct ns_common	ns;
64 
65 	struct proc_dir_entry 	*proc_net;
66 	struct proc_dir_entry 	*proc_net_stat;
67 
68 #ifdef CONFIG_SYSCTL
69 	struct ctl_table_set	sysctls;
70 #endif
71 
72 	struct sock 		*rtnl;			/* rtnetlink socket */
73 	struct sock		*genl_sock;
74 
75 	struct list_head 	dev_base_head;
76 	struct hlist_head 	*dev_name_head;
77 	struct hlist_head	*dev_index_head;
78 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
79 	int			ifindex;
80 	unsigned int		dev_unreg_count;
81 
82 	/* core fib_rules */
83 	struct list_head	rules_ops;
84 
85 
86 	struct net_device       *loopback_dev;          /* The loopback */
87 	struct netns_core	core;
88 	struct netns_mib	mib;
89 	struct netns_packet	packet;
90 	struct netns_unix	unx;
91 	struct netns_ipv4	ipv4;
92 #if IS_ENABLED(CONFIG_IPV6)
93 	struct netns_ipv6	ipv6;
94 #endif
95 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
96 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
97 #endif
98 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
99 	struct netns_sctp	sctp;
100 #endif
101 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
102 	struct netns_dccp	dccp;
103 #endif
104 #ifdef CONFIG_NETFILTER
105 	struct netns_nf		nf;
106 	struct netns_xt		xt;
107 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
108 	struct netns_ct		ct;
109 #endif
110 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
111 	struct netns_nftables	nft;
112 #endif
113 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
114 	struct netns_nf_frag	nf_frag;
115 #endif
116 	struct sock		*nfnl;
117 	struct sock		*nfnl_stash;
118 #endif
119 #ifdef CONFIG_WEXT_CORE
120 	struct sk_buff_head	wext_nlevents;
121 #endif
122 	struct net_generic __rcu	*gen;
123 
124 	/* Note : following structs are cache line aligned */
125 #ifdef CONFIG_XFRM
126 	struct netns_xfrm	xfrm;
127 #endif
128 #if IS_ENABLED(CONFIG_IP_VS)
129 	struct netns_ipvs	*ipvs;
130 #endif
131 #if IS_ENABLED(CONFIG_MPLS)
132 	struct netns_mpls	mpls;
133 #endif
134 	struct sock		*diag_nlsk;
135 	atomic_t		fnhe_genid;
136 };
137 
138 #include <linux/seq_file_net.h>
139 
140 /* Init's network namespace */
141 extern struct net init_net;
142 
143 #ifdef CONFIG_NET_NS
144 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
145 			struct net *old_net);
146 
147 #else /* CONFIG_NET_NS */
148 #include <linux/sched.h>
149 #include <linux/nsproxy.h>
150 static inline struct net *copy_net_ns(unsigned long flags,
151 	struct user_namespace *user_ns, struct net *old_net)
152 {
153 	if (flags & CLONE_NEWNET)
154 		return ERR_PTR(-EINVAL);
155 	return old_net;
156 }
157 #endif /* CONFIG_NET_NS */
158 
159 
160 extern struct list_head net_namespace_list;
161 
162 struct net *get_net_ns_by_pid(pid_t pid);
163 struct net *get_net_ns_by_fd(int pid);
164 
165 #ifdef CONFIG_SYSCTL
166 void ipx_register_sysctl(void);
167 void ipx_unregister_sysctl(void);
168 #else
169 #define ipx_register_sysctl()
170 #define ipx_unregister_sysctl()
171 #endif
172 
173 #ifdef CONFIG_NET_NS
174 void __put_net(struct net *net);
175 
176 static inline struct net *get_net(struct net *net)
177 {
178 	atomic_inc(&net->count);
179 	return net;
180 }
181 
182 static inline struct net *maybe_get_net(struct net *net)
183 {
184 	/* Used when we know struct net exists but we
185 	 * aren't guaranteed a previous reference count
186 	 * exists.  If the reference count is zero this
187 	 * function fails and returns NULL.
188 	 */
189 	if (!atomic_inc_not_zero(&net->count))
190 		net = NULL;
191 	return net;
192 }
193 
194 static inline void put_net(struct net *net)
195 {
196 	if (atomic_dec_and_test(&net->count))
197 		__put_net(net);
198 }
199 
200 static inline
201 int net_eq(const struct net *net1, const struct net *net2)
202 {
203 	return net1 == net2;
204 }
205 
206 void net_drop_ns(void *);
207 
208 #else
209 
210 static inline struct net *get_net(struct net *net)
211 {
212 	return net;
213 }
214 
215 static inline void put_net(struct net *net)
216 {
217 }
218 
219 static inline struct net *maybe_get_net(struct net *net)
220 {
221 	return net;
222 }
223 
224 static inline
225 int net_eq(const struct net *net1, const struct net *net2)
226 {
227 	return 1;
228 }
229 
230 #define net_drop_ns NULL
231 #endif
232 
233 
234 typedef struct {
235 #ifdef CONFIG_NET_NS
236 	struct net *net;
237 #endif
238 } possible_net_t;
239 
240 static inline void write_pnet(possible_net_t *pnet, struct net *net)
241 {
242 #ifdef CONFIG_NET_NS
243 	pnet->net = net;
244 #endif
245 }
246 
247 static inline struct net *read_pnet(const possible_net_t *pnet)
248 {
249 #ifdef CONFIG_NET_NS
250 	return pnet->net;
251 #else
252 	return &init_net;
253 #endif
254 }
255 
256 #define for_each_net(VAR)				\
257 	list_for_each_entry(VAR, &net_namespace_list, list)
258 
259 #define for_each_net_rcu(VAR)				\
260 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
261 
262 #ifdef CONFIG_NET_NS
263 #define __net_init
264 #define __net_exit
265 #define __net_initdata
266 #define __net_initconst
267 #else
268 #define __net_init	__init
269 #define __net_exit	__exit_refok
270 #define __net_initdata	__initdata
271 #define __net_initconst	__initconst
272 #endif
273 
274 int peernet2id(struct net *net, struct net *peer);
275 struct net *get_net_ns_by_id(struct net *net, int id);
276 
277 struct pernet_operations {
278 	struct list_head list;
279 	int (*init)(struct net *net);
280 	void (*exit)(struct net *net);
281 	void (*exit_batch)(struct list_head *net_exit_list);
282 	int *id;
283 	size_t size;
284 };
285 
286 /*
287  * Use these carefully.  If you implement a network device and it
288  * needs per network namespace operations use device pernet operations,
289  * otherwise use pernet subsys operations.
290  *
291  * Network interfaces need to be removed from a dying netns _before_
292  * subsys notifiers can be called, as most of the network code cleanup
293  * (which is done from subsys notifiers) runs with the assumption that
294  * dev_remove_pack has been called so no new packets will arrive during
295  * and after the cleanup functions have been called.  dev_remove_pack
296  * is not per namespace so instead the guarantee of no more packets
297  * arriving in a network namespace is provided by ensuring that all
298  * network devices and all sockets have left the network namespace
299  * before the cleanup methods are called.
300  *
301  * For the longest time the ipv4 icmp code was registered as a pernet
302  * device which caused kernel oops, and panics during network
303  * namespace cleanup.   So please don't get this wrong.
304  */
305 int register_pernet_subsys(struct pernet_operations *);
306 void unregister_pernet_subsys(struct pernet_operations *);
307 int register_pernet_device(struct pernet_operations *);
308 void unregister_pernet_device(struct pernet_operations *);
309 
310 struct ctl_table;
311 struct ctl_table_header;
312 
313 #ifdef CONFIG_SYSCTL
314 int net_sysctl_init(void);
315 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
316 					     struct ctl_table *table);
317 void unregister_net_sysctl_table(struct ctl_table_header *header);
318 #else
319 static inline int net_sysctl_init(void) { return 0; }
320 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
321 	const char *path, struct ctl_table *table)
322 {
323 	return NULL;
324 }
325 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
326 {
327 }
328 #endif
329 
330 static inline int rt_genid_ipv4(struct net *net)
331 {
332 	return atomic_read(&net->ipv4.rt_genid);
333 }
334 
335 static inline void rt_genid_bump_ipv4(struct net *net)
336 {
337 	atomic_inc(&net->ipv4.rt_genid);
338 }
339 
340 extern void (*__fib6_flush_trees)(struct net *net);
341 static inline void rt_genid_bump_ipv6(struct net *net)
342 {
343 	if (__fib6_flush_trees)
344 		__fib6_flush_trees(net);
345 }
346 
347 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
348 static inline struct netns_ieee802154_lowpan *
349 net_ieee802154_lowpan(struct net *net)
350 {
351 	return &net->ieee802154_lowpan;
352 }
353 #endif
354 
355 /* For callers who don't really care about whether it's IPv4 or IPv6 */
356 static inline void rt_genid_bump_all(struct net *net)
357 {
358 	rt_genid_bump_ipv4(net);
359 	rt_genid_bump_ipv6(net);
360 }
361 
362 static inline int fnhe_genid(struct net *net)
363 {
364 	return atomic_read(&net->fnhe_genid);
365 }
366 
367 static inline void fnhe_genid_bump(struct net *net)
368 {
369 	atomic_inc(&net->fnhe_genid);
370 }
371 
372 #endif /* __NET_NET_NAMESPACE_H */
373