nsproxy.c 6.1 KB
Newer Older
S
Serge E. Hallyn 已提交
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.
K
Kirill Korotaev 已提交
10 11 12 13
 *
 *  Jun 2006 - namespaces support
 *             OpenVZ, SWsoft Inc.
 *             Pavel Emelianov <xemul@openvz.org>
S
Serge E. Hallyn 已提交
14 15
 */

16
#include <linux/slab.h>
17
#include <linux/export.h>
S
Serge E. Hallyn 已提交
18
#include <linux/nsproxy.h>
19
#include <linux/init_task.h>
20
#include <linux/mnt_namespace.h>
21
#include <linux/utsname.h>
C
Cedric Le Goater 已提交
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

29 30
static struct kmem_cache *nsproxy_cachep;

A
Alexey Dobriyan 已提交
31
struct nsproxy init_nsproxy = {
32 33
	.count			= ATOMIC_INIT(1),
	.uts_ns			= &init_uts_ns,
A
Alexey Dobriyan 已提交
34
#if defined(CONFIG_POSIX_MQUEUE) || defined(CONFIG_SYSVIPC)
35
	.ipc_ns			= &init_ipc_ns,
A
Alexey Dobriyan 已提交
36
#endif
37 38
	.mnt_ns			= NULL,
	.pid_ns_for_children	= &init_pid_ns,
A
Alexey Dobriyan 已提交
39
#ifdef CONFIG_NET
40
	.net_ns			= &init_net,
A
Alexey Dobriyan 已提交
41 42
#endif
};
S
Serge E. Hallyn 已提交
43

44
static inline struct nsproxy *create_nsproxy(void)
S
Serge E. Hallyn 已提交
45
{
46
	struct nsproxy *nsproxy;
S
Serge E. Hallyn 已提交
47

48 49 50 51
	nsproxy = kmem_cache_alloc(nsproxy_cachep, GFP_KERNEL);
	if (nsproxy)
		atomic_set(&nsproxy->count, 1);
	return nsproxy;
S
Serge E. Hallyn 已提交
52 53 54
}

/*
55 56 57
 * 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.
S
Serge E. Hallyn 已提交
58
 */
59
static struct nsproxy *create_new_namespaces(unsigned long flags,
60 61
	struct task_struct *tsk, struct user_namespace *user_ns,
	struct fs_struct *new_fs)
S
Serge E. Hallyn 已提交
62
{
63
	struct nsproxy *new_nsp;
64
	int err;
S
Serge E. Hallyn 已提交
65

66
	new_nsp = create_nsproxy();
67 68
	if (!new_nsp)
		return ERR_PTR(-ENOMEM);
69

70
	new_nsp->mnt_ns = copy_mnt_ns(flags, tsk->nsproxy->mnt_ns, user_ns, new_fs);
71 72
	if (IS_ERR(new_nsp->mnt_ns)) {
		err = PTR_ERR(new_nsp->mnt_ns);
73
		goto out_ns;
74
	}
75

76
	new_nsp->uts_ns = copy_utsname(flags, user_ns, tsk->nsproxy->uts_ns);
77 78
	if (IS_ERR(new_nsp->uts_ns)) {
		err = PTR_ERR(new_nsp->uts_ns);
79
		goto out_uts;
80
	}
81

82
	new_nsp->ipc_ns = copy_ipcs(flags, user_ns, tsk->nsproxy->ipc_ns);
83 84
	if (IS_ERR(new_nsp->ipc_ns)) {
		err = PTR_ERR(new_nsp->ipc_ns);
85
		goto out_ipc;
86
	}
87

88 89 90 91
	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);
92
		goto out_pid;
93
	}
94

95
	new_nsp->net_ns = copy_net_ns(flags, user_ns, tsk->nsproxy->net_ns);
96 97 98 99 100
	if (IS_ERR(new_nsp->net_ns)) {
		err = PTR_ERR(new_nsp->net_ns);
		goto out_net;
	}

101 102
	return new_nsp;

103
out_net:
104 105
	if (new_nsp->pid_ns_for_children)
		put_pid_ns(new_nsp->pid_ns_for_children);
106 107 108 109 110 111 112 113 114 115
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:
116
	kmem_cache_free(nsproxy_cachep, new_nsp);
117
	return ERR_PTR(err);
S
Serge E. Hallyn 已提交
118 119 120 121 122 123
}

/*
 * called from clone.  This now handles copy for nsproxy and all
 * namespaces therein.
 */
124
int copy_namespaces(unsigned long flags, struct task_struct *tsk)
S
Serge E. Hallyn 已提交
125 126
{
	struct nsproxy *old_ns = tsk->nsproxy;
127
	struct user_namespace *user_ns = task_cred_xxx(tsk, user_ns);
128
	struct nsproxy *new_ns;
S
Serge E. Hallyn 已提交
129

130 131 132
	if (likely(!(flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC |
			      CLONE_NEWPID | CLONE_NEWNET)))) {
		get_nsproxy(old_ns);
S
Serge E. Hallyn 已提交
133
		return 0;
134 135
	}

136 137 138
	if (!ns_capable(user_ns, CAP_SYS_ADMIN))
		return -EPERM;

139 140 141 142 143 144 145
	/*
	 * 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.
	 */
146
	if ((flags & (CLONE_NEWIPC | CLONE_SYSVSEM)) ==
147 148
		(CLONE_NEWIPC | CLONE_SYSVSEM)) 
		return -EINVAL;
149

150
	new_ns = create_new_namespaces(flags, tsk, user_ns, tsk->fs);
151 152
	if (IS_ERR(new_ns))
		return  PTR_ERR(new_ns);
C
Cedric Le Goater 已提交
153

154
	tsk->nsproxy = new_ns;
155
	return 0;
S
Serge E. Hallyn 已提交
156 157 158 159
}

void free_nsproxy(struct nsproxy *ns)
{
C
Cedric Le Goater 已提交
160 161 162 163 164 165
	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);
166 167
	if (ns->pid_ns_for_children)
		put_pid_ns(ns->pid_ns_for_children);
168
	put_net(ns->net_ns);
169
	kmem_cache_free(nsproxy_cachep, ns);
S
Serge E. Hallyn 已提交
170
}
171 172 173

/*
 * Called from unshare. Unshare all the namespaces part of nsproxy.
174
 * On success, returns the new nsproxy.
175 176
 */
int unshare_nsproxy_namespaces(unsigned long unshare_flags,
177
	struct nsproxy **new_nsp, struct cred *new_cred, struct fs_struct *new_fs)
178
{
179
	struct user_namespace *user_ns;
180 181
	int err = 0;

S
Serge E. Hallyn 已提交
182
	if (!(unshare_flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC |
183
			       CLONE_NEWNET | CLONE_NEWPID)))
184 185
		return 0;

186 187
	user_ns = new_cred ? new_cred->user_ns : current_user_ns();
	if (!ns_capable(user_ns, CAP_SYS_ADMIN))
188 189
		return -EPERM;

190
	*new_nsp = create_new_namespaces(unshare_flags, current, user_ns,
191
					 new_fs ? new_fs : current->fs);
192
	if (IS_ERR(*new_nsp)) {
193
		err = PTR_ERR(*new_nsp);
194 195 196 197
		goto out;
	}

out:
198 199
	return err;
}
200

201 202 203 204 205 206
void switch_task_namespaces(struct task_struct *p, struct nsproxy *new)
{
	struct nsproxy *ns;

	might_sleep();

207
	task_lock(p);
208
	ns = p->nsproxy;
209 210
	p->nsproxy = new;
	task_unlock(p);
211

212
	if (ns && atomic_dec_and_test(&ns->count))
213 214 215 216 217 218 219 220
		free_nsproxy(ns);
}

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

221 222 223 224
SYSCALL_DEFINE2(setns, int, fd, int, nstype)
{
	struct task_struct *tsk = current;
	struct nsproxy *new_nsproxy;
225
	struct proc_ns *ei;
226
	struct file *file;
227
	struct ns_common *ns;
228 229 230 231 232 233 234
	int err;

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

	err = -EINVAL;
235
	ei = get_proc_ns(file_inode(file));
236 237
	ns = ei->ns;
	if (nstype && (ns->ops->type != nstype))
238 239
		goto out;

240
	new_nsproxy = create_new_namespaces(0, tsk, current_user_ns(), tsk->fs);
241 242 243 244 245
	if (IS_ERR(new_nsproxy)) {
		err = PTR_ERR(new_nsproxy);
		goto out;
	}

246
	err = ns->ops->install(new_nsproxy, ns);
247 248 249 250 251 252 253 254 255 256
	if (err) {
		free_nsproxy(new_nsproxy);
		goto out;
	}
	switch_task_namespaces(tsk, new_nsproxy);
out:
	fput(file);
	return err;
}

257
int __init nsproxy_cache_init(void)
258
{
259
	nsproxy_cachep = KMEM_CACHE(nsproxy, SLAB_PANIC);
260 261
	return 0;
}