]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Copyright (C) 2006 IBM Corporation | |
3 | * | |
4 | * Author: Serge Hallyn <serue@us.ibm.com> | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License as | |
8 | * published by the Free Software Foundation, version 2 of the | |
9 | * License. | |
10 | * | |
11 | * Jun 2006 - namespaces support | |
12 | * OpenVZ, SWsoft Inc. | |
13 | * Pavel Emelianov <xemul@openvz.org> | |
14 | */ | |
15 | ||
16 | #include <linux/slab.h> | |
17 | #include <linux/module.h> | |
18 | #include <linux/nsproxy.h> | |
19 | #include <linux/init_task.h> | |
20 | #include <linux/mnt_namespace.h> | |
21 | #include <linux/utsname.h> | |
22 | #include <linux/pid_namespace.h> | |
23 | #include <net/net_namespace.h> | |
24 | #include <linux/ipc_namespace.h> | |
25 | ||
26 | static struct kmem_cache *nsproxy_cachep; | |
27 | ||
28 | struct nsproxy init_nsproxy = { | |
29 | .count = ATOMIC_INIT(1), | |
30 | .uts_ns = &init_uts_ns, | |
31 | #if defined(CONFIG_POSIX_MQUEUE) || defined(CONFIG_SYSVIPC) | |
32 | .ipc_ns = &init_ipc_ns, | |
33 | #endif | |
34 | .mnt_ns = NULL, | |
35 | .pid_ns = &init_pid_ns, | |
36 | #ifdef CONFIG_NET | |
37 | .net_ns = &init_net, | |
38 | #endif | |
39 | }; | |
40 | ||
41 | static inline struct nsproxy *create_nsproxy(void) | |
42 | { | |
43 | struct nsproxy *nsproxy; | |
44 | ||
45 | nsproxy = kmem_cache_alloc(nsproxy_cachep, GFP_KERNEL); | |
46 | if (nsproxy) | |
47 | atomic_set(&nsproxy->count, 1); | |
48 | return nsproxy; | |
49 | } | |
50 | ||
51 | /* | |
52 | * Create new nsproxy and all of its the associated namespaces. | |
53 | * Return the newly created nsproxy. Do not attach this to the task, | |
54 | * leave it to the caller to do proper locking and attach it to task. | |
55 | */ | |
56 | static struct nsproxy *create_new_namespaces(unsigned long flags, | |
57 | struct task_struct *tsk, struct fs_struct *new_fs) | |
58 | { | |
59 | struct nsproxy *new_nsp; | |
60 | int err; | |
61 | ||
62 | new_nsp = create_nsproxy(); | |
63 | if (!new_nsp) | |
64 | return ERR_PTR(-ENOMEM); | |
65 | ||
66 | new_nsp->mnt_ns = copy_mnt_ns(flags, tsk->nsproxy->mnt_ns, new_fs); | |
67 | if (IS_ERR(new_nsp->mnt_ns)) { | |
68 | err = PTR_ERR(new_nsp->mnt_ns); | |
69 | goto out_ns; | |
70 | } | |
71 | ||
72 | new_nsp->uts_ns = copy_utsname(flags, tsk->nsproxy->uts_ns); | |
73 | if (IS_ERR(new_nsp->uts_ns)) { | |
74 | err = PTR_ERR(new_nsp->uts_ns); | |
75 | goto out_uts; | |
76 | } | |
77 | ||
78 | new_nsp->ipc_ns = copy_ipcs(flags, tsk->nsproxy->ipc_ns); | |
79 | if (IS_ERR(new_nsp->ipc_ns)) { | |
80 | err = PTR_ERR(new_nsp->ipc_ns); | |
81 | goto out_ipc; | |
82 | } | |
83 | ||
84 | new_nsp->pid_ns = copy_pid_ns(flags, task_active_pid_ns(tsk)); | |
85 | if (IS_ERR(new_nsp->pid_ns)) { | |
86 | err = PTR_ERR(new_nsp->pid_ns); | |
87 | goto out_pid; | |
88 | } | |
89 | ||
90 | new_nsp->net_ns = copy_net_ns(flags, tsk->nsproxy->net_ns); | |
91 | if (IS_ERR(new_nsp->net_ns)) { | |
92 | err = PTR_ERR(new_nsp->net_ns); | |
93 | goto out_net; | |
94 | } | |
95 | ||
96 | return new_nsp; | |
97 | ||
98 | out_net: | |
99 | if (new_nsp->pid_ns) | |
100 | put_pid_ns(new_nsp->pid_ns); | |
101 | out_pid: | |
102 | if (new_nsp->ipc_ns) | |
103 | put_ipc_ns(new_nsp->ipc_ns); | |
104 | out_ipc: | |
105 | if (new_nsp->uts_ns) | |
106 | put_uts_ns(new_nsp->uts_ns); | |
107 | out_uts: | |
108 | if (new_nsp->mnt_ns) | |
109 | put_mnt_ns(new_nsp->mnt_ns); | |
110 | out_ns: | |
111 | kmem_cache_free(nsproxy_cachep, new_nsp); | |
112 | return ERR_PTR(err); | |
113 | } | |
114 | ||
115 | /* | |
116 | * called from clone. This now handles copy for nsproxy and all | |
117 | * namespaces therein. | |
118 | */ | |
119 | int copy_namespaces(unsigned long flags, struct task_struct *tsk) | |
120 | { | |
121 | struct nsproxy *old_ns = tsk->nsproxy; | |
122 | struct nsproxy *new_ns; | |
123 | int err = 0; | |
124 | ||
125 | if (!old_ns) | |
126 | return 0; | |
127 | ||
128 | get_nsproxy(old_ns); | |
129 | ||
130 | if (!(flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC | | |
131 | CLONE_NEWPID | CLONE_NEWNET))) | |
132 | return 0; | |
133 | ||
134 | if (!capable(CAP_SYS_ADMIN)) { | |
135 | err = -EPERM; | |
136 | goto out; | |
137 | } | |
138 | ||
139 | /* | |
140 | * CLONE_NEWIPC must detach from the undolist: after switching | |
141 | * to a new ipc namespace, the semaphore arrays from the old | |
142 | * namespace are unreachable. In clone parlance, CLONE_SYSVSEM | |
143 | * means share undolist with parent, so we must forbid using | |
144 | * it along with CLONE_NEWIPC. | |
145 | */ | |
146 | if ((flags & CLONE_NEWIPC) && (flags & CLONE_SYSVSEM)) { | |
147 | err = -EINVAL; | |
148 | goto out; | |
149 | } | |
150 | ||
151 | new_ns = create_new_namespaces(flags, tsk, tsk->fs); | |
152 | if (IS_ERR(new_ns)) { | |
153 | err = PTR_ERR(new_ns); | |
154 | goto out; | |
155 | } | |
156 | ||
157 | tsk->nsproxy = new_ns; | |
158 | ||
159 | out: | |
160 | put_nsproxy(old_ns); | |
161 | return err; | |
162 | } | |
163 | ||
164 | void free_nsproxy(struct nsproxy *ns) | |
165 | { | |
166 | if (ns->mnt_ns) | |
167 | put_mnt_ns(ns->mnt_ns); | |
168 | if (ns->uts_ns) | |
169 | put_uts_ns(ns->uts_ns); | |
170 | if (ns->ipc_ns) | |
171 | put_ipc_ns(ns->ipc_ns); | |
172 | if (ns->pid_ns) | |
173 | put_pid_ns(ns->pid_ns); | |
174 | put_net(ns->net_ns); | |
175 | kmem_cache_free(nsproxy_cachep, ns); | |
176 | } | |
177 | ||
178 | /* | |
179 | * Called from unshare. Unshare all the namespaces part of nsproxy. | |
180 | * On success, returns the new nsproxy. | |
181 | */ | |
182 | int unshare_nsproxy_namespaces(unsigned long unshare_flags, | |
183 | struct nsproxy **new_nsp, struct fs_struct *new_fs) | |
184 | { | |
185 | int err = 0; | |
186 | ||
187 | if (!(unshare_flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC | | |
188 | CLONE_NEWNET))) | |
189 | return 0; | |
190 | ||
191 | if (!capable(CAP_SYS_ADMIN)) | |
192 | return -EPERM; | |
193 | ||
194 | *new_nsp = create_new_namespaces(unshare_flags, current, | |
195 | new_fs ? new_fs : current->fs); | |
196 | if (IS_ERR(*new_nsp)) { | |
197 | err = PTR_ERR(*new_nsp); | |
198 | goto out; | |
199 | } | |
200 | ||
201 | err = ns_cgroup_clone(current, task_pid(current)); | |
202 | if (err) | |
203 | put_nsproxy(*new_nsp); | |
204 | ||
205 | out: | |
206 | return err; | |
207 | } | |
208 | ||
209 | void switch_task_namespaces(struct task_struct *p, struct nsproxy *new) | |
210 | { | |
211 | struct nsproxy *ns; | |
212 | ||
213 | might_sleep(); | |
214 | ||
215 | ns = p->nsproxy; | |
216 | ||
217 | rcu_assign_pointer(p->nsproxy, new); | |
218 | ||
219 | if (ns && atomic_dec_and_test(&ns->count)) { | |
220 | /* | |
221 | * wait for others to get what they want from this nsproxy. | |
222 | * | |
223 | * cannot release this nsproxy via the call_rcu() since | |
224 | * put_mnt_ns() will want to sleep | |
225 | */ | |
226 | synchronize_rcu(); | |
227 | free_nsproxy(ns); | |
228 | } | |
229 | } | |
230 | ||
231 | void exit_task_namespaces(struct task_struct *p) | |
232 | { | |
233 | switch_task_namespaces(p, NULL); | |
234 | } | |
235 | ||
236 | static int __init nsproxy_cache_init(void) | |
237 | { | |
238 | nsproxy_cachep = KMEM_CACHE(nsproxy, SLAB_PANIC); | |
239 | return 0; | |
240 | } | |
241 | ||
242 | module_init(nsproxy_cache_init); |