2 * This program is free software; you can redistribute it and/or
3 * modify it under the terms of the GNU General Public License as
4 * published by the Free Software Foundation, version 2 of the
8 #include <linux/export.h>
9 #include <linux/nsproxy.h>
10 #include <linux/slab.h>
11 #include <linux/user_namespace.h>
12 #include <linux/proc_ns.h>
13 #include <linux/highuid.h>
14 #include <linux/cred.h>
15 #include <linux/securebits.h>
16 #include <linux/keyctl.h>
17 #include <linux/key-type.h>
18 #include <keys/user-type.h>
19 #include <linux/seq_file.h>
21 #include <linux/uaccess.h>
22 #include <linux/ctype.h>
23 #include <linux/projid.h>
24 #include <linux/fs_struct.h>
26 static struct kmem_cache
*user_ns_cachep __read_mostly
;
27 static DEFINE_MUTEX(userns_state_mutex
);
29 static bool new_idmap_permitted(const struct file
*file
,
30 struct user_namespace
*ns
, int cap_setid
,
31 struct uid_gid_map
*map
);
33 static void set_cred_user_ns(struct cred
*cred
, struct user_namespace
*user_ns
)
35 /* Start with the same capabilities as init but useless for doing
36 * anything as the capabilities are bound to the new user namespace.
38 cred
->securebits
= SECUREBITS_DEFAULT
;
39 cred
->cap_inheritable
= CAP_EMPTY_SET
;
40 cred
->cap_permitted
= CAP_FULL_SET
;
41 cred
->cap_effective
= CAP_FULL_SET
;
42 cred
->cap_bset
= CAP_FULL_SET
;
44 key_put(cred
->request_key_auth
);
45 cred
->request_key_auth
= NULL
;
47 /* tgcred will be cleared in our caller bc CLONE_THREAD won't be set */
48 cred
->user_ns
= user_ns
;
52 * Create a new user namespace, deriving the creator from the user in the
53 * passed credentials, and replacing that user with the new root user for the
56 * This is called by copy_creds(), which will finish setting the target task's
59 int create_user_ns(struct cred
*new)
61 struct user_namespace
*ns
, *parent_ns
= new->user_ns
;
62 kuid_t owner
= new->euid
;
63 kgid_t group
= new->egid
;
66 if (parent_ns
->level
> 32)
70 * Verify that we can not violate the policy of which files
71 * may be accessed that is specified by the root directory,
72 * by verifing that the root directory is at the root of the
73 * mount namespace which allows all files to be accessed.
75 if (current_chrooted())
78 /* The creator needs a mapping in the parent user namespace
79 * or else we won't be able to reasonably tell userspace who
80 * created a user_namespace.
82 if (!kuid_has_mapping(parent_ns
, owner
) ||
83 !kgid_has_mapping(parent_ns
, group
))
86 ns
= kmem_cache_zalloc(user_ns_cachep
, GFP_KERNEL
);
90 ret
= proc_alloc_inum(&ns
->proc_inum
);
92 kmem_cache_free(user_ns_cachep
, ns
);
96 atomic_set(&ns
->count
, 1);
97 /* Leave the new->user_ns reference with the new user namespace. */
98 ns
->parent
= parent_ns
;
99 ns
->level
= parent_ns
->level
+ 1;
103 /* Inherit USERNS_SETGROUPS_ALLOWED from our parent */
104 mutex_lock(&userns_state_mutex
);
105 ns
->flags
= parent_ns
->flags
;
106 mutex_unlock(&userns_state_mutex
);
108 set_cred_user_ns(new, ns
);
110 update_mnt_policy(ns
);
115 int unshare_userns(unsigned long unshare_flags
, struct cred
**new_cred
)
120 if (!(unshare_flags
& CLONE_NEWUSER
))
123 cred
= prepare_creds();
125 err
= create_user_ns(cred
);
135 void free_user_ns(struct user_namespace
*ns
)
137 struct user_namespace
*parent
;
141 proc_free_inum(ns
->proc_inum
);
142 kmem_cache_free(user_ns_cachep
, ns
);
144 } while (atomic_dec_and_test(&parent
->count
));
146 EXPORT_SYMBOL(free_user_ns
);
148 static u32
map_id_range_down(struct uid_gid_map
*map
, u32 id
, u32 count
)
150 unsigned idx
, extents
;
151 u32 first
, last
, id2
;
153 id2
= id
+ count
- 1;
155 /* Find the matching extent */
156 extents
= map
->nr_extents
;
158 for (idx
= 0; idx
< extents
; idx
++) {
159 first
= map
->extent
[idx
].first
;
160 last
= first
+ map
->extent
[idx
].count
- 1;
161 if (id
>= first
&& id
<= last
&&
162 (id2
>= first
&& id2
<= last
))
165 /* Map the id or note failure */
167 id
= (id
- first
) + map
->extent
[idx
].lower_first
;
174 static u32
map_id_down(struct uid_gid_map
*map
, u32 id
)
176 unsigned idx
, extents
;
179 /* Find the matching extent */
180 extents
= map
->nr_extents
;
182 for (idx
= 0; idx
< extents
; idx
++) {
183 first
= map
->extent
[idx
].first
;
184 last
= first
+ map
->extent
[idx
].count
- 1;
185 if (id
>= first
&& id
<= last
)
188 /* Map the id or note failure */
190 id
= (id
- first
) + map
->extent
[idx
].lower_first
;
197 static u32
map_id_up(struct uid_gid_map
*map
, u32 id
)
199 unsigned idx
, extents
;
202 /* Find the matching extent */
203 extents
= map
->nr_extents
;
205 for (idx
= 0; idx
< extents
; idx
++) {
206 first
= map
->extent
[idx
].lower_first
;
207 last
= first
+ map
->extent
[idx
].count
- 1;
208 if (id
>= first
&& id
<= last
)
211 /* Map the id or note failure */
213 id
= (id
- first
) + map
->extent
[idx
].first
;
221 * make_kuid - Map a user-namespace uid pair into a kuid.
222 * @ns: User namespace that the uid is in
223 * @uid: User identifier
225 * Maps a user-namespace uid pair into a kernel internal kuid,
226 * and returns that kuid.
228 * When there is no mapping defined for the user-namespace uid
229 * pair INVALID_UID is returned. Callers are expected to test
230 * for and handle handle INVALID_UID being returned. INVALID_UID
231 * may be tested for using uid_valid().
233 kuid_t
make_kuid(struct user_namespace
*ns
, uid_t uid
)
235 /* Map the uid to a global kernel uid */
236 return KUIDT_INIT(map_id_down(&ns
->uid_map
, uid
));
238 EXPORT_SYMBOL(make_kuid
);
241 * from_kuid - Create a uid from a kuid user-namespace pair.
242 * @targ: The user namespace we want a uid in.
243 * @kuid: The kernel internal uid to start with.
245 * Map @kuid into the user-namespace specified by @targ and
246 * return the resulting uid.
248 * There is always a mapping into the initial user_namespace.
250 * If @kuid has no mapping in @targ (uid_t)-1 is returned.
252 uid_t
from_kuid(struct user_namespace
*targ
, kuid_t kuid
)
254 /* Map the uid from a global kernel uid */
255 return map_id_up(&targ
->uid_map
, __kuid_val(kuid
));
257 EXPORT_SYMBOL(from_kuid
);
260 * from_kuid_munged - Create a uid from a kuid user-namespace pair.
261 * @targ: The user namespace we want a uid in.
262 * @kuid: The kernel internal uid to start with.
264 * Map @kuid into the user-namespace specified by @targ and
265 * return the resulting uid.
267 * There is always a mapping into the initial user_namespace.
269 * Unlike from_kuid from_kuid_munged never fails and always
270 * returns a valid uid. This makes from_kuid_munged appropriate
271 * for use in syscalls like stat and getuid where failing the
272 * system call and failing to provide a valid uid are not an
275 * If @kuid has no mapping in @targ overflowuid is returned.
277 uid_t
from_kuid_munged(struct user_namespace
*targ
, kuid_t kuid
)
280 uid
= from_kuid(targ
, kuid
);
282 if (uid
== (uid_t
) -1)
286 EXPORT_SYMBOL(from_kuid_munged
);
289 * make_kgid - Map a user-namespace gid pair into a kgid.
290 * @ns: User namespace that the gid is in
291 * @uid: group identifier
293 * Maps a user-namespace gid pair into a kernel internal kgid,
294 * and returns that kgid.
296 * When there is no mapping defined for the user-namespace gid
297 * pair INVALID_GID is returned. Callers are expected to test
298 * for and handle INVALID_GID being returned. INVALID_GID may be
299 * tested for using gid_valid().
301 kgid_t
make_kgid(struct user_namespace
*ns
, gid_t gid
)
303 /* Map the gid to a global kernel gid */
304 return KGIDT_INIT(map_id_down(&ns
->gid_map
, gid
));
306 EXPORT_SYMBOL(make_kgid
);
309 * from_kgid - Create a gid from a kgid user-namespace pair.
310 * @targ: The user namespace we want a gid in.
311 * @kgid: The kernel internal gid to start with.
313 * Map @kgid into the user-namespace specified by @targ and
314 * return the resulting gid.
316 * There is always a mapping into the initial user_namespace.
318 * If @kgid has no mapping in @targ (gid_t)-1 is returned.
320 gid_t
from_kgid(struct user_namespace
*targ
, kgid_t kgid
)
322 /* Map the gid from a global kernel gid */
323 return map_id_up(&targ
->gid_map
, __kgid_val(kgid
));
325 EXPORT_SYMBOL(from_kgid
);
328 * from_kgid_munged - Create a gid from a kgid user-namespace pair.
329 * @targ: The user namespace we want a gid in.
330 * @kgid: The kernel internal gid to start with.
332 * Map @kgid into the user-namespace specified by @targ and
333 * return the resulting gid.
335 * There is always a mapping into the initial user_namespace.
337 * Unlike from_kgid from_kgid_munged never fails and always
338 * returns a valid gid. This makes from_kgid_munged appropriate
339 * for use in syscalls like stat and getgid where failing the
340 * system call and failing to provide a valid gid are not options.
342 * If @kgid has no mapping in @targ overflowgid is returned.
344 gid_t
from_kgid_munged(struct user_namespace
*targ
, kgid_t kgid
)
347 gid
= from_kgid(targ
, kgid
);
349 if (gid
== (gid_t
) -1)
353 EXPORT_SYMBOL(from_kgid_munged
);
356 * make_kprojid - Map a user-namespace projid pair into a kprojid.
357 * @ns: User namespace that the projid is in
358 * @projid: Project identifier
360 * Maps a user-namespace uid pair into a kernel internal kuid,
361 * and returns that kuid.
363 * When there is no mapping defined for the user-namespace projid
364 * pair INVALID_PROJID is returned. Callers are expected to test
365 * for and handle handle INVALID_PROJID being returned. INVALID_PROJID
366 * may be tested for using projid_valid().
368 kprojid_t
make_kprojid(struct user_namespace
*ns
, projid_t projid
)
370 /* Map the uid to a global kernel uid */
371 return KPROJIDT_INIT(map_id_down(&ns
->projid_map
, projid
));
373 EXPORT_SYMBOL(make_kprojid
);
376 * from_kprojid - Create a projid from a kprojid user-namespace pair.
377 * @targ: The user namespace we want a projid in.
378 * @kprojid: The kernel internal project identifier to start with.
380 * Map @kprojid into the user-namespace specified by @targ and
381 * return the resulting projid.
383 * There is always a mapping into the initial user_namespace.
385 * If @kprojid has no mapping in @targ (projid_t)-1 is returned.
387 projid_t
from_kprojid(struct user_namespace
*targ
, kprojid_t kprojid
)
389 /* Map the uid from a global kernel uid */
390 return map_id_up(&targ
->projid_map
, __kprojid_val(kprojid
));
392 EXPORT_SYMBOL(from_kprojid
);
395 * from_kprojid_munged - Create a projiid from a kprojid user-namespace pair.
396 * @targ: The user namespace we want a projid in.
397 * @kprojid: The kernel internal projid to start with.
399 * Map @kprojid into the user-namespace specified by @targ and
400 * return the resulting projid.
402 * There is always a mapping into the initial user_namespace.
404 * Unlike from_kprojid from_kprojid_munged never fails and always
405 * returns a valid projid. This makes from_kprojid_munged
406 * appropriate for use in syscalls like stat and where
407 * failing the system call and failing to provide a valid projid are
410 * If @kprojid has no mapping in @targ OVERFLOW_PROJID is returned.
412 projid_t
from_kprojid_munged(struct user_namespace
*targ
, kprojid_t kprojid
)
415 projid
= from_kprojid(targ
, kprojid
);
417 if (projid
== (projid_t
) -1)
418 projid
= OVERFLOW_PROJID
;
421 EXPORT_SYMBOL(from_kprojid_munged
);
424 static int uid_m_show(struct seq_file
*seq
, void *v
)
426 struct user_namespace
*ns
= seq
->private;
427 struct uid_gid_extent
*extent
= v
;
428 struct user_namespace
*lower_ns
;
431 lower_ns
= seq_user_ns(seq
);
432 if ((lower_ns
== ns
) && lower_ns
->parent
)
433 lower_ns
= lower_ns
->parent
;
435 lower
= from_kuid(lower_ns
, KUIDT_INIT(extent
->lower_first
));
437 seq_printf(seq
, "%10u %10u %10u\n",
445 static int gid_m_show(struct seq_file
*seq
, void *v
)
447 struct user_namespace
*ns
= seq
->private;
448 struct uid_gid_extent
*extent
= v
;
449 struct user_namespace
*lower_ns
;
452 lower_ns
= seq_user_ns(seq
);
453 if ((lower_ns
== ns
) && lower_ns
->parent
)
454 lower_ns
= lower_ns
->parent
;
456 lower
= from_kgid(lower_ns
, KGIDT_INIT(extent
->lower_first
));
458 seq_printf(seq
, "%10u %10u %10u\n",
466 static int projid_m_show(struct seq_file
*seq
, void *v
)
468 struct user_namespace
*ns
= seq
->private;
469 struct uid_gid_extent
*extent
= v
;
470 struct user_namespace
*lower_ns
;
473 lower_ns
= seq_user_ns(seq
);
474 if ((lower_ns
== ns
) && lower_ns
->parent
)
475 lower_ns
= lower_ns
->parent
;
477 lower
= from_kprojid(lower_ns
, KPROJIDT_INIT(extent
->lower_first
));
479 seq_printf(seq
, "%10u %10u %10u\n",
487 static void *m_start(struct seq_file
*seq
, loff_t
*ppos
, struct uid_gid_map
*map
)
489 struct uid_gid_extent
*extent
= NULL
;
492 if (pos
< map
->nr_extents
)
493 extent
= &map
->extent
[pos
];
498 static void *uid_m_start(struct seq_file
*seq
, loff_t
*ppos
)
500 struct user_namespace
*ns
= seq
->private;
502 return m_start(seq
, ppos
, &ns
->uid_map
);
505 static void *gid_m_start(struct seq_file
*seq
, loff_t
*ppos
)
507 struct user_namespace
*ns
= seq
->private;
509 return m_start(seq
, ppos
, &ns
->gid_map
);
512 static void *projid_m_start(struct seq_file
*seq
, loff_t
*ppos
)
514 struct user_namespace
*ns
= seq
->private;
516 return m_start(seq
, ppos
, &ns
->projid_map
);
519 static void *m_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
522 return seq
->op
->start(seq
, pos
);
525 static void m_stop(struct seq_file
*seq
, void *v
)
530 struct seq_operations proc_uid_seq_operations
= {
531 .start
= uid_m_start
,
537 struct seq_operations proc_gid_seq_operations
= {
538 .start
= gid_m_start
,
544 struct seq_operations proc_projid_seq_operations
= {
545 .start
= projid_m_start
,
548 .show
= projid_m_show
,
551 static bool mappings_overlap(struct uid_gid_map
*new_map
, struct uid_gid_extent
*extent
)
553 u32 upper_first
, lower_first
, upper_last
, lower_last
;
556 upper_first
= extent
->first
;
557 lower_first
= extent
->lower_first
;
558 upper_last
= upper_first
+ extent
->count
- 1;
559 lower_last
= lower_first
+ extent
->count
- 1;
561 for (idx
= 0; idx
< new_map
->nr_extents
; idx
++) {
562 u32 prev_upper_first
, prev_lower_first
;
563 u32 prev_upper_last
, prev_lower_last
;
564 struct uid_gid_extent
*prev
;
566 prev
= &new_map
->extent
[idx
];
568 prev_upper_first
= prev
->first
;
569 prev_lower_first
= prev
->lower_first
;
570 prev_upper_last
= prev_upper_first
+ prev
->count
- 1;
571 prev_lower_last
= prev_lower_first
+ prev
->count
- 1;
573 /* Does the upper range intersect a previous extent? */
574 if ((prev_upper_first
<= upper_last
) &&
575 (prev_upper_last
>= upper_first
))
578 /* Does the lower range intersect a previous extent? */
579 if ((prev_lower_first
<= lower_last
) &&
580 (prev_lower_last
>= lower_first
))
586 static ssize_t
map_write(struct file
*file
, const char __user
*buf
,
587 size_t count
, loff_t
*ppos
,
589 struct uid_gid_map
*map
,
590 struct uid_gid_map
*parent_map
)
592 struct seq_file
*seq
= file
->private_data
;
593 struct user_namespace
*ns
= seq
->private;
594 struct uid_gid_map new_map
;
596 struct uid_gid_extent
*extent
= NULL
;
597 unsigned long page
= 0;
598 char *kbuf
, *pos
, *next_line
;
599 ssize_t ret
= -EINVAL
;
602 * The userns_state_mutex serializes all writes to any given map.
604 * Any map is only ever written once.
606 * An id map fits within 1 cache line on most architectures.
608 * On read nothing needs to be done unless you are on an
609 * architecture with a crazy cache coherency model like alpha.
611 * There is a one time data dependency between reading the
612 * count of the extents and the values of the extents. The
613 * desired behavior is to see the values of the extents that
614 * were written before the count of the extents.
616 * To achieve this smp_wmb() is used on guarantee the write
617 * order and smp_rmb() is guaranteed that we don't have crazy
618 * architectures returning stale data.
620 mutex_lock(&userns_state_mutex
);
623 /* Only allow one successful write to the map */
624 if (map
->nr_extents
!= 0)
628 * Adjusting namespace settings requires capabilities on the target.
630 if (cap_valid(cap_setid
) && !file_ns_capable(file
, ns
, CAP_SYS_ADMIN
))
635 page
= __get_free_page(GFP_TEMPORARY
);
636 kbuf
= (char *) page
;
640 /* Only allow <= page size writes at the beginning of the file */
642 if ((*ppos
!= 0) || (count
>= PAGE_SIZE
))
645 /* Slurp in the user data */
647 if (copy_from_user(kbuf
, buf
, count
))
651 /* Parse the user data */
654 new_map
.nr_extents
= 0;
655 for (;pos
; pos
= next_line
) {
656 extent
= &new_map
.extent
[new_map
.nr_extents
];
658 /* Find the end of line and ensure I don't look past it */
659 next_line
= strchr(pos
, '\n');
663 if (*next_line
== '\0')
667 pos
= skip_spaces(pos
);
668 extent
->first
= simple_strtoul(pos
, &pos
, 10);
672 pos
= skip_spaces(pos
);
673 extent
->lower_first
= simple_strtoul(pos
, &pos
, 10);
677 pos
= skip_spaces(pos
);
678 extent
->count
= simple_strtoul(pos
, &pos
, 10);
679 if (*pos
&& !isspace(*pos
))
682 /* Verify there is not trailing junk on the line */
683 pos
= skip_spaces(pos
);
687 /* Verify we have been given valid starting values */
688 if ((extent
->first
== (u32
) -1) ||
689 (extent
->lower_first
== (u32
) -1 ))
692 /* Verify count is not zero and does not cause the extent to wrap */
693 if ((extent
->first
+ extent
->count
) <= extent
->first
)
695 if ((extent
->lower_first
+ extent
->count
) <= extent
->lower_first
)
698 /* Do the ranges in extent overlap any previous extents? */
699 if (mappings_overlap(&new_map
, extent
))
702 new_map
.nr_extents
++;
704 /* Fail if the file contains too many extents */
705 if ((new_map
.nr_extents
== UID_GID_MAP_MAX_EXTENTS
) &&
709 /* Be very certaint the new map actually exists */
710 if (new_map
.nr_extents
== 0)
714 /* Validate the user is allowed to use user id's mapped to. */
715 if (!new_idmap_permitted(file
, ns
, cap_setid
, &new_map
))
718 /* Map the lower ids from the parent user namespace to the
719 * kernel global id space.
721 for (idx
= 0; idx
< new_map
.nr_extents
; idx
++) {
723 extent
= &new_map
.extent
[idx
];
725 lower_first
= map_id_range_down(parent_map
,
729 /* Fail if we can not map the specified extent to
730 * the kernel global id space.
732 if (lower_first
== (u32
) -1)
735 extent
->lower_first
= lower_first
;
738 /* Install the map */
739 memcpy(map
->extent
, new_map
.extent
,
740 new_map
.nr_extents
*sizeof(new_map
.extent
[0]));
742 map
->nr_extents
= new_map
.nr_extents
;
747 mutex_unlock(&userns_state_mutex
);
753 ssize_t
proc_uid_map_write(struct file
*file
, const char __user
*buf
, size_t size
, loff_t
*ppos
)
755 struct seq_file
*seq
= file
->private_data
;
756 struct user_namespace
*ns
= seq
->private;
757 struct user_namespace
*seq_ns
= seq_user_ns(seq
);
762 if ((seq_ns
!= ns
) && (seq_ns
!= ns
->parent
))
765 return map_write(file
, buf
, size
, ppos
, CAP_SETUID
,
766 &ns
->uid_map
, &ns
->parent
->uid_map
);
769 ssize_t
proc_gid_map_write(struct file
*file
, const char __user
*buf
, size_t size
, loff_t
*ppos
)
771 struct seq_file
*seq
= file
->private_data
;
772 struct user_namespace
*ns
= seq
->private;
773 struct user_namespace
*seq_ns
= seq_user_ns(seq
);
778 if ((seq_ns
!= ns
) && (seq_ns
!= ns
->parent
))
781 return map_write(file
, buf
, size
, ppos
, CAP_SETGID
,
782 &ns
->gid_map
, &ns
->parent
->gid_map
);
785 ssize_t
proc_projid_map_write(struct file
*file
, const char __user
*buf
, size_t size
, loff_t
*ppos
)
787 struct seq_file
*seq
= file
->private_data
;
788 struct user_namespace
*ns
= seq
->private;
789 struct user_namespace
*seq_ns
= seq_user_ns(seq
);
794 if ((seq_ns
!= ns
) && (seq_ns
!= ns
->parent
))
797 /* Anyone can set any valid project id no capability needed */
798 return map_write(file
, buf
, size
, ppos
, -1,
799 &ns
->projid_map
, &ns
->parent
->projid_map
);
802 static bool new_idmap_permitted(const struct file
*file
,
803 struct user_namespace
*ns
, int cap_setid
,
804 struct uid_gid_map
*new_map
)
806 const struct cred
*cred
= file
->f_cred
;
807 /* Don't allow mappings that would allow anything that wouldn't
808 * be allowed without the establishment of unprivileged mappings.
810 if ((new_map
->nr_extents
== 1) && (new_map
->extent
[0].count
== 1) &&
811 uid_eq(ns
->owner
, cred
->euid
)) {
812 u32 id
= new_map
->extent
[0].lower_first
;
813 if (cap_setid
== CAP_SETUID
) {
814 kuid_t uid
= make_kuid(ns
->parent
, id
);
815 if (uid_eq(uid
, cred
->euid
))
820 /* Allow anyone to set a mapping that doesn't require privilege */
821 if (!cap_valid(cap_setid
))
824 /* Allow the specified ids if we have the appropriate capability
825 * (CAP_SETUID or CAP_SETGID) over the parent user namespace.
826 * And the opener of the id file also had the approprpiate capability.
828 if (ns_capable(ns
->parent
, cap_setid
) &&
829 file_ns_capable(file
, ns
->parent
, cap_setid
))
835 int proc_setgroups_show(struct seq_file
*seq
, void *v
)
837 struct user_namespace
*ns
= seq
->private;
838 unsigned long userns_flags
= ACCESS_ONCE(ns
->flags
);
840 seq_printf(seq
, "%s\n",
841 (userns_flags
& USERNS_SETGROUPS_ALLOWED
) ?
846 ssize_t
proc_setgroups_write(struct file
*file
, const char __user
*buf
,
847 size_t count
, loff_t
*ppos
)
849 struct seq_file
*seq
= file
->private_data
;
850 struct user_namespace
*ns
= seq
->private;
852 bool setgroups_allowed
;
855 /* Only allow a very narrow range of strings to be written */
857 if ((*ppos
!= 0) || (count
>= sizeof(kbuf
)))
860 /* What was written? */
862 if (copy_from_user(kbuf
, buf
, count
))
867 /* What is being requested? */
869 if (strncmp(pos
, "allow", 5) == 0) {
871 setgroups_allowed
= true;
873 else if (strncmp(pos
, "deny", 4) == 0) {
875 setgroups_allowed
= false;
880 /* Verify there is not trailing junk on the line */
881 pos
= skip_spaces(pos
);
886 mutex_lock(&userns_state_mutex
);
887 if (setgroups_allowed
) {
888 /* Enabling setgroups after setgroups has been disabled
891 if (!(ns
->flags
& USERNS_SETGROUPS_ALLOWED
))
894 /* Permanently disabling setgroups after setgroups has
895 * been enabled by writing the gid_map is not allowed.
897 if (ns
->gid_map
.nr_extents
!= 0)
899 ns
->flags
&= ~USERNS_SETGROUPS_ALLOWED
;
901 mutex_unlock(&userns_state_mutex
);
903 /* Report a successful write */
909 mutex_unlock(&userns_state_mutex
);
913 bool userns_may_setgroups(const struct user_namespace
*ns
)
917 mutex_lock(&userns_state_mutex
);
918 /* It is not safe to use setgroups until a gid mapping in
919 * the user namespace has been established.
921 allowed
= ns
->gid_map
.nr_extents
!= 0;
922 /* Is setgroups allowed? */
923 allowed
= allowed
&& (ns
->flags
& USERNS_SETGROUPS_ALLOWED
);
924 mutex_unlock(&userns_state_mutex
);
929 static void *userns_get(struct task_struct
*task
)
931 struct user_namespace
*user_ns
;
934 user_ns
= get_user_ns(__task_cred(task
)->user_ns
);
940 static void userns_put(void *ns
)
945 static int userns_install(struct nsproxy
*nsproxy
, void *ns
)
947 struct user_namespace
*user_ns
= ns
;
950 /* Don't allow gaining capabilities by reentering
951 * the same user namespace.
953 if (user_ns
== current_user_ns())
956 /* Threaded processes may not enter a different user namespace */
957 if (atomic_read(¤t
->mm
->mm_users
) > 1)
960 if (current
->fs
->users
!= 1)
963 if (!ns_capable(user_ns
, CAP_SYS_ADMIN
))
966 cred
= prepare_creds();
970 put_user_ns(cred
->user_ns
);
971 set_cred_user_ns(cred
, get_user_ns(user_ns
));
973 return commit_creds(cred
);
976 static unsigned int userns_inum(void *ns
)
978 struct user_namespace
*user_ns
= ns
;
979 return user_ns
->proc_inum
;
982 const struct proc_ns_operations userns_operations
= {
984 .type
= CLONE_NEWUSER
,
987 .install
= userns_install
,
991 static __init
int user_namespaces_init(void)
993 user_ns_cachep
= KMEM_CACHE(user_namespace
, SLAB_PANIC
);
996 module_init(user_namespaces_init
);