]>
Commit | Line | Data |
---|---|---|
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 | 29 | static struct kmem_cache *user_ns_cachep __read_mostly; |
f0d62aec | 30 | static DEFINE_MUTEX(userns_state_mutex); |
6164281a | 31 | |
6708075f EB |
32 | static bool new_idmap_permitted(const struct file *file, |
33 | struct user_namespace *ns, int cap_setid, | |
22d917d8 | 34 | struct uid_gid_map *map); |
b032132c | 35 | static void free_user_ns(struct work_struct *work); |
22d917d8 | 36 | |
25f9c081 EB |
37 | static struct ucounts *inc_user_namespaces(struct user_namespace *ns, kuid_t uid) |
38 | { | |
39 | return inc_ucount(ns, uid, UCOUNT_USER_NAMESPACES); | |
40 | } | |
41 | ||
42 | static void dec_user_namespaces(struct ucounts *ucounts) | |
43 | { | |
44 | return dec_ucount(ucounts, UCOUNT_USER_NAMESPACES); | |
45 | } | |
46 | ||
cde1975b EB |
47 | static 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 | 74 | int 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 |
145 | fail_keyring: |
146 | #ifdef CONFIG_PERSISTENT_KEYRINGS | |
147 | key_put(ns->persistent_keyring_register); | |
148 | #endif | |
149 | ns_free_inum(&ns->ns); | |
b376c3e1 | 150 | fail_free: |
dbec2846 | 151 | kmem_cache_free(user_ns_cachep, ns); |
b376c3e1 | 152 | fail_dec: |
f6b2db1a | 153 | dec_user_namespaces(ucounts); |
b376c3e1 | 154 | fail: |
dbec2846 | 155 | return ret; |
acce292c CLG |
156 | } |
157 | ||
b2e0d987 EB |
158 | int 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 | 178 | static 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 | |
209 | void __put_user_ns(struct user_namespace *ns) | |
210 | { | |
211 | schedule_work(&ns->work); | |
212 | } | |
213 | EXPORT_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 | */ | |
219 | struct 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 | */ | |
229 | static 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 | ||
235 | /* handle map_id_range_down() */ | |
236 | if (key->count) | |
237 | id2 = key->id + key->count - 1; | |
238 | else | |
239 | id2 = key->id; | |
240 | ||
241 | /* handle map_id_{down,up}() */ | |
242 | if (key->map_up) | |
243 | first = el->lower_first; | |
244 | else | |
245 | first = el->first; | |
246 | ||
247 | last = first + el->count - 1; | |
248 | ||
249 | if (key->id >= first && key->id <= last && | |
250 | (id2 >= first && id2 <= last)) | |
251 | return 0; | |
252 | ||
253 | if (key->id < first || id2 < first) | |
254 | return -1; | |
255 | ||
256 | return 1; | |
257 | } | |
258 | ||
259 | /** | |
260 | * map_id_range_down_max - Find idmap via binary search in ordered idmap array. | |
261 | * Can only be called if number of mappings exceeds UID_GID_MAP_MAX_BASE_EXTENTS. | |
262 | */ | |
263 | static u32 map_id_range_down_max(struct uid_gid_map *map, u32 id, u32 count) | |
264 | { | |
265 | u32 extents; | |
266 | struct uid_gid_extent *extent; | |
267 | struct idmap_key key; | |
268 | ||
269 | key.map_up = false; | |
270 | key.count = count; | |
271 | key.id = id; | |
272 | ||
273 | extents = map->nr_extents; | |
274 | smp_rmb(); | |
275 | ||
276 | extent = bsearch(&key, map->forward, extents, | |
277 | sizeof(struct uid_gid_extent), cmp_map_id); | |
278 | /* Map the id or note failure */ | |
279 | if (extent) | |
280 | id = (id - extent->first) + extent->lower_first; | |
281 | else | |
282 | id = (u32) -1; | |
283 | ||
284 | return id; | |
285 | } | |
286 | ||
287 | /** | |
288 | * map_id_range_down_base - Find idmap via binary search in static extent array. | |
289 | * Can only be called if number of mappings is equal or less than | |
290 | * UID_GID_MAP_MAX_BASE_EXTENTS. | |
291 | */ | |
292 | static u32 map_id_range_down_base(struct uid_gid_map *map, u32 id, u32 count) | |
5c1469de | 293 | { |
22d917d8 EB |
294 | unsigned idx, extents; |
295 | u32 first, last, id2; | |
5c1469de | 296 | |
22d917d8 | 297 | id2 = id + count - 1; |
5c1469de | 298 | |
22d917d8 EB |
299 | /* Find the matching extent */ |
300 | extents = map->nr_extents; | |
e79323bd | 301 | smp_rmb(); |
22d917d8 EB |
302 | for (idx = 0; idx < extents; idx++) { |
303 | first = map->extent[idx].first; | |
304 | last = first + map->extent[idx].count - 1; | |
305 | if (id >= first && id <= last && | |
306 | (id2 >= first && id2 <= last)) | |
307 | break; | |
308 | } | |
309 | /* Map the id or note failure */ | |
310 | if (idx < extents) | |
311 | id = (id - first) + map->extent[idx].lower_first; | |
312 | else | |
313 | id = (u32) -1; | |
314 | ||
315 | return id; | |
316 | } | |
317 | ||
6397fac4 CB |
318 | static u32 map_id_range_down(struct uid_gid_map *map, u32 id, u32 count) |
319 | { | |
320 | u32 extents = map->nr_extents; | |
321 | smp_rmb(); | |
322 | ||
323 | if (extents <= UID_GID_MAP_MAX_BASE_EXTENTS) | |
324 | return map_id_range_down_base(map, id, count); | |
325 | ||
326 | return map_id_range_down_max(map, id, count); | |
327 | } | |
328 | ||
329 | /** | |
330 | * map_id_down_base - Find idmap via binary search in static extent array. | |
331 | * Can only be called if number of mappings is equal or less than | |
332 | * UID_GID_MAP_MAX_BASE_EXTENTS. | |
333 | */ | |
334 | static u32 map_id_down_base(struct uid_gid_map *map, u32 id) | |
22d917d8 EB |
335 | { |
336 | unsigned idx, extents; | |
337 | u32 first, last; | |
338 | ||
339 | /* Find the matching extent */ | |
340 | extents = map->nr_extents; | |
e79323bd | 341 | smp_rmb(); |
22d917d8 EB |
342 | for (idx = 0; idx < extents; idx++) { |
343 | first = map->extent[idx].first; | |
344 | last = first + map->extent[idx].count - 1; | |
345 | if (id >= first && id <= last) | |
346 | break; | |
347 | } | |
348 | /* Map the id or note failure */ | |
349 | if (idx < extents) | |
350 | id = (id - first) + map->extent[idx].lower_first; | |
351 | else | |
352 | id = (u32) -1; | |
353 | ||
354 | return id; | |
355 | } | |
356 | ||
6397fac4 CB |
357 | static u32 map_id_down(struct uid_gid_map *map, u32 id) |
358 | { | |
359 | u32 extents = map->nr_extents; | |
360 | smp_rmb(); | |
361 | ||
362 | if (extents <= UID_GID_MAP_MAX_BASE_EXTENTS) | |
363 | return map_id_down_base(map, id); | |
364 | ||
365 | return map_id_range_down_max(map, id, 0); | |
366 | } | |
367 | ||
368 | /** | |
369 | * map_id_up_base - Find idmap via binary search in static extent array. | |
370 | * Can only be called if number of mappings is equal or less than | |
371 | * UID_GID_MAP_MAX_BASE_EXTENTS. | |
372 | */ | |
373 | static u32 map_id_up_base(struct uid_gid_map *map, u32 id) | |
22d917d8 EB |
374 | { |
375 | unsigned idx, extents; | |
376 | u32 first, last; | |
377 | ||
378 | /* Find the matching extent */ | |
379 | extents = map->nr_extents; | |
e79323bd | 380 | smp_rmb(); |
22d917d8 EB |
381 | for (idx = 0; idx < extents; idx++) { |
382 | first = map->extent[idx].lower_first; | |
383 | last = first + map->extent[idx].count - 1; | |
384 | if (id >= first && id <= last) | |
385 | break; | |
5c1469de | 386 | } |
22d917d8 EB |
387 | /* Map the id or note failure */ |
388 | if (idx < extents) | |
389 | id = (id - first) + map->extent[idx].first; | |
390 | else | |
391 | id = (u32) -1; | |
392 | ||
393 | return id; | |
394 | } | |
395 | ||
6397fac4 CB |
396 | /** |
397 | * map_id_up_max - Find idmap via binary search in ordered idmap array. | |
398 | * Can only be called if number of mappings exceeds UID_GID_MAP_MAX_BASE_EXTENTS. | |
399 | */ | |
400 | static u32 map_id_up_max(struct uid_gid_map *map, u32 id) | |
401 | { | |
402 | u32 extents; | |
403 | struct uid_gid_extent *extent; | |
404 | struct idmap_key key; | |
405 | ||
406 | key.map_up = true; | |
407 | key.count = 0; | |
408 | key.id = id; | |
409 | ||
410 | extents = map->nr_extents; | |
411 | smp_rmb(); | |
412 | ||
413 | extent = bsearch(&key, map->reverse, extents, | |
414 | sizeof(struct uid_gid_extent), cmp_map_id); | |
415 | /* Map the id or note failure */ | |
416 | if (extent) | |
417 | id = (id - extent->lower_first) + extent->first; | |
418 | else | |
419 | id = (u32) -1; | |
420 | ||
421 | return id; | |
422 | } | |
423 | ||
424 | static u32 map_id_up(struct uid_gid_map *map, u32 id) | |
425 | { | |
426 | u32 extents = map->nr_extents; | |
427 | smp_rmb(); | |
428 | ||
429 | if (extents <= UID_GID_MAP_MAX_BASE_EXTENTS) | |
430 | return map_id_up_base(map, id); | |
431 | ||
432 | return map_id_up_max(map, id); | |
433 | } | |
434 | ||
22d917d8 EB |
435 | /** |
436 | * make_kuid - Map a user-namespace uid pair into a kuid. | |
437 | * @ns: User namespace that the uid is in | |
438 | * @uid: User identifier | |
439 | * | |
440 | * Maps a user-namespace uid pair into a kernel internal kuid, | |
441 | * and returns that kuid. | |
442 | * | |
443 | * When there is no mapping defined for the user-namespace uid | |
444 | * pair INVALID_UID is returned. Callers are expected to test | |
b080e047 | 445 | * for and handle INVALID_UID being returned. INVALID_UID |
22d917d8 EB |
446 | * may be tested for using uid_valid(). |
447 | */ | |
448 | kuid_t make_kuid(struct user_namespace *ns, uid_t uid) | |
449 | { | |
450 | /* Map the uid to a global kernel uid */ | |
451 | return KUIDT_INIT(map_id_down(&ns->uid_map, uid)); | |
452 | } | |
453 | EXPORT_SYMBOL(make_kuid); | |
454 | ||
455 | /** | |
456 | * from_kuid - Create a uid from a kuid user-namespace pair. | |
457 | * @targ: The user namespace we want a uid in. | |
458 | * @kuid: The kernel internal uid to start with. | |
459 | * | |
460 | * Map @kuid into the user-namespace specified by @targ and | |
461 | * return the resulting uid. | |
462 | * | |
463 | * There is always a mapping into the initial user_namespace. | |
464 | * | |
465 | * If @kuid has no mapping in @targ (uid_t)-1 is returned. | |
466 | */ | |
467 | uid_t from_kuid(struct user_namespace *targ, kuid_t kuid) | |
468 | { | |
469 | /* Map the uid from a global kernel uid */ | |
470 | return map_id_up(&targ->uid_map, __kuid_val(kuid)); | |
471 | } | |
472 | EXPORT_SYMBOL(from_kuid); | |
473 | ||
474 | /** | |
475 | * from_kuid_munged - Create a uid from a kuid user-namespace pair. | |
476 | * @targ: The user namespace we want a uid in. | |
477 | * @kuid: The kernel internal uid to start with. | |
478 | * | |
479 | * Map @kuid into the user-namespace specified by @targ and | |
480 | * return the resulting uid. | |
481 | * | |
482 | * There is always a mapping into the initial user_namespace. | |
483 | * | |
484 | * Unlike from_kuid from_kuid_munged never fails and always | |
485 | * returns a valid uid. This makes from_kuid_munged appropriate | |
486 | * for use in syscalls like stat and getuid where failing the | |
487 | * system call and failing to provide a valid uid are not an | |
488 | * options. | |
489 | * | |
490 | * If @kuid has no mapping in @targ overflowuid is returned. | |
491 | */ | |
492 | uid_t from_kuid_munged(struct user_namespace *targ, kuid_t kuid) | |
493 | { | |
494 | uid_t uid; | |
495 | uid = from_kuid(targ, kuid); | |
496 | ||
497 | if (uid == (uid_t) -1) | |
498 | uid = overflowuid; | |
499 | return uid; | |
500 | } | |
501 | EXPORT_SYMBOL(from_kuid_munged); | |
502 | ||
503 | /** | |
504 | * make_kgid - Map a user-namespace gid pair into a kgid. | |
505 | * @ns: User namespace that the gid is in | |
68a9a435 | 506 | * @gid: group identifier |
22d917d8 EB |
507 | * |
508 | * Maps a user-namespace gid pair into a kernel internal kgid, | |
509 | * and returns that kgid. | |
510 | * | |
511 | * When there is no mapping defined for the user-namespace gid | |
512 | * pair INVALID_GID is returned. Callers are expected to test | |
513 | * for and handle INVALID_GID being returned. INVALID_GID may be | |
514 | * tested for using gid_valid(). | |
515 | */ | |
516 | kgid_t make_kgid(struct user_namespace *ns, gid_t gid) | |
517 | { | |
518 | /* Map the gid to a global kernel gid */ | |
519 | return KGIDT_INIT(map_id_down(&ns->gid_map, gid)); | |
520 | } | |
521 | EXPORT_SYMBOL(make_kgid); | |
522 | ||
523 | /** | |
524 | * from_kgid - Create a gid from a kgid user-namespace pair. | |
525 | * @targ: The user namespace we want a gid in. | |
526 | * @kgid: The kernel internal gid to start with. | |
527 | * | |
528 | * Map @kgid into the user-namespace specified by @targ and | |
529 | * return the resulting gid. | |
530 | * | |
531 | * There is always a mapping into the initial user_namespace. | |
532 | * | |
533 | * If @kgid has no mapping in @targ (gid_t)-1 is returned. | |
534 | */ | |
535 | gid_t from_kgid(struct user_namespace *targ, kgid_t kgid) | |
536 | { | |
537 | /* Map the gid from a global kernel gid */ | |
538 | return map_id_up(&targ->gid_map, __kgid_val(kgid)); | |
539 | } | |
540 | EXPORT_SYMBOL(from_kgid); | |
541 | ||
542 | /** | |
543 | * from_kgid_munged - Create a gid from a kgid user-namespace pair. | |
544 | * @targ: The user namespace we want a gid in. | |
545 | * @kgid: The kernel internal gid to start with. | |
546 | * | |
547 | * Map @kgid into the user-namespace specified by @targ and | |
548 | * return the resulting gid. | |
549 | * | |
550 | * There is always a mapping into the initial user_namespace. | |
551 | * | |
552 | * Unlike from_kgid from_kgid_munged never fails and always | |
553 | * returns a valid gid. This makes from_kgid_munged appropriate | |
554 | * for use in syscalls like stat and getgid where failing the | |
555 | * system call and failing to provide a valid gid are not options. | |
556 | * | |
557 | * If @kgid has no mapping in @targ overflowgid is returned. | |
558 | */ | |
559 | gid_t from_kgid_munged(struct user_namespace *targ, kgid_t kgid) | |
560 | { | |
561 | gid_t gid; | |
562 | gid = from_kgid(targ, kgid); | |
563 | ||
564 | if (gid == (gid_t) -1) | |
565 | gid = overflowgid; | |
566 | return gid; | |
567 | } | |
568 | EXPORT_SYMBOL(from_kgid_munged); | |
569 | ||
f76d207a EB |
570 | /** |
571 | * make_kprojid - Map a user-namespace projid pair into a kprojid. | |
572 | * @ns: User namespace that the projid is in | |
573 | * @projid: Project identifier | |
574 | * | |
575 | * Maps a user-namespace uid pair into a kernel internal kuid, | |
576 | * and returns that kuid. | |
577 | * | |
578 | * When there is no mapping defined for the user-namespace projid | |
579 | * pair INVALID_PROJID is returned. Callers are expected to test | |
580 | * for and handle handle INVALID_PROJID being returned. INVALID_PROJID | |
581 | * may be tested for using projid_valid(). | |
582 | */ | |
583 | kprojid_t make_kprojid(struct user_namespace *ns, projid_t projid) | |
584 | { | |
585 | /* Map the uid to a global kernel uid */ | |
586 | return KPROJIDT_INIT(map_id_down(&ns->projid_map, projid)); | |
587 | } | |
588 | EXPORT_SYMBOL(make_kprojid); | |
589 | ||
590 | /** | |
591 | * from_kprojid - Create a projid from a kprojid user-namespace pair. | |
592 | * @targ: The user namespace we want a projid in. | |
593 | * @kprojid: The kernel internal project identifier to start with. | |
594 | * | |
595 | * Map @kprojid into the user-namespace specified by @targ and | |
596 | * return the resulting projid. | |
597 | * | |
598 | * There is always a mapping into the initial user_namespace. | |
599 | * | |
600 | * If @kprojid has no mapping in @targ (projid_t)-1 is returned. | |
601 | */ | |
602 | projid_t from_kprojid(struct user_namespace *targ, kprojid_t kprojid) | |
603 | { | |
604 | /* Map the uid from a global kernel uid */ | |
605 | return map_id_up(&targ->projid_map, __kprojid_val(kprojid)); | |
606 | } | |
607 | EXPORT_SYMBOL(from_kprojid); | |
608 | ||
609 | /** | |
610 | * from_kprojid_munged - Create a projiid from a kprojid user-namespace pair. | |
611 | * @targ: The user namespace we want a projid in. | |
612 | * @kprojid: The kernel internal projid to start with. | |
613 | * | |
614 | * Map @kprojid into the user-namespace specified by @targ and | |
615 | * return the resulting projid. | |
616 | * | |
617 | * There is always a mapping into the initial user_namespace. | |
618 | * | |
619 | * Unlike from_kprojid from_kprojid_munged never fails and always | |
620 | * returns a valid projid. This makes from_kprojid_munged | |
621 | * appropriate for use in syscalls like stat and where | |
622 | * failing the system call and failing to provide a valid projid are | |
623 | * not an options. | |
624 | * | |
625 | * If @kprojid has no mapping in @targ OVERFLOW_PROJID is returned. | |
626 | */ | |
627 | projid_t from_kprojid_munged(struct user_namespace *targ, kprojid_t kprojid) | |
628 | { | |
629 | projid_t projid; | |
630 | projid = from_kprojid(targ, kprojid); | |
631 | ||
632 | if (projid == (projid_t) -1) | |
633 | projid = OVERFLOW_PROJID; | |
634 | return projid; | |
635 | } | |
636 | EXPORT_SYMBOL(from_kprojid_munged); | |
637 | ||
638 | ||
22d917d8 EB |
639 | static int uid_m_show(struct seq_file *seq, void *v) |
640 | { | |
641 | struct user_namespace *ns = seq->private; | |
642 | struct uid_gid_extent *extent = v; | |
643 | struct user_namespace *lower_ns; | |
644 | uid_t lower; | |
5c1469de | 645 | |
c450f371 | 646 | lower_ns = seq_user_ns(seq); |
22d917d8 EB |
647 | if ((lower_ns == ns) && lower_ns->parent) |
648 | lower_ns = lower_ns->parent; | |
649 | ||
650 | lower = from_kuid(lower_ns, KUIDT_INIT(extent->lower_first)); | |
651 | ||
652 | seq_printf(seq, "%10u %10u %10u\n", | |
653 | extent->first, | |
654 | lower, | |
655 | extent->count); | |
656 | ||
657 | return 0; | |
5c1469de EB |
658 | } |
659 | ||
22d917d8 | 660 | static int gid_m_show(struct seq_file *seq, void *v) |
5c1469de | 661 | { |
22d917d8 EB |
662 | struct user_namespace *ns = seq->private; |
663 | struct uid_gid_extent *extent = v; | |
664 | struct user_namespace *lower_ns; | |
665 | gid_t lower; | |
5c1469de | 666 | |
c450f371 | 667 | lower_ns = seq_user_ns(seq); |
22d917d8 EB |
668 | if ((lower_ns == ns) && lower_ns->parent) |
669 | lower_ns = lower_ns->parent; | |
5c1469de | 670 | |
22d917d8 EB |
671 | lower = from_kgid(lower_ns, KGIDT_INIT(extent->lower_first)); |
672 | ||
673 | seq_printf(seq, "%10u %10u %10u\n", | |
674 | extent->first, | |
675 | lower, | |
676 | extent->count); | |
677 | ||
678 | return 0; | |
679 | } | |
680 | ||
f76d207a EB |
681 | static int projid_m_show(struct seq_file *seq, void *v) |
682 | { | |
683 | struct user_namespace *ns = seq->private; | |
684 | struct uid_gid_extent *extent = v; | |
685 | struct user_namespace *lower_ns; | |
686 | projid_t lower; | |
687 | ||
688 | lower_ns = seq_user_ns(seq); | |
689 | if ((lower_ns == ns) && lower_ns->parent) | |
690 | lower_ns = lower_ns->parent; | |
691 | ||
692 | lower = from_kprojid(lower_ns, KPROJIDT_INIT(extent->lower_first)); | |
693 | ||
694 | seq_printf(seq, "%10u %10u %10u\n", | |
695 | extent->first, | |
696 | lower, | |
697 | extent->count); | |
698 | ||
699 | return 0; | |
700 | } | |
701 | ||
68a9a435 FF |
702 | static void *m_start(struct seq_file *seq, loff_t *ppos, |
703 | struct uid_gid_map *map) | |
22d917d8 | 704 | { |
22d917d8 EB |
705 | loff_t pos = *ppos; |
706 | ||
6397fac4 CB |
707 | if (pos >= map->nr_extents) |
708 | return NULL; | |
22d917d8 | 709 | |
6397fac4 CB |
710 | if (map->nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) |
711 | return &map->extent[pos]; | |
712 | ||
713 | return &map->forward[pos]; | |
22d917d8 EB |
714 | } |
715 | ||
716 | static void *uid_m_start(struct seq_file *seq, loff_t *ppos) | |
717 | { | |
718 | struct user_namespace *ns = seq->private; | |
719 | ||
720 | return m_start(seq, ppos, &ns->uid_map); | |
721 | } | |
722 | ||
723 | static void *gid_m_start(struct seq_file *seq, loff_t *ppos) | |
724 | { | |
725 | struct user_namespace *ns = seq->private; | |
726 | ||
727 | return m_start(seq, ppos, &ns->gid_map); | |
728 | } | |
729 | ||
f76d207a EB |
730 | static void *projid_m_start(struct seq_file *seq, loff_t *ppos) |
731 | { | |
732 | struct user_namespace *ns = seq->private; | |
733 | ||
734 | return m_start(seq, ppos, &ns->projid_map); | |
735 | } | |
736 | ||
22d917d8 EB |
737 | static void *m_next(struct seq_file *seq, void *v, loff_t *pos) |
738 | { | |
739 | (*pos)++; | |
740 | return seq->op->start(seq, pos); | |
741 | } | |
742 | ||
743 | static void m_stop(struct seq_file *seq, void *v) | |
744 | { | |
745 | return; | |
746 | } | |
747 | ||
ccf94f1b | 748 | const struct seq_operations proc_uid_seq_operations = { |
22d917d8 EB |
749 | .start = uid_m_start, |
750 | .stop = m_stop, | |
751 | .next = m_next, | |
752 | .show = uid_m_show, | |
753 | }; | |
754 | ||
ccf94f1b | 755 | const struct seq_operations proc_gid_seq_operations = { |
22d917d8 EB |
756 | .start = gid_m_start, |
757 | .stop = m_stop, | |
758 | .next = m_next, | |
759 | .show = gid_m_show, | |
760 | }; | |
761 | ||
ccf94f1b | 762 | const struct seq_operations proc_projid_seq_operations = { |
f76d207a EB |
763 | .start = projid_m_start, |
764 | .stop = m_stop, | |
765 | .next = m_next, | |
766 | .show = projid_m_show, | |
767 | }; | |
768 | ||
68a9a435 FF |
769 | static bool mappings_overlap(struct uid_gid_map *new_map, |
770 | struct uid_gid_extent *extent) | |
0bd14b4f EB |
771 | { |
772 | u32 upper_first, lower_first, upper_last, lower_last; | |
773 | unsigned idx; | |
774 | ||
775 | upper_first = extent->first; | |
776 | lower_first = extent->lower_first; | |
777 | upper_last = upper_first + extent->count - 1; | |
778 | lower_last = lower_first + extent->count - 1; | |
779 | ||
780 | for (idx = 0; idx < new_map->nr_extents; idx++) { | |
781 | u32 prev_upper_first, prev_lower_first; | |
782 | u32 prev_upper_last, prev_lower_last; | |
783 | struct uid_gid_extent *prev; | |
784 | ||
6397fac4 CB |
785 | if (new_map->nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) |
786 | prev = &new_map->extent[idx]; | |
787 | else | |
788 | prev = &new_map->forward[idx]; | |
0bd14b4f EB |
789 | |
790 | prev_upper_first = prev->first; | |
791 | prev_lower_first = prev->lower_first; | |
792 | prev_upper_last = prev_upper_first + prev->count - 1; | |
793 | prev_lower_last = prev_lower_first + prev->count - 1; | |
794 | ||
795 | /* Does the upper range intersect a previous extent? */ | |
796 | if ((prev_upper_first <= upper_last) && | |
797 | (prev_upper_last >= upper_first)) | |
798 | return true; | |
799 | ||
800 | /* Does the lower range intersect a previous extent? */ | |
801 | if ((prev_lower_first <= lower_last) && | |
802 | (prev_lower_last >= lower_first)) | |
803 | return true; | |
804 | } | |
805 | return false; | |
806 | } | |
807 | ||
6397fac4 CB |
808 | /** |
809 | * insert_extent - Safely insert a new idmap extent into struct uid_gid_map. | |
810 | * Takes care to allocate a 4K block of memory if the number of mappings exceeds | |
811 | * UID_GID_MAP_MAX_BASE_EXTENTS. | |
812 | */ | |
813 | static int insert_extent(struct uid_gid_map *map, struct uid_gid_extent *extent) | |
814 | { | |
815 | if (map->nr_extents < UID_GID_MAP_MAX_BASE_EXTENTS) { | |
816 | map->extent[map->nr_extents].first = extent->first; | |
817 | map->extent[map->nr_extents].lower_first = extent->lower_first; | |
818 | map->extent[map->nr_extents].count = extent->count; | |
819 | return 0; | |
820 | } | |
821 | ||
822 | if (map->nr_extents == UID_GID_MAP_MAX_BASE_EXTENTS) { | |
823 | struct uid_gid_extent *forward; | |
824 | ||
825 | /* Allocate memory for 340 mappings. */ | |
826 | forward = kmalloc(sizeof(struct uid_gid_extent) * | |
827 | UID_GID_MAP_MAX_EXTENTS, GFP_KERNEL); | |
828 | if (!forward) | |
829 | return -ENOMEM; | |
830 | ||
831 | /* Copy over memory. Only set up memory for the forward pointer. | |
832 | * Defer the memory setup for the reverse pointer. | |
833 | */ | |
834 | memcpy(forward, map->extent, | |
835 | map->nr_extents * sizeof(map->extent[0])); | |
836 | ||
837 | map->forward = forward; | |
838 | map->reverse = NULL; | |
839 | } | |
840 | ||
841 | map->forward[map->nr_extents].first = extent->first; | |
842 | map->forward[map->nr_extents].lower_first = extent->lower_first; | |
843 | map->forward[map->nr_extents].count = extent->count; | |
844 | return 0; | |
845 | } | |
846 | ||
847 | /* cmp function to sort() forward mappings */ | |
848 | static int cmp_extents_forward(const void *a, const void *b) | |
849 | { | |
850 | const struct uid_gid_extent *e1 = a; | |
851 | const struct uid_gid_extent *e2 = b; | |
852 | ||
853 | if (e1->first < e2->first) | |
854 | return -1; | |
855 | ||
856 | if (e1->first > e2->first) | |
857 | return 1; | |
858 | ||
859 | return 0; | |
860 | } | |
861 | ||
862 | /* cmp function to sort() reverse mappings */ | |
863 | static int cmp_extents_reverse(const void *a, const void *b) | |
864 | { | |
865 | const struct uid_gid_extent *e1 = a; | |
866 | const struct uid_gid_extent *e2 = b; | |
867 | ||
868 | if (e1->lower_first < e2->lower_first) | |
869 | return -1; | |
870 | ||
871 | if (e1->lower_first > e2->lower_first) | |
872 | return 1; | |
873 | ||
874 | return 0; | |
875 | } | |
876 | ||
877 | /** | |
878 | * sort_idmaps - Sorts an array of idmap entries. | |
879 | * Can only be called if number of mappings exceeds UID_GID_MAP_MAX_BASE_EXTENTS. | |
880 | */ | |
881 | static int sort_idmaps(struct uid_gid_map *map) | |
882 | { | |
883 | if (map->nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) | |
884 | return 0; | |
885 | ||
886 | /* Sort forward array. */ | |
887 | sort(map->forward, map->nr_extents, sizeof(struct uid_gid_extent), | |
888 | cmp_extents_forward, NULL); | |
889 | ||
890 | /* Only copy the memory from forward we actually need. */ | |
891 | map->reverse = kmemdup(map->forward, | |
892 | map->nr_extents * sizeof(struct uid_gid_extent), | |
893 | GFP_KERNEL); | |
894 | if (!map->reverse) | |
895 | return -ENOMEM; | |
896 | ||
897 | /* Sort reverse array. */ | |
898 | sort(map->reverse, map->nr_extents, sizeof(struct uid_gid_extent), | |
899 | cmp_extents_reverse, NULL); | |
900 | ||
901 | return 0; | |
902 | } | |
903 | ||
22d917d8 EB |
904 | static ssize_t map_write(struct file *file, const char __user *buf, |
905 | size_t count, loff_t *ppos, | |
906 | int cap_setid, | |
907 | struct uid_gid_map *map, | |
908 | struct uid_gid_map *parent_map) | |
909 | { | |
910 | struct seq_file *seq = file->private_data; | |
911 | struct user_namespace *ns = seq->private; | |
912 | struct uid_gid_map new_map; | |
913 | unsigned idx; | |
6397fac4 | 914 | struct uid_gid_extent extent; |
70f6cbb6 | 915 | char *kbuf = NULL, *pos, *next_line; |
22d917d8 EB |
916 | ssize_t ret = -EINVAL; |
917 | ||
918 | /* | |
f0d62aec | 919 | * The userns_state_mutex serializes all writes to any given map. |
22d917d8 EB |
920 | * |
921 | * Any map is only ever written once. | |
922 | * | |
923 | * An id map fits within 1 cache line on most architectures. | |
924 | * | |
925 | * On read nothing needs to be done unless you are on an | |
926 | * architecture with a crazy cache coherency model like alpha. | |
927 | * | |
928 | * There is a one time data dependency between reading the | |
929 | * count of the extents and the values of the extents. The | |
930 | * desired behavior is to see the values of the extents that | |
931 | * were written before the count of the extents. | |
932 | * | |
933 | * To achieve this smp_wmb() is used on guarantee the write | |
e79323bd MP |
934 | * order and smp_rmb() is guaranteed that we don't have crazy |
935 | * architectures returning stale data. | |
22d917d8 | 936 | */ |
f0d62aec | 937 | mutex_lock(&userns_state_mutex); |
22d917d8 | 938 | |
6397fac4 CB |
939 | memset(&new_map, 0, sizeof(struct uid_gid_map)); |
940 | ||
22d917d8 EB |
941 | ret = -EPERM; |
942 | /* Only allow one successful write to the map */ | |
943 | if (map->nr_extents != 0) | |
944 | goto out; | |
945 | ||
41c21e35 AL |
946 | /* |
947 | * Adjusting namespace settings requires capabilities on the target. | |
5c1469de | 948 | */ |
41c21e35 | 949 | if (cap_valid(cap_setid) && !file_ns_capable(file, ns, CAP_SYS_ADMIN)) |
22d917d8 EB |
950 | goto out; |
951 | ||
36476bea | 952 | /* Only allow < page size writes at the beginning of the file */ |
22d917d8 EB |
953 | ret = -EINVAL; |
954 | if ((*ppos != 0) || (count >= PAGE_SIZE)) | |
955 | goto out; | |
956 | ||
957 | /* Slurp in the user data */ | |
70f6cbb6 AV |
958 | kbuf = memdup_user_nul(buf, count); |
959 | if (IS_ERR(kbuf)) { | |
960 | ret = PTR_ERR(kbuf); | |
961 | kbuf = NULL; | |
22d917d8 | 962 | goto out; |
70f6cbb6 | 963 | } |
22d917d8 EB |
964 | |
965 | /* Parse the user data */ | |
966 | ret = -EINVAL; | |
967 | pos = kbuf; | |
68a9a435 | 968 | for (; pos; pos = next_line) { |
22d917d8 EB |
969 | |
970 | /* Find the end of line and ensure I don't look past it */ | |
971 | next_line = strchr(pos, '\n'); | |
972 | if (next_line) { | |
973 | *next_line = '\0'; | |
974 | next_line++; | |
975 | if (*next_line == '\0') | |
976 | next_line = NULL; | |
5c1469de | 977 | } |
22d917d8 EB |
978 | |
979 | pos = skip_spaces(pos); | |
6397fac4 | 980 | extent.first = simple_strtoul(pos, &pos, 10); |
22d917d8 EB |
981 | if (!isspace(*pos)) |
982 | goto out; | |
983 | ||
984 | pos = skip_spaces(pos); | |
6397fac4 | 985 | extent.lower_first = simple_strtoul(pos, &pos, 10); |
22d917d8 EB |
986 | if (!isspace(*pos)) |
987 | goto out; | |
988 | ||
989 | pos = skip_spaces(pos); | |
6397fac4 | 990 | extent.count = simple_strtoul(pos, &pos, 10); |
22d917d8 EB |
991 | if (*pos && !isspace(*pos)) |
992 | goto out; | |
993 | ||
994 | /* Verify there is not trailing junk on the line */ | |
995 | pos = skip_spaces(pos); | |
996 | if (*pos != '\0') | |
997 | goto out; | |
998 | ||
999 | /* Verify we have been given valid starting values */ | |
6397fac4 CB |
1000 | if ((extent.first == (u32) -1) || |
1001 | (extent.lower_first == (u32) -1)) | |
22d917d8 EB |
1002 | goto out; |
1003 | ||
68a9a435 FF |
1004 | /* Verify count is not zero and does not cause the |
1005 | * extent to wrap | |
1006 | */ | |
6397fac4 | 1007 | if ((extent.first + extent.count) <= extent.first) |
22d917d8 | 1008 | goto out; |
6397fac4 CB |
1009 | if ((extent.lower_first + extent.count) <= |
1010 | extent.lower_first) | |
22d917d8 EB |
1011 | goto out; |
1012 | ||
0bd14b4f | 1013 | /* Do the ranges in extent overlap any previous extents? */ |
6397fac4 | 1014 | if (mappings_overlap(&new_map, &extent)) |
22d917d8 EB |
1015 | goto out; |
1016 | ||
6397fac4 | 1017 | if ((new_map.nr_extents + 1) == UID_GID_MAP_MAX_EXTENTS && |
22d917d8 EB |
1018 | (next_line != NULL)) |
1019 | goto out; | |
6397fac4 CB |
1020 | |
1021 | ret = insert_extent(&new_map, &extent); | |
1022 | if (ret < 0) | |
1023 | goto out; | |
1024 | ret = -EINVAL; | |
1025 | ||
1026 | new_map.nr_extents++; | |
5c1469de | 1027 | } |
22d917d8 EB |
1028 | /* Be very certaint the new map actually exists */ |
1029 | if (new_map.nr_extents == 0) | |
1030 | goto out; | |
1031 | ||
1032 | ret = -EPERM; | |
1033 | /* Validate the user is allowed to use user id's mapped to. */ | |
6708075f | 1034 | if (!new_idmap_permitted(file, ns, cap_setid, &new_map)) |
22d917d8 EB |
1035 | goto out; |
1036 | ||
6397fac4 CB |
1037 | ret = sort_idmaps(&new_map); |
1038 | if (ret < 0) | |
1039 | goto out; | |
1040 | ||
1041 | ret = -EPERM; | |
22d917d8 EB |
1042 | /* Map the lower ids from the parent user namespace to the |
1043 | * kernel global id space. | |
1044 | */ | |
1045 | for (idx = 0; idx < new_map.nr_extents; idx++) { | |
6397fac4 | 1046 | struct uid_gid_extent *e; |
22d917d8 | 1047 | u32 lower_first; |
6397fac4 CB |
1048 | |
1049 | if (new_map.nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) | |
1050 | e = &new_map.extent[idx]; | |
1051 | else | |
1052 | e = &new_map.forward[idx]; | |
22d917d8 EB |
1053 | |
1054 | lower_first = map_id_range_down(parent_map, | |
6397fac4 CB |
1055 | e->lower_first, |
1056 | e->count); | |
22d917d8 EB |
1057 | |
1058 | /* Fail if we can not map the specified extent to | |
1059 | * the kernel global id space. | |
1060 | */ | |
1061 | if (lower_first == (u32) -1) | |
1062 | goto out; | |
1063 | ||
6397fac4 | 1064 | e->lower_first = lower_first; |
22d917d8 EB |
1065 | } |
1066 | ||
1067 | /* Install the map */ | |
6397fac4 CB |
1068 | if (new_map.nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) { |
1069 | memcpy(map->extent, new_map.extent, | |
1070 | new_map.nr_extents * sizeof(new_map.extent[0])); | |
1071 | } else { | |
1072 | map->forward = new_map.forward; | |
1073 | map->reverse = new_map.reverse; | |
1074 | } | |
22d917d8 EB |
1075 | smp_wmb(); |
1076 | map->nr_extents = new_map.nr_extents; | |
1077 | ||
1078 | *ppos = count; | |
1079 | ret = count; | |
1080 | out: | |
6397fac4 CB |
1081 | if (ret < 0 && new_map.nr_extents > UID_GID_MAP_MAX_BASE_EXTENTS) { |
1082 | kfree(new_map.forward); | |
1083 | kfree(new_map.reverse); | |
1084 | map->forward = NULL; | |
1085 | map->reverse = NULL; | |
1086 | map->nr_extents = 0; | |
1087 | } | |
1088 | ||
f0d62aec | 1089 | mutex_unlock(&userns_state_mutex); |
70f6cbb6 | 1090 | kfree(kbuf); |
22d917d8 EB |
1091 | return ret; |
1092 | } | |
1093 | ||
68a9a435 FF |
1094 | ssize_t proc_uid_map_write(struct file *file, const char __user *buf, |
1095 | size_t size, loff_t *ppos) | |
22d917d8 EB |
1096 | { |
1097 | struct seq_file *seq = file->private_data; | |
1098 | struct user_namespace *ns = seq->private; | |
c450f371 | 1099 | struct user_namespace *seq_ns = seq_user_ns(seq); |
22d917d8 EB |
1100 | |
1101 | if (!ns->parent) | |
1102 | return -EPERM; | |
1103 | ||
c450f371 EB |
1104 | if ((seq_ns != ns) && (seq_ns != ns->parent)) |
1105 | return -EPERM; | |
1106 | ||
22d917d8 EB |
1107 | return map_write(file, buf, size, ppos, CAP_SETUID, |
1108 | &ns->uid_map, &ns->parent->uid_map); | |
1109 | } | |
1110 | ||
68a9a435 FF |
1111 | ssize_t proc_gid_map_write(struct file *file, const char __user *buf, |
1112 | size_t size, loff_t *ppos) | |
22d917d8 EB |
1113 | { |
1114 | struct seq_file *seq = file->private_data; | |
1115 | struct user_namespace *ns = seq->private; | |
c450f371 | 1116 | struct user_namespace *seq_ns = seq_user_ns(seq); |
22d917d8 EB |
1117 | |
1118 | if (!ns->parent) | |
1119 | return -EPERM; | |
1120 | ||
c450f371 EB |
1121 | if ((seq_ns != ns) && (seq_ns != ns->parent)) |
1122 | return -EPERM; | |
1123 | ||
22d917d8 EB |
1124 | return map_write(file, buf, size, ppos, CAP_SETGID, |
1125 | &ns->gid_map, &ns->parent->gid_map); | |
1126 | } | |
1127 | ||
68a9a435 FF |
1128 | ssize_t proc_projid_map_write(struct file *file, const char __user *buf, |
1129 | size_t size, loff_t *ppos) | |
f76d207a EB |
1130 | { |
1131 | struct seq_file *seq = file->private_data; | |
1132 | struct user_namespace *ns = seq->private; | |
1133 | struct user_namespace *seq_ns = seq_user_ns(seq); | |
1134 | ||
1135 | if (!ns->parent) | |
1136 | return -EPERM; | |
1137 | ||
1138 | if ((seq_ns != ns) && (seq_ns != ns->parent)) | |
1139 | return -EPERM; | |
1140 | ||
1141 | /* Anyone can set any valid project id no capability needed */ | |
1142 | return map_write(file, buf, size, ppos, -1, | |
1143 | &ns->projid_map, &ns->parent->projid_map); | |
1144 | } | |
1145 | ||
68a9a435 | 1146 | static bool new_idmap_permitted(const struct file *file, |
6708075f | 1147 | struct user_namespace *ns, int cap_setid, |
22d917d8 EB |
1148 | struct uid_gid_map *new_map) |
1149 | { | |
f95d7918 | 1150 | const struct cred *cred = file->f_cred; |
0542f17b EB |
1151 | /* Don't allow mappings that would allow anything that wouldn't |
1152 | * be allowed without the establishment of unprivileged mappings. | |
1153 | */ | |
f95d7918 EB |
1154 | if ((new_map->nr_extents == 1) && (new_map->extent[0].count == 1) && |
1155 | uid_eq(ns->owner, cred->euid)) { | |
37657da3 EB |
1156 | u32 id = new_map->extent[0].lower_first; |
1157 | if (cap_setid == CAP_SETUID) { | |
1158 | kuid_t uid = make_kuid(ns->parent, id); | |
f95d7918 | 1159 | if (uid_eq(uid, cred->euid)) |
37657da3 | 1160 | return true; |
68a9a435 | 1161 | } else if (cap_setid == CAP_SETGID) { |
37657da3 | 1162 | kgid_t gid = make_kgid(ns->parent, id); |
66d2f338 EB |
1163 | if (!(ns->flags & USERNS_SETGROUPS_ALLOWED) && |
1164 | gid_eq(gid, cred->egid)) | |
37657da3 EB |
1165 | return true; |
1166 | } | |
1167 | } | |
1168 | ||
f76d207a EB |
1169 | /* Allow anyone to set a mapping that doesn't require privilege */ |
1170 | if (!cap_valid(cap_setid)) | |
1171 | return true; | |
1172 | ||
22d917d8 EB |
1173 | /* Allow the specified ids if we have the appropriate capability |
1174 | * (CAP_SETUID or CAP_SETGID) over the parent user namespace. | |
6708075f | 1175 | * And the opener of the id file also had the approprpiate capability. |
22d917d8 | 1176 | */ |
6708075f EB |
1177 | if (ns_capable(ns->parent, cap_setid) && |
1178 | file_ns_capable(file, ns->parent, cap_setid)) | |
22d917d8 | 1179 | return true; |
5c1469de | 1180 | |
22d917d8 | 1181 | return false; |
5c1469de | 1182 | } |
6164281a | 1183 | |
9cc46516 EB |
1184 | int proc_setgroups_show(struct seq_file *seq, void *v) |
1185 | { | |
1186 | struct user_namespace *ns = seq->private; | |
1187 | unsigned long userns_flags = ACCESS_ONCE(ns->flags); | |
1188 | ||
1189 | seq_printf(seq, "%s\n", | |
1190 | (userns_flags & USERNS_SETGROUPS_ALLOWED) ? | |
1191 | "allow" : "deny"); | |
1192 | return 0; | |
1193 | } | |
1194 | ||
1195 | ssize_t proc_setgroups_write(struct file *file, const char __user *buf, | |
1196 | size_t count, loff_t *ppos) | |
1197 | { | |
1198 | struct seq_file *seq = file->private_data; | |
1199 | struct user_namespace *ns = seq->private; | |
1200 | char kbuf[8], *pos; | |
1201 | bool setgroups_allowed; | |
1202 | ssize_t ret; | |
1203 | ||
1204 | /* Only allow a very narrow range of strings to be written */ | |
1205 | ret = -EINVAL; | |
1206 | if ((*ppos != 0) || (count >= sizeof(kbuf))) | |
1207 | goto out; | |
1208 | ||
1209 | /* What was written? */ | |
1210 | ret = -EFAULT; | |
1211 | if (copy_from_user(kbuf, buf, count)) | |
1212 | goto out; | |
1213 | kbuf[count] = '\0'; | |
1214 | pos = kbuf; | |
1215 | ||
1216 | /* What is being requested? */ | |
1217 | ret = -EINVAL; | |
1218 | if (strncmp(pos, "allow", 5) == 0) { | |
1219 | pos += 5; | |
1220 | setgroups_allowed = true; | |
1221 | } | |
1222 | else if (strncmp(pos, "deny", 4) == 0) { | |
1223 | pos += 4; | |
1224 | setgroups_allowed = false; | |
1225 | } | |
1226 | else | |
1227 | goto out; | |
1228 | ||
1229 | /* Verify there is not trailing junk on the line */ | |
1230 | pos = skip_spaces(pos); | |
1231 | if (*pos != '\0') | |
1232 | goto out; | |
1233 | ||
1234 | ret = -EPERM; | |
1235 | mutex_lock(&userns_state_mutex); | |
1236 | if (setgroups_allowed) { | |
1237 | /* Enabling setgroups after setgroups has been disabled | |
1238 | * is not allowed. | |
1239 | */ | |
1240 | if (!(ns->flags & USERNS_SETGROUPS_ALLOWED)) | |
1241 | goto out_unlock; | |
1242 | } else { | |
1243 | /* Permanently disabling setgroups after setgroups has | |
1244 | * been enabled by writing the gid_map is not allowed. | |
1245 | */ | |
1246 | if (ns->gid_map.nr_extents != 0) | |
1247 | goto out_unlock; | |
1248 | ns->flags &= ~USERNS_SETGROUPS_ALLOWED; | |
1249 | } | |
1250 | mutex_unlock(&userns_state_mutex); | |
1251 | ||
1252 | /* Report a successful write */ | |
1253 | *ppos = count; | |
1254 | ret = count; | |
1255 | out: | |
1256 | return ret; | |
1257 | out_unlock: | |
1258 | mutex_unlock(&userns_state_mutex); | |
1259 | goto out; | |
1260 | } | |
1261 | ||
273d2c67 EB |
1262 | bool userns_may_setgroups(const struct user_namespace *ns) |
1263 | { | |
1264 | bool allowed; | |
1265 | ||
f0d62aec | 1266 | mutex_lock(&userns_state_mutex); |
273d2c67 EB |
1267 | /* It is not safe to use setgroups until a gid mapping in |
1268 | * the user namespace has been established. | |
1269 | */ | |
1270 | allowed = ns->gid_map.nr_extents != 0; | |
9cc46516 EB |
1271 | /* Is setgroups allowed? */ |
1272 | allowed = allowed && (ns->flags & USERNS_SETGROUPS_ALLOWED); | |
f0d62aec | 1273 | mutex_unlock(&userns_state_mutex); |
273d2c67 EB |
1274 | |
1275 | return allowed; | |
1276 | } | |
1277 | ||
d07b846f | 1278 | /* |
a2b42626 EB |
1279 | * Returns true if @child is the same namespace or a descendant of |
1280 | * @ancestor. | |
d07b846f | 1281 | */ |
a2b42626 EB |
1282 | bool in_userns(const struct user_namespace *ancestor, |
1283 | const struct user_namespace *child) | |
1284 | { | |
1285 | const struct user_namespace *ns; | |
1286 | for (ns = child; ns->level > ancestor->level; ns = ns->parent) | |
1287 | ; | |
1288 | return (ns == ancestor); | |
1289 | } | |
1290 | ||
d07b846f SF |
1291 | bool current_in_userns(const struct user_namespace *target_ns) |
1292 | { | |
a2b42626 | 1293 | return in_userns(target_ns, current_user_ns()); |
d07b846f SF |
1294 | } |
1295 | ||
3c041184 AV |
1296 | static inline struct user_namespace *to_user_ns(struct ns_common *ns) |
1297 | { | |
1298 | return container_of(ns, struct user_namespace, ns); | |
1299 | } | |
1300 | ||
64964528 | 1301 | static struct ns_common *userns_get(struct task_struct *task) |
cde1975b EB |
1302 | { |
1303 | struct user_namespace *user_ns; | |
1304 | ||
1305 | rcu_read_lock(); | |
1306 | user_ns = get_user_ns(__task_cred(task)->user_ns); | |
1307 | rcu_read_unlock(); | |
1308 | ||
3c041184 | 1309 | return user_ns ? &user_ns->ns : NULL; |
cde1975b EB |
1310 | } |
1311 | ||
64964528 | 1312 | static void userns_put(struct ns_common *ns) |
cde1975b | 1313 | { |
3c041184 | 1314 | put_user_ns(to_user_ns(ns)); |
cde1975b EB |
1315 | } |
1316 | ||
64964528 | 1317 | static int userns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
cde1975b | 1318 | { |
3c041184 | 1319 | struct user_namespace *user_ns = to_user_ns(ns); |
cde1975b EB |
1320 | struct cred *cred; |
1321 | ||
1322 | /* Don't allow gaining capabilities by reentering | |
1323 | * the same user namespace. | |
1324 | */ | |
1325 | if (user_ns == current_user_ns()) | |
1326 | return -EINVAL; | |
1327 | ||
faf00da5 EB |
1328 | /* Tasks that share a thread group must share a user namespace */ |
1329 | if (!thread_group_empty(current)) | |
cde1975b EB |
1330 | return -EINVAL; |
1331 | ||
e66eded8 EB |
1332 | if (current->fs->users != 1) |
1333 | return -EINVAL; | |
1334 | ||
cde1975b EB |
1335 | if (!ns_capable(user_ns, CAP_SYS_ADMIN)) |
1336 | return -EPERM; | |
1337 | ||
1338 | cred = prepare_creds(); | |
1339 | if (!cred) | |
1340 | return -ENOMEM; | |
1341 | ||
1342 | put_user_ns(cred->user_ns); | |
1343 | set_cred_user_ns(cred, get_user_ns(user_ns)); | |
1344 | ||
1345 | return commit_creds(cred); | |
1346 | } | |
1347 | ||
bcac25a5 AV |
1348 | struct ns_common *ns_get_owner(struct ns_common *ns) |
1349 | { | |
1350 | struct user_namespace *my_user_ns = current_user_ns(); | |
1351 | struct user_namespace *owner, *p; | |
1352 | ||
1353 | /* See if the owner is in the current user namespace */ | |
1354 | owner = p = ns->ops->owner(ns); | |
1355 | for (;;) { | |
1356 | if (!p) | |
1357 | return ERR_PTR(-EPERM); | |
1358 | if (p == my_user_ns) | |
1359 | break; | |
1360 | p = p->parent; | |
1361 | } | |
1362 | ||
1363 | return &get_user_ns(owner)->ns; | |
1364 | } | |
1365 | ||
1366 | static struct user_namespace *userns_owner(struct ns_common *ns) | |
1367 | { | |
1368 | return to_user_ns(ns)->parent; | |
1369 | } | |
1370 | ||
cde1975b EB |
1371 | const struct proc_ns_operations userns_operations = { |
1372 | .name = "user", | |
1373 | .type = CLONE_NEWUSER, | |
1374 | .get = userns_get, | |
1375 | .put = userns_put, | |
1376 | .install = userns_install, | |
bcac25a5 | 1377 | .owner = userns_owner, |
a7306ed8 | 1378 | .get_parent = ns_get_owner, |
cde1975b EB |
1379 | }; |
1380 | ||
6164281a PE |
1381 | static __init int user_namespaces_init(void) |
1382 | { | |
1383 | user_ns_cachep = KMEM_CACHE(user_namespace, SLAB_PANIC); | |
1384 | return 0; | |
1385 | } | |
c96d6660 | 1386 | subsys_initcall(user_namespaces_init); |