nsproxy.c 6.49 KB
Newer Older
1 2 3 4 5 6 7 8 9
/*
 *  Copyright (C) 2006 IBM Corporation
 *
 *  Author: Serge Hallyn <serue@us.ibm.com>
 *
 *  This program is free software; you can redistribute it and/or
 *  modify it under the terms of the GNU General Public License as
 *  published by the Free Software Foundation, version 2 of the
 *  License.
Kirill Korotaev's avatar
Kirill Korotaev committed
10 11 12 13
 *
 *  Jun 2006 - namespaces support
 *             OpenVZ, SWsoft Inc.
 *             Pavel Emelianov <xemul@openvz.org>
14 15
 */

16
#include <linux/slab.h>
17
#include <linux/export.h>
18
#include <linux/nsproxy.h>
19
#include <linux/init_task.h>
20
#include <linux/mnt_namespace.h>
21
#include <linux/utsname.h>
Cedric Le Goater's avatar
Cedric Le Goater committed
22
#include <linux/pid_namespace.h>
23
#include <net/net_namespace.h>
24
#include <linux/ipc_namespace.h>
25
#include <linux/proc_ns.h>
26 27
#include <linux/file.h>
#include <linux/syscalls.h>
28
#include <linux/cgroup.h>
29
#include <linux/perf_event.h>
30

31 32
static struct kmem_cache *nsproxy_cachep;

33
struct nsproxy init_nsproxy = {
34 35
	.count			= ATOMIC_INIT(1),
	.uts_ns			= &init_uts_ns,
36
#if defined(CONFIG_POSIX_MQUEUE) || defined(CONFIG_SYSVIPC)
37
	.ipc_ns			= &init_ipc_ns,
38
#endif
39 40
	.mnt_ns			= NULL,
	.pid_ns_for_children	= &init_pid_ns,
41
#ifdef CONFIG_NET
42
	.net_ns			= &init_net,
43
#endif
44 45 46
#ifdef CONFIG_CGROUPS
	.cgroup_ns		= &init_cgroup_ns,
#endif
47
};
48

49
static inline struct nsproxy *create_nsproxy(void)
50
{
51
	struct nsproxy *nsproxy;
52

53 54 55 56
	nsproxy = kmem_cache_alloc(nsproxy_cachep, GFP_KERNEL);
	if (nsproxy)
		atomic_set(&nsproxy->count, 1);
	return nsproxy;
57 58 59
}

/*
60 61 62
 * Create new nsproxy and all of its the associated namespaces.
 * Return the newly created nsproxy.  Do not attach this to the task,
 * leave it to the caller to do proper locking and attach it to task.
63
 */
64
static struct nsproxy *create_new_namespaces(unsigned long flags,
65 66
	struct task_struct *tsk, struct user_namespace *user_ns,
	struct fs_struct *new_fs)
67
{
68
	struct nsproxy *new_nsp;
69
	int err;
70

71
	new_nsp = create_nsproxy();
72 73
	if (!new_nsp)
		return ERR_PTR(-ENOMEM);
74

75
	new_nsp->mnt_ns = copy_mnt_ns(flags, tsk->nsproxy->mnt_ns, user_ns, new_fs);
76 77
	if (IS_ERR(new_nsp->mnt_ns)) {
		err = PTR_ERR(new_nsp->mnt_ns);
78
		goto out_ns;
79
	}
80

81
	new_nsp->uts_ns = copy_utsname(flags, user_ns, tsk->nsproxy->uts_ns);
82 83
	if (IS_ERR(new_nsp->uts_ns)) {
		err = PTR_ERR(new_nsp->uts_ns);
84
		goto out_uts;
85
	}
86

87
	new_nsp->ipc_ns = copy_ipcs(flags, user_ns, tsk->nsproxy->ipc_ns);
88 89
	if (IS_ERR(new_nsp->ipc_ns)) {
		err = PTR_ERR(new_nsp->ipc_ns);
90
		goto out_ipc;
91
	}
92

93 94 95 96
	new_nsp->pid_ns_for_children =
		copy_pid_ns(flags, user_ns, tsk->nsproxy->pid_ns_for_children);
	if (IS_ERR(new_nsp->pid_ns_for_children)) {
		err = PTR_ERR(new_nsp->pid_ns_for_children);
97
		goto out_pid;
98
	}
99

100 101 102 103 104 105 106
	new_nsp->cgroup_ns = copy_cgroup_ns(flags, user_ns,
					    tsk->nsproxy->cgroup_ns);
	if (IS_ERR(new_nsp->cgroup_ns)) {
		err = PTR_ERR(new_nsp->cgroup_ns);
		goto out_cgroup;
	}

107
	new_nsp->net_ns = copy_net_ns(flags, user_ns, tsk->nsproxy->net_ns);
108 109 110 111 112
	if (IS_ERR(new_nsp->net_ns)) {
		err = PTR_ERR(new_nsp->net_ns);
		goto out_net;
	}

113 114
	return new_nsp;

115
out_net:
116 117
	put_cgroup_ns(new_nsp->cgroup_ns);
out_cgroup:
118 119
	if (new_nsp->pid_ns_for_children)
		put_pid_ns(new_nsp->pid_ns_for_children);
120 121 122 123 124 125 126 127 128 129
out_pid:
	if (new_nsp->ipc_ns)
		put_ipc_ns(new_nsp->ipc_ns);
out_ipc:
	if (new_nsp->uts_ns)
		put_uts_ns(new_nsp->uts_ns);
out_uts:
	if (new_nsp->mnt_ns)
		put_mnt_ns(new_nsp->mnt_ns);
out_ns:
130
	kmem_cache_free(nsproxy_cachep, new_nsp);
131
	return ERR_PTR(err);
132 133 134 135 136 137
}

/*
 * called from clone.  This now handles copy for nsproxy and all
 * namespaces therein.
 */
138
int copy_namespaces(unsigned long flags, struct task_struct *tsk)
139 140
{
	struct nsproxy *old_ns = tsk->nsproxy;
141
	struct user_namespace *user_ns = task_cred_xxx(tsk, user_ns);
142
	struct nsproxy *new_ns;
143

144
	if (likely(!(flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC |
145 146
			      CLONE_NEWPID | CLONE_NEWNET |
			      CLONE_NEWCGROUP)))) {
147
		get_nsproxy(old_ns);
148
		return 0;
149 150
	}

151 152 153
	if (!ns_capable(user_ns, CAP_SYS_ADMIN))
		return -EPERM;

154 155 156 157 158 159 160
	/*
	 * CLONE_NEWIPC must detach from the undolist: after switching
	 * to a new ipc namespace, the semaphore arrays from the old
	 * namespace are unreachable.  In clone parlance, CLONE_SYSVSEM
	 * means share undolist with parent, so we must forbid using
	 * it along with CLONE_NEWIPC.
	 */
161
	if ((flags & (CLONE_NEWIPC | CLONE_SYSVSEM)) ==
162 163
		(CLONE_NEWIPC | CLONE_SYSVSEM)) 
		return -EINVAL;
164

165
	new_ns = create_new_namespaces(flags, tsk, user_ns, tsk->fs);
166 167
	if (IS_ERR(new_ns))
		return  PTR_ERR(new_ns);
Cedric Le Goater's avatar
Cedric Le Goater committed
168

169
	tsk->nsproxy = new_ns;
170
	return 0;
171 172 173 174
}

void free_nsproxy(struct nsproxy *ns)
{
Cedric Le Goater's avatar
Cedric Le Goater committed
175 176 177 178 179 180
	if (ns->mnt_ns)
		put_mnt_ns(ns->mnt_ns);
	if (ns->uts_ns)
		put_uts_ns(ns->uts_ns);
	if (ns->ipc_ns)
		put_ipc_ns(ns->ipc_ns);
181 182
	if (ns->pid_ns_for_children)
		put_pid_ns(ns->pid_ns_for_children);
183
	put_cgroup_ns(ns->cgroup_ns);
184
	put_net(ns->net_ns);
185
	kmem_cache_free(nsproxy_cachep, ns);
186
}
187 188 189

/*
 * Called from unshare. Unshare all the namespaces part of nsproxy.
190
 * On success, returns the new nsproxy.
191 192
 */
int unshare_nsproxy_namespaces(unsigned long unshare_flags,
193
	struct nsproxy **new_nsp, struct cred *new_cred, struct fs_struct *new_fs)
194
{
195
	struct user_namespace *user_ns;
196 197
	int err = 0;

Serge E. Hallyn's avatar
Serge E. Hallyn committed
198
	if (!(unshare_flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC |
199
			       CLONE_NEWNET | CLONE_NEWPID | CLONE_NEWCGROUP)))
200 201
		return 0;

202 203
	user_ns = new_cred ? new_cred->user_ns : current_user_ns();
	if (!ns_capable(user_ns, CAP_SYS_ADMIN))
204 205
		return -EPERM;

206
	*new_nsp = create_new_namespaces(unshare_flags, current, user_ns,
207
					 new_fs ? new_fs : current->fs);
208
	if (IS_ERR(*new_nsp)) {
209
		err = PTR_ERR(*new_nsp);
210 211 212 213
		goto out;
	}

out:
214 215
	return err;
}
216

217 218 219 220 221 222
void switch_task_namespaces(struct task_struct *p, struct nsproxy *new)
{
	struct nsproxy *ns;

	might_sleep();

223
	task_lock(p);
224
	ns = p->nsproxy;
225 226
	p->nsproxy = new;
	task_unlock(p);
227

228
	if (ns && atomic_dec_and_test(&ns->count))
229 230 231 232 233 234 235 236
		free_nsproxy(ns);
}

void exit_task_namespaces(struct task_struct *p)
{
	switch_task_namespaces(p, NULL);
}

237 238 239 240 241
SYSCALL_DEFINE2(setns, int, fd, int, nstype)
{
	struct task_struct *tsk = current;
	struct nsproxy *new_nsproxy;
	struct file *file;
242
	struct ns_common *ns;
243 244 245 246 247 248 249
	int err;

	file = proc_ns_fget(fd);
	if (IS_ERR(file))
		return PTR_ERR(file);

	err = -EINVAL;
Al Viro's avatar
Al Viro committed
250
	ns = get_proc_ns(file_inode(file));
251
	if (nstype && (ns->ops->type != nstype))
252 253
		goto out;

254
	new_nsproxy = create_new_namespaces(0, tsk, current_user_ns(), tsk->fs);
255 256 257 258 259
	if (IS_ERR(new_nsproxy)) {
		err = PTR_ERR(new_nsproxy);
		goto out;
	}

260
	err = ns->ops->install(new_nsproxy, ns);
261 262 263 264 265
	if (err) {
		free_nsproxy(new_nsproxy);
		goto out;
	}
	switch_task_namespaces(tsk, new_nsproxy);
266 267

	perf_event_namespaces(tsk);
268 269 270 271 272
out:
	fput(file);
	return err;
}

273
int __init nsproxy_cache_init(void)
274
{
275
	nsproxy_cachep = KMEM_CACHE(nsproxy, SLAB_PANIC);
276 277
	return 0;
}