]>
Commit | Line | Data |
---|---|---|
b886d83c | 1 | // SPDX-License-Identifier: GPL-2.0-only |
acce292c | 2 | |
9984de1a | 3 | #include <linux/export.h> |
acce292c | 4 | #include <linux/nsproxy.h> |
1aeb272c | 5 | #include <linux/slab.h> |
3f07c014 | 6 | #include <linux/sched/signal.h> |
acce292c | 7 | #include <linux/user_namespace.h> |
0bb80f24 | 8 | #include <linux/proc_ns.h> |
5c1469de | 9 | #include <linux/highuid.h> |
18b6e041 | 10 | #include <linux/cred.h> |
973c5914 | 11 | #include <linux/securebits.h> |
22d917d8 EB |
12 | #include <linux/keyctl.h> |
13 | #include <linux/key-type.h> | |
14 | #include <keys/user-type.h> | |
15 | #include <linux/seq_file.h> | |
16 | #include <linux/fs.h> | |
17 | #include <linux/uaccess.h> | |
18 | #include <linux/ctype.h> | |
f76d207a | 19 | #include <linux/projid.h> |
e66eded8 | 20 | #include <linux/fs_struct.h> |
6397fac4 CB |
21 | #include <linux/bsearch.h> |
22 | #include <linux/sort.h> | |
acce292c | 23 | |
34227646 SH |
24 | /* |
25 | * sysctl determining whether unprivileged users may unshare a new | |
26 | * userns. Allowed by default | |
27 | */ | |
28 | int unprivileged_userns_clone = 1; | |
29 | ||
6164281a | 30 | static struct kmem_cache *user_ns_cachep __read_mostly; |
f0d62aec | 31 | static DEFINE_MUTEX(userns_state_mutex); |
6164281a | 32 | |
6708075f EB |
33 | static bool new_idmap_permitted(const struct file *file, |
34 | struct user_namespace *ns, int cap_setid, | |
22d917d8 | 35 | struct uid_gid_map *map); |
b032132c | 36 | static void free_user_ns(struct work_struct *work); |
22d917d8 | 37 | |
25f9c081 EB |
38 | static struct ucounts *inc_user_namespaces(struct user_namespace *ns, kuid_t uid) |
39 | { | |
40 | return inc_ucount(ns, uid, UCOUNT_USER_NAMESPACES); | |
41 | } | |
42 | ||
43 | static void dec_user_namespaces(struct ucounts *ucounts) | |
44 | { | |
45 | return dec_ucount(ucounts, UCOUNT_USER_NAMESPACES); | |
46 | } | |
47 | ||
cde1975b EB |
48 | static void set_cred_user_ns(struct cred *cred, struct user_namespace *user_ns) |
49 | { | |
50 | /* Start with the same capabilities as init but useless for doing | |
51 | * anything as the capabilities are bound to the new user namespace. | |
52 | */ | |
53 | cred->securebits = SECUREBITS_DEFAULT; | |
54 | cred->cap_inheritable = CAP_EMPTY_SET; | |
55 | cred->cap_permitted = CAP_FULL_SET; | |
56 | cred->cap_effective = CAP_FULL_SET; | |
58319057 | 57 | cred->cap_ambient = CAP_EMPTY_SET; |
cde1975b EB |
58 | cred->cap_bset = CAP_FULL_SET; |
59 | #ifdef CONFIG_KEYS | |
60 | key_put(cred->request_key_auth); | |
61 | cred->request_key_auth = NULL; | |
62 | #endif | |
63 | /* tgcred will be cleared in our caller bc CLONE_THREAD won't be set */ | |
64 | cred->user_ns = user_ns; | |
65 | } | |
66 | ||
77ec739d | 67 | /* |
18b6e041 SH |
68 | * Create a new user namespace, deriving the creator from the user in the |
69 | * passed credentials, and replacing that user with the new root user for the | |
70 | * new namespace. | |
71 | * | |
72 | * This is called by copy_creds(), which will finish setting the target task's | |
73 | * credentials. | |
77ec739d | 74 | */ |
18b6e041 | 75 | int create_user_ns(struct cred *new) |
77ec739d | 76 | { |
0093ccb6 | 77 | struct user_namespace *ns, *parent_ns = new->user_ns; |
078de5f7 EB |
78 | kuid_t owner = new->euid; |
79 | kgid_t group = new->egid; | |
f6b2db1a | 80 | struct ucounts *ucounts; |
25f9c081 | 81 | int ret, i; |
783291e6 | 82 | |
df75e774 | 83 | ret = -ENOSPC; |
8742f229 | 84 | if (parent_ns->level > 32) |
b376c3e1 EB |
85 | goto fail; |
86 | ||
f6b2db1a EB |
87 | ucounts = inc_user_namespaces(parent_ns, owner); |
88 | if (!ucounts) | |
b376c3e1 | 89 | goto fail; |
8742f229 | 90 | |
3151527e EB |
91 | /* |
92 | * Verify that we can not violate the policy of which files | |
93 | * may be accessed that is specified by the root directory, | |
a12f4f85 | 94 | * by verifying that the root directory is at the root of the |
3151527e EB |
95 | * mount namespace which allows all files to be accessed. |
96 | */ | |
b376c3e1 | 97 | ret = -EPERM; |
3151527e | 98 | if (current_chrooted()) |
b376c3e1 | 99 | goto fail_dec; |
3151527e | 100 | |
783291e6 EB |
101 | /* The creator needs a mapping in the parent user namespace |
102 | * or else we won't be able to reasonably tell userspace who | |
103 | * created a user_namespace. | |
104 | */ | |
b376c3e1 | 105 | ret = -EPERM; |
783291e6 EB |
106 | if (!kuid_has_mapping(parent_ns, owner) || |
107 | !kgid_has_mapping(parent_ns, group)) | |
b376c3e1 | 108 | goto fail_dec; |
77ec739d | 109 | |
b376c3e1 | 110 | ret = -ENOMEM; |
22d917d8 | 111 | ns = kmem_cache_zalloc(user_ns_cachep, GFP_KERNEL); |
77ec739d | 112 | if (!ns) |
b376c3e1 | 113 | goto fail_dec; |
77ec739d | 114 | |
db2e718a | 115 | ns->parent_could_setfcap = cap_raised(new->cap_effective, CAP_SETFCAP); |
6344c433 | 116 | ret = ns_alloc_inum(&ns->ns); |
b376c3e1 EB |
117 | if (ret) |
118 | goto fail_free; | |
33c42940 | 119 | ns->ns.ops = &userns_operations; |
98f842e6 | 120 | |
265cbd62 | 121 | refcount_set(&ns->ns.count, 1); |
cde1975b | 122 | /* Leave the new->user_ns reference with the new user namespace. */ |
aeb3ae9d | 123 | ns->parent = parent_ns; |
8742f229 | 124 | ns->level = parent_ns->level + 1; |
783291e6 EB |
125 | ns->owner = owner; |
126 | ns->group = group; | |
b032132c | 127 | INIT_WORK(&ns->work, free_user_ns); |
21d1c5e3 | 128 | for (i = 0; i < MAX_PER_NAMESPACE_UCOUNTS; i++) { |
25f9c081 EB |
129 | ns->ucount_max[i] = INT_MAX; |
130 | } | |
c1ada3dc AG |
131 | set_rlimit_ucount_max(ns, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)); |
132 | set_rlimit_ucount_max(ns, UCOUNT_RLIMIT_MSGQUEUE, rlimit(RLIMIT_MSGQUEUE)); | |
133 | set_rlimit_ucount_max(ns, UCOUNT_RLIMIT_SIGPENDING, rlimit(RLIMIT_SIGPENDING)); | |
134 | set_rlimit_ucount_max(ns, UCOUNT_RLIMIT_MEMLOCK, rlimit(RLIMIT_MEMLOCK)); | |
f6b2db1a | 135 | ns->ucounts = ucounts; |
22d917d8 | 136 | |
9cc46516 EB |
137 | /* Inherit USERNS_SETGROUPS_ALLOWED from our parent */ |
138 | mutex_lock(&userns_state_mutex); | |
139 | ns->flags = parent_ns->flags; | |
140 | mutex_unlock(&userns_state_mutex); | |
141 | ||
b206f281 DH |
142 | #ifdef CONFIG_KEYS |
143 | INIT_LIST_HEAD(&ns->keyring_name_list); | |
0f44e4d9 | 144 | init_rwsem(&ns->keyring_sem); |
f36f8c75 | 145 | #endif |
dbec2846 EB |
146 | ret = -ENOMEM; |
147 | if (!setup_userns_sysctls(ns)) | |
148 | goto fail_keyring; | |
149 | ||
150 | set_cred_user_ns(new, ns); | |
18b6e041 | 151 | return 0; |
dbec2846 EB |
152 | fail_keyring: |
153 | #ifdef CONFIG_PERSISTENT_KEYRINGS | |
154 | key_put(ns->persistent_keyring_register); | |
155 | #endif | |
156 | ns_free_inum(&ns->ns); | |
b376c3e1 | 157 | fail_free: |
dbec2846 | 158 | kmem_cache_free(user_ns_cachep, ns); |
b376c3e1 | 159 | fail_dec: |
f6b2db1a | 160 | dec_user_namespaces(ucounts); |
b376c3e1 | 161 | fail: |
dbec2846 | 162 | return ret; |
acce292c CLG |
163 | } |
164 | ||
b2e0d987 EB |
165 | int unshare_userns(unsigned long unshare_flags, struct cred **new_cred) |
166 | { | |
167 | struct cred *cred; | |
6160968c | 168 | int err = -ENOMEM; |
b2e0d987 EB |
169 | |
170 | if (!(unshare_flags & CLONE_NEWUSER)) | |
171 | return 0; | |
172 | ||
173 | cred = prepare_creds(); | |
6160968c ON |
174 | if (cred) { |
175 | err = create_user_ns(cred); | |
176 | if (err) | |
177 | put_cred(cred); | |
178 | else | |
179 | *new_cred = cred; | |
180 | } | |
b2e0d987 | 181 | |
6160968c | 182 | return err; |
b2e0d987 EB |
183 | } |
184 | ||
b032132c | 185 | static void free_user_ns(struct work_struct *work) |
acce292c | 186 | { |
b032132c EB |
187 | struct user_namespace *parent, *ns = |
188 | container_of(work, struct user_namespace, work); | |
783291e6 | 189 | |
c61a2810 | 190 | do { |
f6b2db1a | 191 | struct ucounts *ucounts = ns->ucounts; |
c61a2810 | 192 | parent = ns->parent; |
6397fac4 CB |
193 | if (ns->gid_map.nr_extents > UID_GID_MAP_MAX_BASE_EXTENTS) { |
194 | kfree(ns->gid_map.forward); | |
195 | kfree(ns->gid_map.reverse); | |
196 | } | |
197 | if (ns->uid_map.nr_extents > UID_GID_MAP_MAX_BASE_EXTENTS) { | |
198 | kfree(ns->uid_map.forward); | |
199 | kfree(ns->uid_map.reverse); | |
200 | } | |
201 | if (ns->projid_map.nr_extents > UID_GID_MAP_MAX_BASE_EXTENTS) { | |
202 | kfree(ns->projid_map.forward); | |
203 | kfree(ns->projid_map.reverse); | |
204 | } | |
dbec2846 | 205 | retire_userns_sysctls(ns); |
b206f281 | 206 | key_free_user_ns(ns); |
6344c433 | 207 | ns_free_inum(&ns->ns); |
c61a2810 | 208 | kmem_cache_free(user_ns_cachep, ns); |
f6b2db1a | 209 | dec_user_namespaces(ucounts); |
c61a2810 | 210 | ns = parent; |
265cbd62 | 211 | } while (refcount_dec_and_test(&parent->ns.count)); |
acce292c | 212 | } |
b032132c EB |
213 | |
214 | void __put_user_ns(struct user_namespace *ns) | |
215 | { | |
216 | schedule_work(&ns->work); | |
217 | } | |
218 | EXPORT_SYMBOL(__put_user_ns); | |
5c1469de | 219 | |
6397fac4 CB |
220 | /** |
221 | * idmap_key struct holds the information necessary to find an idmapping in a | |
222 | * sorted idmap array. It is passed to cmp_map_id() as first argument. | |
223 | */ | |
224 | struct idmap_key { | |
225 | bool map_up; /* true -> id from kid; false -> kid from id */ | |
226 | u32 id; /* id to find */ | |
227 | u32 count; /* == 0 unless used with map_id_range_down() */ | |
228 | }; | |
229 | ||
230 | /** | |
231 | * cmp_map_id - Function to be passed to bsearch() to find the requested | |
232 | * idmapping. Expects struct idmap_key to be passed via @k. | |
233 | */ | |
234 | static int cmp_map_id(const void *k, const void *e) | |
235 | { | |
236 | u32 first, last, id2; | |
237 | const struct idmap_key *key = k; | |
238 | const struct uid_gid_extent *el = e; | |
239 | ||
11a8b927 | 240 | id2 = key->id + key->count - 1; |
6397fac4 CB |
241 | |
242 | /* handle map_id_{down,up}() */ | |
243 | if (key->map_up) | |
244 | first = el->lower_first; | |
245 | else | |
246 | first = el->first; | |
247 | ||
248 | last = first + el->count - 1; | |
249 | ||
250 | if (key->id >= first && key->id <= last && | |
251 | (id2 >= first && id2 <= last)) | |
252 | return 0; | |
253 | ||
254 | if (key->id < first || id2 < first) | |
255 | return -1; | |
256 | ||
257 | return 1; | |
258 | } | |
259 | ||
260 | /** | |
261 | * map_id_range_down_max - Find idmap via binary search in ordered idmap array. | |
262 | * Can only be called if number of mappings exceeds UID_GID_MAP_MAX_BASE_EXTENTS. | |
263 | */ | |
3edf652f EB |
264 | static struct uid_gid_extent * |
265 | map_id_range_down_max(unsigned extents, struct uid_gid_map *map, u32 id, u32 count) | |
5c1469de | 266 | { |
6397fac4 CB |
267 | struct idmap_key key; |
268 | ||
269 | key.map_up = false; | |
270 | key.count = count; | |
271 | key.id = id; | |
272 | ||
3edf652f EB |
273 | return bsearch(&key, map->forward, extents, |
274 | sizeof(struct uid_gid_extent), cmp_map_id); | |
6397fac4 CB |
275 | } |
276 | ||
277 | /** | |
278 | * map_id_range_down_base - Find idmap via binary search in static extent array. | |
279 | * Can only be called if number of mappings is equal or less than | |
280 | * UID_GID_MAP_MAX_BASE_EXTENTS. | |
281 | */ | |
3edf652f EB |
282 | static struct uid_gid_extent * |
283 | map_id_range_down_base(unsigned extents, struct uid_gid_map *map, u32 id, u32 count) | |
5c1469de | 284 | { |
3edf652f | 285 | unsigned idx; |
22d917d8 | 286 | u32 first, last, id2; |
5c1469de | 287 | |
22d917d8 | 288 | id2 = id + count - 1; |
5c1469de | 289 | |
22d917d8 | 290 | /* Find the matching extent */ |
22d917d8 EB |
291 | for (idx = 0; idx < extents; idx++) { |
292 | first = map->extent[idx].first; | |
293 | last = first + map->extent[idx].count - 1; | |
294 | if (id >= first && id <= last && | |
295 | (id2 >= first && id2 <= last)) | |
3edf652f | 296 | return &map->extent[idx]; |
22d917d8 | 297 | } |
3edf652f | 298 | return NULL; |
22d917d8 EB |
299 | } |
300 | ||
6397fac4 CB |
301 | static u32 map_id_range_down(struct uid_gid_map *map, u32 id, u32 count) |
302 | { | |
3edf652f EB |
303 | struct uid_gid_extent *extent; |
304 | unsigned extents = map->nr_extents; | |
6397fac4 CB |
305 | smp_rmb(); |
306 | ||
307 | if (extents <= UID_GID_MAP_MAX_BASE_EXTENTS) | |
3edf652f EB |
308 | extent = map_id_range_down_base(extents, map, id, count); |
309 | else | |
310 | extent = map_id_range_down_max(extents, map, id, count); | |
311 | ||
22d917d8 | 312 | /* Map the id or note failure */ |
3edf652f EB |
313 | if (extent) |
314 | id = (id - extent->first) + extent->lower_first; | |
22d917d8 EB |
315 | else |
316 | id = (u32) -1; | |
317 | ||
318 | return id; | |
319 | } | |
320 | ||
321 | static u32 map_id_down(struct uid_gid_map *map, u32 id) | |
322 | { | |
ece66133 | 323 | return map_id_range_down(map, id, 1); |
6397fac4 CB |
324 | } |
325 | ||
326 | /** | |
327 | * map_id_up_base - Find idmap via binary search in static extent array. | |
328 | * Can only be called if number of mappings is equal or less than | |
329 | * UID_GID_MAP_MAX_BASE_EXTENTS. | |
330 | */ | |
3edf652f EB |
331 | static struct uid_gid_extent * |
332 | map_id_up_base(unsigned extents, struct uid_gid_map *map, u32 id) | |
22d917d8 | 333 | { |
3edf652f | 334 | unsigned idx; |
22d917d8 EB |
335 | u32 first, last; |
336 | ||
337 | /* Find the matching extent */ | |
22d917d8 | 338 | for (idx = 0; idx < extents; idx++) { |
22d917d8 | 339 | first = map->extent[idx].lower_first; |
22d917d8 EB |
340 | last = first + map->extent[idx].count - 1; |
341 | if (id >= first && id <= last) | |
3edf652f | 342 | return &map->extent[idx]; |
22d917d8 | 343 | } |
3edf652f | 344 | return NULL; |
22d917d8 | 345 | } |
22d917d8 | 346 | |
6397fac4 CB |
347 | /** |
348 | * map_id_up_max - Find idmap via binary search in ordered idmap array. | |
349 | * Can only be called if number of mappings exceeds UID_GID_MAP_MAX_BASE_EXTENTS. | |
350 | */ | |
3edf652f EB |
351 | static struct uid_gid_extent * |
352 | map_id_up_max(unsigned extents, struct uid_gid_map *map, u32 id) | |
6397fac4 | 353 | { |
6397fac4 CB |
354 | struct idmap_key key; |
355 | ||
356 | key.map_up = true; | |
11a8b927 | 357 | key.count = 1; |
6397fac4 CB |
358 | key.id = id; |
359 | ||
3edf652f EB |
360 | return bsearch(&key, map->reverse, extents, |
361 | sizeof(struct uid_gid_extent), cmp_map_id); | |
22d917d8 EB |
362 | } |
363 | ||
364 | static u32 map_id_up(struct uid_gid_map *map, u32 id) | |
365 | { | |
3edf652f EB |
366 | struct uid_gid_extent *extent; |
367 | unsigned extents = map->nr_extents; | |
e79323bd | 368 | smp_rmb(); |
6397fac4 | 369 | |
3edf652f EB |
370 | if (extents <= UID_GID_MAP_MAX_BASE_EXTENTS) |
371 | extent = map_id_up_base(extents, map, id); | |
372 | else | |
373 | extent = map_id_up_max(extents, map, id); | |
374 | ||
22d917d8 | 375 | /* Map the id or note failure */ |
6397fac4 CB |
376 | if (extent) |
377 | id = (id - extent->lower_first) + extent->first; | |
22d917d8 EB |
378 | else |
379 | id = (u32) -1; | |
380 | ||
381 | return id; | |
382 | } | |
383 | ||
384 | /** | |
385 | * make_kuid - Map a user-namespace uid pair into a kuid. | |
386 | * @ns: User namespace that the uid is in | |
387 | * @uid: User identifier | |
388 | * | |
389 | * Maps a user-namespace uid pair into a kernel internal kuid, | |
390 | * and returns that kuid. | |
391 | * | |
392 | * When there is no mapping defined for the user-namespace uid | |
393 | * pair INVALID_UID is returned. Callers are expected to test | |
b080e047 | 394 | * for and handle INVALID_UID being returned. INVALID_UID |
22d917d8 EB |
395 | * may be tested for using uid_valid(). |
396 | */ | |
397 | kuid_t make_kuid(struct user_namespace *ns, uid_t uid) | |
398 | { | |
399 | /* Map the uid to a global kernel uid */ | |
400 | return KUIDT_INIT(map_id_down(&ns->uid_map, uid)); | |
401 | } | |
402 | EXPORT_SYMBOL(make_kuid); | |
403 | ||
404 | /** | |
405 | * from_kuid - Create a uid from a kuid user-namespace pair. | |
406 | * @targ: The user namespace we want a uid in. | |
407 | * @kuid: The kernel internal uid to start with. | |
408 | * | |
409 | * Map @kuid into the user-namespace specified by @targ and | |
410 | * return the resulting uid. | |
411 | * | |
412 | * There is always a mapping into the initial user_namespace. | |
413 | * | |
414 | * If @kuid has no mapping in @targ (uid_t)-1 is returned. | |
415 | */ | |
416 | uid_t from_kuid(struct user_namespace *targ, kuid_t kuid) | |
417 | { | |
418 | /* Map the uid from a global kernel uid */ | |
419 | return map_id_up(&targ->uid_map, __kuid_val(kuid)); | |
420 | } | |
421 | EXPORT_SYMBOL(from_kuid); | |
422 | ||
423 | /** | |
424 | * from_kuid_munged - Create a uid from a kuid user-namespace pair. | |
425 | * @targ: The user namespace we want a uid in. | |
426 | * @kuid: The kernel internal uid to start with. | |
427 | * | |
428 | * Map @kuid into the user-namespace specified by @targ and | |
429 | * return the resulting uid. | |
430 | * | |
431 | * There is always a mapping into the initial user_namespace. | |
432 | * | |
433 | * Unlike from_kuid from_kuid_munged never fails and always | |
434 | * returns a valid uid. This makes from_kuid_munged appropriate | |
435 | * for use in syscalls like stat and getuid where failing the | |
436 | * system call and failing to provide a valid uid are not an | |
437 | * options. | |
438 | * | |
439 | * If @kuid has no mapping in @targ overflowuid is returned. | |
440 | */ | |
441 | uid_t from_kuid_munged(struct user_namespace *targ, kuid_t kuid) | |
442 | { | |
443 | uid_t uid; | |
444 | uid = from_kuid(targ, kuid); | |
445 | ||
446 | if (uid == (uid_t) -1) | |
447 | uid = overflowuid; | |
448 | return uid; | |
449 | } | |
450 | EXPORT_SYMBOL(from_kuid_munged); | |
451 | ||
452 | /** | |
453 | * make_kgid - Map a user-namespace gid pair into a kgid. | |
454 | * @ns: User namespace that the gid is in | |
68a9a435 | 455 | * @gid: group identifier |
22d917d8 EB |
456 | * |
457 | * Maps a user-namespace gid pair into a kernel internal kgid, | |
458 | * and returns that kgid. | |
459 | * | |
460 | * When there is no mapping defined for the user-namespace gid | |
461 | * pair INVALID_GID is returned. Callers are expected to test | |
462 | * for and handle INVALID_GID being returned. INVALID_GID may be | |
463 | * tested for using gid_valid(). | |
464 | */ | |
465 | kgid_t make_kgid(struct user_namespace *ns, gid_t gid) | |
466 | { | |
467 | /* Map the gid to a global kernel gid */ | |
468 | return KGIDT_INIT(map_id_down(&ns->gid_map, gid)); | |
469 | } | |
470 | EXPORT_SYMBOL(make_kgid); | |
471 | ||
472 | /** | |
473 | * from_kgid - Create a gid from a kgid user-namespace pair. | |
474 | * @targ: The user namespace we want a gid in. | |
475 | * @kgid: The kernel internal gid to start with. | |
476 | * | |
477 | * Map @kgid into the user-namespace specified by @targ and | |
478 | * return the resulting gid. | |
479 | * | |
480 | * There is always a mapping into the initial user_namespace. | |
481 | * | |
482 | * If @kgid has no mapping in @targ (gid_t)-1 is returned. | |
483 | */ | |
484 | gid_t from_kgid(struct user_namespace *targ, kgid_t kgid) | |
485 | { | |
486 | /* Map the gid from a global kernel gid */ | |
487 | return map_id_up(&targ->gid_map, __kgid_val(kgid)); | |
488 | } | |
489 | EXPORT_SYMBOL(from_kgid); | |
490 | ||
491 | /** | |
492 | * from_kgid_munged - Create a gid from a kgid user-namespace pair. | |
493 | * @targ: The user namespace we want a gid in. | |
494 | * @kgid: The kernel internal gid to start with. | |
495 | * | |
496 | * Map @kgid into the user-namespace specified by @targ and | |
497 | * return the resulting gid. | |
498 | * | |
499 | * There is always a mapping into the initial user_namespace. | |
500 | * | |
501 | * Unlike from_kgid from_kgid_munged never fails and always | |
502 | * returns a valid gid. This makes from_kgid_munged appropriate | |
503 | * for use in syscalls like stat and getgid where failing the | |
504 | * system call and failing to provide a valid gid are not options. | |
505 | * | |
506 | * If @kgid has no mapping in @targ overflowgid is returned. | |
507 | */ | |
508 | gid_t from_kgid_munged(struct user_namespace *targ, kgid_t kgid) | |
509 | { | |
510 | gid_t gid; | |
511 | gid = from_kgid(targ, kgid); | |
512 | ||
513 | if (gid == (gid_t) -1) | |
514 | gid = overflowgid; | |
515 | return gid; | |
516 | } | |
517 | EXPORT_SYMBOL(from_kgid_munged); | |
518 | ||
f76d207a EB |
519 | /** |
520 | * make_kprojid - Map a user-namespace projid pair into a kprojid. | |
521 | * @ns: User namespace that the projid is in | |
522 | * @projid: Project identifier | |
523 | * | |
524 | * Maps a user-namespace uid pair into a kernel internal kuid, | |
525 | * and returns that kuid. | |
526 | * | |
527 | * When there is no mapping defined for the user-namespace projid | |
528 | * pair INVALID_PROJID is returned. Callers are expected to test | |
7b7b8a2c | 529 | * for and handle INVALID_PROJID being returned. INVALID_PROJID |
f76d207a EB |
530 | * may be tested for using projid_valid(). |
531 | */ | |
532 | kprojid_t make_kprojid(struct user_namespace *ns, projid_t projid) | |
533 | { | |
534 | /* Map the uid to a global kernel uid */ | |
535 | return KPROJIDT_INIT(map_id_down(&ns->projid_map, projid)); | |
536 | } | |
537 | EXPORT_SYMBOL(make_kprojid); | |
538 | ||
539 | /** | |
540 | * from_kprojid - Create a projid from a kprojid user-namespace pair. | |
541 | * @targ: The user namespace we want a projid in. | |
542 | * @kprojid: The kernel internal project identifier to start with. | |
543 | * | |
544 | * Map @kprojid into the user-namespace specified by @targ and | |
545 | * return the resulting projid. | |
546 | * | |
547 | * There is always a mapping into the initial user_namespace. | |
548 | * | |
549 | * If @kprojid has no mapping in @targ (projid_t)-1 is returned. | |
550 | */ | |
551 | projid_t from_kprojid(struct user_namespace *targ, kprojid_t kprojid) | |
552 | { | |
553 | /* Map the uid from a global kernel uid */ | |
554 | return map_id_up(&targ->projid_map, __kprojid_val(kprojid)); | |
555 | } | |
556 | EXPORT_SYMBOL(from_kprojid); | |
557 | ||
558 | /** | |
559 | * from_kprojid_munged - Create a projiid from a kprojid user-namespace pair. | |
560 | * @targ: The user namespace we want a projid in. | |
561 | * @kprojid: The kernel internal projid to start with. | |
562 | * | |
563 | * Map @kprojid into the user-namespace specified by @targ and | |
564 | * return the resulting projid. | |
565 | * | |
566 | * There is always a mapping into the initial user_namespace. | |
567 | * | |
568 | * Unlike from_kprojid from_kprojid_munged never fails and always | |
569 | * returns a valid projid. This makes from_kprojid_munged | |
570 | * appropriate for use in syscalls like stat and where | |
571 | * failing the system call and failing to provide a valid projid are | |
572 | * not an options. | |
573 | * | |
574 | * If @kprojid has no mapping in @targ OVERFLOW_PROJID is returned. | |
575 | */ | |
576 | projid_t from_kprojid_munged(struct user_namespace *targ, kprojid_t kprojid) | |
577 | { | |
578 | projid_t projid; | |
579 | projid = from_kprojid(targ, kprojid); | |
580 | ||
581 | if (projid == (projid_t) -1) | |
582 | projid = OVERFLOW_PROJID; | |
583 | return projid; | |
584 | } | |
585 | EXPORT_SYMBOL(from_kprojid_munged); | |
586 | ||
587 | ||
22d917d8 EB |
588 | static int uid_m_show(struct seq_file *seq, void *v) |
589 | { | |
590 | struct user_namespace *ns = seq->private; | |
591 | struct uid_gid_extent *extent = v; | |
592 | struct user_namespace *lower_ns; | |
593 | uid_t lower; | |
5c1469de | 594 | |
c450f371 | 595 | lower_ns = seq_user_ns(seq); |
22d917d8 EB |
596 | if ((lower_ns == ns) && lower_ns->parent) |
597 | lower_ns = lower_ns->parent; | |
598 | ||
599 | lower = from_kuid(lower_ns, KUIDT_INIT(extent->lower_first)); | |
600 | ||
601 | seq_printf(seq, "%10u %10u %10u\n", | |
602 | extent->first, | |
603 | lower, | |
604 | extent->count); | |
605 | ||
606 | return 0; | |
5c1469de EB |
607 | } |
608 | ||
22d917d8 | 609 | static int gid_m_show(struct seq_file *seq, void *v) |
5c1469de | 610 | { |
22d917d8 EB |
611 | struct user_namespace *ns = seq->private; |
612 | struct uid_gid_extent *extent = v; | |
613 | struct user_namespace *lower_ns; | |
614 | gid_t lower; | |
5c1469de | 615 | |
c450f371 | 616 | lower_ns = seq_user_ns(seq); |
22d917d8 EB |
617 | if ((lower_ns == ns) && lower_ns->parent) |
618 | lower_ns = lower_ns->parent; | |
5c1469de | 619 | |
22d917d8 EB |
620 | lower = from_kgid(lower_ns, KGIDT_INIT(extent->lower_first)); |
621 | ||
622 | seq_printf(seq, "%10u %10u %10u\n", | |
623 | extent->first, | |
624 | lower, | |
625 | extent->count); | |
626 | ||
627 | return 0; | |
628 | } | |
629 | ||
f76d207a EB |
630 | static int projid_m_show(struct seq_file *seq, void *v) |
631 | { | |
632 | struct user_namespace *ns = seq->private; | |
633 | struct uid_gid_extent *extent = v; | |
634 | struct user_namespace *lower_ns; | |
635 | projid_t lower; | |
636 | ||
637 | lower_ns = seq_user_ns(seq); | |
638 | if ((lower_ns == ns) && lower_ns->parent) | |
639 | lower_ns = lower_ns->parent; | |
640 | ||
641 | lower = from_kprojid(lower_ns, KPROJIDT_INIT(extent->lower_first)); | |
642 | ||
643 | seq_printf(seq, "%10u %10u %10u\n", | |
644 | extent->first, | |
645 | lower, | |
646 | extent->count); | |
647 | ||
648 | return 0; | |
649 | } | |
650 | ||
68a9a435 FF |
651 | static void *m_start(struct seq_file *seq, loff_t *ppos, |
652 | struct uid_gid_map *map) | |
22d917d8 | 653 | { |
22d917d8 | 654 | loff_t pos = *ppos; |
d5e7b3c5 EB |
655 | unsigned extents = map->nr_extents; |
656 | smp_rmb(); | |
22d917d8 | 657 | |
d5e7b3c5 | 658 | if (pos >= extents) |
6397fac4 | 659 | return NULL; |
22d917d8 | 660 | |
d5e7b3c5 | 661 | if (extents <= UID_GID_MAP_MAX_BASE_EXTENTS) |
6397fac4 | 662 | return &map->extent[pos]; |
22d917d8 | 663 | |
6397fac4 | 664 | return &map->forward[pos]; |
22d917d8 EB |
665 | } |
666 | ||
667 | static void *uid_m_start(struct seq_file *seq, loff_t *ppos) | |
668 | { | |
669 | struct user_namespace *ns = seq->private; | |
670 | ||
671 | return m_start(seq, ppos, &ns->uid_map); | |
672 | } | |
673 | ||
674 | static void *gid_m_start(struct seq_file *seq, loff_t *ppos) | |
675 | { | |
676 | struct user_namespace *ns = seq->private; | |
677 | ||
678 | return m_start(seq, ppos, &ns->gid_map); | |
679 | } | |
680 | ||
f76d207a EB |
681 | static void *projid_m_start(struct seq_file *seq, loff_t *ppos) |
682 | { | |
683 | struct user_namespace *ns = seq->private; | |
684 | ||
685 | return m_start(seq, ppos, &ns->projid_map); | |
686 | } | |
687 | ||
22d917d8 EB |
688 | static void *m_next(struct seq_file *seq, void *v, loff_t *pos) |
689 | { | |
690 | (*pos)++; | |
691 | return seq->op->start(seq, pos); | |
692 | } | |
693 | ||
694 | static void m_stop(struct seq_file *seq, void *v) | |
695 | { | |
696 | return; | |
697 | } | |
698 | ||
ccf94f1b | 699 | const struct seq_operations proc_uid_seq_operations = { |
22d917d8 EB |
700 | .start = uid_m_start, |
701 | .stop = m_stop, | |
702 | .next = m_next, | |
703 | .show = uid_m_show, | |
704 | }; | |
705 | ||
ccf94f1b | 706 | const struct seq_operations proc_gid_seq_operations = { |
22d917d8 EB |
707 | .start = gid_m_start, |
708 | .stop = m_stop, | |
709 | .next = m_next, | |
710 | .show = gid_m_show, | |
711 | }; | |
712 | ||
ccf94f1b | 713 | const struct seq_operations proc_projid_seq_operations = { |
f76d207a EB |
714 | .start = projid_m_start, |
715 | .stop = m_stop, | |
716 | .next = m_next, | |
717 | .show = projid_m_show, | |
718 | }; | |
719 | ||
68a9a435 FF |
720 | static bool mappings_overlap(struct uid_gid_map *new_map, |
721 | struct uid_gid_extent *extent) | |
0bd14b4f EB |
722 | { |
723 | u32 upper_first, lower_first, upper_last, lower_last; | |
724 | unsigned idx; | |
725 | ||
726 | upper_first = extent->first; | |
727 | lower_first = extent->lower_first; | |
728 | upper_last = upper_first + extent->count - 1; | |
729 | lower_last = lower_first + extent->count - 1; | |
730 | ||
731 | for (idx = 0; idx < new_map->nr_extents; idx++) { | |
732 | u32 prev_upper_first, prev_lower_first; | |
733 | u32 prev_upper_last, prev_lower_last; | |
734 | struct uid_gid_extent *prev; | |
735 | ||
6397fac4 CB |
736 | if (new_map->nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) |
737 | prev = &new_map->extent[idx]; | |
738 | else | |
739 | prev = &new_map->forward[idx]; | |
0bd14b4f EB |
740 | |
741 | prev_upper_first = prev->first; | |
742 | prev_lower_first = prev->lower_first; | |
743 | prev_upper_last = prev_upper_first + prev->count - 1; | |
744 | prev_lower_last = prev_lower_first + prev->count - 1; | |
745 | ||
746 | /* Does the upper range intersect a previous extent? */ | |
747 | if ((prev_upper_first <= upper_last) && | |
748 | (prev_upper_last >= upper_first)) | |
749 | return true; | |
750 | ||
751 | /* Does the lower range intersect a previous extent? */ | |
752 | if ((prev_lower_first <= lower_last) && | |
753 | (prev_lower_last >= lower_first)) | |
754 | return true; | |
755 | } | |
756 | return false; | |
757 | } | |
758 | ||
6397fac4 CB |
759 | /** |
760 | * insert_extent - Safely insert a new idmap extent into struct uid_gid_map. | |
761 | * Takes care to allocate a 4K block of memory if the number of mappings exceeds | |
762 | * UID_GID_MAP_MAX_BASE_EXTENTS. | |
763 | */ | |
764 | static int insert_extent(struct uid_gid_map *map, struct uid_gid_extent *extent) | |
765 | { | |
3fda0e73 | 766 | struct uid_gid_extent *dest; |
6397fac4 CB |
767 | |
768 | if (map->nr_extents == UID_GID_MAP_MAX_BASE_EXTENTS) { | |
769 | struct uid_gid_extent *forward; | |
770 | ||
771 | /* Allocate memory for 340 mappings. */ | |
6da2ec56 KC |
772 | forward = kmalloc_array(UID_GID_MAP_MAX_EXTENTS, |
773 | sizeof(struct uid_gid_extent), | |
774 | GFP_KERNEL); | |
6397fac4 CB |
775 | if (!forward) |
776 | return -ENOMEM; | |
777 | ||
778 | /* Copy over memory. Only set up memory for the forward pointer. | |
779 | * Defer the memory setup for the reverse pointer. | |
780 | */ | |
781 | memcpy(forward, map->extent, | |
782 | map->nr_extents * sizeof(map->extent[0])); | |
783 | ||
784 | map->forward = forward; | |
785 | map->reverse = NULL; | |
786 | } | |
787 | ||
3fda0e73 EB |
788 | if (map->nr_extents < UID_GID_MAP_MAX_BASE_EXTENTS) |
789 | dest = &map->extent[map->nr_extents]; | |
790 | else | |
791 | dest = &map->forward[map->nr_extents]; | |
792 | ||
793 | *dest = *extent; | |
794 | map->nr_extents++; | |
6397fac4 CB |
795 | return 0; |
796 | } | |
797 | ||
798 | /* cmp function to sort() forward mappings */ | |
799 | static int cmp_extents_forward(const void *a, const void *b) | |
800 | { | |
801 | const struct uid_gid_extent *e1 = a; | |
802 | const struct uid_gid_extent *e2 = b; | |
803 | ||
804 | if (e1->first < e2->first) | |
805 | return -1; | |
806 | ||
807 | if (e1->first > e2->first) | |
808 | return 1; | |
809 | ||
810 | return 0; | |
811 | } | |
812 | ||
813 | /* cmp function to sort() reverse mappings */ | |
814 | static int cmp_extents_reverse(const void *a, const void *b) | |
815 | { | |
816 | const struct uid_gid_extent *e1 = a; | |
817 | const struct uid_gid_extent *e2 = b; | |
818 | ||
819 | if (e1->lower_first < e2->lower_first) | |
820 | return -1; | |
821 | ||
822 | if (e1->lower_first > e2->lower_first) | |
823 | return 1; | |
824 | ||
825 | return 0; | |
826 | } | |
827 | ||
828 | /** | |
829 | * sort_idmaps - Sorts an array of idmap entries. | |
830 | * Can only be called if number of mappings exceeds UID_GID_MAP_MAX_BASE_EXTENTS. | |
831 | */ | |
832 | static int sort_idmaps(struct uid_gid_map *map) | |
833 | { | |
834 | if (map->nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) | |
835 | return 0; | |
836 | ||
837 | /* Sort forward array. */ | |
838 | sort(map->forward, map->nr_extents, sizeof(struct uid_gid_extent), | |
839 | cmp_extents_forward, NULL); | |
840 | ||
841 | /* Only copy the memory from forward we actually need. */ | |
842 | map->reverse = kmemdup(map->forward, | |
843 | map->nr_extents * sizeof(struct uid_gid_extent), | |
844 | GFP_KERNEL); | |
845 | if (!map->reverse) | |
846 | return -ENOMEM; | |
847 | ||
848 | /* Sort reverse array. */ | |
849 | sort(map->reverse, map->nr_extents, sizeof(struct uid_gid_extent), | |
850 | cmp_extents_reverse, NULL); | |
851 | ||
852 | return 0; | |
853 | } | |
854 | ||
db2e718a SH |
855 | /** |
856 | * verify_root_map() - check the uid 0 mapping | |
857 | * @file: idmapping file | |
858 | * @map_ns: user namespace of the target process | |
859 | * @new_map: requested idmap | |
860 | * | |
861 | * If a process requests mapping parent uid 0 into the new ns, verify that the | |
862 | * process writing the map had the CAP_SETFCAP capability as the target process | |
863 | * will be able to write fscaps that are valid in ancestor user namespaces. | |
864 | * | |
865 | * Return: true if the mapping is allowed, false if not. | |
866 | */ | |
867 | static bool verify_root_map(const struct file *file, | |
868 | struct user_namespace *map_ns, | |
869 | struct uid_gid_map *new_map) | |
870 | { | |
871 | int idx; | |
872 | const struct user_namespace *file_ns = file->f_cred->user_ns; | |
873 | struct uid_gid_extent *extent0 = NULL; | |
874 | ||
875 | for (idx = 0; idx < new_map->nr_extents; idx++) { | |
876 | if (new_map->nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) | |
877 | extent0 = &new_map->extent[idx]; | |
878 | else | |
879 | extent0 = &new_map->forward[idx]; | |
880 | if (extent0->lower_first == 0) | |
881 | break; | |
882 | ||
883 | extent0 = NULL; | |
884 | } | |
885 | ||
886 | if (!extent0) | |
887 | return true; | |
888 | ||
889 | if (map_ns == file_ns) { | |
890 | /* The process unshared its ns and is writing to its own | |
891 | * /proc/self/uid_map. User already has full capabilites in | |
892 | * the new namespace. Verify that the parent had CAP_SETFCAP | |
893 | * when it unshared. | |
894 | * */ | |
895 | if (!file_ns->parent_could_setfcap) | |
896 | return false; | |
897 | } else { | |
898 | /* Process p1 is writing to uid_map of p2, who is in a child | |
899 | * user namespace to p1's. Verify that the opener of the map | |
900 | * file has CAP_SETFCAP against the parent of the new map | |
901 | * namespace */ | |
902 | if (!file_ns_capable(file, map_ns->parent, CAP_SETFCAP)) | |
903 | return false; | |
904 | } | |
905 | ||
906 | return true; | |
907 | } | |
908 | ||
22d917d8 EB |
909 | static ssize_t map_write(struct file *file, const char __user *buf, |
910 | size_t count, loff_t *ppos, | |
911 | int cap_setid, | |
912 | struct uid_gid_map *map, | |
913 | struct uid_gid_map *parent_map) | |
914 | { | |
915 | struct seq_file *seq = file->private_data; | |
db2e718a | 916 | struct user_namespace *map_ns = seq->private; |
22d917d8 EB |
917 | struct uid_gid_map new_map; |
918 | unsigned idx; | |
6397fac4 | 919 | struct uid_gid_extent extent; |
70f6cbb6 | 920 | char *kbuf = NULL, *pos, *next_line; |
5820f140 JH |
921 | ssize_t ret; |
922 | ||
923 | /* Only allow < page size writes at the beginning of the file */ | |
924 | if ((*ppos != 0) || (count >= PAGE_SIZE)) | |
925 | return -EINVAL; | |
926 | ||
927 | /* Slurp in the user data */ | |
928 | kbuf = memdup_user_nul(buf, count); | |
929 | if (IS_ERR(kbuf)) | |
930 | return PTR_ERR(kbuf); | |
22d917d8 EB |
931 | |
932 | /* | |
f0d62aec | 933 | * The userns_state_mutex serializes all writes to any given map. |
22d917d8 EB |
934 | * |
935 | * Any map is only ever written once. | |
936 | * | |
937 | * An id map fits within 1 cache line on most architectures. | |
938 | * | |
939 | * On read nothing needs to be done unless you are on an | |
940 | * architecture with a crazy cache coherency model like alpha. | |
941 | * | |
942 | * There is a one time data dependency between reading the | |
943 | * count of the extents and the values of the extents. The | |
944 | * desired behavior is to see the values of the extents that | |
945 | * were written before the count of the extents. | |
946 | * | |
947 | * To achieve this smp_wmb() is used on guarantee the write | |
e79323bd MP |
948 | * order and smp_rmb() is guaranteed that we don't have crazy |
949 | * architectures returning stale data. | |
22d917d8 | 950 | */ |
f0d62aec | 951 | mutex_lock(&userns_state_mutex); |
22d917d8 | 952 | |
6397fac4 CB |
953 | memset(&new_map, 0, sizeof(struct uid_gid_map)); |
954 | ||
22d917d8 EB |
955 | ret = -EPERM; |
956 | /* Only allow one successful write to the map */ | |
957 | if (map->nr_extents != 0) | |
958 | goto out; | |
959 | ||
41c21e35 AL |
960 | /* |
961 | * Adjusting namespace settings requires capabilities on the target. | |
5c1469de | 962 | */ |
db2e718a | 963 | if (cap_valid(cap_setid) && !file_ns_capable(file, map_ns, CAP_SYS_ADMIN)) |
22d917d8 EB |
964 | goto out; |
965 | ||
22d917d8 EB |
966 | /* Parse the user data */ |
967 | ret = -EINVAL; | |
968 | pos = kbuf; | |
68a9a435 | 969 | for (; pos; pos = next_line) { |
22d917d8 EB |
970 | |
971 | /* Find the end of line and ensure I don't look past it */ | |
972 | next_line = strchr(pos, '\n'); | |
973 | if (next_line) { | |
974 | *next_line = '\0'; | |
975 | next_line++; | |
976 | if (*next_line == '\0') | |
977 | next_line = NULL; | |
5c1469de | 978 | } |
22d917d8 EB |
979 | |
980 | pos = skip_spaces(pos); | |
6397fac4 | 981 | extent.first = simple_strtoul(pos, &pos, 10); |
22d917d8 EB |
982 | if (!isspace(*pos)) |
983 | goto out; | |
984 | ||
985 | pos = skip_spaces(pos); | |
6397fac4 | 986 | extent.lower_first = simple_strtoul(pos, &pos, 10); |
22d917d8 EB |
987 | if (!isspace(*pos)) |
988 | goto out; | |
989 | ||
990 | pos = skip_spaces(pos); | |
6397fac4 | 991 | extent.count = simple_strtoul(pos, &pos, 10); |
22d917d8 EB |
992 | if (*pos && !isspace(*pos)) |
993 | goto out; | |
994 | ||
995 | /* Verify there is not trailing junk on the line */ | |
996 | pos = skip_spaces(pos); | |
997 | if (*pos != '\0') | |
998 | goto out; | |
999 | ||
1000 | /* Verify we have been given valid starting values */ | |
6397fac4 CB |
1001 | if ((extent.first == (u32) -1) || |
1002 | (extent.lower_first == (u32) -1)) | |
22d917d8 EB |
1003 | goto out; |
1004 | ||
68a9a435 FF |
1005 | /* Verify count is not zero and does not cause the |
1006 | * extent to wrap | |
1007 | */ | |
6397fac4 | 1008 | if ((extent.first + extent.count) <= extent.first) |
22d917d8 | 1009 | goto out; |
6397fac4 CB |
1010 | if ((extent.lower_first + extent.count) <= |
1011 | extent.lower_first) | |
22d917d8 EB |
1012 | goto out; |
1013 | ||
0bd14b4f | 1014 | /* Do the ranges in extent overlap any previous extents? */ |
6397fac4 | 1015 | if (mappings_overlap(&new_map, &extent)) |
22d917d8 EB |
1016 | goto out; |
1017 | ||
6397fac4 | 1018 | if ((new_map.nr_extents + 1) == UID_GID_MAP_MAX_EXTENTS && |
22d917d8 EB |
1019 | (next_line != NULL)) |
1020 | goto out; | |
6397fac4 CB |
1021 | |
1022 | ret = insert_extent(&new_map, &extent); | |
1023 | if (ret < 0) | |
1024 | goto out; | |
1025 | ret = -EINVAL; | |
5c1469de | 1026 | } |
a12f4f85 | 1027 | /* Be very certain the new map actually exists */ |
22d917d8 EB |
1028 | if (new_map.nr_extents == 0) |
1029 | goto out; | |
1030 | ||
1031 | ret = -EPERM; | |
1032 | /* Validate the user is allowed to use user id's mapped to. */ | |
db2e718a | 1033 | if (!new_idmap_permitted(file, map_ns, cap_setid, &new_map)) |
22d917d8 EB |
1034 | goto out; |
1035 | ||
6397fac4 | 1036 | ret = -EPERM; |
22d917d8 EB |
1037 | /* Map the lower ids from the parent user namespace to the |
1038 | * kernel global id space. | |
1039 | */ | |
1040 | for (idx = 0; idx < new_map.nr_extents; idx++) { | |
6397fac4 | 1041 | struct uid_gid_extent *e; |
22d917d8 | 1042 | u32 lower_first; |
6397fac4 CB |
1043 | |
1044 | if (new_map.nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) | |
1045 | e = &new_map.extent[idx]; | |
1046 | else | |
1047 | e = &new_map.forward[idx]; | |
22d917d8 EB |
1048 | |
1049 | lower_first = map_id_range_down(parent_map, | |
6397fac4 CB |
1050 | e->lower_first, |
1051 | e->count); | |
22d917d8 EB |
1052 | |
1053 | /* Fail if we can not map the specified extent to | |
1054 | * the kernel global id space. | |
1055 | */ | |
1056 | if (lower_first == (u32) -1) | |
1057 | goto out; | |
1058 | ||
6397fac4 | 1059 | e->lower_first = lower_first; |
22d917d8 EB |
1060 | } |
1061 | ||
d2f007db JH |
1062 | /* |
1063 | * If we want to use binary search for lookup, this clones the extent | |
1064 | * array and sorts both copies. | |
1065 | */ | |
1066 | ret = sort_idmaps(&new_map); | |
1067 | if (ret < 0) | |
1068 | goto out; | |
1069 | ||
22d917d8 | 1070 | /* Install the map */ |
6397fac4 CB |
1071 | if (new_map.nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) { |
1072 | memcpy(map->extent, new_map.extent, | |
1073 | new_map.nr_extents * sizeof(new_map.extent[0])); | |
1074 | } else { | |
1075 | map->forward = new_map.forward; | |
1076 | map->reverse = new_map.reverse; | |
1077 | } | |
22d917d8 EB |
1078 | smp_wmb(); |
1079 | map->nr_extents = new_map.nr_extents; | |
1080 | ||
1081 | *ppos = count; | |
1082 | ret = count; | |
1083 | out: | |
6397fac4 CB |
1084 | if (ret < 0 && new_map.nr_extents > UID_GID_MAP_MAX_BASE_EXTENTS) { |
1085 | kfree(new_map.forward); | |
1086 | kfree(new_map.reverse); | |
1087 | map->forward = NULL; | |
1088 | map->reverse = NULL; | |
1089 | map->nr_extents = 0; | |
1090 | } | |
1091 | ||
f0d62aec | 1092 | mutex_unlock(&userns_state_mutex); |
70f6cbb6 | 1093 | kfree(kbuf); |
22d917d8 EB |
1094 | return ret; |
1095 | } | |
1096 | ||
68a9a435 FF |
1097 | ssize_t proc_uid_map_write(struct file *file, const char __user *buf, |
1098 | size_t size, loff_t *ppos) | |
22d917d8 EB |
1099 | { |
1100 | struct seq_file *seq = file->private_data; | |
1101 | struct user_namespace *ns = seq->private; | |
c450f371 | 1102 | struct user_namespace *seq_ns = seq_user_ns(seq); |
22d917d8 EB |
1103 | |
1104 | if (!ns->parent) | |
1105 | return -EPERM; | |
1106 | ||
c450f371 EB |
1107 | if ((seq_ns != ns) && (seq_ns != ns->parent)) |
1108 | return -EPERM; | |
1109 | ||
22d917d8 EB |
1110 | return map_write(file, buf, size, ppos, CAP_SETUID, |
1111 | &ns->uid_map, &ns->parent->uid_map); | |
1112 | } | |
1113 | ||
68a9a435 FF |
1114 | ssize_t proc_gid_map_write(struct file *file, const char __user *buf, |
1115 | size_t size, loff_t *ppos) | |
22d917d8 EB |
1116 | { |
1117 | struct seq_file *seq = file->private_data; | |
1118 | struct user_namespace *ns = seq->private; | |
c450f371 | 1119 | struct user_namespace *seq_ns = seq_user_ns(seq); |
22d917d8 EB |
1120 | |
1121 | if (!ns->parent) | |
1122 | return -EPERM; | |
1123 | ||
c450f371 EB |
1124 | if ((seq_ns != ns) && (seq_ns != ns->parent)) |
1125 | return -EPERM; | |
1126 | ||
22d917d8 EB |
1127 | return map_write(file, buf, size, ppos, CAP_SETGID, |
1128 | &ns->gid_map, &ns->parent->gid_map); | |
1129 | } | |
1130 | ||
68a9a435 FF |
1131 | ssize_t proc_projid_map_write(struct file *file, const char __user *buf, |
1132 | size_t size, loff_t *ppos) | |
f76d207a EB |
1133 | { |
1134 | struct seq_file *seq = file->private_data; | |
1135 | struct user_namespace *ns = seq->private; | |
1136 | struct user_namespace *seq_ns = seq_user_ns(seq); | |
1137 | ||
1138 | if (!ns->parent) | |
1139 | return -EPERM; | |
1140 | ||
1141 | if ((seq_ns != ns) && (seq_ns != ns->parent)) | |
1142 | return -EPERM; | |
1143 | ||
1144 | /* Anyone can set any valid project id no capability needed */ | |
1145 | return map_write(file, buf, size, ppos, -1, | |
1146 | &ns->projid_map, &ns->parent->projid_map); | |
1147 | } | |
1148 | ||
68a9a435 | 1149 | static bool new_idmap_permitted(const struct file *file, |
6708075f | 1150 | struct user_namespace *ns, int cap_setid, |
22d917d8 EB |
1151 | struct uid_gid_map *new_map) |
1152 | { | |
f95d7918 | 1153 | const struct cred *cred = file->f_cred; |
db2e718a SH |
1154 | |
1155 | if (cap_setid == CAP_SETUID && !verify_root_map(file, ns, new_map)) | |
1156 | return false; | |
1157 | ||
0542f17b EB |
1158 | /* Don't allow mappings that would allow anything that wouldn't |
1159 | * be allowed without the establishment of unprivileged mappings. | |
1160 | */ | |
f95d7918 EB |
1161 | if ((new_map->nr_extents == 1) && (new_map->extent[0].count == 1) && |
1162 | uid_eq(ns->owner, cred->euid)) { | |
37657da3 EB |
1163 | u32 id = new_map->extent[0].lower_first; |
1164 | if (cap_setid == CAP_SETUID) { | |
1165 | kuid_t uid = make_kuid(ns->parent, id); | |
f95d7918 | 1166 | if (uid_eq(uid, cred->euid)) |
37657da3 | 1167 | return true; |
68a9a435 | 1168 | } else if (cap_setid == CAP_SETGID) { |
37657da3 | 1169 | kgid_t gid = make_kgid(ns->parent, id); |
66d2f338 EB |
1170 | if (!(ns->flags & USERNS_SETGROUPS_ALLOWED) && |
1171 | gid_eq(gid, cred->egid)) | |
37657da3 EB |
1172 | return true; |
1173 | } | |
1174 | } | |
1175 | ||
f76d207a EB |
1176 | /* Allow anyone to set a mapping that doesn't require privilege */ |
1177 | if (!cap_valid(cap_setid)) | |
1178 | return true; | |
1179 | ||
22d917d8 EB |
1180 | /* Allow the specified ids if we have the appropriate capability |
1181 | * (CAP_SETUID or CAP_SETGID) over the parent user namespace. | |
a12f4f85 | 1182 | * And the opener of the id file also has the appropriate capability. |
22d917d8 | 1183 | */ |
6708075f EB |
1184 | if (ns_capable(ns->parent, cap_setid) && |
1185 | file_ns_capable(file, ns->parent, cap_setid)) | |
22d917d8 | 1186 | return true; |
5c1469de | 1187 | |
22d917d8 | 1188 | return false; |
5c1469de | 1189 | } |
6164281a | 1190 | |
9cc46516 EB |
1191 | int proc_setgroups_show(struct seq_file *seq, void *v) |
1192 | { | |
1193 | struct user_namespace *ns = seq->private; | |
6aa7de05 | 1194 | unsigned long userns_flags = READ_ONCE(ns->flags); |
9cc46516 EB |
1195 | |
1196 | seq_printf(seq, "%s\n", | |
1197 | (userns_flags & USERNS_SETGROUPS_ALLOWED) ? | |
1198 | "allow" : "deny"); | |
1199 | return 0; | |
1200 | } | |
1201 | ||
1202 | ssize_t proc_setgroups_write(struct file *file, const char __user *buf, | |
1203 | size_t count, loff_t *ppos) | |
1204 | { | |
1205 | struct seq_file *seq = file->private_data; | |
1206 | struct user_namespace *ns = seq->private; | |
1207 | char kbuf[8], *pos; | |
1208 | bool setgroups_allowed; | |
1209 | ssize_t ret; | |
1210 | ||
1211 | /* Only allow a very narrow range of strings to be written */ | |
1212 | ret = -EINVAL; | |
1213 | if ((*ppos != 0) || (count >= sizeof(kbuf))) | |
1214 | goto out; | |
1215 | ||
1216 | /* What was written? */ | |
1217 | ret = -EFAULT; | |
1218 | if (copy_from_user(kbuf, buf, count)) | |
1219 | goto out; | |
1220 | kbuf[count] = '\0'; | |
1221 | pos = kbuf; | |
1222 | ||
1223 | /* What is being requested? */ | |
1224 | ret = -EINVAL; | |
1225 | if (strncmp(pos, "allow", 5) == 0) { | |
1226 | pos += 5; | |
1227 | setgroups_allowed = true; | |
1228 | } | |
1229 | else if (strncmp(pos, "deny", 4) == 0) { | |
1230 | pos += 4; | |
1231 | setgroups_allowed = false; | |
1232 | } | |
1233 | else | |
1234 | goto out; | |
1235 | ||
1236 | /* Verify there is not trailing junk on the line */ | |
1237 | pos = skip_spaces(pos); | |
1238 | if (*pos != '\0') | |
1239 | goto out; | |
1240 | ||
1241 | ret = -EPERM; | |
1242 | mutex_lock(&userns_state_mutex); | |
1243 | if (setgroups_allowed) { | |
1244 | /* Enabling setgroups after setgroups has been disabled | |
1245 | * is not allowed. | |
1246 | */ | |
1247 | if (!(ns->flags & USERNS_SETGROUPS_ALLOWED)) | |
1248 | goto out_unlock; | |
1249 | } else { | |
1250 | /* Permanently disabling setgroups after setgroups has | |
1251 | * been enabled by writing the gid_map is not allowed. | |
1252 | */ | |
1253 | if (ns->gid_map.nr_extents != 0) | |
1254 | goto out_unlock; | |
1255 | ns->flags &= ~USERNS_SETGROUPS_ALLOWED; | |
1256 | } | |
1257 | mutex_unlock(&userns_state_mutex); | |
1258 | ||
1259 | /* Report a successful write */ | |
1260 | *ppos = count; | |
1261 | ret = count; | |
1262 | out: | |
1263 | return ret; | |
1264 | out_unlock: | |
1265 | mutex_unlock(&userns_state_mutex); | |
1266 | goto out; | |
1267 | } | |
1268 | ||
273d2c67 EB |
1269 | bool userns_may_setgroups(const struct user_namespace *ns) |
1270 | { | |
1271 | bool allowed; | |
1272 | ||
f0d62aec | 1273 | mutex_lock(&userns_state_mutex); |
273d2c67 EB |
1274 | /* It is not safe to use setgroups until a gid mapping in |
1275 | * the user namespace has been established. | |
1276 | */ | |
1277 | allowed = ns->gid_map.nr_extents != 0; | |
9cc46516 EB |
1278 | /* Is setgroups allowed? */ |
1279 | allowed = allowed && (ns->flags & USERNS_SETGROUPS_ALLOWED); | |
f0d62aec | 1280 | mutex_unlock(&userns_state_mutex); |
273d2c67 EB |
1281 | |
1282 | return allowed; | |
1283 | } | |
1284 | ||
d07b846f | 1285 | /* |
a2b42626 EB |
1286 | * Returns true if @child is the same namespace or a descendant of |
1287 | * @ancestor. | |
d07b846f | 1288 | */ |
a2b42626 EB |
1289 | bool in_userns(const struct user_namespace *ancestor, |
1290 | const struct user_namespace *child) | |
1291 | { | |
1292 | const struct user_namespace *ns; | |
1293 | for (ns = child; ns->level > ancestor->level; ns = ns->parent) | |
1294 | ; | |
1295 | return (ns == ancestor); | |
1296 | } | |
1297 | ||
d07b846f SF |
1298 | bool current_in_userns(const struct user_namespace *target_ns) |
1299 | { | |
a2b42626 | 1300 | return in_userns(target_ns, current_user_ns()); |
d07b846f | 1301 | } |
73f03c2b | 1302 | EXPORT_SYMBOL(current_in_userns); |
d07b846f | 1303 | |
3c041184 AV |
1304 | static inline struct user_namespace *to_user_ns(struct ns_common *ns) |
1305 | { | |
1306 | return container_of(ns, struct user_namespace, ns); | |
1307 | } | |
1308 | ||
64964528 | 1309 | static struct ns_common *userns_get(struct task_struct *task) |
cde1975b EB |
1310 | { |
1311 | struct user_namespace *user_ns; | |
1312 | ||
1313 | rcu_read_lock(); | |
1314 | user_ns = get_user_ns(__task_cred(task)->user_ns); | |
1315 | rcu_read_unlock(); | |
1316 | ||
3c041184 | 1317 | return user_ns ? &user_ns->ns : NULL; |
cde1975b EB |
1318 | } |
1319 | ||
64964528 | 1320 | static void userns_put(struct ns_common *ns) |
cde1975b | 1321 | { |
3c041184 | 1322 | put_user_ns(to_user_ns(ns)); |
cde1975b EB |
1323 | } |
1324 | ||
f2a8d52e | 1325 | static int userns_install(struct nsset *nsset, struct ns_common *ns) |
cde1975b | 1326 | { |
3c041184 | 1327 | struct user_namespace *user_ns = to_user_ns(ns); |
cde1975b EB |
1328 | struct cred *cred; |
1329 | ||
1330 | /* Don't allow gaining capabilities by reentering | |
1331 | * the same user namespace. | |
1332 | */ | |
1333 | if (user_ns == current_user_ns()) | |
1334 | return -EINVAL; | |
1335 | ||
faf00da5 EB |
1336 | /* Tasks that share a thread group must share a user namespace */ |
1337 | if (!thread_group_empty(current)) | |
cde1975b EB |
1338 | return -EINVAL; |
1339 | ||
e66eded8 EB |
1340 | if (current->fs->users != 1) |
1341 | return -EINVAL; | |
1342 | ||
cde1975b EB |
1343 | if (!ns_capable(user_ns, CAP_SYS_ADMIN)) |
1344 | return -EPERM; | |
1345 | ||
f2a8d52e | 1346 | cred = nsset_cred(nsset); |
cde1975b | 1347 | if (!cred) |
f2a8d52e | 1348 | return -EINVAL; |
cde1975b EB |
1349 | |
1350 | put_user_ns(cred->user_ns); | |
1351 | set_cred_user_ns(cred, get_user_ns(user_ns)); | |
1352 | ||
905ae01c AG |
1353 | if (set_cred_ucounts(cred) < 0) |
1354 | return -EINVAL; | |
1355 | ||
f2a8d52e | 1356 | return 0; |
cde1975b EB |
1357 | } |
1358 | ||
bcac25a5 AV |
1359 | struct ns_common *ns_get_owner(struct ns_common *ns) |
1360 | { | |
1361 | struct user_namespace *my_user_ns = current_user_ns(); | |
1362 | struct user_namespace *owner, *p; | |
1363 | ||
1364 | /* See if the owner is in the current user namespace */ | |
1365 | owner = p = ns->ops->owner(ns); | |
1366 | for (;;) { | |
1367 | if (!p) | |
1368 | return ERR_PTR(-EPERM); | |
1369 | if (p == my_user_ns) | |
1370 | break; | |
1371 | p = p->parent; | |
1372 | } | |
1373 | ||
1374 | return &get_user_ns(owner)->ns; | |
1375 | } | |
1376 | ||
1377 | static struct user_namespace *userns_owner(struct ns_common *ns) | |
1378 | { | |
1379 | return to_user_ns(ns)->parent; | |
1380 | } | |
1381 | ||
cde1975b EB |
1382 | const struct proc_ns_operations userns_operations = { |
1383 | .name = "user", | |
1384 | .type = CLONE_NEWUSER, | |
1385 | .get = userns_get, | |
1386 | .put = userns_put, | |
1387 | .install = userns_install, | |
bcac25a5 | 1388 | .owner = userns_owner, |
a7306ed8 | 1389 | .get_parent = ns_get_owner, |
cde1975b EB |
1390 | }; |
1391 | ||
6164281a PE |
1392 | static __init int user_namespaces_init(void) |
1393 | { | |
30acd0bd | 1394 | user_ns_cachep = KMEM_CACHE(user_namespace, SLAB_PANIC | SLAB_ACCOUNT); |
6164281a PE |
1395 | return 0; |
1396 | } | |
c96d6660 | 1397 | subsys_initcall(user_namespaces_init); |