nsproxy.c 6.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281
  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. #include <linux/slab.h>
  16. #include <linux/module.h>
  17. #include <linux/nsproxy.h>
  18. #include <linux/init_task.h>
  19. #include <linux/mnt_namespace.h>
  20. #include <linux/utsname.h>
  21. #include <linux/pid_namespace.h>
  22. #include <net/net_namespace.h>
  23. #include <linux/ipc_namespace.h>
  24. #include <linux/proc_fs.h>
  25. #include <linux/file.h>
  26. #include <linux/syscalls.h>
  27. static struct kmem_cache *nsproxy_cachep;
  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. static inline struct nsproxy *create_nsproxy(void)
  41. {
  42. struct nsproxy *nsproxy;
  43. nsproxy = kmem_cache_alloc(nsproxy_cachep, GFP_KERNEL);
  44. if (nsproxy)
  45. atomic_set(&nsproxy->count, 1);
  46. return nsproxy;
  47. }
  48. /*
  49. * Create new nsproxy and all of its the associated namespaces.
  50. * Return the newly created nsproxy. Do not attach this to the task,
  51. * leave it to the caller to do proper locking and attach it to task.
  52. */
  53. static struct nsproxy *create_new_namespaces(unsigned long flags,
  54. struct task_struct *tsk, struct fs_struct *new_fs)
  55. {
  56. struct nsproxy *new_nsp;
  57. int err;
  58. new_nsp = create_nsproxy();
  59. if (!new_nsp)
  60. return ERR_PTR(-ENOMEM);
  61. new_nsp->mnt_ns = copy_mnt_ns(flags, tsk->nsproxy->mnt_ns, new_fs);
  62. if (IS_ERR(new_nsp->mnt_ns)) {
  63. err = PTR_ERR(new_nsp->mnt_ns);
  64. goto out_ns;
  65. }
  66. new_nsp->uts_ns = copy_utsname(flags, tsk);
  67. if (IS_ERR(new_nsp->uts_ns)) {
  68. err = PTR_ERR(new_nsp->uts_ns);
  69. goto out_uts;
  70. }
  71. new_nsp->ipc_ns = copy_ipcs(flags, tsk);
  72. if (IS_ERR(new_nsp->ipc_ns)) {
  73. err = PTR_ERR(new_nsp->ipc_ns);
  74. goto out_ipc;
  75. }
  76. new_nsp->pid_ns = copy_pid_ns(flags, task_active_pid_ns(tsk));
  77. if (IS_ERR(new_nsp->pid_ns)) {
  78. err = PTR_ERR(new_nsp->pid_ns);
  79. goto out_pid;
  80. }
  81. new_nsp->net_ns = copy_net_ns(flags, tsk->nsproxy->net_ns);
  82. if (IS_ERR(new_nsp->net_ns)) {
  83. err = PTR_ERR(new_nsp->net_ns);
  84. goto out_net;
  85. }
  86. return new_nsp;
  87. out_net:
  88. if (new_nsp->pid_ns)
  89. put_pid_ns(new_nsp->pid_ns);
  90. out_pid:
  91. if (new_nsp->ipc_ns)
  92. put_ipc_ns(new_nsp->ipc_ns);
  93. out_ipc:
  94. if (new_nsp->uts_ns)
  95. put_uts_ns(new_nsp->uts_ns);
  96. out_uts:
  97. if (new_nsp->mnt_ns)
  98. put_mnt_ns(new_nsp->mnt_ns);
  99. out_ns:
  100. kmem_cache_free(nsproxy_cachep, new_nsp);
  101. return ERR_PTR(err);
  102. }
  103. /*
  104. * called from clone. This now handles copy for nsproxy and all
  105. * namespaces therein.
  106. */
  107. int copy_namespaces(unsigned long flags, struct task_struct *tsk)
  108. {
  109. struct nsproxy *old_ns = tsk->nsproxy;
  110. struct nsproxy *new_ns;
  111. int err = 0;
  112. if (!old_ns)
  113. return 0;
  114. get_nsproxy(old_ns);
  115. if (!(flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC |
  116. CLONE_NEWPID | CLONE_NEWNET)))
  117. return 0;
  118. if (!capable(CAP_SYS_ADMIN)) {
  119. err = -EPERM;
  120. goto out;
  121. }
  122. /*
  123. * CLONE_NEWIPC must detach from the undolist: after switching
  124. * to a new ipc namespace, the semaphore arrays from the old
  125. * namespace are unreachable. In clone parlance, CLONE_SYSVSEM
  126. * means share undolist with parent, so we must forbid using
  127. * it along with CLONE_NEWIPC.
  128. */
  129. if ((flags & CLONE_NEWIPC) && (flags & CLONE_SYSVSEM)) {
  130. err = -EINVAL;
  131. goto out;
  132. }
  133. new_ns = create_new_namespaces(flags, tsk, tsk->fs);
  134. if (IS_ERR(new_ns)) {
  135. err = PTR_ERR(new_ns);
  136. goto out;
  137. }
  138. tsk->nsproxy = new_ns;
  139. out:
  140. put_nsproxy(old_ns);
  141. return err;
  142. }
  143. void free_nsproxy(struct nsproxy *ns)
  144. {
  145. if (ns->mnt_ns)
  146. put_mnt_ns(ns->mnt_ns);
  147. if (ns->uts_ns)
  148. put_uts_ns(ns->uts_ns);
  149. if (ns->ipc_ns)
  150. put_ipc_ns(ns->ipc_ns);
  151. if (ns->pid_ns)
  152. put_pid_ns(ns->pid_ns);
  153. put_net(ns->net_ns);
  154. kmem_cache_free(nsproxy_cachep, ns);
  155. }
  156. /*
  157. * Called from unshare. Unshare all the namespaces part of nsproxy.
  158. * On success, returns the new nsproxy.
  159. */
  160. int unshare_nsproxy_namespaces(unsigned long unshare_flags,
  161. struct nsproxy **new_nsp, struct fs_struct *new_fs)
  162. {
  163. int err = 0;
  164. if (!(unshare_flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC |
  165. CLONE_NEWNET)))
  166. return 0;
  167. if (!capable(CAP_SYS_ADMIN))
  168. return -EPERM;
  169. *new_nsp = create_new_namespaces(unshare_flags, current,
  170. new_fs ? new_fs : current->fs);
  171. if (IS_ERR(*new_nsp)) {
  172. err = PTR_ERR(*new_nsp);
  173. goto out;
  174. }
  175. out:
  176. return err;
  177. }
  178. void switch_task_namespaces(struct task_struct *p, struct nsproxy *new)
  179. {
  180. struct nsproxy *ns;
  181. might_sleep();
  182. ns = p->nsproxy;
  183. rcu_assign_pointer(p->nsproxy, new);
  184. if (ns && atomic_dec_and_test(&ns->count)) {
  185. /*
  186. * wait for others to get what they want from this nsproxy.
  187. *
  188. * cannot release this nsproxy via the call_rcu() since
  189. * put_mnt_ns() will want to sleep
  190. */
  191. synchronize_rcu();
  192. free_nsproxy(ns);
  193. }
  194. }
  195. void exit_task_namespaces(struct task_struct *p)
  196. {
  197. switch_task_namespaces(p, NULL);
  198. }
  199. SYSCALL_DEFINE2(setns, int, fd, int, nstype)
  200. {
  201. const struct proc_ns_operations *ops;
  202. struct task_struct *tsk = current;
  203. struct nsproxy *new_nsproxy;
  204. struct proc_inode *ei;
  205. struct file *file;
  206. int err;
  207. if (!capable(CAP_SYS_ADMIN))
  208. return -EPERM;
  209. file = proc_ns_fget(fd);
  210. if (IS_ERR(file))
  211. return PTR_ERR(file);
  212. err = -EINVAL;
  213. ei = PROC_I(file->f_dentry->d_inode);
  214. ops = ei->ns_ops;
  215. if (nstype && (ops->type != nstype))
  216. goto out;
  217. new_nsproxy = create_new_namespaces(0, tsk, tsk->fs);
  218. if (IS_ERR(new_nsproxy)) {
  219. err = PTR_ERR(new_nsproxy);
  220. goto out;
  221. }
  222. err = ops->install(new_nsproxy, ei->ns);
  223. if (err) {
  224. free_nsproxy(new_nsproxy);
  225. goto out;
  226. }
  227. switch_task_namespaces(tsk, new_nsproxy);
  228. out:
  229. fput(file);
  230. return err;
  231. }
  232. static int __init nsproxy_cache_init(void)
  233. {
  234. nsproxy_cachep = KMEM_CACHE(nsproxy, SLAB_PANIC);
  235. return 0;
  236. }
  237. module_init(nsproxy_cache_init);