Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 1 | #include <linux/workqueue.h> |
| 2 | #include <linux/rtnetlink.h> |
| 3 | #include <linux/cache.h> |
| 4 | #include <linux/slab.h> |
| 5 | #include <linux/list.h> |
| 6 | #include <linux/delay.h> |
Eric W. Biederman | 9dd776b | 2007-09-26 22:04:26 -0700 | [diff] [blame] | 7 | #include <linux/sched.h> |
Pavel Emelyanov | c93cf61 | 2008-04-15 00:35:23 -0700 | [diff] [blame] | 8 | #include <linux/idr.h> |
Johannes Berg | 11a28d3 | 2009-07-10 09:51:33 +0000 | [diff] [blame] | 9 | #include <linux/rculist.h> |
Johannes Berg | 30ffee8 | 2009-07-10 09:51:35 +0000 | [diff] [blame] | 10 | #include <linux/nsproxy.h> |
Eric W. Biederman | f063052 | 2011-05-04 17:51:50 -0700 | [diff] [blame] | 11 | #include <linux/proc_fs.h> |
| 12 | #include <linux/file.h> |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 13 | #include <net/net_namespace.h> |
Pavel Emelyanov | dec827d | 2008-04-15 00:36:08 -0700 | [diff] [blame] | 14 | #include <net/netns/generic.h> |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 15 | |
| 16 | /* |
| 17 | * Our network namespace constructor/destructor lists |
| 18 | */ |
| 19 | |
| 20 | static LIST_HEAD(pernet_list); |
| 21 | static struct list_head *first_device = &pernet_list; |
| 22 | static DEFINE_MUTEX(net_mutex); |
| 23 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 24 | LIST_HEAD(net_namespace_list); |
Alexey Dobriyan | b76a461 | 2008-10-08 11:35:06 +0200 | [diff] [blame] | 25 | EXPORT_SYMBOL_GPL(net_namespace_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 26 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 27 | struct net init_net; |
Denis V. Lunev | ff4b950 | 2008-01-22 22:05:33 -0800 | [diff] [blame] | 28 | EXPORT_SYMBOL(init_net); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 29 | |
Pavel Emelyanov | dec827d | 2008-04-15 00:36:08 -0700 | [diff] [blame] | 30 | #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */ |
| 31 | |
Jiri Pirko | 05fceb4 | 2010-04-23 01:40:47 +0000 | [diff] [blame] | 32 | static int net_assign_generic(struct net *net, int id, void *data) |
| 33 | { |
| 34 | struct net_generic *ng, *old_ng; |
| 35 | |
| 36 | BUG_ON(!mutex_is_locked(&net_mutex)); |
| 37 | BUG_ON(id == 0); |
| 38 | |
Eric Dumazet | 1c87733 | 2010-10-25 03:20:11 +0000 | [diff] [blame] | 39 | old_ng = rcu_dereference_protected(net->gen, |
| 40 | lockdep_is_held(&net_mutex)); |
| 41 | ng = old_ng; |
Jiri Pirko | 05fceb4 | 2010-04-23 01:40:47 +0000 | [diff] [blame] | 42 | if (old_ng->len >= id) |
| 43 | goto assign; |
| 44 | |
| 45 | ng = kzalloc(sizeof(struct net_generic) + |
| 46 | id * sizeof(void *), GFP_KERNEL); |
| 47 | if (ng == NULL) |
| 48 | return -ENOMEM; |
| 49 | |
| 50 | /* |
| 51 | * Some synchronisation notes: |
| 52 | * |
| 53 | * The net_generic explores the net->gen array inside rcu |
| 54 | * read section. Besides once set the net->gen->ptr[x] |
| 55 | * pointer never changes (see rules in netns/generic.h). |
| 56 | * |
| 57 | * That said, we simply duplicate this array and schedule |
| 58 | * the old copy for kfree after a grace period. |
| 59 | */ |
| 60 | |
| 61 | ng->len = id; |
| 62 | memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*)); |
| 63 | |
| 64 | rcu_assign_pointer(net->gen, ng); |
Lai Jiangshan | 04d4dfe | 2011-03-18 12:06:32 +0800 | [diff] [blame] | 65 | kfree_rcu(old_ng, rcu); |
Jiri Pirko | 05fceb4 | 2010-04-23 01:40:47 +0000 | [diff] [blame] | 66 | assign: |
| 67 | ng->ptr[id - 1] = data; |
| 68 | return 0; |
| 69 | } |
| 70 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 71 | static int ops_init(const struct pernet_operations *ops, struct net *net) |
| 72 | { |
| 73 | int err; |
| 74 | if (ops->id && ops->size) { |
| 75 | void *data = kzalloc(ops->size, GFP_KERNEL); |
| 76 | if (!data) |
| 77 | return -ENOMEM; |
| 78 | |
| 79 | err = net_assign_generic(net, *ops->id, data); |
| 80 | if (err) { |
| 81 | kfree(data); |
| 82 | return err; |
| 83 | } |
| 84 | } |
| 85 | if (ops->init) |
| 86 | return ops->init(net); |
| 87 | return 0; |
| 88 | } |
| 89 | |
| 90 | static void ops_free(const struct pernet_operations *ops, struct net *net) |
| 91 | { |
| 92 | if (ops->id && ops->size) { |
| 93 | int id = *ops->id; |
| 94 | kfree(net_generic(net, id)); |
| 95 | } |
| 96 | } |
| 97 | |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 98 | static void ops_exit_list(const struct pernet_operations *ops, |
| 99 | struct list_head *net_exit_list) |
| 100 | { |
| 101 | struct net *net; |
| 102 | if (ops->exit) { |
| 103 | list_for_each_entry(net, net_exit_list, exit_list) |
| 104 | ops->exit(net); |
| 105 | } |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 106 | if (ops->exit_batch) |
| 107 | ops->exit_batch(net_exit_list); |
| 108 | } |
| 109 | |
| 110 | static void ops_free_list(const struct pernet_operations *ops, |
| 111 | struct list_head *net_exit_list) |
| 112 | { |
| 113 | struct net *net; |
| 114 | if (ops->size && ops->id) { |
| 115 | list_for_each_entry(net, net_exit_list, exit_list) |
| 116 | ops_free(ops, net); |
| 117 | } |
| 118 | } |
| 119 | |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 120 | /* |
| 121 | * setup_net runs the initializers for the network namespace object. |
| 122 | */ |
Pavel Emelyanov | 1a2ee93 | 2007-11-01 00:45:59 -0700 | [diff] [blame] | 123 | static __net_init int setup_net(struct net *net) |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 124 | { |
| 125 | /* Must be called with net_mutex held */ |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 126 | const struct pernet_operations *ops, *saved_ops; |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 127 | int error = 0; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 128 | LIST_HEAD(net_exit_list); |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 129 | |
| 130 | atomic_set(&net->count, 1); |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 131 | |
Denis V. Lunev | 5d1e446 | 2008-04-16 01:58:04 -0700 | [diff] [blame] | 132 | #ifdef NETNS_REFCNT_DEBUG |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 133 | atomic_set(&net->use_count, 0); |
Denis V. Lunev | 5d1e446 | 2008-04-16 01:58:04 -0700 | [diff] [blame] | 134 | #endif |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 135 | |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 136 | list_for_each_entry(ops, &pernet_list, list) { |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 137 | error = ops_init(ops, net); |
| 138 | if (error < 0) |
| 139 | goto out_undo; |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 140 | } |
| 141 | out: |
| 142 | return error; |
| 143 | |
| 144 | out_undo: |
| 145 | /* Walk through the list backwards calling the exit functions |
| 146 | * for the pernet modules whose init functions did not fail. |
| 147 | */ |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 148 | list_add(&net->exit_list, &net_exit_list); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 149 | saved_ops = ops; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 150 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
| 151 | ops_exit_list(ops, &net_exit_list); |
| 152 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 153 | ops = saved_ops; |
| 154 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 155 | ops_free_list(ops, &net_exit_list); |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 156 | |
| 157 | rcu_barrier(); |
| 158 | goto out; |
| 159 | } |
| 160 | |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 161 | static struct net_generic *net_alloc_generic(void) |
| 162 | { |
| 163 | struct net_generic *ng; |
| 164 | size_t generic_size = sizeof(struct net_generic) + |
| 165 | INITIAL_NET_GEN_PTRS * sizeof(void *); |
| 166 | |
| 167 | ng = kzalloc(generic_size, GFP_KERNEL); |
| 168 | if (ng) |
| 169 | ng->len = INITIAL_NET_GEN_PTRS; |
| 170 | |
| 171 | return ng; |
| 172 | } |
| 173 | |
Clemens Noss | ebe47d4 | 2009-02-23 15:37:35 -0800 | [diff] [blame] | 174 | #ifdef CONFIG_NET_NS |
| 175 | static struct kmem_cache *net_cachep; |
| 176 | static struct workqueue_struct *netns_wq; |
| 177 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 178 | static struct net *net_alloc(void) |
| 179 | { |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 180 | struct net *net = NULL; |
| 181 | struct net_generic *ng; |
| 182 | |
| 183 | ng = net_alloc_generic(); |
| 184 | if (!ng) |
| 185 | goto out; |
| 186 | |
| 187 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); |
| 188 | if (!net) |
| 189 | goto out_free; |
| 190 | |
| 191 | rcu_assign_pointer(net->gen, ng); |
| 192 | out: |
| 193 | return net; |
| 194 | |
| 195 | out_free: |
| 196 | kfree(ng); |
| 197 | goto out; |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 198 | } |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 199 | |
Johann Felix Soden | 45a19b0 | 2007-11-07 01:30:30 -0800 | [diff] [blame] | 200 | static void net_free(struct net *net) |
| 201 | { |
Denis V. Lunev | 5d1e446 | 2008-04-16 01:58:04 -0700 | [diff] [blame] | 202 | #ifdef NETNS_REFCNT_DEBUG |
Johann Felix Soden | 45a19b0 | 2007-11-07 01:30:30 -0800 | [diff] [blame] | 203 | if (unlikely(atomic_read(&net->use_count) != 0)) { |
| 204 | printk(KERN_EMERG "network namespace not free! Usage: %d\n", |
| 205 | atomic_read(&net->use_count)); |
| 206 | return; |
| 207 | } |
Denis V. Lunev | 5d1e446 | 2008-04-16 01:58:04 -0700 | [diff] [blame] | 208 | #endif |
Alexey Dobriyan | 4ef079c | 2008-10-14 22:54:48 -0700 | [diff] [blame] | 209 | kfree(net->gen); |
Johann Felix Soden | 45a19b0 | 2007-11-07 01:30:30 -0800 | [diff] [blame] | 210 | kmem_cache_free(net_cachep, net); |
| 211 | } |
| 212 | |
Rob Landley | 911cb19 | 2011-04-15 02:26:25 +0000 | [diff] [blame] | 213 | struct net *copy_net_ns(unsigned long flags, struct net *old_net) |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 214 | { |
Alexey Dobriyan | 088eb2d | 2009-05-04 11:12:14 -0700 | [diff] [blame] | 215 | struct net *net; |
| 216 | int rv; |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 217 | |
Rob Landley | 911cb19 | 2011-04-15 02:26:25 +0000 | [diff] [blame] | 218 | if (!(flags & CLONE_NEWNET)) |
| 219 | return get_net(old_net); |
| 220 | |
Alexey Dobriyan | 088eb2d | 2009-05-04 11:12:14 -0700 | [diff] [blame] | 221 | net = net_alloc(); |
| 222 | if (!net) |
| 223 | return ERR_PTR(-ENOMEM); |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 224 | mutex_lock(&net_mutex); |
Alexey Dobriyan | 088eb2d | 2009-05-04 11:12:14 -0700 | [diff] [blame] | 225 | rv = setup_net(net); |
| 226 | if (rv == 0) { |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 227 | rtnl_lock(); |
Johannes Berg | 11a28d3 | 2009-07-10 09:51:33 +0000 | [diff] [blame] | 228 | list_add_tail_rcu(&net->list, &net_namespace_list); |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 229 | rtnl_unlock(); |
| 230 | } |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 231 | mutex_unlock(&net_mutex); |
Alexey Dobriyan | 088eb2d | 2009-05-04 11:12:14 -0700 | [diff] [blame] | 232 | if (rv < 0) { |
| 233 | net_free(net); |
| 234 | return ERR_PTR(rv); |
| 235 | } |
| 236 | return net; |
| 237 | } |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 238 | |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 239 | static DEFINE_SPINLOCK(cleanup_list_lock); |
| 240 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ |
| 241 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 242 | static void cleanup_net(struct work_struct *work) |
| 243 | { |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 244 | const struct pernet_operations *ops; |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 245 | struct net *net, *tmp; |
| 246 | LIST_HEAD(net_kill_list); |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 247 | LIST_HEAD(net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 248 | |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 249 | /* Atomically snapshot the list of namespaces to cleanup */ |
| 250 | spin_lock_irq(&cleanup_list_lock); |
| 251 | list_replace_init(&cleanup_list, &net_kill_list); |
| 252 | spin_unlock_irq(&cleanup_list_lock); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 253 | |
| 254 | mutex_lock(&net_mutex); |
| 255 | |
| 256 | /* Don't let anyone else find us. */ |
Eric W. Biederman | f4618d3 | 2007-09-26 22:40:08 -0700 | [diff] [blame] | 257 | rtnl_lock(); |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 258 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 259 | list_del_rcu(&net->list); |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 260 | list_add_tail(&net->exit_list, &net_exit_list); |
| 261 | } |
Eric W. Biederman | f4618d3 | 2007-09-26 22:40:08 -0700 | [diff] [blame] | 262 | rtnl_unlock(); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 263 | |
Johannes Berg | 11a28d3 | 2009-07-10 09:51:33 +0000 | [diff] [blame] | 264 | /* |
| 265 | * Another CPU might be rcu-iterating the list, wait for it. |
| 266 | * This needs to be before calling the exit() notifiers, so |
| 267 | * the rcu_barrier() below isn't sufficient alone. |
| 268 | */ |
| 269 | synchronize_rcu(); |
| 270 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 271 | /* Run all of the network namespace exit methods */ |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 272 | list_for_each_entry_reverse(ops, &pernet_list, list) |
| 273 | ops_exit_list(ops, &net_exit_list); |
| 274 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 275 | /* Free the net generic variables */ |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 276 | list_for_each_entry_reverse(ops, &pernet_list, list) |
| 277 | ops_free_list(ops, &net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 278 | |
| 279 | mutex_unlock(&net_mutex); |
| 280 | |
| 281 | /* Ensure there are no outstanding rcu callbacks using this |
| 282 | * network namespace. |
| 283 | */ |
| 284 | rcu_barrier(); |
| 285 | |
| 286 | /* Finally it is safe to free my network namespace structure */ |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 287 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
| 288 | list_del_init(&net->exit_list); |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 289 | net_free(net); |
| 290 | } |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 291 | } |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 292 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 293 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 294 | void __put_net(struct net *net) |
| 295 | { |
| 296 | /* Cleanup the network namespace in process context */ |
Eric W. Biederman | 2b035b3 | 2009-11-29 22:25:27 +0000 | [diff] [blame] | 297 | unsigned long flags; |
| 298 | |
| 299 | spin_lock_irqsave(&cleanup_list_lock, flags); |
| 300 | list_add(&net->cleanup_list, &cleanup_list); |
| 301 | spin_unlock_irqrestore(&cleanup_list_lock, flags); |
| 302 | |
| 303 | queue_work(netns_wq, &net_cleanup_work); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 304 | } |
| 305 | EXPORT_SYMBOL_GPL(__put_net); |
| 306 | |
Stephen Rothwell | 956c920 | 2011-05-12 13:51:13 +1000 | [diff] [blame] | 307 | struct net *get_net_ns_by_fd(int fd) |
| 308 | { |
| 309 | struct proc_inode *ei; |
| 310 | struct file *file; |
| 311 | struct net *net; |
| 312 | |
| 313 | net = ERR_PTR(-EINVAL); |
| 314 | file = proc_ns_fget(fd); |
| 315 | if (!file) |
| 316 | goto out; |
| 317 | |
| 318 | ei = PROC_I(file->f_dentry->d_inode); |
| 319 | if (ei->ns_ops != &netns_operations) |
| 320 | goto out; |
| 321 | |
| 322 | net = get_net(ei->ns); |
| 323 | out: |
| 324 | if (file) |
| 325 | fput(file); |
| 326 | return net; |
| 327 | } |
| 328 | |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 329 | #else |
Eric W. Biederman | 9dd776b | 2007-09-26 22:04:26 -0700 | [diff] [blame] | 330 | struct net *copy_net_ns(unsigned long flags, struct net *old_net) |
| 331 | { |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 332 | if (flags & CLONE_NEWNET) |
| 333 | return ERR_PTR(-EINVAL); |
| 334 | return old_net; |
Eric W. Biederman | 9dd776b | 2007-09-26 22:04:26 -0700 | [diff] [blame] | 335 | } |
Stephen Rothwell | 956c920 | 2011-05-12 13:51:13 +1000 | [diff] [blame] | 336 | |
| 337 | struct net *get_net_ns_by_fd(int fd) |
| 338 | { |
| 339 | return ERR_PTR(-EINVAL); |
| 340 | } |
Pavel Emelyanov | 6a1a3b9 | 2007-11-01 00:44:50 -0700 | [diff] [blame] | 341 | #endif |
Eric W. Biederman | 9dd776b | 2007-09-26 22:04:26 -0700 | [diff] [blame] | 342 | |
Johannes Berg | 30ffee8 | 2009-07-10 09:51:35 +0000 | [diff] [blame] | 343 | struct net *get_net_ns_by_pid(pid_t pid) |
| 344 | { |
| 345 | struct task_struct *tsk; |
| 346 | struct net *net; |
| 347 | |
| 348 | /* Lookup the network namespace */ |
| 349 | net = ERR_PTR(-ESRCH); |
| 350 | rcu_read_lock(); |
| 351 | tsk = find_task_by_vpid(pid); |
| 352 | if (tsk) { |
| 353 | struct nsproxy *nsproxy; |
| 354 | nsproxy = task_nsproxy(tsk); |
| 355 | if (nsproxy) |
| 356 | net = get_net(nsproxy->net_ns); |
| 357 | } |
| 358 | rcu_read_unlock(); |
| 359 | return net; |
| 360 | } |
| 361 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); |
| 362 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 363 | static int __init net_ns_init(void) |
| 364 | { |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 365 | struct net_generic *ng; |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 366 | |
Pavel Emelyanov | d57a921 | 2007-11-01 00:46:50 -0700 | [diff] [blame] | 367 | #ifdef CONFIG_NET_NS |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 368 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
| 369 | SMP_CACHE_BYTES, |
| 370 | SLAB_PANIC, NULL); |
Benjamin Thery | 3ef1355 | 2007-11-19 23:18:16 -0800 | [diff] [blame] | 371 | |
| 372 | /* Create workqueue for cleanup */ |
| 373 | netns_wq = create_singlethread_workqueue("netns"); |
| 374 | if (!netns_wq) |
| 375 | panic("Could not create netns workq"); |
Pavel Emelyanov | d57a921 | 2007-11-01 00:46:50 -0700 | [diff] [blame] | 376 | #endif |
Benjamin Thery | 3ef1355 | 2007-11-19 23:18:16 -0800 | [diff] [blame] | 377 | |
Daniel Lezcano | 486a87f | 2009-02-22 00:07:53 -0800 | [diff] [blame] | 378 | ng = net_alloc_generic(); |
| 379 | if (!ng) |
| 380 | panic("Could not allocate generic netns"); |
| 381 | |
| 382 | rcu_assign_pointer(init_net.gen, ng); |
| 383 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 384 | mutex_lock(&net_mutex); |
Stephen Hemminger | ca0f311 | 2009-05-21 15:10:31 -0700 | [diff] [blame] | 385 | if (setup_net(&init_net)) |
| 386 | panic("Could not setup the initial network namespace"); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 387 | |
Eric W. Biederman | f4618d3 | 2007-09-26 22:40:08 -0700 | [diff] [blame] | 388 | rtnl_lock(); |
Johannes Berg | 11a28d3 | 2009-07-10 09:51:33 +0000 | [diff] [blame] | 389 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
Eric W. Biederman | f4618d3 | 2007-09-26 22:40:08 -0700 | [diff] [blame] | 390 | rtnl_unlock(); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 391 | |
| 392 | mutex_unlock(&net_mutex); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 393 | |
| 394 | return 0; |
| 395 | } |
| 396 | |
| 397 | pure_initcall(net_ns_init); |
| 398 | |
Denis V. Lunev | ed160e8 | 2007-11-13 03:23:21 -0800 | [diff] [blame] | 399 | #ifdef CONFIG_NET_NS |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 400 | static int __register_pernet_operations(struct list_head *list, |
| 401 | struct pernet_operations *ops) |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 402 | { |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 403 | struct net *net; |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 404 | int error; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 405 | LIST_HEAD(net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 406 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 407 | list_add_tail(&ops->list, list); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 408 | if (ops->init || (ops->id && ops->size)) { |
Pavel Emelyanov | 1dba323 | 2007-11-01 00:42:43 -0700 | [diff] [blame] | 409 | for_each_net(net) { |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 410 | error = ops_init(ops, net); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 411 | if (error) |
| 412 | goto out_undo; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 413 | list_add_tail(&net->exit_list, &net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 414 | } |
| 415 | } |
Pavel Emelyanov | 1dba323 | 2007-11-01 00:42:43 -0700 | [diff] [blame] | 416 | return 0; |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 417 | |
| 418 | out_undo: |
| 419 | /* If I have an error cleanup all namespaces I initialized */ |
| 420 | list_del(&ops->list); |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 421 | ops_exit_list(ops, &net_exit_list); |
| 422 | ops_free_list(ops, &net_exit_list); |
Pavel Emelyanov | 1dba323 | 2007-11-01 00:42:43 -0700 | [diff] [blame] | 423 | return error; |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 424 | } |
| 425 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 426 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 427 | { |
| 428 | struct net *net; |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 429 | LIST_HEAD(net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 430 | |
| 431 | list_del(&ops->list); |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 432 | for_each_net(net) |
| 433 | list_add_tail(&net->exit_list, &net_exit_list); |
| 434 | ops_exit_list(ops, &net_exit_list); |
| 435 | ops_free_list(ops, &net_exit_list); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 436 | } |
| 437 | |
Denis V. Lunev | ed160e8 | 2007-11-13 03:23:21 -0800 | [diff] [blame] | 438 | #else |
| 439 | |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 440 | static int __register_pernet_operations(struct list_head *list, |
| 441 | struct pernet_operations *ops) |
| 442 | { |
| 443 | int err = 0; |
| 444 | err = ops_init(ops, &init_net); |
| 445 | if (err) |
| 446 | ops_free(ops, &init_net); |
| 447 | return err; |
| 448 | |
| 449 | } |
| 450 | |
| 451 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
| 452 | { |
Eric W. Biederman | 72ad937 | 2009-12-03 02:29:03 +0000 | [diff] [blame] | 453 | LIST_HEAD(net_exit_list); |
| 454 | list_add(&init_net.exit_list, &net_exit_list); |
| 455 | ops_exit_list(ops, &net_exit_list); |
| 456 | ops_free_list(ops, &net_exit_list); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 457 | } |
| 458 | |
| 459 | #endif /* CONFIG_NET_NS */ |
| 460 | |
| 461 | static DEFINE_IDA(net_generic_ids); |
| 462 | |
Denis V. Lunev | ed160e8 | 2007-11-13 03:23:21 -0800 | [diff] [blame] | 463 | static int register_pernet_operations(struct list_head *list, |
| 464 | struct pernet_operations *ops) |
| 465 | { |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 466 | int error; |
| 467 | |
| 468 | if (ops->id) { |
| 469 | again: |
| 470 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); |
| 471 | if (error < 0) { |
| 472 | if (error == -EAGAIN) { |
| 473 | ida_pre_get(&net_generic_ids, GFP_KERNEL); |
| 474 | goto again; |
| 475 | } |
| 476 | return error; |
| 477 | } |
| 478 | } |
| 479 | error = __register_pernet_operations(list, ops); |
Eric W. Biederman | 3a765ed | 2009-12-03 02:29:06 +0000 | [diff] [blame] | 480 | if (error) { |
| 481 | rcu_barrier(); |
| 482 | if (ops->id) |
| 483 | ida_remove(&net_generic_ids, *ops->id); |
| 484 | } |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 485 | |
| 486 | return error; |
Denis V. Lunev | ed160e8 | 2007-11-13 03:23:21 -0800 | [diff] [blame] | 487 | } |
| 488 | |
| 489 | static void unregister_pernet_operations(struct pernet_operations *ops) |
| 490 | { |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 491 | |
| 492 | __unregister_pernet_operations(ops); |
Eric W. Biederman | 3a765ed | 2009-12-03 02:29:06 +0000 | [diff] [blame] | 493 | rcu_barrier(); |
Eric W. Biederman | f875bae | 2009-11-29 22:25:28 +0000 | [diff] [blame] | 494 | if (ops->id) |
| 495 | ida_remove(&net_generic_ids, *ops->id); |
Denis V. Lunev | ed160e8 | 2007-11-13 03:23:21 -0800 | [diff] [blame] | 496 | } |
Pavel Emelyanov | c93cf61 | 2008-04-15 00:35:23 -0700 | [diff] [blame] | 497 | |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 498 | /** |
| 499 | * register_pernet_subsys - register a network namespace subsystem |
| 500 | * @ops: pernet operations structure for the subsystem |
| 501 | * |
| 502 | * Register a subsystem which has init and exit functions |
| 503 | * that are called when network namespaces are created and |
| 504 | * destroyed respectively. |
| 505 | * |
| 506 | * When registered all network namespace init functions are |
| 507 | * called for every existing network namespace. Allowing kernel |
| 508 | * modules to have a race free view of the set of network namespaces. |
| 509 | * |
| 510 | * When a new network namespace is created all of the init |
| 511 | * methods are called in the order in which they were registered. |
| 512 | * |
| 513 | * When a network namespace is destroyed all of the exit methods |
| 514 | * are called in the reverse of the order with which they were |
| 515 | * registered. |
| 516 | */ |
| 517 | int register_pernet_subsys(struct pernet_operations *ops) |
| 518 | { |
| 519 | int error; |
| 520 | mutex_lock(&net_mutex); |
| 521 | error = register_pernet_operations(first_device, ops); |
| 522 | mutex_unlock(&net_mutex); |
| 523 | return error; |
| 524 | } |
| 525 | EXPORT_SYMBOL_GPL(register_pernet_subsys); |
| 526 | |
| 527 | /** |
| 528 | * unregister_pernet_subsys - unregister a network namespace subsystem |
| 529 | * @ops: pernet operations structure to manipulate |
| 530 | * |
| 531 | * Remove the pernet operations structure from the list to be |
Oliver Pinter | 53379e5 | 2008-02-03 17:56:48 +0200 | [diff] [blame] | 532 | * used when network namespaces are created or destroyed. In |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 533 | * addition run the exit method for all existing network |
| 534 | * namespaces. |
| 535 | */ |
Jiri Pirko | b3c981d | 2010-04-25 00:49:56 -0700 | [diff] [blame] | 536 | void unregister_pernet_subsys(struct pernet_operations *ops) |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 537 | { |
| 538 | mutex_lock(&net_mutex); |
Jiri Pirko | b3c981d | 2010-04-25 00:49:56 -0700 | [diff] [blame] | 539 | unregister_pernet_operations(ops); |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 540 | mutex_unlock(&net_mutex); |
| 541 | } |
| 542 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); |
| 543 | |
| 544 | /** |
| 545 | * register_pernet_device - register a network namespace device |
| 546 | * @ops: pernet operations structure for the subsystem |
| 547 | * |
| 548 | * Register a device which has init and exit functions |
| 549 | * that are called when network namespaces are created and |
| 550 | * destroyed respectively. |
| 551 | * |
| 552 | * When registered all network namespace init functions are |
| 553 | * called for every existing network namespace. Allowing kernel |
| 554 | * modules to have a race free view of the set of network namespaces. |
| 555 | * |
| 556 | * When a new network namespace is created all of the init |
| 557 | * methods are called in the order in which they were registered. |
| 558 | * |
| 559 | * When a network namespace is destroyed all of the exit methods |
| 560 | * are called in the reverse of the order with which they were |
| 561 | * registered. |
| 562 | */ |
| 563 | int register_pernet_device(struct pernet_operations *ops) |
| 564 | { |
| 565 | int error; |
| 566 | mutex_lock(&net_mutex); |
| 567 | error = register_pernet_operations(&pernet_list, ops); |
| 568 | if (!error && (first_device == &pernet_list)) |
| 569 | first_device = &ops->list; |
| 570 | mutex_unlock(&net_mutex); |
| 571 | return error; |
| 572 | } |
| 573 | EXPORT_SYMBOL_GPL(register_pernet_device); |
| 574 | |
| 575 | /** |
| 576 | * unregister_pernet_device - unregister a network namespace netdevice |
| 577 | * @ops: pernet operations structure to manipulate |
| 578 | * |
| 579 | * Remove the pernet operations structure from the list to be |
Oliver Pinter | 53379e5 | 2008-02-03 17:56:48 +0200 | [diff] [blame] | 580 | * used when network namespaces are created or destroyed. In |
Eric W. Biederman | 5f256be | 2007-09-12 11:50:50 +0200 | [diff] [blame] | 581 | * addition run the exit method for all existing network |
| 582 | * namespaces. |
| 583 | */ |
| 584 | void unregister_pernet_device(struct pernet_operations *ops) |
| 585 | { |
| 586 | mutex_lock(&net_mutex); |
| 587 | if (&ops->list == first_device) |
| 588 | first_device = first_device->next; |
| 589 | unregister_pernet_operations(ops); |
| 590 | mutex_unlock(&net_mutex); |
| 591 | } |
| 592 | EXPORT_SYMBOL_GPL(unregister_pernet_device); |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 593 | |
| 594 | #ifdef CONFIG_NET_NS |
| 595 | static void *netns_get(struct task_struct *task) |
| 596 | { |
Eric W. Biederman | f063052 | 2011-05-04 17:51:50 -0700 | [diff] [blame] | 597 | struct net *net = NULL; |
| 598 | struct nsproxy *nsproxy; |
| 599 | |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 600 | rcu_read_lock(); |
Eric W. Biederman | f063052 | 2011-05-04 17:51:50 -0700 | [diff] [blame] | 601 | nsproxy = task_nsproxy(task); |
| 602 | if (nsproxy) |
| 603 | net = get_net(nsproxy->net_ns); |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 604 | rcu_read_unlock(); |
Eric W. Biederman | f063052 | 2011-05-04 17:51:50 -0700 | [diff] [blame] | 605 | |
Eric W. Biederman | 13b6f57 | 2010-03-07 18:14:23 -0800 | [diff] [blame] | 606 | return net; |
| 607 | } |
| 608 | |
| 609 | static void netns_put(void *ns) |
| 610 | { |
| 611 | put_net(ns); |
| 612 | } |
| 613 | |
| 614 | static int netns_install(struct nsproxy *nsproxy, void *ns) |
| 615 | { |
| 616 | put_net(nsproxy->net_ns); |
| 617 | nsproxy->net_ns = get_net(ns); |
| 618 | return 0; |
| 619 | } |
| 620 | |
| 621 | const struct proc_ns_operations netns_operations = { |
| 622 | .name = "net", |
| 623 | .type = CLONE_NEWNET, |
| 624 | .get = netns_get, |
| 625 | .put = netns_put, |
| 626 | .install = netns_install, |
| 627 | }; |
| 628 | #endif |