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_ambient
= CAP_EMPTY_SET
;
43 cred
->cap_bset
= CAP_FULL_SET
;
45 key_put(cred
->request_key_auth
);
46 cred
->request_key_auth
= NULL
;
48 /* tgcred will be cleared in our caller bc CLONE_THREAD won't be set */
49 cred
->user_ns
= user_ns
;
53 * Create a new user namespace, deriving the creator from the user in the
54 * passed credentials, and replacing that user with the new root user for the
57 * This is called by copy_creds(), which will finish setting the target task's
60 int create_user_ns(struct cred
*new)
62 struct user_namespace
*ns
, *parent_ns
= new->user_ns
;
63 kuid_t owner
= new->euid
;
64 kgid_t group
= new->egid
;
67 if (parent_ns
->level
> 32)
71 * Verify that we can not violate the policy of which files
72 * may be accessed that is specified by the root directory,
73 * by verifing that the root directory is at the root of the
74 * mount namespace which allows all files to be accessed.
76 if (current_chrooted())
79 /* The creator needs a mapping in the parent user namespace
80 * or else we won't be able to reasonably tell userspace who
81 * created a user_namespace.
83 if (!kuid_has_mapping(parent_ns
, owner
) ||
84 !kgid_has_mapping(parent_ns
, group
))
87 ns
= kmem_cache_zalloc(user_ns_cachep
, GFP_KERNEL
);
91 ret
= proc_alloc_inum(&ns
->proc_inum
);
93 kmem_cache_free(user_ns_cachep
, ns
);
97 atomic_set(&ns
->count
, 1);
98 /* Leave the new->user_ns reference with the new user namespace. */
99 ns
->parent
= parent_ns
;
100 ns
->level
= parent_ns
->level
+ 1;
104 /* Inherit USERNS_SETGROUPS_ALLOWED from our parent */
105 mutex_lock(&userns_state_mutex
);
106 ns
->flags
= parent_ns
->flags
;
107 mutex_unlock(&userns_state_mutex
);
109 set_cred_user_ns(new, ns
);
111 update_mnt_policy(ns
);
116 int unshare_userns(unsigned long unshare_flags
, struct cred
**new_cred
)
121 if (!(unshare_flags
& CLONE_NEWUSER
))
124 cred
= prepare_creds();
126 err
= create_user_ns(cred
);
136 void free_user_ns(struct user_namespace
*ns
)
138 struct user_namespace
*parent
;
142 proc_free_inum(ns
->proc_inum
);
143 kmem_cache_free(user_ns_cachep
, ns
);
145 } while (atomic_dec_and_test(&parent
->count
));
147 EXPORT_SYMBOL(free_user_ns
);
149 static u32
map_id_range_down(struct uid_gid_map
*map
, u32 id
, u32 count
)
151 unsigned idx
, extents
;
152 u32 first
, last
, id2
;
154 id2
= id
+ count
- 1;
156 /* Find the matching extent */
157 extents
= map
->nr_extents
;
159 for (idx
= 0; idx
< extents
; idx
++) {
160 first
= map
->extent
[idx
].first
;
161 last
= first
+ map
->extent
[idx
].count
- 1;
162 if (id
>= first
&& id
<= last
&&
163 (id2
>= first
&& id2
<= last
))
166 /* Map the id or note failure */
168 id
= (id
- first
) + map
->extent
[idx
].lower_first
;
175 static u32
map_id_down(struct uid_gid_map
*map
, u32 id
)
177 unsigned idx
, extents
;
180 /* Find the matching extent */
181 extents
= map
->nr_extents
;
183 for (idx
= 0; idx
< extents
; idx
++) {
184 first
= map
->extent
[idx
].first
;
185 last
= first
+ map
->extent
[idx
].count
- 1;
186 if (id
>= first
&& id
<= last
)
189 /* Map the id or note failure */
191 id
= (id
- first
) + map
->extent
[idx
].lower_first
;
198 static u32
map_id_up(struct uid_gid_map
*map
, u32 id
)
200 unsigned idx
, extents
;
203 /* Find the matching extent */
204 extents
= map
->nr_extents
;
206 for (idx
= 0; idx
< extents
; idx
++) {
207 first
= map
->extent
[idx
].lower_first
;
208 last
= first
+ map
->extent
[idx
].count
- 1;
209 if (id
>= first
&& id
<= last
)
212 /* Map the id or note failure */
214 id
= (id
- first
) + map
->extent
[idx
].first
;
222 * make_kuid - Map a user-namespace uid pair into a kuid.
223 * @ns: User namespace that the uid is in
224 * @uid: User identifier
226 * Maps a user-namespace uid pair into a kernel internal kuid,
227 * and returns that kuid.
229 * When there is no mapping defined for the user-namespace uid
230 * pair INVALID_UID is returned. Callers are expected to test
231 * for and handle handle INVALID_UID being returned. INVALID_UID
232 * may be tested for using uid_valid().
234 kuid_t
make_kuid(struct user_namespace
*ns
, uid_t uid
)
236 /* Map the uid to a global kernel uid */
237 return KUIDT_INIT(map_id_down(&ns
->uid_map
, uid
));
239 EXPORT_SYMBOL(make_kuid
);
242 * from_kuid - Create a uid from a kuid user-namespace pair.
243 * @targ: The user namespace we want a uid in.
244 * @kuid: The kernel internal uid to start with.
246 * Map @kuid into the user-namespace specified by @targ and
247 * return the resulting uid.
249 * There is always a mapping into the initial user_namespace.
251 * If @kuid has no mapping in @targ (uid_t)-1 is returned.
253 uid_t
from_kuid(struct user_namespace
*targ
, kuid_t kuid
)
255 /* Map the uid from a global kernel uid */
256 return map_id_up(&targ
->uid_map
, __kuid_val(kuid
));
258 EXPORT_SYMBOL(from_kuid
);
261 * from_kuid_munged - Create a uid from a kuid user-namespace pair.
262 * @targ: The user namespace we want a uid in.
263 * @kuid: The kernel internal uid to start with.
265 * Map @kuid into the user-namespace specified by @targ and
266 * return the resulting uid.
268 * There is always a mapping into the initial user_namespace.
270 * Unlike from_kuid from_kuid_munged never fails and always
271 * returns a valid uid. This makes from_kuid_munged appropriate
272 * for use in syscalls like stat and getuid where failing the
273 * system call and failing to provide a valid uid are not an
276 * If @kuid has no mapping in @targ overflowuid is returned.
278 uid_t
from_kuid_munged(struct user_namespace
*targ
, kuid_t kuid
)
281 uid
= from_kuid(targ
, kuid
);
283 if (uid
== (uid_t
) -1)
287 EXPORT_SYMBOL(from_kuid_munged
);
290 * make_kgid - Map a user-namespace gid pair into a kgid.
291 * @ns: User namespace that the gid is in
292 * @uid: group identifier
294 * Maps a user-namespace gid pair into a kernel internal kgid,
295 * and returns that kgid.
297 * When there is no mapping defined for the user-namespace gid
298 * pair INVALID_GID is returned. Callers are expected to test
299 * for and handle INVALID_GID being returned. INVALID_GID may be
300 * tested for using gid_valid().
302 kgid_t
make_kgid(struct user_namespace
*ns
, gid_t gid
)
304 /* Map the gid to a global kernel gid */
305 return KGIDT_INIT(map_id_down(&ns
->gid_map
, gid
));
307 EXPORT_SYMBOL(make_kgid
);
310 * from_kgid - Create a gid from a kgid user-namespace pair.
311 * @targ: The user namespace we want a gid in.
312 * @kgid: The kernel internal gid to start with.
314 * Map @kgid into the user-namespace specified by @targ and
315 * return the resulting gid.
317 * There is always a mapping into the initial user_namespace.
319 * If @kgid has no mapping in @targ (gid_t)-1 is returned.
321 gid_t
from_kgid(struct user_namespace
*targ
, kgid_t kgid
)
323 /* Map the gid from a global kernel gid */
324 return map_id_up(&targ
->gid_map
, __kgid_val(kgid
));
326 EXPORT_SYMBOL(from_kgid
);
329 * from_kgid_munged - Create a gid from a kgid user-namespace pair.
330 * @targ: The user namespace we want a gid in.
331 * @kgid: The kernel internal gid to start with.
333 * Map @kgid into the user-namespace specified by @targ and
334 * return the resulting gid.
336 * There is always a mapping into the initial user_namespace.
338 * Unlike from_kgid from_kgid_munged never fails and always
339 * returns a valid gid. This makes from_kgid_munged appropriate
340 * for use in syscalls like stat and getgid where failing the
341 * system call and failing to provide a valid gid are not options.
343 * If @kgid has no mapping in @targ overflowgid is returned.
345 gid_t
from_kgid_munged(struct user_namespace
*targ
, kgid_t kgid
)
348 gid
= from_kgid(targ
, kgid
);
350 if (gid
== (gid_t
) -1)
354 EXPORT_SYMBOL(from_kgid_munged
);
357 * make_kprojid - Map a user-namespace projid pair into a kprojid.
358 * @ns: User namespace that the projid is in
359 * @projid: Project identifier
361 * Maps a user-namespace uid pair into a kernel internal kuid,
362 * and returns that kuid.
364 * When there is no mapping defined for the user-namespace projid
365 * pair INVALID_PROJID is returned. Callers are expected to test
366 * for and handle handle INVALID_PROJID being returned. INVALID_PROJID
367 * may be tested for using projid_valid().
369 kprojid_t
make_kprojid(struct user_namespace
*ns
, projid_t projid
)
371 /* Map the uid to a global kernel uid */
372 return KPROJIDT_INIT(map_id_down(&ns
->projid_map
, projid
));
374 EXPORT_SYMBOL(make_kprojid
);
377 * from_kprojid - Create a projid from a kprojid user-namespace pair.
378 * @targ: The user namespace we want a projid in.
379 * @kprojid: The kernel internal project identifier to start with.
381 * Map @kprojid into the user-namespace specified by @targ and
382 * return the resulting projid.
384 * There is always a mapping into the initial user_namespace.
386 * If @kprojid has no mapping in @targ (projid_t)-1 is returned.
388 projid_t
from_kprojid(struct user_namespace
*targ
, kprojid_t kprojid
)
390 /* Map the uid from a global kernel uid */
391 return map_id_up(&targ
->projid_map
, __kprojid_val(kprojid
));
393 EXPORT_SYMBOL(from_kprojid
);
396 * from_kprojid_munged - Create a projiid from a kprojid user-namespace pair.
397 * @targ: The user namespace we want a projid in.
398 * @kprojid: The kernel internal projid to start with.
400 * Map @kprojid into the user-namespace specified by @targ and
401 * return the resulting projid.
403 * There is always a mapping into the initial user_namespace.
405 * Unlike from_kprojid from_kprojid_munged never fails and always
406 * returns a valid projid. This makes from_kprojid_munged
407 * appropriate for use in syscalls like stat and where
408 * failing the system call and failing to provide a valid projid are
411 * If @kprojid has no mapping in @targ OVERFLOW_PROJID is returned.
413 projid_t
from_kprojid_munged(struct user_namespace
*targ
, kprojid_t kprojid
)
416 projid
= from_kprojid(targ
, kprojid
);
418 if (projid
== (projid_t
) -1)
419 projid
= OVERFLOW_PROJID
;
422 EXPORT_SYMBOL(from_kprojid_munged
);
425 static int uid_m_show(struct seq_file
*seq
, void *v
)
427 struct user_namespace
*ns
= seq
->private;
428 struct uid_gid_extent
*extent
= v
;
429 struct user_namespace
*lower_ns
;
432 lower_ns
= seq_user_ns(seq
);
433 if ((lower_ns
== ns
) && lower_ns
->parent
)
434 lower_ns
= lower_ns
->parent
;
436 lower
= from_kuid(lower_ns
, KUIDT_INIT(extent
->lower_first
));
438 seq_printf(seq
, "%10u %10u %10u\n",
446 static int gid_m_show(struct seq_file
*seq
, void *v
)
448 struct user_namespace
*ns
= seq
->private;
449 struct uid_gid_extent
*extent
= v
;
450 struct user_namespace
*lower_ns
;
453 lower_ns
= seq_user_ns(seq
);
454 if ((lower_ns
== ns
) && lower_ns
->parent
)
455 lower_ns
= lower_ns
->parent
;
457 lower
= from_kgid(lower_ns
, KGIDT_INIT(extent
->lower_first
));
459 seq_printf(seq
, "%10u %10u %10u\n",
467 static int projid_m_show(struct seq_file
*seq
, void *v
)
469 struct user_namespace
*ns
= seq
->private;
470 struct uid_gid_extent
*extent
= v
;
471 struct user_namespace
*lower_ns
;
474 lower_ns
= seq_user_ns(seq
);
475 if ((lower_ns
== ns
) && lower_ns
->parent
)
476 lower_ns
= lower_ns
->parent
;
478 lower
= from_kprojid(lower_ns
, KPROJIDT_INIT(extent
->lower_first
));
480 seq_printf(seq
, "%10u %10u %10u\n",
488 static void *m_start(struct seq_file
*seq
, loff_t
*ppos
, struct uid_gid_map
*map
)
490 struct uid_gid_extent
*extent
= NULL
;
493 if (pos
< map
->nr_extents
)
494 extent
= &map
->extent
[pos
];
499 static void *uid_m_start(struct seq_file
*seq
, loff_t
*ppos
)
501 struct user_namespace
*ns
= seq
->private;
503 return m_start(seq
, ppos
, &ns
->uid_map
);
506 static void *gid_m_start(struct seq_file
*seq
, loff_t
*ppos
)
508 struct user_namespace
*ns
= seq
->private;
510 return m_start(seq
, ppos
, &ns
->gid_map
);
513 static void *projid_m_start(struct seq_file
*seq
, loff_t
*ppos
)
515 struct user_namespace
*ns
= seq
->private;
517 return m_start(seq
, ppos
, &ns
->projid_map
);
520 static void *m_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
523 return seq
->op
->start(seq
, pos
);
526 static void m_stop(struct seq_file
*seq
, void *v
)
531 struct seq_operations proc_uid_seq_operations
= {
532 .start
= uid_m_start
,
538 struct seq_operations proc_gid_seq_operations
= {
539 .start
= gid_m_start
,
545 struct seq_operations proc_projid_seq_operations
= {
546 .start
= projid_m_start
,
549 .show
= projid_m_show
,
552 static bool mappings_overlap(struct uid_gid_map
*new_map
, struct uid_gid_extent
*extent
)
554 u32 upper_first
, lower_first
, upper_last
, lower_last
;
557 upper_first
= extent
->first
;
558 lower_first
= extent
->lower_first
;
559 upper_last
= upper_first
+ extent
->count
- 1;
560 lower_last
= lower_first
+ extent
->count
- 1;
562 for (idx
= 0; idx
< new_map
->nr_extents
; idx
++) {
563 u32 prev_upper_first
, prev_lower_first
;
564 u32 prev_upper_last
, prev_lower_last
;
565 struct uid_gid_extent
*prev
;
567 prev
= &new_map
->extent
[idx
];
569 prev_upper_first
= prev
->first
;
570 prev_lower_first
= prev
->lower_first
;
571 prev_upper_last
= prev_upper_first
+ prev
->count
- 1;
572 prev_lower_last
= prev_lower_first
+ prev
->count
- 1;
574 /* Does the upper range intersect a previous extent? */
575 if ((prev_upper_first
<= upper_last
) &&
576 (prev_upper_last
>= upper_first
))
579 /* Does the lower range intersect a previous extent? */
580 if ((prev_lower_first
<= lower_last
) &&
581 (prev_lower_last
>= lower_first
))
587 static ssize_t
map_write(struct file
*file
, const char __user
*buf
,
588 size_t count
, loff_t
*ppos
,
590 struct uid_gid_map
*map
,
591 struct uid_gid_map
*parent_map
)
593 struct seq_file
*seq
= file
->private_data
;
594 struct user_namespace
*ns
= seq
->private;
595 struct uid_gid_map new_map
;
597 struct uid_gid_extent
*extent
= NULL
;
598 unsigned long page
= 0;
599 char *kbuf
, *pos
, *next_line
;
600 ssize_t ret
= -EINVAL
;
603 * The userns_state_mutex serializes all writes to any given map.
605 * Any map is only ever written once.
607 * An id map fits within 1 cache line on most architectures.
609 * On read nothing needs to be done unless you are on an
610 * architecture with a crazy cache coherency model like alpha.
612 * There is a one time data dependency between reading the
613 * count of the extents and the values of the extents. The
614 * desired behavior is to see the values of the extents that
615 * were written before the count of the extents.
617 * To achieve this smp_wmb() is used on guarantee the write
618 * order and smp_rmb() is guaranteed that we don't have crazy
619 * architectures returning stale data.
621 mutex_lock(&userns_state_mutex
);
624 /* Only allow one successful write to the map */
625 if (map
->nr_extents
!= 0)
629 * Adjusting namespace settings requires capabilities on the target.
631 if (cap_valid(cap_setid
) && !file_ns_capable(file
, ns
, CAP_SYS_ADMIN
))
636 page
= __get_free_page(GFP_TEMPORARY
);
637 kbuf
= (char *) page
;
641 /* Only allow <= page size writes at the beginning of the file */
643 if ((*ppos
!= 0) || (count
>= PAGE_SIZE
))
646 /* Slurp in the user data */
648 if (copy_from_user(kbuf
, buf
, count
))
652 /* Parse the user data */
655 new_map
.nr_extents
= 0;
656 for (;pos
; pos
= next_line
) {
657 extent
= &new_map
.extent
[new_map
.nr_extents
];
659 /* Find the end of line and ensure I don't look past it */
660 next_line
= strchr(pos
, '\n');
664 if (*next_line
== '\0')
668 pos
= skip_spaces(pos
);
669 extent
->first
= simple_strtoul(pos
, &pos
, 10);
673 pos
= skip_spaces(pos
);
674 extent
->lower_first
= simple_strtoul(pos
, &pos
, 10);
678 pos
= skip_spaces(pos
);
679 extent
->count
= simple_strtoul(pos
, &pos
, 10);
680 if (*pos
&& !isspace(*pos
))
683 /* Verify there is not trailing junk on the line */
684 pos
= skip_spaces(pos
);
688 /* Verify we have been given valid starting values */
689 if ((extent
->first
== (u32
) -1) ||
690 (extent
->lower_first
== (u32
) -1 ))
693 /* Verify count is not zero and does not cause the extent to wrap */
694 if ((extent
->first
+ extent
->count
) <= extent
->first
)
696 if ((extent
->lower_first
+ extent
->count
) <= extent
->lower_first
)
699 /* Do the ranges in extent overlap any previous extents? */
700 if (mappings_overlap(&new_map
, extent
))
703 new_map
.nr_extents
++;
705 /* Fail if the file contains too many extents */
706 if ((new_map
.nr_extents
== UID_GID_MAP_MAX_EXTENTS
) &&
710 /* Be very certaint the new map actually exists */
711 if (new_map
.nr_extents
== 0)
715 /* Validate the user is allowed to use user id's mapped to. */
716 if (!new_idmap_permitted(file
, ns
, cap_setid
, &new_map
))
719 /* Map the lower ids from the parent user namespace to the
720 * kernel global id space.
722 for (idx
= 0; idx
< new_map
.nr_extents
; idx
++) {
724 extent
= &new_map
.extent
[idx
];
726 lower_first
= map_id_range_down(parent_map
,
730 /* Fail if we can not map the specified extent to
731 * the kernel global id space.
733 if (lower_first
== (u32
) -1)
736 extent
->lower_first
= lower_first
;
739 /* Install the map */
740 memcpy(map
->extent
, new_map
.extent
,
741 new_map
.nr_extents
*sizeof(new_map
.extent
[0]));
743 map
->nr_extents
= new_map
.nr_extents
;
748 mutex_unlock(&userns_state_mutex
);
754 ssize_t
proc_uid_map_write(struct file
*file
, const char __user
*buf
, size_t size
, loff_t
*ppos
)
756 struct seq_file
*seq
= file
->private_data
;
757 struct user_namespace
*ns
= seq
->private;
758 struct user_namespace
*seq_ns
= seq_user_ns(seq
);
763 if ((seq_ns
!= ns
) && (seq_ns
!= ns
->parent
))
766 return map_write(file
, buf
, size
, ppos
, CAP_SETUID
,
767 &ns
->uid_map
, &ns
->parent
->uid_map
);
770 ssize_t
proc_gid_map_write(struct file
*file
, const char __user
*buf
, size_t size
, loff_t
*ppos
)
772 struct seq_file
*seq
= file
->private_data
;
773 struct user_namespace
*ns
= seq
->private;
774 struct user_namespace
*seq_ns
= seq_user_ns(seq
);
779 if ((seq_ns
!= ns
) && (seq_ns
!= ns
->parent
))
782 return map_write(file
, buf
, size
, ppos
, CAP_SETGID
,
783 &ns
->gid_map
, &ns
->parent
->gid_map
);
786 ssize_t
proc_projid_map_write(struct file
*file
, const char __user
*buf
, size_t size
, loff_t
*ppos
)
788 struct seq_file
*seq
= file
->private_data
;
789 struct user_namespace
*ns
= seq
->private;
790 struct user_namespace
*seq_ns
= seq_user_ns(seq
);
795 if ((seq_ns
!= ns
) && (seq_ns
!= ns
->parent
))
798 /* Anyone can set any valid project id no capability needed */
799 return map_write(file
, buf
, size
, ppos
, -1,
800 &ns
->projid_map
, &ns
->parent
->projid_map
);
803 static bool new_idmap_permitted(const struct file
*file
,
804 struct user_namespace
*ns
, int cap_setid
,
805 struct uid_gid_map
*new_map
)
807 const struct cred
*cred
= file
->f_cred
;
808 /* Don't allow mappings that would allow anything that wouldn't
809 * be allowed without the establishment of unprivileged mappings.
811 if ((new_map
->nr_extents
== 1) && (new_map
->extent
[0].count
== 1) &&
812 uid_eq(ns
->owner
, cred
->euid
)) {
813 u32 id
= new_map
->extent
[0].lower_first
;
814 if (cap_setid
== CAP_SETUID
) {
815 kuid_t uid
= make_kuid(ns
->parent
, id
);
816 if (uid_eq(uid
, cred
->euid
))
818 } else if (cap_setid
== CAP_SETGID
) {
819 kgid_t gid
= make_kgid(ns
->parent
, id
);
820 if (!(ns
->flags
& USERNS_SETGROUPS_ALLOWED
) &&
821 gid_eq(gid
, cred
->egid
))
826 /* Allow anyone to set a mapping that doesn't require privilege */
827 if (!cap_valid(cap_setid
))
830 /* Allow the specified ids if we have the appropriate capability
831 * (CAP_SETUID or CAP_SETGID) over the parent user namespace.
832 * And the opener of the id file also had the approprpiate capability.
834 if (ns_capable(ns
->parent
, cap_setid
) &&
835 file_ns_capable(file
, ns
->parent
, cap_setid
))
841 int proc_setgroups_show(struct seq_file
*seq
, void *v
)
843 struct user_namespace
*ns
= seq
->private;
844 unsigned long userns_flags
= ACCESS_ONCE(ns
->flags
);
846 seq_printf(seq
, "%s\n",
847 (userns_flags
& USERNS_SETGROUPS_ALLOWED
) ?
852 ssize_t
proc_setgroups_write(struct file
*file
, const char __user
*buf
,
853 size_t count
, loff_t
*ppos
)
855 struct seq_file
*seq
= file
->private_data
;
856 struct user_namespace
*ns
= seq
->private;
858 bool setgroups_allowed
;
861 /* Only allow a very narrow range of strings to be written */
863 if ((*ppos
!= 0) || (count
>= sizeof(kbuf
)))
866 /* What was written? */
868 if (copy_from_user(kbuf
, buf
, count
))
873 /* What is being requested? */
875 if (strncmp(pos
, "allow", 5) == 0) {
877 setgroups_allowed
= true;
879 else if (strncmp(pos
, "deny", 4) == 0) {
881 setgroups_allowed
= false;
886 /* Verify there is not trailing junk on the line */
887 pos
= skip_spaces(pos
);
892 mutex_lock(&userns_state_mutex
);
893 if (setgroups_allowed
) {
894 /* Enabling setgroups after setgroups has been disabled
897 if (!(ns
->flags
& USERNS_SETGROUPS_ALLOWED
))
900 /* Permanently disabling setgroups after setgroups has
901 * been enabled by writing the gid_map is not allowed.
903 if (ns
->gid_map
.nr_extents
!= 0)
905 ns
->flags
&= ~USERNS_SETGROUPS_ALLOWED
;
907 mutex_unlock(&userns_state_mutex
);
909 /* Report a successful write */
915 mutex_unlock(&userns_state_mutex
);
919 bool userns_may_setgroups(const struct user_namespace
*ns
)
923 mutex_lock(&userns_state_mutex
);
924 /* It is not safe to use setgroups until a gid mapping in
925 * the user namespace has been established.
927 allowed
= ns
->gid_map
.nr_extents
!= 0;
928 /* Is setgroups allowed? */
929 allowed
= allowed
&& (ns
->flags
& USERNS_SETGROUPS_ALLOWED
);
930 mutex_unlock(&userns_state_mutex
);
935 static void *userns_get(struct task_struct
*task
)
937 struct user_namespace
*user_ns
;
940 user_ns
= get_user_ns(__task_cred(task
)->user_ns
);
946 static void userns_put(void *ns
)
951 static int userns_install(struct nsproxy
*nsproxy
, void *ns
)
953 struct user_namespace
*user_ns
= ns
;
956 /* Don't allow gaining capabilities by reentering
957 * the same user namespace.
959 if (user_ns
== current_user_ns())
962 /* Threaded processes may not enter a different user namespace */
963 if (atomic_read(¤t
->mm
->mm_users
) > 1)
966 if (current
->fs
->users
!= 1)
969 if (!ns_capable(user_ns
, CAP_SYS_ADMIN
))
972 cred
= prepare_creds();
976 put_user_ns(cred
->user_ns
);
977 set_cred_user_ns(cred
, get_user_ns(user_ns
));
979 return commit_creds(cred
);
982 static unsigned int userns_inum(void *ns
)
984 struct user_namespace
*user_ns
= ns
;
985 return user_ns
->proc_inum
;
988 const struct proc_ns_operations userns_operations
= {
990 .type
= CLONE_NEWUSER
,
993 .install
= userns_install
,
997 static __init
int user_namespaces_init(void)
999 user_ns_cachep
= KMEM_CACHE(user_namespace
, SLAB_PANIC
);
1002 module_init(user_namespaces_init
);