Joe Perches | e005d19 | 2012-05-16 19:58:40 +0000 | [diff] [blame] | 1 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
| 2 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 3 | #include <linux/workqueue.h> |
| 4 | #include <linux/rtnetlink.h> |
| 5 | #include <linux/cache.h> |
| 6 | #include <linux/slab.h> |
| 7 | #include <linux/list.h> |
| 8 | #include <linux/delay.h> |
Eric W. Biederman | 9dd776b | 2007-09-26 22:04:26 -0700 | [diff] [blame] | 9 | #include <linux/sched.h> |
Pavel Emelyanov | c93cf61 | 2008-04-15 00:35:23 -0700 | [diff] [blame] | 10 | #include <linux/idr.h> |
Johannes Berg | 11a28d3 | 2009-07-10 09:51:33 +0000 | [diff] [blame] | 11 | #include <linux/rculist.h> |
Johannes Berg | 30ffee8 | 2009-07-10 09:51:35 +0000 | [diff] [blame] | 12 | #include <linux/nsproxy.h> |
David Howells | 0bb80f2 | 2013-04-12 01:50:06 +0100 | [diff] [blame] | 13 | #include <linux/fs.h> |
| 14 | #include <linux/proc_ns.h> |
Eric W. Biederman | f063052 | 2011-05-04 17:51:50 -0700 | [diff] [blame] | 15 | #include <linux/file.h> |
Paul Gortmaker | bc3b2d7 | 2011-07-15 11:47:34 -0400 | [diff] [blame] | 16 | #include <linux/export.h> |
Eric W. Biederman | 038e733 | 2012-06-14 02:31:10 -0700 | [diff] [blame] | 17 | #include <linux/user_namespace.h> |
Nicolas Dichtel | 0c7aecd | 2015-01-15 15:11:15 +0100 | [diff] [blame] | 18 | #include <linux/net_namespace.h> |
| 19 | #include <linux/rtnetlink.h> |
| 20 | #include <net/sock.h> |
| 21 | #include <net/netlink.h> |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 22 | #include <net/net_namespace.h> |
Pavel Emelyanov | dec827d | 2008-04-15 00:36:08 -0700 | [diff] [blame] | 23 | #include <net/netns/generic.h> |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 24 | |
| 25 | /* |
| 26 | * Our network namespace constructor/destructor lists |
| 27 | */ |
| 28 | |
| 29 | static LIST_HEAD(pernet_list); |
| 30 | static struct list_head *first_device = &pernet_list; |
Cong Wang | 200b916 | 2014-05-12 15:11:20 -0700 | [diff] [blame] | 31 | DEFINE_MUTEX(net_mutex); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 32 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 33 | LIST_HEAD(net_namespace_list); |
Alexey Dobriyan | b76a461 | 2008-10-08 11:35:06 +0200 | [diff] [blame] | 34 | EXPORT_SYMBOL_GPL(net_namespace_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 35 | |
Rustad, Mark D | 734b654 | 2012-07-18 09:06:07 +0000 | [diff] [blame] | 36 | struct net init_net = { |
| 37 | .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head), |
| 38 | }; |
Denis V. Lunev | ff4b950 | 2008-01-22 22:05:33 -0800 | [diff] [blame] | 39 | EXPORT_SYMBOL(init_net); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 40 | |
Pavel Emelyanov | dec827d | 2008-04-15 00:36:08 -0700 | [diff] [blame] | 41 | #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */ |
| 42 | |
Eric Dumazet | 073862b | 2012-01-26 00:41:38 +0000 | [diff] [blame] | 43 | static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS; |
| 44 | |
| 45 | static struct net_generic *net_alloc_generic(void) |
| 46 | { |
| 47 | struct net_generic *ng; |
| 48 | size_t generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]); |
| 49 | |
| 50 | ng = kzalloc(generic_size, GFP_KERNEL); |
| 51 | if (ng) |
| 52 | ng->len = max_gen_ptrs; |
| 53 | |
| 54 | return ng; |
| 55 | } |
| 56 | |
Jiri Pirko | 05fceb4 | 2010-04-23 01:40:47 +0000 | [diff] [blame] | 57 | static int net_assign_generic(struct net *net, int id, void *data) |
| 58 | { |
| 59 | struct net_generic *ng, *old_ng; |
| 60 | |
| 61 | BUG_ON(!mutex_is_locked(&net_mutex)); |
| 62 | BUG_ON(id == 0); |
| 63 | |
Eric Dumazet | 1c87733 | 2010-10-25 03:20:11 +0000 | [diff] [blame] | 64 | old_ng = rcu_dereference_protected(net->gen, |
| 65 | lockdep_is_held(&net_mutex)); |
| 66 | ng = old_ng; |
Jiri Pirko | 05fceb4 | 2010-04-23 01:40:47 +0000 | [diff] [blame] | 67 | if (old_ng->len >= id) |
| 68 | goto assign; |
| 69 | |
Eric Dumazet | 073862b | 2012-01-26 00:41:38 +0000 | [diff] [blame] | 70 | ng = net_alloc_generic(); |
Jiri Pirko | 05fceb4 | 2010-04-23 01:40:47 +0000 | [diff] [blame] | 71 | if (ng == NULL) |
| 72 | return -ENOMEM; |
| 73 | |
| 74 | /* |
| 75 | * Some synchronisation notes: |
| 76 | * |
| 77 | * The net_generic explores the net->gen array inside rcu |
| 78 | * read section. Besides once set the net->gen->ptr[x] |
| 79 | * pointer never changes (see rules in netns/generic.h). |
| 80 | * |
| 81 | * That said, we simply duplicate this array and schedule |
| 82 | * the old copy for kfree after a grace period. |
| 83 | */ |
| 84 | |
Jiri Pirko | 05fceb4 | 2010-04-23 01:40:47 +0000 | [diff] [blame] | 85 | memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*)); |
| 86 | |
| 87 | rcu_assign_pointer(net->gen, ng); |
Lai Jiangshan | 04d4dfe | 2011-03-18 12:06:32 +0800 | [diff] [blame] | 88 | kfree_rcu(old_ng, rcu); |
Jiri Pirko | 05fceb4 | 2010-04-23 01:40:47 +0000 | [diff] [blame] | 89 | assign: |
| 90 | ng->ptr[id - 1] = data; |
| 91 | return 0; |
| 92 | } |
| 93 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 94 | static int ops_init(const struct pernet_operations *ops, struct net *net) |
| 95 | { |
Julian Anastasov | b922934 | 2012-04-16 04:43:15 +0000 | [diff] [blame] | 96 | int err = -ENOMEM; |
| 97 | void *data = NULL; |
| 98 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 99 | if (ops->id && ops->size) { |
Julian Anastasov | b922934 | 2012-04-16 04:43:15 +0000 | [diff] [blame] | 100 | data = kzalloc(ops->size, GFP_KERNEL); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 101 | if (!data) |
Julian Anastasov | b922934 | 2012-04-16 04:43:15 +0000 | [diff] [blame] | 102 | goto out; |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 103 | |
| 104 | err = net_assign_generic(net, *ops->id, data); |
Julian Anastasov | b922934 | 2012-04-16 04:43:15 +0000 | [diff] [blame] | 105 | if (err) |
| 106 | goto cleanup; |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 107 | } |
Julian Anastasov | b922934 | 2012-04-16 04:43:15 +0000 | [diff] [blame] | 108 | err = 0; |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 109 | if (ops->init) |
Julian Anastasov | b922934 | 2012-04-16 04:43:15 +0000 | [diff] [blame] | 110 | err = ops->init(net); |
| 111 | if (!err) |
| 112 | return 0; |
| 113 | |
| 114 | cleanup: |
| 115 | kfree(data); |
| 116 | |
| 117 | out: |
| 118 | return err; |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 119 | } |
| 120 | |
| 121 | static void ops_free(const struct pernet_operations *ops, struct net *net) |
| 122 | { |
| 123 | if (ops->id && ops->size) { |
| 124 | int id = *ops->id; |
| 125 | kfree(net_generic(net, id)); |
| 126 | } |
| 127 | } |
| 128 | |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 129 | static void ops_exit_list(const struct pernet_operations *ops, |
| 130 | struct list_head *net_exit_list) |
| 131 | { |
| 132 | struct net *net; |
| 133 | if (ops->exit) { |
| 134 | list_for_each_entry(net, net_exit_list, exit_list) |
| 135 | ops->exit(net); |
| 136 | } |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 137 | if (ops->exit_batch) |
| 138 | ops->exit_batch(net_exit_list); |
| 139 | } |
| 140 | |
| 141 | static void ops_free_list(const struct pernet_operations *ops, |
| 142 | struct list_head *net_exit_list) |
| 143 | { |
| 144 | struct net *net; |
| 145 | if (ops->size && ops->id) { |
| 146 | list_for_each_entry(net, net_exit_list, exit_list) |
| 147 | ops_free(ops, net); |
| 148 | } |
| 149 | } |
| 150 | |
Nicolas Dichtel | 0c7aecd | 2015-01-15 15:11:15 +0100 | [diff] [blame] | 151 | static int alloc_netid(struct net *net, struct net *peer, int reqid) |
| 152 | { |
| 153 | int min = 0, max = 0; |
| 154 | |
| 155 | ASSERT_RTNL(); |
| 156 | |
| 157 | if (reqid >= 0) { |
| 158 | min = reqid; |
| 159 | max = reqid + 1; |
| 160 | } |
| 161 | |
| 162 | return idr_alloc(&net->netns_ids, peer, min, max, GFP_KERNEL); |
| 163 | } |
| 164 | |
| 165 | /* This function is used by idr_for_each(). If net is equal to peer, the |
| 166 | * function returns the id so that idr_for_each() stops. Because we cannot |
| 167 | * returns the id 0 (idr_for_each() will not stop), we return the magic value |
| 168 | * NET_ID_ZERO (-1) for it. |
| 169 | */ |
| 170 | #define NET_ID_ZERO -1 |
| 171 | static int net_eq_idr(int id, void *net, void *peer) |
| 172 | { |
| 173 | if (net_eq(net, peer)) |
| 174 | return id ? : NET_ID_ZERO; |
| 175 | return 0; |
| 176 | } |
| 177 | |
| 178 | static int __peernet2id(struct net *net, struct net *peer, bool alloc) |
| 179 | { |
| 180 | int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); |
| 181 | |
| 182 | ASSERT_RTNL(); |
| 183 | |
| 184 | /* Magic value for id 0. */ |
| 185 | if (id == NET_ID_ZERO) |
| 186 | return 0; |
| 187 | if (id > 0) |
| 188 | return id; |
| 189 | |
| 190 | if (alloc) |
| 191 | return alloc_netid(net, peer, -1); |
| 192 | |
| 193 | return -ENOENT; |
| 194 | } |
| 195 | |
| 196 | /* This function returns the id of a peer netns. If no id is assigned, one will |
| 197 | * be allocated and returned. |
| 198 | */ |
| 199 | int peernet2id(struct net *net, struct net *peer) |
| 200 | { |
Nicolas Dichtel | 576b7cd | 2015-04-03 12:02:37 +0200 | [diff] [blame^] | 201 | bool alloc = atomic_read(&peer->count) == 0 ? false : true; |
| 202 | int id; |
Nicolas Dichtel | 0c7aecd | 2015-01-15 15:11:15 +0100 | [diff] [blame] | 203 | |
Nicolas Dichtel | 576b7cd | 2015-04-03 12:02:37 +0200 | [diff] [blame^] | 204 | id = __peernet2id(net, peer, alloc); |
Nicolas Dichtel | 0c7aecd | 2015-01-15 15:11:15 +0100 | [diff] [blame] | 205 | return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; |
| 206 | } |
Nicolas Dichtel | 193523b | 2015-01-20 15:15:47 +0100 | [diff] [blame] | 207 | EXPORT_SYMBOL(peernet2id); |
Nicolas Dichtel | 0c7aecd | 2015-01-15 15:11:15 +0100 | [diff] [blame] | 208 | |
| 209 | struct net *get_net_ns_by_id(struct net *net, int id) |
| 210 | { |
| 211 | struct net *peer; |
| 212 | |
| 213 | if (id < 0) |
| 214 | return NULL; |
| 215 | |
| 216 | rcu_read_lock(); |
| 217 | peer = idr_find(&net->netns_ids, id); |
| 218 | if (peer) |
| 219 | get_net(peer); |
| 220 | rcu_read_unlock(); |
| 221 | |
| 222 | return peer; |
| 223 | } |
| 224 | |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 225 | /* |
| 226 | * setup_net runs the initializers for the network namespace object. |
| 227 | */ |
Eric W. Biederman | 038e733 | 2012-06-14 02:31:10 -0700 | [diff] [blame] | 228 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 229 | { |
| 230 | /* Must be called with net_mutex held */ |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 231 | const struct pernet_operations *ops, *saved_ops; |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 232 | int error = 0; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 233 | LIST_HEAD(net_exit_list); |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 234 | |
| 235 | atomic_set(&net->count, 1); |
Al Viro | a685e08 | 2011-06-08 21:13:01 -0400 | [diff] [blame] | 236 | atomic_set(&net->passive, 1); |
Thomas Graf | 4e985ad | 2011-06-21 03:11:20 +0000 | [diff] [blame] | 237 | net->dev_base_seq = 1; |
Eric W. Biederman | 038e733 | 2012-06-14 02:31:10 -0700 | [diff] [blame] | 238 | net->user_ns = user_ns; |
Nicolas Dichtel | 0c7aecd | 2015-01-15 15:11:15 +0100 | [diff] [blame] | 239 | idr_init(&net->netns_ids); |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 240 | |
Denis V. Lunev | 5d1e446 | 2008-04-16 01:58:04 -0700 | [diff] [blame] | 241 | #ifdef NETNS_REFCNT_DEBUG |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 242 | atomic_set(&net->use_count, 0); |
Denis V. Lunev | 5d1e446 | 2008-04-16 01:58:04 -0700 | [diff] [blame] | 243 | #endif |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 244 | |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 245 | list_for_each_entry(ops, &pernet_list, list) { |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 246 | error = ops_init(ops, net); |
| 247 | if (error < 0) |
| 248 | goto out_undo; |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 249 | } |
| 250 | out: |
| 251 | return error; |
| 252 | |
| 253 | out_undo: |
| 254 | /* Walk through the list backwards calling the exit functions |
| 255 | * for the pernet modules whose init functions did not fail. |
| 256 | */ |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 257 | list_add(&net->exit_list, &net_exit_list); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 258 | saved_ops = ops; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 259 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
| 260 | ops_exit_list(ops, &net_exit_list); |
| 261 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 262 | ops = saved_ops; |
| 263 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 264 | ops_free_list(ops, &net_exit_list); |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 265 | |
| 266 | rcu_barrier(); |
| 267 | goto out; |
| 268 | } |
| 269 | |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 270 | |
Clemens Noss | ebe47d4 | 2009-02-23 15:37:35 -0800 | [diff] [blame] | 271 | #ifdef CONFIG_NET_NS |
| 272 | static struct kmem_cache *net_cachep; |
| 273 | static struct workqueue_struct *netns_wq; |
| 274 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 275 | static struct net *net_alloc(void) |
| 276 | { |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 277 | struct net *net = NULL; |
| 278 | struct net_generic *ng; |
| 279 | |
| 280 | ng = net_alloc_generic(); |
| 281 | if (!ng) |
| 282 | goto out; |
| 283 | |
| 284 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); |
| 285 | if (!net) |
| 286 | goto out_free; |
| 287 | |
| 288 | rcu_assign_pointer(net->gen, ng); |
| 289 | out: |
| 290 | return net; |
| 291 | |
| 292 | out_free: |
| 293 | kfree(ng); |
| 294 | goto out; |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 295 | } |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 296 | |
Johann Felix Soden | 45a19b0 | 2007-11-07 01:30:30 -0800 | [diff] [blame] | 297 | static void net_free(struct net *net) |
| 298 | { |
Denis V. Lunev | 5d1e446 | 2008-04-16 01:58:04 -0700 | [diff] [blame] | 299 | #ifdef NETNS_REFCNT_DEBUG |
Johann Felix Soden | 45a19b0 | 2007-11-07 01:30:30 -0800 | [diff] [blame] | 300 | if (unlikely(atomic_read(&net->use_count) != 0)) { |
Joe Perches | e005d19 | 2012-05-16 19:58:40 +0000 | [diff] [blame] | 301 | pr_emerg("network namespace not free! Usage: %d\n", |
| 302 | atomic_read(&net->use_count)); |
Johann Felix Soden | 45a19b0 | 2007-11-07 01:30:30 -0800 | [diff] [blame] | 303 | return; |
| 304 | } |
Denis V. Lunev | 5d1e446 | 2008-04-16 01:58:04 -0700 | [diff] [blame] | 305 | #endif |
Eric Dumazet | 416c51e | 2014-09-09 08:24:53 -0700 | [diff] [blame] | 306 | kfree(rcu_access_pointer(net->gen)); |
Johann Felix Soden | 45a19b0 | 2007-11-07 01:30:30 -0800 | [diff] [blame] | 307 | kmem_cache_free(net_cachep, net); |
| 308 | } |
| 309 | |
Al Viro | a685e08 | 2011-06-08 21:13:01 -0400 | [diff] [blame] | 310 | void net_drop_ns(void *p) |
| 311 | { |
| 312 | struct net *ns = p; |
| 313 | if (ns && atomic_dec_and_test(&ns->passive)) |
| 314 | net_free(ns); |
| 315 | } |
| 316 | |
Eric W. Biederman | 038e733 | 2012-06-14 02:31:10 -0700 | [diff] [blame] | 317 | struct net *copy_net_ns(unsigned long flags, |
| 318 | struct user_namespace *user_ns, struct net *old_net) |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 319 | { |
Alexey Dobriyan | 088eb2d | 2009-05-04 11:12:14 -0700 | [diff] [blame] | 320 | struct net *net; |
| 321 | int rv; |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 322 | |
Rob Landley | 911cb19 | 2011-04-15 02:26:25 +0000 | [diff] [blame] | 323 | if (!(flags & CLONE_NEWNET)) |
| 324 | return get_net(old_net); |
| 325 | |
Alexey Dobriyan | 088eb2d | 2009-05-04 11:12:14 -0700 | [diff] [blame] | 326 | net = net_alloc(); |
| 327 | if (!net) |
| 328 | return ERR_PTR(-ENOMEM); |
Eric W. Biederman | 038e733 | 2012-06-14 02:31:10 -0700 | [diff] [blame] | 329 | |
| 330 | get_user_ns(user_ns); |
| 331 | |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 332 | mutex_lock(&net_mutex); |
Eric W. Biederman | 038e733 | 2012-06-14 02:31:10 -0700 | [diff] [blame] | 333 | rv = setup_net(net, user_ns); |
Alexey Dobriyan | 088eb2d | 2009-05-04 11:12:14 -0700 | [diff] [blame] | 334 | if (rv == 0) { |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 335 | rtnl_lock(); |
Johannes Berg | 11a28d3 | 2009-07-10 09:51:33 +0000 | [diff] [blame] | 336 | list_add_tail_rcu(&net->list, &net_namespace_list); |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 337 | rtnl_unlock(); |
| 338 | } |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 339 | mutex_unlock(&net_mutex); |
Alexey Dobriyan | 088eb2d | 2009-05-04 11:12:14 -0700 | [diff] [blame] | 340 | if (rv < 0) { |
Eric W. Biederman | 038e733 | 2012-06-14 02:31:10 -0700 | [diff] [blame] | 341 | put_user_ns(user_ns); |
Al Viro | a685e08 | 2011-06-08 21:13:01 -0400 | [diff] [blame] | 342 | net_drop_ns(net); |
Alexey Dobriyan | 088eb2d | 2009-05-04 11:12:14 -0700 | [diff] [blame] | 343 | return ERR_PTR(rv); |
| 344 | } |
| 345 | return net; |
| 346 | } |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 347 | |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 348 | static DEFINE_SPINLOCK(cleanup_list_lock); |
| 349 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ |
| 350 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 351 | static void cleanup_net(struct work_struct *work) |
| 352 | { |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 353 | const struct pernet_operations *ops; |
Nicolas Dichtel | 6d458f5 | 2015-04-03 12:02:36 +0200 | [diff] [blame] | 354 | struct net *net, *tmp; |
xiao jin | 1818ce4 | 2014-04-25 08:50:54 +0800 | [diff] [blame] | 355 | struct list_head net_kill_list; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 356 | LIST_HEAD(net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 357 | |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 358 | /* Atomically snapshot the list of namespaces to cleanup */ |
| 359 | spin_lock_irq(&cleanup_list_lock); |
| 360 | list_replace_init(&cleanup_list, &net_kill_list); |
| 361 | spin_unlock_irq(&cleanup_list_lock); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 362 | |
| 363 | mutex_lock(&net_mutex); |
| 364 | |
| 365 | /* Don't let anyone else find us. */ |
Eric W. Biederman | f4618d3 | 2007-09-26 22:40:08 -0700 | [diff] [blame] | 366 | rtnl_lock(); |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 367 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 368 | list_del_rcu(&net->list); |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 369 | list_add_tail(&net->exit_list, &net_exit_list); |
Nicolas Dichtel | 6d458f5 | 2015-04-03 12:02:36 +0200 | [diff] [blame] | 370 | for_each_net(tmp) { |
| 371 | int id = __peernet2id(tmp, net, false); |
| 372 | |
| 373 | if (id >= 0) |
| 374 | idr_remove(&tmp->netns_ids, id); |
| 375 | } |
| 376 | idr_destroy(&net->netns_ids); |
| 377 | |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 378 | } |
Eric W. Biederman | f4618d3 | 2007-09-26 22:40:08 -0700 | [diff] [blame] | 379 | rtnl_unlock(); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 380 | |
Johannes Berg | 11a28d3 | 2009-07-10 09:51:33 +0000 | [diff] [blame] | 381 | /* |
| 382 | * Another CPU might be rcu-iterating the list, wait for it. |
| 383 | * This needs to be before calling the exit() notifiers, so |
| 384 | * the rcu_barrier() below isn't sufficient alone. |
| 385 | */ |
| 386 | synchronize_rcu(); |
| 387 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 388 | /* Run all of the network namespace exit methods */ |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 389 | list_for_each_entry_reverse(ops, &pernet_list, list) |
| 390 | ops_exit_list(ops, &net_exit_list); |
| 391 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 392 | /* Free the net generic variables */ |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 393 | list_for_each_entry_reverse(ops, &pernet_list, list) |
| 394 | ops_free_list(ops, &net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 395 | |
| 396 | mutex_unlock(&net_mutex); |
| 397 | |
| 398 | /* Ensure there are no outstanding rcu callbacks using this |
| 399 | * network namespace. |
| 400 | */ |
| 401 | rcu_barrier(); |
| 402 | |
| 403 | /* Finally it is safe to free my network namespace structure */ |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 404 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
| 405 | list_del_init(&net->exit_list); |
Eric W. Biederman | 038e733 | 2012-06-14 02:31:10 -0700 | [diff] [blame] | 406 | put_user_ns(net->user_ns); |
Al Viro | a685e08 | 2011-06-08 21:13:01 -0400 | [diff] [blame] | 407 | net_drop_ns(net); |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 408 | } |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 409 | } |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 410 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 411 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 412 | void __put_net(struct net *net) |
| 413 | { |
| 414 | /* Cleanup the network namespace in process context */ |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 415 | unsigned long flags; |
| 416 | |
| 417 | spin_lock_irqsave(&cleanup_list_lock, flags); |
| 418 | list_add(&net->cleanup_list, &cleanup_list); |
| 419 | spin_unlock_irqrestore(&cleanup_list_lock, flags); |
| 420 | |
| 421 | queue_work(netns_wq, &net_cleanup_work); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 422 | } |
| 423 | EXPORT_SYMBOL_GPL(__put_net); |
| 424 | |
Stephen Rothwell | 956c920 | 2011-05-12 13:51:13 +1000 | [diff] [blame] | 425 | struct net *get_net_ns_by_fd(int fd) |
| 426 | { |
Stephen Rothwell | 956c920 | 2011-05-12 13:51:13 +1000 | [diff] [blame] | 427 | struct file *file; |
Al Viro | 33c4294 | 2014-11-01 02:32:53 -0400 | [diff] [blame] | 428 | struct ns_common *ns; |
Stephen Rothwell | 956c920 | 2011-05-12 13:51:13 +1000 | [diff] [blame] | 429 | struct net *net; |
| 430 | |
Stephen Rothwell | 956c920 | 2011-05-12 13:51:13 +1000 | [diff] [blame] | 431 | file = proc_ns_fget(fd); |
Al Viro | c316e6a | 2011-06-05 00:37:35 +0000 | [diff] [blame] | 432 | if (IS_ERR(file)) |
| 433 | return ERR_CAST(file); |
Stephen Rothwell | 956c920 | 2011-05-12 13:51:13 +1000 | [diff] [blame] | 434 | |
Al Viro | f77c801 | 2014-11-01 03:13:17 -0400 | [diff] [blame] | 435 | ns = get_proc_ns(file_inode(file)); |
Al Viro | 33c4294 | 2014-11-01 02:32:53 -0400 | [diff] [blame] | 436 | if (ns->ops == &netns_operations) |
| 437 | net = get_net(container_of(ns, struct net, ns)); |
Al Viro | c316e6a | 2011-06-05 00:37:35 +0000 | [diff] [blame] | 438 | else |
| 439 | net = ERR_PTR(-EINVAL); |
Stephen Rothwell | 956c920 | 2011-05-12 13:51:13 +1000 | [diff] [blame] | 440 | |
Al Viro | c316e6a | 2011-06-05 00:37:35 +0000 | [diff] [blame] | 441 | fput(file); |
Stephen Rothwell | 956c920 | 2011-05-12 13:51:13 +1000 | [diff] [blame] | 442 | return net; |
| 443 | } |
| 444 | |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 445 | #else |
Stephen Rothwell | 956c920 | 2011-05-12 13:51:13 +1000 | [diff] [blame] | 446 | struct net *get_net_ns_by_fd(int fd) |
| 447 | { |
| 448 | return ERR_PTR(-EINVAL); |
| 449 | } |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 450 | #endif |
Vadim Kochan | 4b681c8 | 2015-01-12 16:34:05 +0200 | [diff] [blame] | 451 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
Eric W. Biederman | 9dd776b | 2007-09-26 22:04:26 -0700 | [diff] [blame] | 452 | |
Johannes Berg | 30ffee8 | 2009-07-10 09:51:35 +0000 | [diff] [blame] | 453 | struct net *get_net_ns_by_pid(pid_t pid) |
| 454 | { |
| 455 | struct task_struct *tsk; |
| 456 | struct net *net; |
| 457 | |
| 458 | /* Lookup the network namespace */ |
| 459 | net = ERR_PTR(-ESRCH); |
| 460 | rcu_read_lock(); |
| 461 | tsk = find_task_by_vpid(pid); |
| 462 | if (tsk) { |
| 463 | struct nsproxy *nsproxy; |
Eric W. Biederman | 728dba3 | 2014-02-03 19:13:49 -0800 | [diff] [blame] | 464 | task_lock(tsk); |
| 465 | nsproxy = tsk->nsproxy; |
Johannes Berg | 30ffee8 | 2009-07-10 09:51:35 +0000 | [diff] [blame] | 466 | if (nsproxy) |
| 467 | net = get_net(nsproxy->net_ns); |
Eric W. Biederman | 728dba3 | 2014-02-03 19:13:49 -0800 | [diff] [blame] | 468 | task_unlock(tsk); |
Johannes Berg | 30ffee8 | 2009-07-10 09:51:35 +0000 | [diff] [blame] | 469 | } |
| 470 | rcu_read_unlock(); |
| 471 | return net; |
| 472 | } |
| 473 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); |
| 474 | |
Eric W. Biederman | 98f842e | 2011-06-15 10:21:48 -0700 | [diff] [blame] | 475 | static __net_init int net_ns_net_init(struct net *net) |
| 476 | { |
Al Viro | 33c4294 | 2014-11-01 02:32:53 -0400 | [diff] [blame] | 477 | #ifdef CONFIG_NET_NS |
| 478 | net->ns.ops = &netns_operations; |
| 479 | #endif |
Al Viro | 6344c43 | 2014-11-01 00:45:45 -0400 | [diff] [blame] | 480 | return ns_alloc_inum(&net->ns); |
Eric W. Biederman | 98f842e | 2011-06-15 10:21:48 -0700 | [diff] [blame] | 481 | } |
| 482 | |
| 483 | static __net_exit void net_ns_net_exit(struct net *net) |
| 484 | { |
Al Viro | 6344c43 | 2014-11-01 00:45:45 -0400 | [diff] [blame] | 485 | ns_free_inum(&net->ns); |
Eric W. Biederman | 98f842e | 2011-06-15 10:21:48 -0700 | [diff] [blame] | 486 | } |
| 487 | |
| 488 | static struct pernet_operations __net_initdata net_ns_ops = { |
| 489 | .init = net_ns_net_init, |
| 490 | .exit = net_ns_net_exit, |
| 491 | }; |
| 492 | |
Nicolas Dichtel | 0c7aecd | 2015-01-15 15:11:15 +0100 | [diff] [blame] | 493 | static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
| 494 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, |
| 495 | [NETNSA_NSID] = { .type = NLA_S32 }, |
| 496 | [NETNSA_PID] = { .type = NLA_U32 }, |
| 497 | [NETNSA_FD] = { .type = NLA_U32 }, |
| 498 | }; |
| 499 | |
| 500 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) |
| 501 | { |
| 502 | struct net *net = sock_net(skb->sk); |
| 503 | struct nlattr *tb[NETNSA_MAX + 1]; |
| 504 | struct net *peer; |
| 505 | int nsid, err; |
| 506 | |
| 507 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, |
| 508 | rtnl_net_policy); |
| 509 | if (err < 0) |
| 510 | return err; |
| 511 | if (!tb[NETNSA_NSID]) |
| 512 | return -EINVAL; |
| 513 | nsid = nla_get_s32(tb[NETNSA_NSID]); |
| 514 | |
| 515 | if (tb[NETNSA_PID]) |
| 516 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); |
| 517 | else if (tb[NETNSA_FD]) |
| 518 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); |
| 519 | else |
| 520 | return -EINVAL; |
| 521 | if (IS_ERR(peer)) |
| 522 | return PTR_ERR(peer); |
| 523 | |
| 524 | if (__peernet2id(net, peer, false) >= 0) { |
| 525 | err = -EEXIST; |
| 526 | goto out; |
| 527 | } |
| 528 | |
| 529 | err = alloc_netid(net, peer, nsid); |
| 530 | if (err > 0) |
| 531 | err = 0; |
| 532 | out: |
| 533 | put_net(peer); |
| 534 | return err; |
| 535 | } |
| 536 | |
| 537 | static int rtnl_net_get_size(void) |
| 538 | { |
| 539 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) |
| 540 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ |
| 541 | ; |
| 542 | } |
| 543 | |
| 544 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, |
| 545 | int cmd, struct net *net, struct net *peer) |
| 546 | { |
| 547 | struct nlmsghdr *nlh; |
| 548 | struct rtgenmsg *rth; |
| 549 | int id; |
| 550 | |
| 551 | ASSERT_RTNL(); |
| 552 | |
| 553 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); |
| 554 | if (!nlh) |
| 555 | return -EMSGSIZE; |
| 556 | |
| 557 | rth = nlmsg_data(nlh); |
| 558 | rth->rtgen_family = AF_UNSPEC; |
| 559 | |
| 560 | id = __peernet2id(net, peer, false); |
| 561 | if (id < 0) |
| 562 | id = NETNSA_NSID_NOT_ASSIGNED; |
| 563 | if (nla_put_s32(skb, NETNSA_NSID, id)) |
| 564 | goto nla_put_failure; |
| 565 | |
| 566 | nlmsg_end(skb, nlh); |
| 567 | return 0; |
| 568 | |
| 569 | nla_put_failure: |
| 570 | nlmsg_cancel(skb, nlh); |
| 571 | return -EMSGSIZE; |
| 572 | } |
| 573 | |
| 574 | static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) |
| 575 | { |
| 576 | struct net *net = sock_net(skb->sk); |
| 577 | struct nlattr *tb[NETNSA_MAX + 1]; |
| 578 | struct sk_buff *msg; |
| 579 | int err = -ENOBUFS; |
| 580 | struct net *peer; |
| 581 | |
| 582 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, |
| 583 | rtnl_net_policy); |
| 584 | if (err < 0) |
| 585 | return err; |
| 586 | if (tb[NETNSA_PID]) |
| 587 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); |
| 588 | else if (tb[NETNSA_FD]) |
| 589 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); |
| 590 | else |
| 591 | return -EINVAL; |
| 592 | |
| 593 | if (IS_ERR(peer)) |
| 594 | return PTR_ERR(peer); |
| 595 | |
| 596 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); |
| 597 | if (!msg) { |
| 598 | err = -ENOMEM; |
| 599 | goto out; |
| 600 | } |
| 601 | |
| 602 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, |
| 603 | RTM_GETNSID, net, peer); |
| 604 | if (err < 0) |
| 605 | goto err_out; |
| 606 | |
| 607 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); |
| 608 | goto out; |
| 609 | |
| 610 | err_out: |
| 611 | nlmsg_free(msg); |
| 612 | out: |
| 613 | put_net(peer); |
| 614 | return err; |
| 615 | } |
| 616 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 617 | static int __init net_ns_init(void) |
| 618 | { |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 619 | struct net_generic *ng; |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 620 | |
Pavel Emelyanov | d57a921 | 2007-11-01 00:46:50 -0700 | [diff] [blame] | 621 | #ifdef CONFIG_NET_NS |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 622 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
| 623 | SMP_CACHE_BYTES, |
| 624 | SLAB_PANIC, NULL); |
Benjamin Thery | 3ef1355 | 2007-11-19 23:18:16 -0800 | [diff] [blame] | 625 | |
| 626 | /* Create workqueue for cleanup */ |
| 627 | netns_wq = create_singlethread_workqueue("netns"); |
| 628 | if (!netns_wq) |
| 629 | panic("Could not create netns workq"); |
Pavel Emelyanov | d57a921 | 2007-11-01 00:46:50 -0700 | [diff] [blame] | 630 | #endif |
Benjamin Thery | 3ef1355 | 2007-11-19 23:18:16 -0800 | [diff] [blame] | 631 | |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 632 | ng = net_alloc_generic(); |
| 633 | if (!ng) |
| 634 | panic("Could not allocate generic netns"); |
| 635 | |
| 636 | rcu_assign_pointer(init_net.gen, ng); |
| 637 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 638 | mutex_lock(&net_mutex); |
Eric W. Biederman | 038e733 | 2012-06-14 02:31:10 -0700 | [diff] [blame] | 639 | if (setup_net(&init_net, &init_user_ns)) |
Stephen Hemminger | ca0f311 | 2009-05-21 15:10:31 -0700 | [diff] [blame] | 640 | panic("Could not setup the initial network namespace"); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 641 | |
Eric W. Biederman | f4618d3 | 2007-09-26 22:40:08 -0700 | [diff] [blame] | 642 | rtnl_lock(); |
Johannes Berg | 11a28d3 | 2009-07-10 09:51:33 +0000 | [diff] [blame] | 643 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
Eric W. Biederman | f4618d3 | 2007-09-26 22:40:08 -0700 | [diff] [blame] | 644 | rtnl_unlock(); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 645 | |
| 646 | mutex_unlock(&net_mutex); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 647 | |
Eric W. Biederman | 98f842e | 2011-06-15 10:21:48 -0700 | [diff] [blame] | 648 | register_pernet_subsys(&net_ns_ops); |
| 649 | |
Nicolas Dichtel | 0c7aecd | 2015-01-15 15:11:15 +0100 | [diff] [blame] | 650 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); |
| 651 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, NULL, NULL); |
| 652 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 653 | return 0; |
| 654 | } |
| 655 | |
| 656 | pure_initcall(net_ns_init); |
| 657 | |
Denis V. Lunev | ed160e8 | 2007-11-13 03:23:21 -0800 | [diff] [blame] | 658 | #ifdef CONFIG_NET_NS |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 659 | static int __register_pernet_operations(struct list_head *list, |
| 660 | struct pernet_operations *ops) |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 661 | { |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 662 | struct net *net; |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 663 | int error; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 664 | LIST_HEAD(net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 665 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 666 | list_add_tail(&ops->list, list); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 667 | if (ops->init || (ops->id && ops->size)) { |
Pavel Emelyanov | 1dba323 | 2007-11-01 00:42:43 -0700 | [diff] [blame] | 668 | for_each_net(net) { |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 669 | error = ops_init(ops, net); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 670 | if (error) |
| 671 | goto out_undo; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 672 | list_add_tail(&net->exit_list, &net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 673 | } |
| 674 | } |
Pavel Emelyanov | 1dba323 | 2007-11-01 00:42:43 -0700 | [diff] [blame] | 675 | return 0; |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 676 | |
| 677 | out_undo: |
| 678 | /* If I have an error cleanup all namespaces I initialized */ |
| 679 | list_del(&ops->list); |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 680 | ops_exit_list(ops, &net_exit_list); |
| 681 | ops_free_list(ops, &net_exit_list); |
Pavel Emelyanov | 1dba323 | 2007-11-01 00:42:43 -0700 | [diff] [blame] | 682 | return error; |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 683 | } |
| 684 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 685 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 686 | { |
| 687 | struct net *net; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 688 | LIST_HEAD(net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 689 | |
| 690 | list_del(&ops->list); |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 691 | for_each_net(net) |
| 692 | list_add_tail(&net->exit_list, &net_exit_list); |
| 693 | ops_exit_list(ops, &net_exit_list); |
| 694 | ops_free_list(ops, &net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 695 | } |
| 696 | |
Denis V. Lunev | ed160e8 | 2007-11-13 03:23:21 -0800 | [diff] [blame] | 697 | #else |
| 698 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 699 | static int __register_pernet_operations(struct list_head *list, |
| 700 | struct pernet_operations *ops) |
| 701 | { |
Julian Anastasov | b922934 | 2012-04-16 04:43:15 +0000 | [diff] [blame] | 702 | return ops_init(ops, &init_net); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 703 | } |
| 704 | |
| 705 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
| 706 | { |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 707 | LIST_HEAD(net_exit_list); |
| 708 | list_add(&init_net.exit_list, &net_exit_list); |
| 709 | ops_exit_list(ops, &net_exit_list); |
| 710 | ops_free_list(ops, &net_exit_list); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 711 | } |
| 712 | |
| 713 | #endif /* CONFIG_NET_NS */ |
| 714 | |
| 715 | static DEFINE_IDA(net_generic_ids); |
| 716 | |
Denis V. Lunev | ed160e8 | 2007-11-13 03:23:21 -0800 | [diff] [blame] | 717 | static int register_pernet_operations(struct list_head *list, |
| 718 | struct pernet_operations *ops) |
| 719 | { |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 720 | int error; |
| 721 | |
| 722 | if (ops->id) { |
| 723 | again: |
| 724 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); |
| 725 | if (error < 0) { |
| 726 | if (error == -EAGAIN) { |
| 727 | ida_pre_get(&net_generic_ids, GFP_KERNEL); |
| 728 | goto again; |
| 729 | } |
| 730 | return error; |
| 731 | } |
Eric Dumazet | 073862b | 2012-01-26 00:41:38 +0000 | [diff] [blame] | 732 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 733 | } |
| 734 | error = __register_pernet_operations(list, ops); |
Eric W. Biederman | 3a765ed | 2009-12-03 02:29:06 +0000 | [diff] [blame] | 735 | if (error) { |
| 736 | rcu_barrier(); |
| 737 | if (ops->id) |
| 738 | ida_remove(&net_generic_ids, *ops->id); |
| 739 | } |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 740 | |
| 741 | return error; |
Denis V. Lunev | ed160e8 | 2007-11-13 03:23:21 -0800 | [diff] [blame] | 742 | } |
| 743 | |
| 744 | static void unregister_pernet_operations(struct pernet_operations *ops) |
| 745 | { |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 746 | |
| 747 | __unregister_pernet_operations(ops); |
Eric W. Biederman | 3a765ed | 2009-12-03 02:29:06 +0000 | [diff] [blame] | 748 | rcu_barrier(); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 749 | if (ops->id) |
| 750 | ida_remove(&net_generic_ids, *ops->id); |
Denis V. Lunev | ed160e8 | 2007-11-13 03:23:21 -0800 | [diff] [blame] | 751 | } |
Pavel Emelyanov | c93cf61 | 2008-04-15 00:35:23 -0700 | [diff] [blame] | 752 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 753 | /** |
| 754 | * register_pernet_subsys - register a network namespace subsystem |
| 755 | * @ops: pernet operations structure for the subsystem |
| 756 | * |
| 757 | * Register a subsystem which has init and exit functions |
| 758 | * that are called when network namespaces are created and |
| 759 | * destroyed respectively. |
| 760 | * |
| 761 | * When registered all network namespace init functions are |
| 762 | * called for every existing network namespace. Allowing kernel |
| 763 | * modules to have a race free view of the set of network namespaces. |
| 764 | * |
| 765 | * When a new network namespace is created all of the init |
| 766 | * methods are called in the order in which they were registered. |
| 767 | * |
| 768 | * When a network namespace is destroyed all of the exit methods |
| 769 | * are called in the reverse of the order with which they were |
| 770 | * registered. |
| 771 | */ |
| 772 | int register_pernet_subsys(struct pernet_operations *ops) |
| 773 | { |
| 774 | int error; |
| 775 | mutex_lock(&net_mutex); |
| 776 | error = register_pernet_operations(first_device, ops); |
| 777 | mutex_unlock(&net_mutex); |
| 778 | return error; |
| 779 | } |
| 780 | EXPORT_SYMBOL_GPL(register_pernet_subsys); |
| 781 | |
| 782 | /** |
| 783 | * unregister_pernet_subsys - unregister a network namespace subsystem |
| 784 | * @ops: pernet operations structure to manipulate |
| 785 | * |
| 786 | * Remove the pernet operations structure from the list to be |
Oliver Pinter | 53379e5 | 2008-02-03 17:56:48 +0200 | [diff] [blame] | 787 | * used when network namespaces are created or destroyed. In |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 788 | * addition run the exit method for all existing network |
| 789 | * namespaces. |
| 790 | */ |
Jiri Pirko | b3c981d | 2010-04-25 00:49:56 -0700 | [diff] [blame] | 791 | void unregister_pernet_subsys(struct pernet_operations *ops) |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 792 | { |
| 793 | mutex_lock(&net_mutex); |
Jiri Pirko | b3c981d | 2010-04-25 00:49:56 -0700 | [diff] [blame] | 794 | unregister_pernet_operations(ops); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 795 | mutex_unlock(&net_mutex); |
| 796 | } |
| 797 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); |
| 798 | |
| 799 | /** |
| 800 | * register_pernet_device - register a network namespace device |
| 801 | * @ops: pernet operations structure for the subsystem |
| 802 | * |
| 803 | * Register a device which has init and exit functions |
| 804 | * that are called when network namespaces are created and |
| 805 | * destroyed respectively. |
| 806 | * |
| 807 | * When registered all network namespace init functions are |
| 808 | * called for every existing network namespace. Allowing kernel |
| 809 | * modules to have a race free view of the set of network namespaces. |
| 810 | * |
| 811 | * When a new network namespace is created all of the init |
| 812 | * methods are called in the order in which they were registered. |
| 813 | * |
| 814 | * When a network namespace is destroyed all of the exit methods |
| 815 | * are called in the reverse of the order with which they were |
| 816 | * registered. |
| 817 | */ |
| 818 | int register_pernet_device(struct pernet_operations *ops) |
| 819 | { |
| 820 | int error; |
| 821 | mutex_lock(&net_mutex); |
| 822 | error = register_pernet_operations(&pernet_list, ops); |
| 823 | if (!error && (first_device == &pernet_list)) |
| 824 | first_device = &ops->list; |
| 825 | mutex_unlock(&net_mutex); |
| 826 | return error; |
| 827 | } |
| 828 | EXPORT_SYMBOL_GPL(register_pernet_device); |
| 829 | |
| 830 | /** |
| 831 | * unregister_pernet_device - unregister a network namespace netdevice |
| 832 | * @ops: pernet operations structure to manipulate |
| 833 | * |
| 834 | * Remove the pernet operations structure from the list to be |
Oliver Pinter | 53379e5 | 2008-02-03 17:56:48 +0200 | [diff] [blame] | 835 | * used when network namespaces are created or destroyed. In |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 836 | * addition run the exit method for all existing network |
| 837 | * namespaces. |
| 838 | */ |
| 839 | void unregister_pernet_device(struct pernet_operations *ops) |
| 840 | { |
| 841 | mutex_lock(&net_mutex); |
| 842 | if (&ops->list == first_device) |
| 843 | first_device = first_device->next; |
| 844 | unregister_pernet_operations(ops); |
| 845 | mutex_unlock(&net_mutex); |
| 846 | } |
| 847 | EXPORT_SYMBOL_GPL(unregister_pernet_device); |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 848 | |
| 849 | #ifdef CONFIG_NET_NS |
Al Viro | 6496452 | 2014-11-01 00:37:32 -0400 | [diff] [blame] | 850 | static struct ns_common *netns_get(struct task_struct *task) |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 851 | { |
Eric W. Biederman | f063052 | 2011-05-04 17:51:50 -0700 | [diff] [blame] | 852 | struct net *net = NULL; |
| 853 | struct nsproxy *nsproxy; |
| 854 | |
Eric W. Biederman | 728dba3 | 2014-02-03 19:13:49 -0800 | [diff] [blame] | 855 | task_lock(task); |
| 856 | nsproxy = task->nsproxy; |
Eric W. Biederman | f063052 | 2011-05-04 17:51:50 -0700 | [diff] [blame] | 857 | if (nsproxy) |
| 858 | net = get_net(nsproxy->net_ns); |
Eric W. Biederman | 728dba3 | 2014-02-03 19:13:49 -0800 | [diff] [blame] | 859 | task_unlock(task); |
Eric W. Biederman | f063052 | 2011-05-04 17:51:50 -0700 | [diff] [blame] | 860 | |
Al Viro | ff24870 | 2014-11-01 00:10:50 -0400 | [diff] [blame] | 861 | return net ? &net->ns : NULL; |
| 862 | } |
| 863 | |
| 864 | static inline struct net *to_net_ns(struct ns_common *ns) |
| 865 | { |
| 866 | return container_of(ns, struct net, ns); |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 867 | } |
| 868 | |
Al Viro | 6496452 | 2014-11-01 00:37:32 -0400 | [diff] [blame] | 869 | static void netns_put(struct ns_common *ns) |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 870 | { |
Al Viro | ff24870 | 2014-11-01 00:10:50 -0400 | [diff] [blame] | 871 | put_net(to_net_ns(ns)); |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 872 | } |
| 873 | |
Al Viro | 6496452 | 2014-11-01 00:37:32 -0400 | [diff] [blame] | 874 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 875 | { |
Al Viro | ff24870 | 2014-11-01 00:10:50 -0400 | [diff] [blame] | 876 | struct net *net = to_net_ns(ns); |
Eric W. Biederman | 142e1d1 | 2012-07-26 01:13:20 -0700 | [diff] [blame] | 877 | |
Eric W. Biederman | 5e4a084 | 2012-12-14 07:55:36 -0800 | [diff] [blame] | 878 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
Eric W. Biederman | c7b96ac | 2013-03-20 12:49:49 -0700 | [diff] [blame] | 879 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
Eric W. Biederman | 142e1d1 | 2012-07-26 01:13:20 -0700 | [diff] [blame] | 880 | return -EPERM; |
| 881 | |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 882 | put_net(nsproxy->net_ns); |
Eric W. Biederman | 142e1d1 | 2012-07-26 01:13:20 -0700 | [diff] [blame] | 883 | nsproxy->net_ns = get_net(net); |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 884 | return 0; |
| 885 | } |
| 886 | |
| 887 | const struct proc_ns_operations netns_operations = { |
| 888 | .name = "net", |
| 889 | .type = CLONE_NEWNET, |
| 890 | .get = netns_get, |
| 891 | .put = netns_put, |
| 892 | .install = netns_install, |
| 893 | }; |
| 894 | #endif |