spi: sh-msiof: Add r8a774a1 support
[linux-2.6-block.git] / kernel / user_namespace.c
CommitLineData
acce292c
CLG
1/*
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
5 * License.
6 */
7
9984de1a 8#include <linux/export.h>
acce292c 9#include <linux/nsproxy.h>
1aeb272c 10#include <linux/slab.h>
3f07c014 11#include <linux/sched/signal.h>
acce292c 12#include <linux/user_namespace.h>
0bb80f24 13#include <linux/proc_ns.h>
5c1469de 14#include <linux/highuid.h>
18b6e041 15#include <linux/cred.h>
973c5914 16#include <linux/securebits.h>
22d917d8
EB
17#include <linux/keyctl.h>
18#include <linux/key-type.h>
19#include <keys/user-type.h>
20#include <linux/seq_file.h>
21#include <linux/fs.h>
22#include <linux/uaccess.h>
23#include <linux/ctype.h>
f76d207a 24#include <linux/projid.h>
e66eded8 25#include <linux/fs_struct.h>
6397fac4
CB
26#include <linux/bsearch.h>
27#include <linux/sort.h>
acce292c 28
6164281a 29static struct kmem_cache *user_ns_cachep __read_mostly;
f0d62aec 30static DEFINE_MUTEX(userns_state_mutex);
6164281a 31
6708075f
EB
32static bool new_idmap_permitted(const struct file *file,
33 struct user_namespace *ns, int cap_setid,
22d917d8 34 struct uid_gid_map *map);
b032132c 35static void free_user_ns(struct work_struct *work);
22d917d8 36
25f9c081
EB
37static struct ucounts *inc_user_namespaces(struct user_namespace *ns, kuid_t uid)
38{
39 return inc_ucount(ns, uid, UCOUNT_USER_NAMESPACES);
40}
41
42static void dec_user_namespaces(struct ucounts *ucounts)
43{
44 return dec_ucount(ucounts, UCOUNT_USER_NAMESPACES);
45}
46
cde1975b
EB
47static void set_cred_user_ns(struct cred *cred, struct user_namespace *user_ns)
48{
49 /* Start with the same capabilities as init but useless for doing
50 * anything as the capabilities are bound to the new user namespace.
51 */
52 cred->securebits = SECUREBITS_DEFAULT;
53 cred->cap_inheritable = CAP_EMPTY_SET;
54 cred->cap_permitted = CAP_FULL_SET;
55 cred->cap_effective = CAP_FULL_SET;
58319057 56 cred->cap_ambient = CAP_EMPTY_SET;
cde1975b
EB
57 cred->cap_bset = CAP_FULL_SET;
58#ifdef CONFIG_KEYS
59 key_put(cred->request_key_auth);
60 cred->request_key_auth = NULL;
61#endif
62 /* tgcred will be cleared in our caller bc CLONE_THREAD won't be set */
63 cred->user_ns = user_ns;
64}
65
77ec739d 66/*
18b6e041
SH
67 * Create a new user namespace, deriving the creator from the user in the
68 * passed credentials, and replacing that user with the new root user for the
69 * new namespace.
70 *
71 * This is called by copy_creds(), which will finish setting the target task's
72 * credentials.
77ec739d 73 */
18b6e041 74int create_user_ns(struct cred *new)
77ec739d 75{
0093ccb6 76 struct user_namespace *ns, *parent_ns = new->user_ns;
078de5f7
EB
77 kuid_t owner = new->euid;
78 kgid_t group = new->egid;
f6b2db1a 79 struct ucounts *ucounts;
25f9c081 80 int ret, i;
783291e6 81
df75e774 82 ret = -ENOSPC;
8742f229 83 if (parent_ns->level > 32)
b376c3e1
EB
84 goto fail;
85
f6b2db1a
EB
86 ucounts = inc_user_namespaces(parent_ns, owner);
87 if (!ucounts)
b376c3e1 88 goto fail;
8742f229 89
3151527e
EB
90 /*
91 * Verify that we can not violate the policy of which files
92 * may be accessed that is specified by the root directory,
93 * by verifing that the root directory is at the root of the
94 * mount namespace which allows all files to be accessed.
95 */
b376c3e1 96 ret = -EPERM;
3151527e 97 if (current_chrooted())
b376c3e1 98 goto fail_dec;
3151527e 99
783291e6
EB
100 /* The creator needs a mapping in the parent user namespace
101 * or else we won't be able to reasonably tell userspace who
102 * created a user_namespace.
103 */
b376c3e1 104 ret = -EPERM;
783291e6
EB
105 if (!kuid_has_mapping(parent_ns, owner) ||
106 !kgid_has_mapping(parent_ns, group))
b376c3e1 107 goto fail_dec;
77ec739d 108
b376c3e1 109 ret = -ENOMEM;
22d917d8 110 ns = kmem_cache_zalloc(user_ns_cachep, GFP_KERNEL);
77ec739d 111 if (!ns)
b376c3e1 112 goto fail_dec;
77ec739d 113
6344c433 114 ret = ns_alloc_inum(&ns->ns);
b376c3e1
EB
115 if (ret)
116 goto fail_free;
33c42940 117 ns->ns.ops = &userns_operations;
98f842e6 118
c61a2810 119 atomic_set(&ns->count, 1);
cde1975b 120 /* Leave the new->user_ns reference with the new user namespace. */
aeb3ae9d 121 ns->parent = parent_ns;
8742f229 122 ns->level = parent_ns->level + 1;
783291e6
EB
123 ns->owner = owner;
124 ns->group = group;
b032132c 125 INIT_WORK(&ns->work, free_user_ns);
25f9c081
EB
126 for (i = 0; i < UCOUNT_COUNTS; i++) {
127 ns->ucount_max[i] = INT_MAX;
128 }
f6b2db1a 129 ns->ucounts = ucounts;
22d917d8 130
9cc46516
EB
131 /* Inherit USERNS_SETGROUPS_ALLOWED from our parent */
132 mutex_lock(&userns_state_mutex);
133 ns->flags = parent_ns->flags;
134 mutex_unlock(&userns_state_mutex);
135
f36f8c75
DH
136#ifdef CONFIG_PERSISTENT_KEYRINGS
137 init_rwsem(&ns->persistent_keyring_register_sem);
138#endif
dbec2846
EB
139 ret = -ENOMEM;
140 if (!setup_userns_sysctls(ns))
141 goto fail_keyring;
142
143 set_cred_user_ns(new, ns);
18b6e041 144 return 0;
dbec2846
EB
145fail_keyring:
146#ifdef CONFIG_PERSISTENT_KEYRINGS
147 key_put(ns->persistent_keyring_register);
148#endif
149 ns_free_inum(&ns->ns);
b376c3e1 150fail_free:
dbec2846 151 kmem_cache_free(user_ns_cachep, ns);
b376c3e1 152fail_dec:
f6b2db1a 153 dec_user_namespaces(ucounts);
b376c3e1 154fail:
dbec2846 155 return ret;
acce292c
CLG
156}
157
b2e0d987
EB
158int unshare_userns(unsigned long unshare_flags, struct cred **new_cred)
159{
160 struct cred *cred;
6160968c 161 int err = -ENOMEM;
b2e0d987
EB
162
163 if (!(unshare_flags & CLONE_NEWUSER))
164 return 0;
165
166 cred = prepare_creds();
6160968c
ON
167 if (cred) {
168 err = create_user_ns(cred);
169 if (err)
170 put_cred(cred);
171 else
172 *new_cred = cred;
173 }
b2e0d987 174
6160968c 175 return err;
b2e0d987
EB
176}
177
b032132c 178static void free_user_ns(struct work_struct *work)
acce292c 179{
b032132c
EB
180 struct user_namespace *parent, *ns =
181 container_of(work, struct user_namespace, work);
783291e6 182
c61a2810 183 do {
f6b2db1a 184 struct ucounts *ucounts = ns->ucounts;
c61a2810 185 parent = ns->parent;
6397fac4
CB
186 if (ns->gid_map.nr_extents > UID_GID_MAP_MAX_BASE_EXTENTS) {
187 kfree(ns->gid_map.forward);
188 kfree(ns->gid_map.reverse);
189 }
190 if (ns->uid_map.nr_extents > UID_GID_MAP_MAX_BASE_EXTENTS) {
191 kfree(ns->uid_map.forward);
192 kfree(ns->uid_map.reverse);
193 }
194 if (ns->projid_map.nr_extents > UID_GID_MAP_MAX_BASE_EXTENTS) {
195 kfree(ns->projid_map.forward);
196 kfree(ns->projid_map.reverse);
197 }
dbec2846 198 retire_userns_sysctls(ns);
f36f8c75
DH
199#ifdef CONFIG_PERSISTENT_KEYRINGS
200 key_put(ns->persistent_keyring_register);
201#endif
6344c433 202 ns_free_inum(&ns->ns);
c61a2810 203 kmem_cache_free(user_ns_cachep, ns);
f6b2db1a 204 dec_user_namespaces(ucounts);
c61a2810
EB
205 ns = parent;
206 } while (atomic_dec_and_test(&parent->count));
acce292c 207}
b032132c
EB
208
209void __put_user_ns(struct user_namespace *ns)
210{
211 schedule_work(&ns->work);
212}
213EXPORT_SYMBOL(__put_user_ns);
5c1469de 214
6397fac4
CB
215/**
216 * idmap_key struct holds the information necessary to find an idmapping in a
217 * sorted idmap array. It is passed to cmp_map_id() as first argument.
218 */
219struct idmap_key {
220 bool map_up; /* true -> id from kid; false -> kid from id */
221 u32 id; /* id to find */
222 u32 count; /* == 0 unless used with map_id_range_down() */
223};
224
225/**
226 * cmp_map_id - Function to be passed to bsearch() to find the requested
227 * idmapping. Expects struct idmap_key to be passed via @k.
228 */
229static int cmp_map_id(const void *k, const void *e)
230{
231 u32 first, last, id2;
232 const struct idmap_key *key = k;
233 const struct uid_gid_extent *el = e;
234
11a8b927 235 id2 = key->id + key->count - 1;
6397fac4
CB
236
237 /* handle map_id_{down,up}() */
238 if (key->map_up)
239 first = el->lower_first;
240 else
241 first = el->first;
242
243 last = first + el->count - 1;
244
245 if (key->id >= first && key->id <= last &&
246 (id2 >= first && id2 <= last))
247 return 0;
248
249 if (key->id < first || id2 < first)
250 return -1;
251
252 return 1;
253}
254
255/**
256 * map_id_range_down_max - Find idmap via binary search in ordered idmap array.
257 * Can only be called if number of mappings exceeds UID_GID_MAP_MAX_BASE_EXTENTS.
258 */
3edf652f
EB
259static struct uid_gid_extent *
260map_id_range_down_max(unsigned extents, struct uid_gid_map *map, u32 id, u32 count)
5c1469de 261{
6397fac4
CB
262 struct idmap_key key;
263
264 key.map_up = false;
265 key.count = count;
266 key.id = id;
267
3edf652f
EB
268 return bsearch(&key, map->forward, extents,
269 sizeof(struct uid_gid_extent), cmp_map_id);
6397fac4
CB
270}
271
272/**
273 * map_id_range_down_base - Find idmap via binary search in static extent array.
274 * Can only be called if number of mappings is equal or less than
275 * UID_GID_MAP_MAX_BASE_EXTENTS.
276 */
3edf652f
EB
277static struct uid_gid_extent *
278map_id_range_down_base(unsigned extents, struct uid_gid_map *map, u32 id, u32 count)
5c1469de 279{
3edf652f 280 unsigned idx;
22d917d8 281 u32 first, last, id2;
5c1469de 282
22d917d8 283 id2 = id + count - 1;
5c1469de 284
22d917d8 285 /* Find the matching extent */
22d917d8
EB
286 for (idx = 0; idx < extents; idx++) {
287 first = map->extent[idx].first;
288 last = first + map->extent[idx].count - 1;
289 if (id >= first && id <= last &&
290 (id2 >= first && id2 <= last))
3edf652f 291 return &map->extent[idx];
22d917d8 292 }
3edf652f 293 return NULL;
22d917d8
EB
294}
295
6397fac4
CB
296static u32 map_id_range_down(struct uid_gid_map *map, u32 id, u32 count)
297{
3edf652f
EB
298 struct uid_gid_extent *extent;
299 unsigned extents = map->nr_extents;
6397fac4
CB
300 smp_rmb();
301
302 if (extents <= UID_GID_MAP_MAX_BASE_EXTENTS)
3edf652f
EB
303 extent = map_id_range_down_base(extents, map, id, count);
304 else
305 extent = map_id_range_down_max(extents, map, id, count);
306
22d917d8 307 /* Map the id or note failure */
3edf652f
EB
308 if (extent)
309 id = (id - extent->first) + extent->lower_first;
22d917d8
EB
310 else
311 id = (u32) -1;
312
313 return id;
314}
315
316static u32 map_id_down(struct uid_gid_map *map, u32 id)
317{
ece66133 318 return map_id_range_down(map, id, 1);
6397fac4
CB
319}
320
321/**
322 * map_id_up_base - Find idmap via binary search in static extent array.
323 * Can only be called if number of mappings is equal or less than
324 * UID_GID_MAP_MAX_BASE_EXTENTS.
325 */
3edf652f
EB
326static struct uid_gid_extent *
327map_id_up_base(unsigned extents, struct uid_gid_map *map, u32 id)
22d917d8 328{
3edf652f 329 unsigned idx;
22d917d8
EB
330 u32 first, last;
331
332 /* Find the matching extent */
22d917d8 333 for (idx = 0; idx < extents; idx++) {
22d917d8 334 first = map->extent[idx].lower_first;
22d917d8
EB
335 last = first + map->extent[idx].count - 1;
336 if (id >= first && id <= last)
3edf652f 337 return &map->extent[idx];
22d917d8 338 }
3edf652f 339 return NULL;
22d917d8 340}
22d917d8 341
6397fac4
CB
342/**
343 * map_id_up_max - Find idmap via binary search in ordered idmap array.
344 * Can only be called if number of mappings exceeds UID_GID_MAP_MAX_BASE_EXTENTS.
345 */
3edf652f
EB
346static struct uid_gid_extent *
347map_id_up_max(unsigned extents, struct uid_gid_map *map, u32 id)
6397fac4 348{
6397fac4
CB
349 struct idmap_key key;
350
351 key.map_up = true;
11a8b927 352 key.count = 1;
6397fac4
CB
353 key.id = id;
354
3edf652f
EB
355 return bsearch(&key, map->reverse, extents,
356 sizeof(struct uid_gid_extent), cmp_map_id);
22d917d8
EB
357}
358
359static u32 map_id_up(struct uid_gid_map *map, u32 id)
360{
3edf652f
EB
361 struct uid_gid_extent *extent;
362 unsigned extents = map->nr_extents;
e79323bd 363 smp_rmb();
6397fac4 364
3edf652f
EB
365 if (extents <= UID_GID_MAP_MAX_BASE_EXTENTS)
366 extent = map_id_up_base(extents, map, id);
367 else
368 extent = map_id_up_max(extents, map, id);
369
22d917d8 370 /* Map the id or note failure */
6397fac4
CB
371 if (extent)
372 id = (id - extent->lower_first) + extent->first;
22d917d8
EB
373 else
374 id = (u32) -1;
375
376 return id;
377}
378
379/**
380 * make_kuid - Map a user-namespace uid pair into a kuid.
381 * @ns: User namespace that the uid is in
382 * @uid: User identifier
383 *
384 * Maps a user-namespace uid pair into a kernel internal kuid,
385 * and returns that kuid.
386 *
387 * When there is no mapping defined for the user-namespace uid
388 * pair INVALID_UID is returned. Callers are expected to test
b080e047 389 * for and handle INVALID_UID being returned. INVALID_UID
22d917d8
EB
390 * may be tested for using uid_valid().
391 */
392kuid_t make_kuid(struct user_namespace *ns, uid_t uid)
393{
394 /* Map the uid to a global kernel uid */
395 return KUIDT_INIT(map_id_down(&ns->uid_map, uid));
396}
397EXPORT_SYMBOL(make_kuid);
398
399/**
400 * from_kuid - Create a uid from a kuid user-namespace pair.
401 * @targ: The user namespace we want a uid in.
402 * @kuid: The kernel internal uid to start with.
403 *
404 * Map @kuid into the user-namespace specified by @targ and
405 * return the resulting uid.
406 *
407 * There is always a mapping into the initial user_namespace.
408 *
409 * If @kuid has no mapping in @targ (uid_t)-1 is returned.
410 */
411uid_t from_kuid(struct user_namespace *targ, kuid_t kuid)
412{
413 /* Map the uid from a global kernel uid */
414 return map_id_up(&targ->uid_map, __kuid_val(kuid));
415}
416EXPORT_SYMBOL(from_kuid);
417
418/**
419 * from_kuid_munged - Create a uid from a kuid user-namespace pair.
420 * @targ: The user namespace we want a uid in.
421 * @kuid: The kernel internal uid to start with.
422 *
423 * Map @kuid into the user-namespace specified by @targ and
424 * return the resulting uid.
425 *
426 * There is always a mapping into the initial user_namespace.
427 *
428 * Unlike from_kuid from_kuid_munged never fails and always
429 * returns a valid uid. This makes from_kuid_munged appropriate
430 * for use in syscalls like stat and getuid where failing the
431 * system call and failing to provide a valid uid are not an
432 * options.
433 *
434 * If @kuid has no mapping in @targ overflowuid is returned.
435 */
436uid_t from_kuid_munged(struct user_namespace *targ, kuid_t kuid)
437{
438 uid_t uid;
439 uid = from_kuid(targ, kuid);
440
441 if (uid == (uid_t) -1)
442 uid = overflowuid;
443 return uid;
444}
445EXPORT_SYMBOL(from_kuid_munged);
446
447/**
448 * make_kgid - Map a user-namespace gid pair into a kgid.
449 * @ns: User namespace that the gid is in
68a9a435 450 * @gid: group identifier
22d917d8
EB
451 *
452 * Maps a user-namespace gid pair into a kernel internal kgid,
453 * and returns that kgid.
454 *
455 * When there is no mapping defined for the user-namespace gid
456 * pair INVALID_GID is returned. Callers are expected to test
457 * for and handle INVALID_GID being returned. INVALID_GID may be
458 * tested for using gid_valid().
459 */
460kgid_t make_kgid(struct user_namespace *ns, gid_t gid)
461{
462 /* Map the gid to a global kernel gid */
463 return KGIDT_INIT(map_id_down(&ns->gid_map, gid));
464}
465EXPORT_SYMBOL(make_kgid);
466
467/**
468 * from_kgid - Create a gid from a kgid user-namespace pair.
469 * @targ: The user namespace we want a gid in.
470 * @kgid: The kernel internal gid to start with.
471 *
472 * Map @kgid into the user-namespace specified by @targ and
473 * return the resulting gid.
474 *
475 * There is always a mapping into the initial user_namespace.
476 *
477 * If @kgid has no mapping in @targ (gid_t)-1 is returned.
478 */
479gid_t from_kgid(struct user_namespace *targ, kgid_t kgid)
480{
481 /* Map the gid from a global kernel gid */
482 return map_id_up(&targ->gid_map, __kgid_val(kgid));
483}
484EXPORT_SYMBOL(from_kgid);
485
486/**
487 * from_kgid_munged - Create a gid from a kgid user-namespace pair.
488 * @targ: The user namespace we want a gid in.
489 * @kgid: The kernel internal gid to start with.
490 *
491 * Map @kgid into the user-namespace specified by @targ and
492 * return the resulting gid.
493 *
494 * There is always a mapping into the initial user_namespace.
495 *
496 * Unlike from_kgid from_kgid_munged never fails and always
497 * returns a valid gid. This makes from_kgid_munged appropriate
498 * for use in syscalls like stat and getgid where failing the
499 * system call and failing to provide a valid gid are not options.
500 *
501 * If @kgid has no mapping in @targ overflowgid is returned.
502 */
503gid_t from_kgid_munged(struct user_namespace *targ, kgid_t kgid)
504{
505 gid_t gid;
506 gid = from_kgid(targ, kgid);
507
508 if (gid == (gid_t) -1)
509 gid = overflowgid;
510 return gid;
511}
512EXPORT_SYMBOL(from_kgid_munged);
513
f76d207a
EB
514/**
515 * make_kprojid - Map a user-namespace projid pair into a kprojid.
516 * @ns: User namespace that the projid is in
517 * @projid: Project identifier
518 *
519 * Maps a user-namespace uid pair into a kernel internal kuid,
520 * and returns that kuid.
521 *
522 * When there is no mapping defined for the user-namespace projid
523 * pair INVALID_PROJID is returned. Callers are expected to test
524 * for and handle handle INVALID_PROJID being returned. INVALID_PROJID
525 * may be tested for using projid_valid().
526 */
527kprojid_t make_kprojid(struct user_namespace *ns, projid_t projid)
528{
529 /* Map the uid to a global kernel uid */
530 return KPROJIDT_INIT(map_id_down(&ns->projid_map, projid));
531}
532EXPORT_SYMBOL(make_kprojid);
533
534/**
535 * from_kprojid - Create a projid from a kprojid user-namespace pair.
536 * @targ: The user namespace we want a projid in.
537 * @kprojid: The kernel internal project identifier to start with.
538 *
539 * Map @kprojid into the user-namespace specified by @targ and
540 * return the resulting projid.
541 *
542 * There is always a mapping into the initial user_namespace.
543 *
544 * If @kprojid has no mapping in @targ (projid_t)-1 is returned.
545 */
546projid_t from_kprojid(struct user_namespace *targ, kprojid_t kprojid)
547{
548 /* Map the uid from a global kernel uid */
549 return map_id_up(&targ->projid_map, __kprojid_val(kprojid));
550}
551EXPORT_SYMBOL(from_kprojid);
552
553/**
554 * from_kprojid_munged - Create a projiid from a kprojid user-namespace pair.
555 * @targ: The user namespace we want a projid in.
556 * @kprojid: The kernel internal projid to start with.
557 *
558 * Map @kprojid into the user-namespace specified by @targ and
559 * return the resulting projid.
560 *
561 * There is always a mapping into the initial user_namespace.
562 *
563 * Unlike from_kprojid from_kprojid_munged never fails and always
564 * returns a valid projid. This makes from_kprojid_munged
565 * appropriate for use in syscalls like stat and where
566 * failing the system call and failing to provide a valid projid are
567 * not an options.
568 *
569 * If @kprojid has no mapping in @targ OVERFLOW_PROJID is returned.
570 */
571projid_t from_kprojid_munged(struct user_namespace *targ, kprojid_t kprojid)
572{
573 projid_t projid;
574 projid = from_kprojid(targ, kprojid);
575
576 if (projid == (projid_t) -1)
577 projid = OVERFLOW_PROJID;
578 return projid;
579}
580EXPORT_SYMBOL(from_kprojid_munged);
581
582
22d917d8
EB
583static int uid_m_show(struct seq_file *seq, void *v)
584{
585 struct user_namespace *ns = seq->private;
586 struct uid_gid_extent *extent = v;
587 struct user_namespace *lower_ns;
588 uid_t lower;
5c1469de 589
c450f371 590 lower_ns = seq_user_ns(seq);
22d917d8
EB
591 if ((lower_ns == ns) && lower_ns->parent)
592 lower_ns = lower_ns->parent;
593
594 lower = from_kuid(lower_ns, KUIDT_INIT(extent->lower_first));
595
596 seq_printf(seq, "%10u %10u %10u\n",
597 extent->first,
598 lower,
599 extent->count);
600
601 return 0;
5c1469de
EB
602}
603
22d917d8 604static int gid_m_show(struct seq_file *seq, void *v)
5c1469de 605{
22d917d8
EB
606 struct user_namespace *ns = seq->private;
607 struct uid_gid_extent *extent = v;
608 struct user_namespace *lower_ns;
609 gid_t lower;
5c1469de 610
c450f371 611 lower_ns = seq_user_ns(seq);
22d917d8
EB
612 if ((lower_ns == ns) && lower_ns->parent)
613 lower_ns = lower_ns->parent;
5c1469de 614
22d917d8
EB
615 lower = from_kgid(lower_ns, KGIDT_INIT(extent->lower_first));
616
617 seq_printf(seq, "%10u %10u %10u\n",
618 extent->first,
619 lower,
620 extent->count);
621
622 return 0;
623}
624
f76d207a
EB
625static int projid_m_show(struct seq_file *seq, void *v)
626{
627 struct user_namespace *ns = seq->private;
628 struct uid_gid_extent *extent = v;
629 struct user_namespace *lower_ns;
630 projid_t lower;
631
632 lower_ns = seq_user_ns(seq);
633 if ((lower_ns == ns) && lower_ns->parent)
634 lower_ns = lower_ns->parent;
635
636 lower = from_kprojid(lower_ns, KPROJIDT_INIT(extent->lower_first));
637
638 seq_printf(seq, "%10u %10u %10u\n",
639 extent->first,
640 lower,
641 extent->count);
642
643 return 0;
644}
645
68a9a435
FF
646static void *m_start(struct seq_file *seq, loff_t *ppos,
647 struct uid_gid_map *map)
22d917d8 648{
22d917d8 649 loff_t pos = *ppos;
d5e7b3c5
EB
650 unsigned extents = map->nr_extents;
651 smp_rmb();
22d917d8 652
d5e7b3c5 653 if (pos >= extents)
6397fac4 654 return NULL;
22d917d8 655
d5e7b3c5 656 if (extents <= UID_GID_MAP_MAX_BASE_EXTENTS)
6397fac4 657 return &map->extent[pos];
22d917d8 658
6397fac4 659 return &map->forward[pos];
22d917d8
EB
660}
661
662static void *uid_m_start(struct seq_file *seq, loff_t *ppos)
663{
664 struct user_namespace *ns = seq->private;
665
666 return m_start(seq, ppos, &ns->uid_map);
667}
668
669static void *gid_m_start(struct seq_file *seq, loff_t *ppos)
670{
671 struct user_namespace *ns = seq->private;
672
673 return m_start(seq, ppos, &ns->gid_map);
674}
675
f76d207a
EB
676static void *projid_m_start(struct seq_file *seq, loff_t *ppos)
677{
678 struct user_namespace *ns = seq->private;
679
680 return m_start(seq, ppos, &ns->projid_map);
681}
682
22d917d8
EB
683static void *m_next(struct seq_file *seq, void *v, loff_t *pos)
684{
685 (*pos)++;
686 return seq->op->start(seq, pos);
687}
688
689static void m_stop(struct seq_file *seq, void *v)
690{
691 return;
692}
693
ccf94f1b 694const struct seq_operations proc_uid_seq_operations = {
22d917d8
EB
695 .start = uid_m_start,
696 .stop = m_stop,
697 .next = m_next,
698 .show = uid_m_show,
699};
700
ccf94f1b 701const struct seq_operations proc_gid_seq_operations = {
22d917d8
EB
702 .start = gid_m_start,
703 .stop = m_stop,
704 .next = m_next,
705 .show = gid_m_show,
706};
707
ccf94f1b 708const struct seq_operations proc_projid_seq_operations = {
f76d207a
EB
709 .start = projid_m_start,
710 .stop = m_stop,
711 .next = m_next,
712 .show = projid_m_show,
713};
714
68a9a435
FF
715static bool mappings_overlap(struct uid_gid_map *new_map,
716 struct uid_gid_extent *extent)
0bd14b4f
EB
717{
718 u32 upper_first, lower_first, upper_last, lower_last;
719 unsigned idx;
720
721 upper_first = extent->first;
722 lower_first = extent->lower_first;
723 upper_last = upper_first + extent->count - 1;
724 lower_last = lower_first + extent->count - 1;
725
726 for (idx = 0; idx < new_map->nr_extents; idx++) {
727 u32 prev_upper_first, prev_lower_first;
728 u32 prev_upper_last, prev_lower_last;
729 struct uid_gid_extent *prev;
730
6397fac4
CB
731 if (new_map->nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS)
732 prev = &new_map->extent[idx];
733 else
734 prev = &new_map->forward[idx];
0bd14b4f
EB
735
736 prev_upper_first = prev->first;
737 prev_lower_first = prev->lower_first;
738 prev_upper_last = prev_upper_first + prev->count - 1;
739 prev_lower_last = prev_lower_first + prev->count - 1;
740
741 /* Does the upper range intersect a previous extent? */
742 if ((prev_upper_first <= upper_last) &&
743 (prev_upper_last >= upper_first))
744 return true;
745
746 /* Does the lower range intersect a previous extent? */
747 if ((prev_lower_first <= lower_last) &&
748 (prev_lower_last >= lower_first))
749 return true;
750 }
751 return false;
752}
753
6397fac4
CB
754/**
755 * insert_extent - Safely insert a new idmap extent into struct uid_gid_map.
756 * Takes care to allocate a 4K block of memory if the number of mappings exceeds
757 * UID_GID_MAP_MAX_BASE_EXTENTS.
758 */
759static int insert_extent(struct uid_gid_map *map, struct uid_gid_extent *extent)
760{
3fda0e73 761 struct uid_gid_extent *dest;
6397fac4
CB
762
763 if (map->nr_extents == UID_GID_MAP_MAX_BASE_EXTENTS) {
764 struct uid_gid_extent *forward;
765
766 /* Allocate memory for 340 mappings. */
6da2ec56
KC
767 forward = kmalloc_array(UID_GID_MAP_MAX_EXTENTS,
768 sizeof(struct uid_gid_extent),
769 GFP_KERNEL);
6397fac4
CB
770 if (!forward)
771 return -ENOMEM;
772
773 /* Copy over memory. Only set up memory for the forward pointer.
774 * Defer the memory setup for the reverse pointer.
775 */
776 memcpy(forward, map->extent,
777 map->nr_extents * sizeof(map->extent[0]));
778
779 map->forward = forward;
780 map->reverse = NULL;
781 }
782
3fda0e73
EB
783 if (map->nr_extents < UID_GID_MAP_MAX_BASE_EXTENTS)
784 dest = &map->extent[map->nr_extents];
785 else
786 dest = &map->forward[map->nr_extents];
787
788 *dest = *extent;
789 map->nr_extents++;
6397fac4
CB
790 return 0;
791}
792
793/* cmp function to sort() forward mappings */
794static int cmp_extents_forward(const void *a, const void *b)
795{
796 const struct uid_gid_extent *e1 = a;
797 const struct uid_gid_extent *e2 = b;
798
799 if (e1->first < e2->first)
800 return -1;
801
802 if (e1->first > e2->first)
803 return 1;
804
805 return 0;
806}
807
808/* cmp function to sort() reverse mappings */
809static int cmp_extents_reverse(const void *a, const void *b)
810{
811 const struct uid_gid_extent *e1 = a;
812 const struct uid_gid_extent *e2 = b;
813
814 if (e1->lower_first < e2->lower_first)
815 return -1;
816
817 if (e1->lower_first > e2->lower_first)
818 return 1;
819
820 return 0;
821}
822
823/**
824 * sort_idmaps - Sorts an array of idmap entries.
825 * Can only be called if number of mappings exceeds UID_GID_MAP_MAX_BASE_EXTENTS.
826 */
827static int sort_idmaps(struct uid_gid_map *map)
828{
829 if (map->nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS)
830 return 0;
831
832 /* Sort forward array. */
833 sort(map->forward, map->nr_extents, sizeof(struct uid_gid_extent),
834 cmp_extents_forward, NULL);
835
836 /* Only copy the memory from forward we actually need. */
837 map->reverse = kmemdup(map->forward,
838 map->nr_extents * sizeof(struct uid_gid_extent),
839 GFP_KERNEL);
840 if (!map->reverse)
841 return -ENOMEM;
842
843 /* Sort reverse array. */
844 sort(map->reverse, map->nr_extents, sizeof(struct uid_gid_extent),
845 cmp_extents_reverse, NULL);
846
847 return 0;
848}
849
22d917d8
EB
850static ssize_t map_write(struct file *file, const char __user *buf,
851 size_t count, loff_t *ppos,
852 int cap_setid,
853 struct uid_gid_map *map,
854 struct uid_gid_map *parent_map)
855{
856 struct seq_file *seq = file->private_data;
857 struct user_namespace *ns = seq->private;
858 struct uid_gid_map new_map;
859 unsigned idx;
6397fac4 860 struct uid_gid_extent extent;
70f6cbb6 861 char *kbuf = NULL, *pos, *next_line;
5820f140
JH
862 ssize_t ret;
863
864 /* Only allow < page size writes at the beginning of the file */
865 if ((*ppos != 0) || (count >= PAGE_SIZE))
866 return -EINVAL;
867
868 /* Slurp in the user data */
869 kbuf = memdup_user_nul(buf, count);
870 if (IS_ERR(kbuf))
871 return PTR_ERR(kbuf);
22d917d8
EB
872
873 /*
f0d62aec 874 * The userns_state_mutex serializes all writes to any given map.
22d917d8
EB
875 *
876 * Any map is only ever written once.
877 *
878 * An id map fits within 1 cache line on most architectures.
879 *
880 * On read nothing needs to be done unless you are on an
881 * architecture with a crazy cache coherency model like alpha.
882 *
883 * There is a one time data dependency between reading the
884 * count of the extents and the values of the extents. The
885 * desired behavior is to see the values of the extents that
886 * were written before the count of the extents.
887 *
888 * To achieve this smp_wmb() is used on guarantee the write
e79323bd
MP
889 * order and smp_rmb() is guaranteed that we don't have crazy
890 * architectures returning stale data.
22d917d8 891 */
f0d62aec 892 mutex_lock(&userns_state_mutex);
22d917d8 893
6397fac4
CB
894 memset(&new_map, 0, sizeof(struct uid_gid_map));
895
22d917d8
EB
896 ret = -EPERM;
897 /* Only allow one successful write to the map */
898 if (map->nr_extents != 0)
899 goto out;
900
41c21e35
AL
901 /*
902 * Adjusting namespace settings requires capabilities on the target.
5c1469de 903 */
41c21e35 904 if (cap_valid(cap_setid) && !file_ns_capable(file, ns, CAP_SYS_ADMIN))
22d917d8
EB
905 goto out;
906
22d917d8
EB
907 /* Parse the user data */
908 ret = -EINVAL;
909 pos = kbuf;
68a9a435 910 for (; pos; pos = next_line) {
22d917d8
EB
911
912 /* Find the end of line and ensure I don't look past it */
913 next_line = strchr(pos, '\n');
914 if (next_line) {
915 *next_line = '\0';
916 next_line++;
917 if (*next_line == '\0')
918 next_line = NULL;
5c1469de 919 }
22d917d8
EB
920
921 pos = skip_spaces(pos);
6397fac4 922 extent.first = simple_strtoul(pos, &pos, 10);
22d917d8
EB
923 if (!isspace(*pos))
924 goto out;
925
926 pos = skip_spaces(pos);
6397fac4 927 extent.lower_first = simple_strtoul(pos, &pos, 10);
22d917d8
EB
928 if (!isspace(*pos))
929 goto out;
930
931 pos = skip_spaces(pos);
6397fac4 932 extent.count = simple_strtoul(pos, &pos, 10);
22d917d8
EB
933 if (*pos && !isspace(*pos))
934 goto out;
935
936 /* Verify there is not trailing junk on the line */
937 pos = skip_spaces(pos);
938 if (*pos != '\0')
939 goto out;
940
941 /* Verify we have been given valid starting values */
6397fac4
CB
942 if ((extent.first == (u32) -1) ||
943 (extent.lower_first == (u32) -1))
22d917d8
EB
944 goto out;
945
68a9a435
FF
946 /* Verify count is not zero and does not cause the
947 * extent to wrap
948 */
6397fac4 949 if ((extent.first + extent.count) <= extent.first)
22d917d8 950 goto out;
6397fac4
CB
951 if ((extent.lower_first + extent.count) <=
952 extent.lower_first)
22d917d8
EB
953 goto out;
954
0bd14b4f 955 /* Do the ranges in extent overlap any previous extents? */
6397fac4 956 if (mappings_overlap(&new_map, &extent))
22d917d8
EB
957 goto out;
958
6397fac4 959 if ((new_map.nr_extents + 1) == UID_GID_MAP_MAX_EXTENTS &&
22d917d8
EB
960 (next_line != NULL))
961 goto out;
6397fac4
CB
962
963 ret = insert_extent(&new_map, &extent);
964 if (ret < 0)
965 goto out;
966 ret = -EINVAL;
5c1469de 967 }
22d917d8
EB
968 /* Be very certaint the new map actually exists */
969 if (new_map.nr_extents == 0)
970 goto out;
971
972 ret = -EPERM;
973 /* Validate the user is allowed to use user id's mapped to. */
6708075f 974 if (!new_idmap_permitted(file, ns, cap_setid, &new_map))
22d917d8
EB
975 goto out;
976
6397fac4
CB
977 ret = sort_idmaps(&new_map);
978 if (ret < 0)
979 goto out;
980
981 ret = -EPERM;
22d917d8
EB
982 /* Map the lower ids from the parent user namespace to the
983 * kernel global id space.
984 */
985 for (idx = 0; idx < new_map.nr_extents; idx++) {
6397fac4 986 struct uid_gid_extent *e;
22d917d8 987 u32 lower_first;
6397fac4
CB
988
989 if (new_map.nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS)
990 e = &new_map.extent[idx];
991 else
992 e = &new_map.forward[idx];
22d917d8
EB
993
994 lower_first = map_id_range_down(parent_map,
6397fac4
CB
995 e->lower_first,
996 e->count);
22d917d8
EB
997
998 /* Fail if we can not map the specified extent to
999 * the kernel global id space.
1000 */
1001 if (lower_first == (u32) -1)
1002 goto out;
1003
6397fac4 1004 e->lower_first = lower_first;
22d917d8
EB
1005 }
1006
1007 /* Install the map */
6397fac4
CB
1008 if (new_map.nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) {
1009 memcpy(map->extent, new_map.extent,
1010 new_map.nr_extents * sizeof(new_map.extent[0]));
1011 } else {
1012 map->forward = new_map.forward;
1013 map->reverse = new_map.reverse;
1014 }
22d917d8
EB
1015 smp_wmb();
1016 map->nr_extents = new_map.nr_extents;
1017
1018 *ppos = count;
1019 ret = count;
1020out:
6397fac4
CB
1021 if (ret < 0 && new_map.nr_extents > UID_GID_MAP_MAX_BASE_EXTENTS) {
1022 kfree(new_map.forward);
1023 kfree(new_map.reverse);
1024 map->forward = NULL;
1025 map->reverse = NULL;
1026 map->nr_extents = 0;
1027 }
1028
f0d62aec 1029 mutex_unlock(&userns_state_mutex);
70f6cbb6 1030 kfree(kbuf);
22d917d8
EB
1031 return ret;
1032}
1033
68a9a435
FF
1034ssize_t proc_uid_map_write(struct file *file, const char __user *buf,
1035 size_t size, loff_t *ppos)
22d917d8
EB
1036{
1037 struct seq_file *seq = file->private_data;
1038 struct user_namespace *ns = seq->private;
c450f371 1039 struct user_namespace *seq_ns = seq_user_ns(seq);
22d917d8
EB
1040
1041 if (!ns->parent)
1042 return -EPERM;
1043
c450f371
EB
1044 if ((seq_ns != ns) && (seq_ns != ns->parent))
1045 return -EPERM;
1046
22d917d8
EB
1047 return map_write(file, buf, size, ppos, CAP_SETUID,
1048 &ns->uid_map, &ns->parent->uid_map);
1049}
1050
68a9a435
FF
1051ssize_t proc_gid_map_write(struct file *file, const char __user *buf,
1052 size_t size, loff_t *ppos)
22d917d8
EB
1053{
1054 struct seq_file *seq = file->private_data;
1055 struct user_namespace *ns = seq->private;
c450f371 1056 struct user_namespace *seq_ns = seq_user_ns(seq);
22d917d8
EB
1057
1058 if (!ns->parent)
1059 return -EPERM;
1060
c450f371
EB
1061 if ((seq_ns != ns) && (seq_ns != ns->parent))
1062 return -EPERM;
1063
22d917d8
EB
1064 return map_write(file, buf, size, ppos, CAP_SETGID,
1065 &ns->gid_map, &ns->parent->gid_map);
1066}
1067
68a9a435
FF
1068ssize_t proc_projid_map_write(struct file *file, const char __user *buf,
1069 size_t size, loff_t *ppos)
f76d207a
EB
1070{
1071 struct seq_file *seq = file->private_data;
1072 struct user_namespace *ns = seq->private;
1073 struct user_namespace *seq_ns = seq_user_ns(seq);
1074
1075 if (!ns->parent)
1076 return -EPERM;
1077
1078 if ((seq_ns != ns) && (seq_ns != ns->parent))
1079 return -EPERM;
1080
1081 /* Anyone can set any valid project id no capability needed */
1082 return map_write(file, buf, size, ppos, -1,
1083 &ns->projid_map, &ns->parent->projid_map);
1084}
1085
68a9a435 1086static bool new_idmap_permitted(const struct file *file,
6708075f 1087 struct user_namespace *ns, int cap_setid,
22d917d8
EB
1088 struct uid_gid_map *new_map)
1089{
f95d7918 1090 const struct cred *cred = file->f_cred;
0542f17b
EB
1091 /* Don't allow mappings that would allow anything that wouldn't
1092 * be allowed without the establishment of unprivileged mappings.
1093 */
f95d7918
EB
1094 if ((new_map->nr_extents == 1) && (new_map->extent[0].count == 1) &&
1095 uid_eq(ns->owner, cred->euid)) {
37657da3
EB
1096 u32 id = new_map->extent[0].lower_first;
1097 if (cap_setid == CAP_SETUID) {
1098 kuid_t uid = make_kuid(ns->parent, id);
f95d7918 1099 if (uid_eq(uid, cred->euid))
37657da3 1100 return true;
68a9a435 1101 } else if (cap_setid == CAP_SETGID) {
37657da3 1102 kgid_t gid = make_kgid(ns->parent, id);
66d2f338
EB
1103 if (!(ns->flags & USERNS_SETGROUPS_ALLOWED) &&
1104 gid_eq(gid, cred->egid))
37657da3
EB
1105 return true;
1106 }
1107 }
1108
f76d207a
EB
1109 /* Allow anyone to set a mapping that doesn't require privilege */
1110 if (!cap_valid(cap_setid))
1111 return true;
1112
22d917d8
EB
1113 /* Allow the specified ids if we have the appropriate capability
1114 * (CAP_SETUID or CAP_SETGID) over the parent user namespace.
6708075f 1115 * And the opener of the id file also had the approprpiate capability.
22d917d8 1116 */
6708075f
EB
1117 if (ns_capable(ns->parent, cap_setid) &&
1118 file_ns_capable(file, ns->parent, cap_setid))
22d917d8 1119 return true;
5c1469de 1120
22d917d8 1121 return false;
5c1469de 1122}
6164281a 1123
9cc46516
EB
1124int proc_setgroups_show(struct seq_file *seq, void *v)
1125{
1126 struct user_namespace *ns = seq->private;
6aa7de05 1127 unsigned long userns_flags = READ_ONCE(ns->flags);
9cc46516
EB
1128
1129 seq_printf(seq, "%s\n",
1130 (userns_flags & USERNS_SETGROUPS_ALLOWED) ?
1131 "allow" : "deny");
1132 return 0;
1133}
1134
1135ssize_t proc_setgroups_write(struct file *file, const char __user *buf,
1136 size_t count, loff_t *ppos)
1137{
1138 struct seq_file *seq = file->private_data;
1139 struct user_namespace *ns = seq->private;
1140 char kbuf[8], *pos;
1141 bool setgroups_allowed;
1142 ssize_t ret;
1143
1144 /* Only allow a very narrow range of strings to be written */
1145 ret = -EINVAL;
1146 if ((*ppos != 0) || (count >= sizeof(kbuf)))
1147 goto out;
1148
1149 /* What was written? */
1150 ret = -EFAULT;
1151 if (copy_from_user(kbuf, buf, count))
1152 goto out;
1153 kbuf[count] = '\0';
1154 pos = kbuf;
1155
1156 /* What is being requested? */
1157 ret = -EINVAL;
1158 if (strncmp(pos, "allow", 5) == 0) {
1159 pos += 5;
1160 setgroups_allowed = true;
1161 }
1162 else if (strncmp(pos, "deny", 4) == 0) {
1163 pos += 4;
1164 setgroups_allowed = false;
1165 }
1166 else
1167 goto out;
1168
1169 /* Verify there is not trailing junk on the line */
1170 pos = skip_spaces(pos);
1171 if (*pos != '\0')
1172 goto out;
1173
1174 ret = -EPERM;
1175 mutex_lock(&userns_state_mutex);
1176 if (setgroups_allowed) {
1177 /* Enabling setgroups after setgroups has been disabled
1178 * is not allowed.
1179 */
1180 if (!(ns->flags & USERNS_SETGROUPS_ALLOWED))
1181 goto out_unlock;
1182 } else {
1183 /* Permanently disabling setgroups after setgroups has
1184 * been enabled by writing the gid_map is not allowed.
1185 */
1186 if (ns->gid_map.nr_extents != 0)
1187 goto out_unlock;
1188 ns->flags &= ~USERNS_SETGROUPS_ALLOWED;
1189 }
1190 mutex_unlock(&userns_state_mutex);
1191
1192 /* Report a successful write */
1193 *ppos = count;
1194 ret = count;
1195out:
1196 return ret;
1197out_unlock:
1198 mutex_unlock(&userns_state_mutex);
1199 goto out;
1200}
1201
273d2c67
EB
1202bool userns_may_setgroups(const struct user_namespace *ns)
1203{
1204 bool allowed;
1205
f0d62aec 1206 mutex_lock(&userns_state_mutex);
273d2c67
EB
1207 /* It is not safe to use setgroups until a gid mapping in
1208 * the user namespace has been established.
1209 */
1210 allowed = ns->gid_map.nr_extents != 0;
9cc46516
EB
1211 /* Is setgroups allowed? */
1212 allowed = allowed && (ns->flags & USERNS_SETGROUPS_ALLOWED);
f0d62aec 1213 mutex_unlock(&userns_state_mutex);
273d2c67
EB
1214
1215 return allowed;
1216}
1217
d07b846f 1218/*
a2b42626
EB
1219 * Returns true if @child is the same namespace or a descendant of
1220 * @ancestor.
d07b846f 1221 */
a2b42626
EB
1222bool in_userns(const struct user_namespace *ancestor,
1223 const struct user_namespace *child)
1224{
1225 const struct user_namespace *ns;
1226 for (ns = child; ns->level > ancestor->level; ns = ns->parent)
1227 ;
1228 return (ns == ancestor);
1229}
1230
d07b846f
SF
1231bool current_in_userns(const struct user_namespace *target_ns)
1232{
a2b42626 1233 return in_userns(target_ns, current_user_ns());
d07b846f 1234}
73f03c2b 1235EXPORT_SYMBOL(current_in_userns);
d07b846f 1236
3c041184
AV
1237static inline struct user_namespace *to_user_ns(struct ns_common *ns)
1238{
1239 return container_of(ns, struct user_namespace, ns);
1240}
1241
64964528 1242static struct ns_common *userns_get(struct task_struct *task)
cde1975b
EB
1243{
1244 struct user_namespace *user_ns;
1245
1246 rcu_read_lock();
1247 user_ns = get_user_ns(__task_cred(task)->user_ns);
1248 rcu_read_unlock();
1249
3c041184 1250 return user_ns ? &user_ns->ns : NULL;
cde1975b
EB
1251}
1252
64964528 1253static void userns_put(struct ns_common *ns)
cde1975b 1254{
3c041184 1255 put_user_ns(to_user_ns(ns));
cde1975b
EB
1256}
1257
64964528 1258static int userns_install(struct nsproxy *nsproxy, struct ns_common *ns)
cde1975b 1259{
3c041184 1260 struct user_namespace *user_ns = to_user_ns(ns);
cde1975b
EB
1261 struct cred *cred;
1262
1263 /* Don't allow gaining capabilities by reentering
1264 * the same user namespace.
1265 */
1266 if (user_ns == current_user_ns())
1267 return -EINVAL;
1268
faf00da5
EB
1269 /* Tasks that share a thread group must share a user namespace */
1270 if (!thread_group_empty(current))
cde1975b
EB
1271 return -EINVAL;
1272
e66eded8
EB
1273 if (current->fs->users != 1)
1274 return -EINVAL;
1275
cde1975b
EB
1276 if (!ns_capable(user_ns, CAP_SYS_ADMIN))
1277 return -EPERM;
1278
1279 cred = prepare_creds();
1280 if (!cred)
1281 return -ENOMEM;
1282
1283 put_user_ns(cred->user_ns);
1284 set_cred_user_ns(cred, get_user_ns(user_ns));
1285
1286 return commit_creds(cred);
1287}
1288
bcac25a5
AV
1289struct ns_common *ns_get_owner(struct ns_common *ns)
1290{
1291 struct user_namespace *my_user_ns = current_user_ns();
1292 struct user_namespace *owner, *p;
1293
1294 /* See if the owner is in the current user namespace */
1295 owner = p = ns->ops->owner(ns);
1296 for (;;) {
1297 if (!p)
1298 return ERR_PTR(-EPERM);
1299 if (p == my_user_ns)
1300 break;
1301 p = p->parent;
1302 }
1303
1304 return &get_user_ns(owner)->ns;
1305}
1306
1307static struct user_namespace *userns_owner(struct ns_common *ns)
1308{
1309 return to_user_ns(ns)->parent;
1310}
1311
cde1975b
EB
1312const struct proc_ns_operations userns_operations = {
1313 .name = "user",
1314 .type = CLONE_NEWUSER,
1315 .get = userns_get,
1316 .put = userns_put,
1317 .install = userns_install,
bcac25a5 1318 .owner = userns_owner,
a7306ed8 1319 .get_parent = ns_get_owner,
cde1975b
EB
1320};
1321
6164281a
PE
1322static __init int user_namespaces_init(void)
1323{
1324 user_ns_cachep = KMEM_CACHE(user_namespace, SLAB_PANIC);
1325 return 0;
1326}
c96d6660 1327subsys_initcall(user_namespaces_init);