Commit | Line | Data |
---|---|---|
ab516013 SH |
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. | |
25b21cb2 KK |
10 | * |
11 | * Jun 2006 - namespaces support | |
12 | * OpenVZ, SWsoft Inc. | |
13 | * Pavel Emelianov <xemul@openvz.org> | |
ab516013 SH |
14 | */ |
15 | ||
5a0e3ad6 | 16 | #include <linux/slab.h> |
9984de1a | 17 | #include <linux/export.h> |
ab516013 | 18 | #include <linux/nsproxy.h> |
0437eb59 | 19 | #include <linux/init_task.h> |
6b3286ed | 20 | #include <linux/mnt_namespace.h> |
4865ecf1 | 21 | #include <linux/utsname.h> |
9a575a92 | 22 | #include <linux/pid_namespace.h> |
9dd776b6 | 23 | #include <net/net_namespace.h> |
ae5e1b22 | 24 | #include <linux/ipc_namespace.h> |
0663c6f8 EB |
25 | #include <linux/proc_fs.h> |
26 | #include <linux/file.h> | |
27 | #include <linux/syscalls.h> | |
0437eb59 | 28 | |
98c0d07c CLG |
29 | static struct kmem_cache *nsproxy_cachep; |
30 | ||
8467005d AD |
31 | struct nsproxy init_nsproxy = { |
32 | .count = ATOMIC_INIT(1), | |
33 | .uts_ns = &init_uts_ns, | |
34 | #if defined(CONFIG_POSIX_MQUEUE) || defined(CONFIG_SYSVIPC) | |
35 | .ipc_ns = &init_ipc_ns, | |
36 | #endif | |
37 | .mnt_ns = NULL, | |
38 | .pid_ns = &init_pid_ns, | |
39 | #ifdef CONFIG_NET | |
40 | .net_ns = &init_net, | |
41 | #endif | |
42 | }; | |
ab516013 | 43 | |
90af90d7 | 44 | static inline struct nsproxy *create_nsproxy(void) |
ab516013 | 45 | { |
90af90d7 | 46 | struct nsproxy *nsproxy; |
ab516013 | 47 | |
90af90d7 AD |
48 | nsproxy = kmem_cache_alloc(nsproxy_cachep, GFP_KERNEL); |
49 | if (nsproxy) | |
50 | atomic_set(&nsproxy->count, 1); | |
51 | return nsproxy; | |
ab516013 SH |
52 | } |
53 | ||
54 | /* | |
e3222c4e BP |
55 | * Create new nsproxy and all of its the associated namespaces. |
56 | * Return the newly created nsproxy. Do not attach this to the task, | |
57 | * leave it to the caller to do proper locking and attach it to task. | |
ab516013 | 58 | */ |
213dd266 | 59 | static struct nsproxy *create_new_namespaces(unsigned long flags, |
bcf58e72 EB |
60 | struct task_struct *tsk, struct user_namespace *user_ns, |
61 | struct fs_struct *new_fs) | |
ab516013 | 62 | { |
e3222c4e | 63 | struct nsproxy *new_nsp; |
467e9f4b | 64 | int err; |
ab516013 | 65 | |
90af90d7 | 66 | new_nsp = create_nsproxy(); |
e3222c4e BP |
67 | if (!new_nsp) |
68 | return ERR_PTR(-ENOMEM); | |
1651e14e | 69 | |
bcf58e72 | 70 | new_nsp->mnt_ns = copy_mnt_ns(flags, tsk->nsproxy->mnt_ns, user_ns, new_fs); |
467e9f4b CLG |
71 | if (IS_ERR(new_nsp->mnt_ns)) { |
72 | err = PTR_ERR(new_nsp->mnt_ns); | |
e3222c4e | 73 | goto out_ns; |
467e9f4b | 74 | } |
e3222c4e | 75 | |
bcf58e72 | 76 | new_nsp->uts_ns = copy_utsname(flags, user_ns, tsk->nsproxy->uts_ns); |
467e9f4b CLG |
77 | if (IS_ERR(new_nsp->uts_ns)) { |
78 | err = PTR_ERR(new_nsp->uts_ns); | |
e3222c4e | 79 | goto out_uts; |
467e9f4b | 80 | } |
e3222c4e | 81 | |
bcf58e72 | 82 | new_nsp->ipc_ns = copy_ipcs(flags, user_ns, tsk->nsproxy->ipc_ns); |
467e9f4b CLG |
83 | if (IS_ERR(new_nsp->ipc_ns)) { |
84 | err = PTR_ERR(new_nsp->ipc_ns); | |
e3222c4e | 85 | goto out_ipc; |
467e9f4b | 86 | } |
e3222c4e | 87 | |
bcf58e72 | 88 | new_nsp->pid_ns = copy_pid_ns(flags, user_ns, tsk->nsproxy->pid_ns); |
467e9f4b CLG |
89 | if (IS_ERR(new_nsp->pid_ns)) { |
90 | err = PTR_ERR(new_nsp->pid_ns); | |
e3222c4e | 91 | goto out_pid; |
467e9f4b | 92 | } |
e3222c4e | 93 | |
bcf58e72 | 94 | new_nsp->net_ns = copy_net_ns(flags, user_ns, tsk->nsproxy->net_ns); |
9dd776b6 EB |
95 | if (IS_ERR(new_nsp->net_ns)) { |
96 | err = PTR_ERR(new_nsp->net_ns); | |
97 | goto out_net; | |
98 | } | |
99 | ||
e3222c4e BP |
100 | return new_nsp; |
101 | ||
9dd776b6 | 102 | out_net: |
acce292c CLG |
103 | if (new_nsp->pid_ns) |
104 | put_pid_ns(new_nsp->pid_ns); | |
e3222c4e BP |
105 | out_pid: |
106 | if (new_nsp->ipc_ns) | |
107 | put_ipc_ns(new_nsp->ipc_ns); | |
108 | out_ipc: | |
109 | if (new_nsp->uts_ns) | |
110 | put_uts_ns(new_nsp->uts_ns); | |
111 | out_uts: | |
112 | if (new_nsp->mnt_ns) | |
113 | put_mnt_ns(new_nsp->mnt_ns); | |
114 | out_ns: | |
98c0d07c | 115 | kmem_cache_free(nsproxy_cachep, new_nsp); |
467e9f4b | 116 | return ERR_PTR(err); |
ab516013 SH |
117 | } |
118 | ||
119 | /* | |
120 | * called from clone. This now handles copy for nsproxy and all | |
121 | * namespaces therein. | |
122 | */ | |
213dd266 | 123 | int copy_namespaces(unsigned long flags, struct task_struct *tsk) |
ab516013 SH |
124 | { |
125 | struct nsproxy *old_ns = tsk->nsproxy; | |
b33c77ef | 126 | struct user_namespace *user_ns = task_cred_xxx(tsk, user_ns); |
1651e14e SH |
127 | struct nsproxy *new_ns; |
128 | int err = 0; | |
ab516013 SH |
129 | |
130 | if (!old_ns) | |
131 | return 0; | |
132 | ||
133 | get_nsproxy(old_ns); | |
134 | ||
30e49c26 | 135 | if (!(flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC | |
18b6e041 | 136 | CLONE_NEWPID | CLONE_NEWNET))) |
1651e14e SH |
137 | return 0; |
138 | ||
b33c77ef | 139 | if (!ns_capable(user_ns, CAP_SYS_ADMIN)) { |
e3222c4e | 140 | err = -EPERM; |
1651e14e SH |
141 | goto out; |
142 | } | |
143 | ||
02fdb36a SH |
144 | /* |
145 | * CLONE_NEWIPC must detach from the undolist: after switching | |
146 | * to a new ipc namespace, the semaphore arrays from the old | |
147 | * namespace are unreachable. In clone parlance, CLONE_SYSVSEM | |
148 | * means share undolist with parent, so we must forbid using | |
149 | * it along with CLONE_NEWIPC. | |
150 | */ | |
151 | if ((flags & CLONE_NEWIPC) && (flags & CLONE_SYSVSEM)) { | |
152 | err = -EINVAL; | |
153 | goto out; | |
154 | } | |
155 | ||
bcf58e72 EB |
156 | new_ns = create_new_namespaces(flags, tsk, |
157 | task_cred_xxx(tsk, user_ns), tsk->fs); | |
e3222c4e BP |
158 | if (IS_ERR(new_ns)) { |
159 | err = PTR_ERR(new_ns); | |
160 | goto out; | |
161 | } | |
9a575a92 | 162 | |
e3222c4e | 163 | tsk->nsproxy = new_ns; |
858d72ea | 164 | |
1651e14e | 165 | out: |
444f378b | 166 | put_nsproxy(old_ns); |
1651e14e | 167 | return err; |
ab516013 SH |
168 | } |
169 | ||
170 | void free_nsproxy(struct nsproxy *ns) | |
171 | { | |
9a575a92 CLG |
172 | if (ns->mnt_ns) |
173 | put_mnt_ns(ns->mnt_ns); | |
174 | if (ns->uts_ns) | |
175 | put_uts_ns(ns->uts_ns); | |
176 | if (ns->ipc_ns) | |
177 | put_ipc_ns(ns->ipc_ns); | |
178 | if (ns->pid_ns) | |
179 | put_pid_ns(ns->pid_ns); | |
9dd776b6 | 180 | put_net(ns->net_ns); |
98c0d07c | 181 | kmem_cache_free(nsproxy_cachep, ns); |
ab516013 | 182 | } |
e3222c4e BP |
183 | |
184 | /* | |
185 | * Called from unshare. Unshare all the namespaces part of nsproxy. | |
4e71e474 | 186 | * On success, returns the new nsproxy. |
e3222c4e BP |
187 | */ |
188 | int unshare_nsproxy_namespaces(unsigned long unshare_flags, | |
b2e0d987 | 189 | struct nsproxy **new_nsp, struct cred *new_cred, struct fs_struct *new_fs) |
e3222c4e | 190 | { |
bcf58e72 | 191 | struct user_namespace *user_ns; |
e3222c4e BP |
192 | int err = 0; |
193 | ||
77ec739d | 194 | if (!(unshare_flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC | |
50804fe3 | 195 | CLONE_NEWNET | CLONE_NEWPID))) |
e3222c4e BP |
196 | return 0; |
197 | ||
b2e0d987 EB |
198 | user_ns = new_cred ? new_cred->user_ns : current_user_ns(); |
199 | if (!ns_capable(user_ns, CAP_SYS_ADMIN)) | |
e3222c4e BP |
200 | return -EPERM; |
201 | ||
bcf58e72 | 202 | *new_nsp = create_new_namespaces(unshare_flags, current, user_ns, |
b2e0d987 | 203 | new_fs ? new_fs : current->fs); |
858d72ea | 204 | if (IS_ERR(*new_nsp)) { |
e3222c4e | 205 | err = PTR_ERR(*new_nsp); |
858d72ea SH |
206 | goto out; |
207 | } | |
208 | ||
858d72ea | 209 | out: |
e3222c4e BP |
210 | return err; |
211 | } | |
98c0d07c | 212 | |
cf7b708c PE |
213 | void switch_task_namespaces(struct task_struct *p, struct nsproxy *new) |
214 | { | |
215 | struct nsproxy *ns; | |
216 | ||
217 | might_sleep(); | |
218 | ||
219 | ns = p->nsproxy; | |
220 | ||
221 | rcu_assign_pointer(p->nsproxy, new); | |
222 | ||
223 | if (ns && atomic_dec_and_test(&ns->count)) { | |
224 | /* | |
225 | * wait for others to get what they want from this nsproxy. | |
226 | * | |
227 | * cannot release this nsproxy via the call_rcu() since | |
228 | * put_mnt_ns() will want to sleep | |
229 | */ | |
230 | synchronize_rcu(); | |
231 | free_nsproxy(ns); | |
232 | } | |
233 | } | |
234 | ||
235 | void exit_task_namespaces(struct task_struct *p) | |
236 | { | |
237 | switch_task_namespaces(p, NULL); | |
238 | } | |
239 | ||
0663c6f8 EB |
240 | SYSCALL_DEFINE2(setns, int, fd, int, nstype) |
241 | { | |
242 | const struct proc_ns_operations *ops; | |
243 | struct task_struct *tsk = current; | |
244 | struct nsproxy *new_nsproxy; | |
245 | struct proc_inode *ei; | |
246 | struct file *file; | |
247 | int err; | |
248 | ||
0663c6f8 EB |
249 | file = proc_ns_fget(fd); |
250 | if (IS_ERR(file)) | |
251 | return PTR_ERR(file); | |
252 | ||
253 | err = -EINVAL; | |
254 | ei = PROC_I(file->f_dentry->d_inode); | |
255 | ops = ei->ns_ops; | |
256 | if (nstype && (ops->type != nstype)) | |
257 | goto out; | |
258 | ||
bcf58e72 | 259 | new_nsproxy = create_new_namespaces(0, tsk, current_user_ns(), tsk->fs); |
0663c6f8 EB |
260 | if (IS_ERR(new_nsproxy)) { |
261 | err = PTR_ERR(new_nsproxy); | |
262 | goto out; | |
263 | } | |
264 | ||
265 | err = ops->install(new_nsproxy, ei->ns); | |
266 | if (err) { | |
267 | free_nsproxy(new_nsproxy); | |
268 | goto out; | |
269 | } | |
270 | switch_task_namespaces(tsk, new_nsproxy); | |
271 | out: | |
272 | fput(file); | |
273 | return err; | |
274 | } | |
275 | ||
66577193 | 276 | int __init nsproxy_cache_init(void) |
98c0d07c | 277 | { |
db8906da | 278 | nsproxy_cachep = KMEM_CACHE(nsproxy, SLAB_PANIC); |
98c0d07c CLG |
279 | return 0; |
280 | } |