xref: /openbmc/linux/include/net/net_namespace.h (revision d0b73b48)
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/netns/core.h>
13 #include <net/netns/mib.h>
14 #include <net/netns/unix.h>
15 #include <net/netns/packet.h>
16 #include <net/netns/ipv4.h>
17 #include <net/netns/ipv6.h>
18 #include <net/netns/sctp.h>
19 #include <net/netns/dccp.h>
20 #include <net/netns/x_tables.h>
21 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
22 #include <net/netns/conntrack.h>
23 #endif
24 #include <net/netns/xfrm.h>
25 
26 struct user_namespace;
27 struct proc_dir_entry;
28 struct net_device;
29 struct sock;
30 struct ctl_table_header;
31 struct net_generic;
32 struct sock;
33 struct netns_ipvs;
34 
35 
36 #define NETDEV_HASHBITS    8
37 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
38 
39 struct net {
40 	atomic_t		passive;	/* To decided when the network
41 						 * namespace should be freed.
42 						 */
43 	atomic_t		count;		/* To decided when the network
44 						 *  namespace should be shut down.
45 						 */
46 #ifdef NETNS_REFCNT_DEBUG
47 	atomic_t		use_count;	/* To track references we
48 						 * destroy on demand
49 						 */
50 #endif
51 	spinlock_t		rules_mod_lock;
52 
53 	struct list_head	list;		/* list of network namespaces */
54 	struct list_head	cleanup_list;	/* namespaces on death row */
55 	struct list_head	exit_list;	/* Use only net_mutex */
56 
57 	struct user_namespace   *user_ns;	/* Owning user namespace */
58 
59 	unsigned int		proc_inum;
60 
61 	struct proc_dir_entry 	*proc_net;
62 	struct proc_dir_entry 	*proc_net_stat;
63 
64 #ifdef CONFIG_SYSCTL
65 	struct ctl_table_set	sysctls;
66 #endif
67 
68 	struct sock 		*rtnl;			/* rtnetlink socket */
69 	struct sock		*genl_sock;
70 
71 	struct list_head 	dev_base_head;
72 	struct hlist_head 	*dev_name_head;
73 	struct hlist_head	*dev_index_head;
74 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
75 	int			ifindex;
76 
77 	/* core fib_rules */
78 	struct list_head	rules_ops;
79 
80 
81 	struct net_device       *loopback_dev;          /* The loopback */
82 	struct netns_core	core;
83 	struct netns_mib	mib;
84 	struct netns_packet	packet;
85 	struct netns_unix	unx;
86 	struct netns_ipv4	ipv4;
87 #if IS_ENABLED(CONFIG_IPV6)
88 	struct netns_ipv6	ipv6;
89 #endif
90 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
91 	struct netns_sctp	sctp;
92 #endif
93 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
94 	struct netns_dccp	dccp;
95 #endif
96 #ifdef CONFIG_NETFILTER
97 	struct netns_xt		xt;
98 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
99 	struct netns_ct		ct;
100 #endif
101 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
102 	struct netns_nf_frag	nf_frag;
103 #endif
104 	struct sock		*nfnl;
105 	struct sock		*nfnl_stash;
106 #endif
107 #ifdef CONFIG_WEXT_CORE
108 	struct sk_buff_head	wext_nlevents;
109 #endif
110 	struct net_generic __rcu	*gen;
111 
112 	/* Note : following structs are cache line aligned */
113 #ifdef CONFIG_XFRM
114 	struct netns_xfrm	xfrm;
115 #endif
116 	struct netns_ipvs	*ipvs;
117 	struct sock		*diag_nlsk;
118 	atomic_t		rt_genid;
119 };
120 
121 /*
122  * ifindex generation is per-net namespace, and loopback is
123  * always the 1st device in ns (see net_dev_init), thus any
124  * loopback device should get ifindex 1
125  */
126 
127 #define LOOPBACK_IFINDEX	1
128 
129 #include <linux/seq_file_net.h>
130 
131 /* Init's network namespace */
132 extern struct net init_net;
133 
134 #ifdef CONFIG_NET_NS
135 extern struct net *copy_net_ns(unsigned long flags,
136 	struct user_namespace *user_ns, struct net *old_net);
137 
138 #else /* CONFIG_NET_NS */
139 #include <linux/sched.h>
140 #include <linux/nsproxy.h>
141 static inline struct net *copy_net_ns(unsigned long flags,
142 	struct user_namespace *user_ns, struct net *old_net)
143 {
144 	if (flags & CLONE_NEWNET)
145 		return ERR_PTR(-EINVAL);
146 	return old_net;
147 }
148 #endif /* CONFIG_NET_NS */
149 
150 
151 extern struct list_head net_namespace_list;
152 
153 extern struct net *get_net_ns_by_pid(pid_t pid);
154 extern struct net *get_net_ns_by_fd(int pid);
155 
156 #ifdef CONFIG_NET_NS
157 extern void __put_net(struct net *net);
158 
159 static inline struct net *get_net(struct net *net)
160 {
161 	atomic_inc(&net->count);
162 	return net;
163 }
164 
165 static inline struct net *maybe_get_net(struct net *net)
166 {
167 	/* Used when we know struct net exists but we
168 	 * aren't guaranteed a previous reference count
169 	 * exists.  If the reference count is zero this
170 	 * function fails and returns NULL.
171 	 */
172 	if (!atomic_inc_not_zero(&net->count))
173 		net = NULL;
174 	return net;
175 }
176 
177 static inline void put_net(struct net *net)
178 {
179 	if (atomic_dec_and_test(&net->count))
180 		__put_net(net);
181 }
182 
183 static inline
184 int net_eq(const struct net *net1, const struct net *net2)
185 {
186 	return net1 == net2;
187 }
188 
189 extern void net_drop_ns(void *);
190 
191 #else
192 
193 static inline struct net *get_net(struct net *net)
194 {
195 	return net;
196 }
197 
198 static inline void put_net(struct net *net)
199 {
200 }
201 
202 static inline struct net *maybe_get_net(struct net *net)
203 {
204 	return net;
205 }
206 
207 static inline
208 int net_eq(const struct net *net1, const struct net *net2)
209 {
210 	return 1;
211 }
212 
213 #define net_drop_ns NULL
214 #endif
215 
216 
217 #ifdef NETNS_REFCNT_DEBUG
218 static inline struct net *hold_net(struct net *net)
219 {
220 	if (net)
221 		atomic_inc(&net->use_count);
222 	return net;
223 }
224 
225 static inline void release_net(struct net *net)
226 {
227 	if (net)
228 		atomic_dec(&net->use_count);
229 }
230 #else
231 static inline struct net *hold_net(struct net *net)
232 {
233 	return net;
234 }
235 
236 static inline void release_net(struct net *net)
237 {
238 }
239 #endif
240 
241 #ifdef CONFIG_NET_NS
242 
243 static inline void write_pnet(struct net **pnet, struct net *net)
244 {
245 	*pnet = net;
246 }
247 
248 static inline struct net *read_pnet(struct net * const *pnet)
249 {
250 	return *pnet;
251 }
252 
253 #else
254 
255 #define write_pnet(pnet, net)	do { (void)(net);} while (0)
256 #define read_pnet(pnet)		(&init_net)
257 
258 #endif
259 
260 #define for_each_net(VAR)				\
261 	list_for_each_entry(VAR, &net_namespace_list, list)
262 
263 #define for_each_net_rcu(VAR)				\
264 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
265 
266 #ifdef CONFIG_NET_NS
267 #define __net_init
268 #define __net_exit
269 #define __net_initdata
270 #define __net_initconst
271 #else
272 #define __net_init	__init
273 #define __net_exit	__exit_refok
274 #define __net_initdata	__initdata
275 #define __net_initconst	__initconst
276 #endif
277 
278 struct pernet_operations {
279 	struct list_head list;
280 	int (*init)(struct net *net);
281 	void (*exit)(struct net *net);
282 	void (*exit_batch)(struct list_head *net_exit_list);
283 	int *id;
284 	size_t size;
285 };
286 
287 /*
288  * Use these carefully.  If you implement a network device and it
289  * needs per network namespace operations use device pernet operations,
290  * otherwise use pernet subsys operations.
291  *
292  * Network interfaces need to be removed from a dying netns _before_
293  * subsys notifiers can be called, as most of the network code cleanup
294  * (which is done from subsys notifiers) runs with the assumption that
295  * dev_remove_pack has been called so no new packets will arrive during
296  * and after the cleanup functions have been called.  dev_remove_pack
297  * is not per namespace so instead the guarantee of no more packets
298  * arriving in a network namespace is provided by ensuring that all
299  * network devices and all sockets have left the network namespace
300  * before the cleanup methods are called.
301  *
302  * For the longest time the ipv4 icmp code was registered as a pernet
303  * device which caused kernel oops, and panics during network
304  * namespace cleanup.   So please don't get this wrong.
305  */
306 extern int register_pernet_subsys(struct pernet_operations *);
307 extern void unregister_pernet_subsys(struct pernet_operations *);
308 extern int register_pernet_device(struct pernet_operations *);
309 extern void unregister_pernet_device(struct pernet_operations *);
310 
311 struct ctl_table;
312 struct ctl_table_header;
313 
314 #ifdef CONFIG_SYSCTL
315 extern int net_sysctl_init(void);
316 extern struct ctl_table_header *register_net_sysctl(struct net *net,
317 	const char *path, struct ctl_table *table);
318 extern void unregister_net_sysctl_table(struct ctl_table_header *header);
319 #else
320 static inline int net_sysctl_init(void) { return 0; }
321 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
322 	const char *path, struct ctl_table *table)
323 {
324 	return NULL;
325 }
326 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
327 {
328 }
329 #endif
330 
331 static inline int rt_genid(struct net *net)
332 {
333 	return atomic_read(&net->rt_genid);
334 }
335 
336 static inline void rt_genid_bump(struct net *net)
337 {
338 	atomic_inc(&net->rt_genid);
339 }
340 
341 #endif /* __NET_NET_NAMESPACE_H */
342