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> |
acce292c | 11 | #include <linux/user_namespace.h> |
0bb80f24 | 12 | #include <linux/proc_ns.h> |
5c1469de | 13 | #include <linux/highuid.h> |
18b6e041 | 14 | #include <linux/cred.h> |
973c5914 | 15 | #include <linux/securebits.h> |
22d917d8 EB |
16 | #include <linux/keyctl.h> |
17 | #include <linux/key-type.h> | |
18 | #include <keys/user-type.h> | |
19 | #include <linux/seq_file.h> | |
20 | #include <linux/fs.h> | |
21 | #include <linux/uaccess.h> | |
22 | #include <linux/ctype.h> | |
f76d207a | 23 | #include <linux/projid.h> |
e66eded8 | 24 | #include <linux/fs_struct.h> |
acce292c | 25 | |
6164281a | 26 | static struct kmem_cache *user_ns_cachep __read_mostly; |
f0d62aec | 27 | static DEFINE_MUTEX(userns_state_mutex); |
6164281a | 28 | |
6708075f EB |
29 | static bool new_idmap_permitted(const struct file *file, |
30 | struct user_namespace *ns, int cap_setid, | |
22d917d8 EB |
31 | struct uid_gid_map *map); |
32 | ||
cde1975b EB |
33 | static void set_cred_user_ns(struct cred *cred, struct user_namespace *user_ns) |
34 | { | |
35 | /* Start with the same capabilities as init but useless for doing | |
36 | * anything as the capabilities are bound to the new user namespace. | |
37 | */ | |
38 | cred->securebits = SECUREBITS_DEFAULT; | |
39 | cred->cap_inheritable = CAP_EMPTY_SET; | |
40 | cred->cap_permitted = CAP_FULL_SET; | |
41 | cred->cap_effective = CAP_FULL_SET; | |
42 | cred->cap_bset = CAP_FULL_SET; | |
43 | #ifdef CONFIG_KEYS | |
44 | key_put(cred->request_key_auth); | |
45 | cred->request_key_auth = NULL; | |
46 | #endif | |
47 | /* tgcred will be cleared in our caller bc CLONE_THREAD won't be set */ | |
48 | cred->user_ns = user_ns; | |
49 | } | |
50 | ||
77ec739d | 51 | /* |
18b6e041 SH |
52 | * Create a new user namespace, deriving the creator from the user in the |
53 | * passed credentials, and replacing that user with the new root user for the | |
54 | * new namespace. | |
55 | * | |
56 | * This is called by copy_creds(), which will finish setting the target task's | |
57 | * credentials. | |
77ec739d | 58 | */ |
18b6e041 | 59 | int create_user_ns(struct cred *new) |
77ec739d | 60 | { |
0093ccb6 | 61 | struct user_namespace *ns, *parent_ns = new->user_ns; |
078de5f7 EB |
62 | kuid_t owner = new->euid; |
63 | kgid_t group = new->egid; | |
98f842e6 | 64 | int ret; |
783291e6 | 65 | |
8742f229 ON |
66 | if (parent_ns->level > 32) |
67 | return -EUSERS; | |
68 | ||
3151527e EB |
69 | /* |
70 | * Verify that we can not violate the policy of which files | |
71 | * may be accessed that is specified by the root directory, | |
72 | * by verifing that the root directory is at the root of the | |
73 | * mount namespace which allows all files to be accessed. | |
74 | */ | |
75 | if (current_chrooted()) | |
76 | return -EPERM; | |
77 | ||
783291e6 EB |
78 | /* The creator needs a mapping in the parent user namespace |
79 | * or else we won't be able to reasonably tell userspace who | |
80 | * created a user_namespace. | |
81 | */ | |
82 | if (!kuid_has_mapping(parent_ns, owner) || | |
83 | !kgid_has_mapping(parent_ns, group)) | |
84 | return -EPERM; | |
77ec739d | 85 | |
22d917d8 | 86 | ns = kmem_cache_zalloc(user_ns_cachep, GFP_KERNEL); |
77ec739d | 87 | if (!ns) |
18b6e041 | 88 | return -ENOMEM; |
77ec739d | 89 | |
98f842e6 EB |
90 | ret = proc_alloc_inum(&ns->proc_inum); |
91 | if (ret) { | |
92 | kmem_cache_free(user_ns_cachep, ns); | |
93 | return ret; | |
94 | } | |
95 | ||
c61a2810 | 96 | atomic_set(&ns->count, 1); |
cde1975b | 97 | /* Leave the new->user_ns reference with the new user namespace. */ |
aeb3ae9d | 98 | ns->parent = parent_ns; |
8742f229 | 99 | ns->level = parent_ns->level + 1; |
783291e6 EB |
100 | ns->owner = owner; |
101 | ns->group = group; | |
22d917d8 | 102 | |
9cc46516 EB |
103 | /* Inherit USERNS_SETGROUPS_ALLOWED from our parent */ |
104 | mutex_lock(&userns_state_mutex); | |
105 | ns->flags = parent_ns->flags; | |
106 | mutex_unlock(&userns_state_mutex); | |
107 | ||
cde1975b | 108 | set_cred_user_ns(new, ns); |
0093ccb6 | 109 | |
f36f8c75 DH |
110 | #ifdef CONFIG_PERSISTENT_KEYRINGS |
111 | init_rwsem(&ns->persistent_keyring_register_sem); | |
112 | #endif | |
18b6e041 | 113 | return 0; |
acce292c CLG |
114 | } |
115 | ||
b2e0d987 EB |
116 | int unshare_userns(unsigned long unshare_flags, struct cred **new_cred) |
117 | { | |
118 | struct cred *cred; | |
6160968c | 119 | int err = -ENOMEM; |
b2e0d987 EB |
120 | |
121 | if (!(unshare_flags & CLONE_NEWUSER)) | |
122 | return 0; | |
123 | ||
124 | cred = prepare_creds(); | |
6160968c ON |
125 | if (cred) { |
126 | err = create_user_ns(cred); | |
127 | if (err) | |
128 | put_cred(cred); | |
129 | else | |
130 | *new_cred = cred; | |
131 | } | |
b2e0d987 | 132 | |
6160968c | 133 | return err; |
b2e0d987 EB |
134 | } |
135 | ||
c61a2810 | 136 | void free_user_ns(struct user_namespace *ns) |
acce292c | 137 | { |
c61a2810 | 138 | struct user_namespace *parent; |
783291e6 | 139 | |
c61a2810 EB |
140 | do { |
141 | parent = ns->parent; | |
f36f8c75 DH |
142 | #ifdef CONFIG_PERSISTENT_KEYRINGS |
143 | key_put(ns->persistent_keyring_register); | |
144 | #endif | |
c61a2810 EB |
145 | proc_free_inum(ns->proc_inum); |
146 | kmem_cache_free(user_ns_cachep, ns); | |
147 | ns = parent; | |
148 | } while (atomic_dec_and_test(&parent->count)); | |
acce292c | 149 | } |
6a3fd92e | 150 | EXPORT_SYMBOL(free_user_ns); |
5c1469de | 151 | |
22d917d8 | 152 | static u32 map_id_range_down(struct uid_gid_map *map, u32 id, u32 count) |
5c1469de | 153 | { |
22d917d8 EB |
154 | unsigned idx, extents; |
155 | u32 first, last, id2; | |
5c1469de | 156 | |
22d917d8 | 157 | id2 = id + count - 1; |
5c1469de | 158 | |
22d917d8 EB |
159 | /* Find the matching extent */ |
160 | extents = map->nr_extents; | |
e79323bd | 161 | smp_rmb(); |
22d917d8 EB |
162 | for (idx = 0; idx < extents; idx++) { |
163 | first = map->extent[idx].first; | |
164 | last = first + map->extent[idx].count - 1; | |
165 | if (id >= first && id <= last && | |
166 | (id2 >= first && id2 <= last)) | |
167 | break; | |
168 | } | |
169 | /* Map the id or note failure */ | |
170 | if (idx < extents) | |
171 | id = (id - first) + map->extent[idx].lower_first; | |
172 | else | |
173 | id = (u32) -1; | |
174 | ||
175 | return id; | |
176 | } | |
177 | ||
178 | static u32 map_id_down(struct uid_gid_map *map, u32 id) | |
179 | { | |
180 | unsigned idx, extents; | |
181 | u32 first, last; | |
182 | ||
183 | /* Find the matching extent */ | |
184 | extents = map->nr_extents; | |
e79323bd | 185 | smp_rmb(); |
22d917d8 EB |
186 | for (idx = 0; idx < extents; idx++) { |
187 | first = map->extent[idx].first; | |
188 | last = first + map->extent[idx].count - 1; | |
189 | if (id >= first && id <= last) | |
190 | break; | |
191 | } | |
192 | /* Map the id or note failure */ | |
193 | if (idx < extents) | |
194 | id = (id - first) + map->extent[idx].lower_first; | |
195 | else | |
196 | id = (u32) -1; | |
197 | ||
198 | return id; | |
199 | } | |
200 | ||
201 | static u32 map_id_up(struct uid_gid_map *map, u32 id) | |
202 | { | |
203 | unsigned idx, extents; | |
204 | u32 first, last; | |
205 | ||
206 | /* Find the matching extent */ | |
207 | extents = map->nr_extents; | |
e79323bd | 208 | smp_rmb(); |
22d917d8 EB |
209 | for (idx = 0; idx < extents; idx++) { |
210 | first = map->extent[idx].lower_first; | |
211 | last = first + map->extent[idx].count - 1; | |
212 | if (id >= first && id <= last) | |
213 | break; | |
5c1469de | 214 | } |
22d917d8 EB |
215 | /* Map the id or note failure */ |
216 | if (idx < extents) | |
217 | id = (id - first) + map->extent[idx].first; | |
218 | else | |
219 | id = (u32) -1; | |
220 | ||
221 | return id; | |
222 | } | |
223 | ||
224 | /** | |
225 | * make_kuid - Map a user-namespace uid pair into a kuid. | |
226 | * @ns: User namespace that the uid is in | |
227 | * @uid: User identifier | |
228 | * | |
229 | * Maps a user-namespace uid pair into a kernel internal kuid, | |
230 | * and returns that kuid. | |
231 | * | |
232 | * When there is no mapping defined for the user-namespace uid | |
233 | * pair INVALID_UID is returned. Callers are expected to test | |
b080e047 | 234 | * for and handle INVALID_UID being returned. INVALID_UID |
22d917d8 EB |
235 | * may be tested for using uid_valid(). |
236 | */ | |
237 | kuid_t make_kuid(struct user_namespace *ns, uid_t uid) | |
238 | { | |
239 | /* Map the uid to a global kernel uid */ | |
240 | return KUIDT_INIT(map_id_down(&ns->uid_map, uid)); | |
241 | } | |
242 | EXPORT_SYMBOL(make_kuid); | |
243 | ||
244 | /** | |
245 | * from_kuid - Create a uid from a kuid user-namespace pair. | |
246 | * @targ: The user namespace we want a uid in. | |
247 | * @kuid: The kernel internal uid to start with. | |
248 | * | |
249 | * Map @kuid into the user-namespace specified by @targ and | |
250 | * return the resulting uid. | |
251 | * | |
252 | * There is always a mapping into the initial user_namespace. | |
253 | * | |
254 | * If @kuid has no mapping in @targ (uid_t)-1 is returned. | |
255 | */ | |
256 | uid_t from_kuid(struct user_namespace *targ, kuid_t kuid) | |
257 | { | |
258 | /* Map the uid from a global kernel uid */ | |
259 | return map_id_up(&targ->uid_map, __kuid_val(kuid)); | |
260 | } | |
261 | EXPORT_SYMBOL(from_kuid); | |
262 | ||
263 | /** | |
264 | * from_kuid_munged - Create a uid from a kuid user-namespace pair. | |
265 | * @targ: The user namespace we want a uid in. | |
266 | * @kuid: The kernel internal uid to start with. | |
267 | * | |
268 | * Map @kuid into the user-namespace specified by @targ and | |
269 | * return the resulting uid. | |
270 | * | |
271 | * There is always a mapping into the initial user_namespace. | |
272 | * | |
273 | * Unlike from_kuid from_kuid_munged never fails and always | |
274 | * returns a valid uid. This makes from_kuid_munged appropriate | |
275 | * for use in syscalls like stat and getuid where failing the | |
276 | * system call and failing to provide a valid uid are not an | |
277 | * options. | |
278 | * | |
279 | * If @kuid has no mapping in @targ overflowuid is returned. | |
280 | */ | |
281 | uid_t from_kuid_munged(struct user_namespace *targ, kuid_t kuid) | |
282 | { | |
283 | uid_t uid; | |
284 | uid = from_kuid(targ, kuid); | |
285 | ||
286 | if (uid == (uid_t) -1) | |
287 | uid = overflowuid; | |
288 | return uid; | |
289 | } | |
290 | EXPORT_SYMBOL(from_kuid_munged); | |
291 | ||
292 | /** | |
293 | * make_kgid - Map a user-namespace gid pair into a kgid. | |
294 | * @ns: User namespace that the gid is in | |
68a9a435 | 295 | * @gid: group identifier |
22d917d8 EB |
296 | * |
297 | * Maps a user-namespace gid pair into a kernel internal kgid, | |
298 | * and returns that kgid. | |
299 | * | |
300 | * When there is no mapping defined for the user-namespace gid | |
301 | * pair INVALID_GID is returned. Callers are expected to test | |
302 | * for and handle INVALID_GID being returned. INVALID_GID may be | |
303 | * tested for using gid_valid(). | |
304 | */ | |
305 | kgid_t make_kgid(struct user_namespace *ns, gid_t gid) | |
306 | { | |
307 | /* Map the gid to a global kernel gid */ | |
308 | return KGIDT_INIT(map_id_down(&ns->gid_map, gid)); | |
309 | } | |
310 | EXPORT_SYMBOL(make_kgid); | |
311 | ||
312 | /** | |
313 | * from_kgid - Create a gid from a kgid user-namespace pair. | |
314 | * @targ: The user namespace we want a gid in. | |
315 | * @kgid: The kernel internal gid to start with. | |
316 | * | |
317 | * Map @kgid into the user-namespace specified by @targ and | |
318 | * return the resulting gid. | |
319 | * | |
320 | * There is always a mapping into the initial user_namespace. | |
321 | * | |
322 | * If @kgid has no mapping in @targ (gid_t)-1 is returned. | |
323 | */ | |
324 | gid_t from_kgid(struct user_namespace *targ, kgid_t kgid) | |
325 | { | |
326 | /* Map the gid from a global kernel gid */ | |
327 | return map_id_up(&targ->gid_map, __kgid_val(kgid)); | |
328 | } | |
329 | EXPORT_SYMBOL(from_kgid); | |
330 | ||
331 | /** | |
332 | * from_kgid_munged - Create a gid from a kgid user-namespace pair. | |
333 | * @targ: The user namespace we want a gid in. | |
334 | * @kgid: The kernel internal gid to start with. | |
335 | * | |
336 | * Map @kgid into the user-namespace specified by @targ and | |
337 | * return the resulting gid. | |
338 | * | |
339 | * There is always a mapping into the initial user_namespace. | |
340 | * | |
341 | * Unlike from_kgid from_kgid_munged never fails and always | |
342 | * returns a valid gid. This makes from_kgid_munged appropriate | |
343 | * for use in syscalls like stat and getgid where failing the | |
344 | * system call and failing to provide a valid gid are not options. | |
345 | * | |
346 | * If @kgid has no mapping in @targ overflowgid is returned. | |
347 | */ | |
348 | gid_t from_kgid_munged(struct user_namespace *targ, kgid_t kgid) | |
349 | { | |
350 | gid_t gid; | |
351 | gid = from_kgid(targ, kgid); | |
352 | ||
353 | if (gid == (gid_t) -1) | |
354 | gid = overflowgid; | |
355 | return gid; | |
356 | } | |
357 | EXPORT_SYMBOL(from_kgid_munged); | |
358 | ||
f76d207a EB |
359 | /** |
360 | * make_kprojid - Map a user-namespace projid pair into a kprojid. | |
361 | * @ns: User namespace that the projid is in | |
362 | * @projid: Project identifier | |
363 | * | |
364 | * Maps a user-namespace uid pair into a kernel internal kuid, | |
365 | * and returns that kuid. | |
366 | * | |
367 | * When there is no mapping defined for the user-namespace projid | |
368 | * pair INVALID_PROJID is returned. Callers are expected to test | |
369 | * for and handle handle INVALID_PROJID being returned. INVALID_PROJID | |
370 | * may be tested for using projid_valid(). | |
371 | */ | |
372 | kprojid_t make_kprojid(struct user_namespace *ns, projid_t projid) | |
373 | { | |
374 | /* Map the uid to a global kernel uid */ | |
375 | return KPROJIDT_INIT(map_id_down(&ns->projid_map, projid)); | |
376 | } | |
377 | EXPORT_SYMBOL(make_kprojid); | |
378 | ||
379 | /** | |
380 | * from_kprojid - Create a projid from a kprojid user-namespace pair. | |
381 | * @targ: The user namespace we want a projid in. | |
382 | * @kprojid: The kernel internal project identifier to start with. | |
383 | * | |
384 | * Map @kprojid into the user-namespace specified by @targ and | |
385 | * return the resulting projid. | |
386 | * | |
387 | * There is always a mapping into the initial user_namespace. | |
388 | * | |
389 | * If @kprojid has no mapping in @targ (projid_t)-1 is returned. | |
390 | */ | |
391 | projid_t from_kprojid(struct user_namespace *targ, kprojid_t kprojid) | |
392 | { | |
393 | /* Map the uid from a global kernel uid */ | |
394 | return map_id_up(&targ->projid_map, __kprojid_val(kprojid)); | |
395 | } | |
396 | EXPORT_SYMBOL(from_kprojid); | |
397 | ||
398 | /** | |
399 | * from_kprojid_munged - Create a projiid from a kprojid user-namespace pair. | |
400 | * @targ: The user namespace we want a projid in. | |
401 | * @kprojid: The kernel internal projid to start with. | |
402 | * | |
403 | * Map @kprojid into the user-namespace specified by @targ and | |
404 | * return the resulting projid. | |
405 | * | |
406 | * There is always a mapping into the initial user_namespace. | |
407 | * | |
408 | * Unlike from_kprojid from_kprojid_munged never fails and always | |
409 | * returns a valid projid. This makes from_kprojid_munged | |
410 | * appropriate for use in syscalls like stat and where | |
411 | * failing the system call and failing to provide a valid projid are | |
412 | * not an options. | |
413 | * | |
414 | * If @kprojid has no mapping in @targ OVERFLOW_PROJID is returned. | |
415 | */ | |
416 | projid_t from_kprojid_munged(struct user_namespace *targ, kprojid_t kprojid) | |
417 | { | |
418 | projid_t projid; | |
419 | projid = from_kprojid(targ, kprojid); | |
420 | ||
421 | if (projid == (projid_t) -1) | |
422 | projid = OVERFLOW_PROJID; | |
423 | return projid; | |
424 | } | |
425 | EXPORT_SYMBOL(from_kprojid_munged); | |
426 | ||
427 | ||
22d917d8 EB |
428 | static int uid_m_show(struct seq_file *seq, void *v) |
429 | { | |
430 | struct user_namespace *ns = seq->private; | |
431 | struct uid_gid_extent *extent = v; | |
432 | struct user_namespace *lower_ns; | |
433 | uid_t lower; | |
5c1469de | 434 | |
c450f371 | 435 | lower_ns = seq_user_ns(seq); |
22d917d8 EB |
436 | if ((lower_ns == ns) && lower_ns->parent) |
437 | lower_ns = lower_ns->parent; | |
438 | ||
439 | lower = from_kuid(lower_ns, KUIDT_INIT(extent->lower_first)); | |
440 | ||
441 | seq_printf(seq, "%10u %10u %10u\n", | |
442 | extent->first, | |
443 | lower, | |
444 | extent->count); | |
445 | ||
446 | return 0; | |
5c1469de EB |
447 | } |
448 | ||
22d917d8 | 449 | static int gid_m_show(struct seq_file *seq, void *v) |
5c1469de | 450 | { |
22d917d8 EB |
451 | struct user_namespace *ns = seq->private; |
452 | struct uid_gid_extent *extent = v; | |
453 | struct user_namespace *lower_ns; | |
454 | gid_t lower; | |
5c1469de | 455 | |
c450f371 | 456 | lower_ns = seq_user_ns(seq); |
22d917d8 EB |
457 | if ((lower_ns == ns) && lower_ns->parent) |
458 | lower_ns = lower_ns->parent; | |
5c1469de | 459 | |
22d917d8 EB |
460 | lower = from_kgid(lower_ns, KGIDT_INIT(extent->lower_first)); |
461 | ||
462 | seq_printf(seq, "%10u %10u %10u\n", | |
463 | extent->first, | |
464 | lower, | |
465 | extent->count); | |
466 | ||
467 | return 0; | |
468 | } | |
469 | ||
f76d207a EB |
470 | static int projid_m_show(struct seq_file *seq, void *v) |
471 | { | |
472 | struct user_namespace *ns = seq->private; | |
473 | struct uid_gid_extent *extent = v; | |
474 | struct user_namespace *lower_ns; | |
475 | projid_t lower; | |
476 | ||
477 | lower_ns = seq_user_ns(seq); | |
478 | if ((lower_ns == ns) && lower_ns->parent) | |
479 | lower_ns = lower_ns->parent; | |
480 | ||
481 | lower = from_kprojid(lower_ns, KPROJIDT_INIT(extent->lower_first)); | |
482 | ||
483 | seq_printf(seq, "%10u %10u %10u\n", | |
484 | extent->first, | |
485 | lower, | |
486 | extent->count); | |
487 | ||
488 | return 0; | |
489 | } | |
490 | ||
68a9a435 FF |
491 | static void *m_start(struct seq_file *seq, loff_t *ppos, |
492 | struct uid_gid_map *map) | |
22d917d8 EB |
493 | { |
494 | struct uid_gid_extent *extent = NULL; | |
495 | loff_t pos = *ppos; | |
496 | ||
497 | if (pos < map->nr_extents) | |
498 | extent = &map->extent[pos]; | |
499 | ||
500 | return extent; | |
501 | } | |
502 | ||
503 | static void *uid_m_start(struct seq_file *seq, loff_t *ppos) | |
504 | { | |
505 | struct user_namespace *ns = seq->private; | |
506 | ||
507 | return m_start(seq, ppos, &ns->uid_map); | |
508 | } | |
509 | ||
510 | static void *gid_m_start(struct seq_file *seq, loff_t *ppos) | |
511 | { | |
512 | struct user_namespace *ns = seq->private; | |
513 | ||
514 | return m_start(seq, ppos, &ns->gid_map); | |
515 | } | |
516 | ||
f76d207a EB |
517 | static void *projid_m_start(struct seq_file *seq, loff_t *ppos) |
518 | { | |
519 | struct user_namespace *ns = seq->private; | |
520 | ||
521 | return m_start(seq, ppos, &ns->projid_map); | |
522 | } | |
523 | ||
22d917d8 EB |
524 | static void *m_next(struct seq_file *seq, void *v, loff_t *pos) |
525 | { | |
526 | (*pos)++; | |
527 | return seq->op->start(seq, pos); | |
528 | } | |
529 | ||
530 | static void m_stop(struct seq_file *seq, void *v) | |
531 | { | |
532 | return; | |
533 | } | |
534 | ||
ccf94f1b | 535 | const struct seq_operations proc_uid_seq_operations = { |
22d917d8 EB |
536 | .start = uid_m_start, |
537 | .stop = m_stop, | |
538 | .next = m_next, | |
539 | .show = uid_m_show, | |
540 | }; | |
541 | ||
ccf94f1b | 542 | const struct seq_operations proc_gid_seq_operations = { |
22d917d8 EB |
543 | .start = gid_m_start, |
544 | .stop = m_stop, | |
545 | .next = m_next, | |
546 | .show = gid_m_show, | |
547 | }; | |
548 | ||
ccf94f1b | 549 | const struct seq_operations proc_projid_seq_operations = { |
f76d207a EB |
550 | .start = projid_m_start, |
551 | .stop = m_stop, | |
552 | .next = m_next, | |
553 | .show = projid_m_show, | |
554 | }; | |
555 | ||
68a9a435 FF |
556 | static bool mappings_overlap(struct uid_gid_map *new_map, |
557 | struct uid_gid_extent *extent) | |
0bd14b4f EB |
558 | { |
559 | u32 upper_first, lower_first, upper_last, lower_last; | |
560 | unsigned idx; | |
561 | ||
562 | upper_first = extent->first; | |
563 | lower_first = extent->lower_first; | |
564 | upper_last = upper_first + extent->count - 1; | |
565 | lower_last = lower_first + extent->count - 1; | |
566 | ||
567 | for (idx = 0; idx < new_map->nr_extents; idx++) { | |
568 | u32 prev_upper_first, prev_lower_first; | |
569 | u32 prev_upper_last, prev_lower_last; | |
570 | struct uid_gid_extent *prev; | |
571 | ||
572 | prev = &new_map->extent[idx]; | |
573 | ||
574 | prev_upper_first = prev->first; | |
575 | prev_lower_first = prev->lower_first; | |
576 | prev_upper_last = prev_upper_first + prev->count - 1; | |
577 | prev_lower_last = prev_lower_first + prev->count - 1; | |
578 | ||
579 | /* Does the upper range intersect a previous extent? */ | |
580 | if ((prev_upper_first <= upper_last) && | |
581 | (prev_upper_last >= upper_first)) | |
582 | return true; | |
583 | ||
584 | /* Does the lower range intersect a previous extent? */ | |
585 | if ((prev_lower_first <= lower_last) && | |
586 | (prev_lower_last >= lower_first)) | |
587 | return true; | |
588 | } | |
589 | return false; | |
590 | } | |
591 | ||
22d917d8 EB |
592 | static ssize_t map_write(struct file *file, const char __user *buf, |
593 | size_t count, loff_t *ppos, | |
594 | int cap_setid, | |
595 | struct uid_gid_map *map, | |
596 | struct uid_gid_map *parent_map) | |
597 | { | |
598 | struct seq_file *seq = file->private_data; | |
599 | struct user_namespace *ns = seq->private; | |
600 | struct uid_gid_map new_map; | |
601 | unsigned idx; | |
0bd14b4f | 602 | struct uid_gid_extent *extent = NULL; |
22d917d8 EB |
603 | unsigned long page = 0; |
604 | char *kbuf, *pos, *next_line; | |
605 | ssize_t ret = -EINVAL; | |
606 | ||
607 | /* | |
f0d62aec | 608 | * The userns_state_mutex serializes all writes to any given map. |
22d917d8 EB |
609 | * |
610 | * Any map is only ever written once. | |
611 | * | |
612 | * An id map fits within 1 cache line on most architectures. | |
613 | * | |
614 | * On read nothing needs to be done unless you are on an | |
615 | * architecture with a crazy cache coherency model like alpha. | |
616 | * | |
617 | * There is a one time data dependency between reading the | |
618 | * count of the extents and the values of the extents. The | |
619 | * desired behavior is to see the values of the extents that | |
620 | * were written before the count of the extents. | |
621 | * | |
622 | * To achieve this smp_wmb() is used on guarantee the write | |
e79323bd MP |
623 | * order and smp_rmb() is guaranteed that we don't have crazy |
624 | * architectures returning stale data. | |
22d917d8 | 625 | */ |
f0d62aec | 626 | mutex_lock(&userns_state_mutex); |
22d917d8 EB |
627 | |
628 | ret = -EPERM; | |
629 | /* Only allow one successful write to the map */ | |
630 | if (map->nr_extents != 0) | |
631 | goto out; | |
632 | ||
41c21e35 AL |
633 | /* |
634 | * Adjusting namespace settings requires capabilities on the target. | |
5c1469de | 635 | */ |
41c21e35 | 636 | if (cap_valid(cap_setid) && !file_ns_capable(file, ns, CAP_SYS_ADMIN)) |
22d917d8 EB |
637 | goto out; |
638 | ||
639 | /* Get a buffer */ | |
640 | ret = -ENOMEM; | |
641 | page = __get_free_page(GFP_TEMPORARY); | |
642 | kbuf = (char *) page; | |
643 | if (!page) | |
644 | goto out; | |
645 | ||
646 | /* Only allow <= page size writes at the beginning of the file */ | |
647 | ret = -EINVAL; | |
648 | if ((*ppos != 0) || (count >= PAGE_SIZE)) | |
649 | goto out; | |
650 | ||
651 | /* Slurp in the user data */ | |
652 | ret = -EFAULT; | |
653 | if (copy_from_user(kbuf, buf, count)) | |
654 | goto out; | |
655 | kbuf[count] = '\0'; | |
656 | ||
657 | /* Parse the user data */ | |
658 | ret = -EINVAL; | |
659 | pos = kbuf; | |
660 | new_map.nr_extents = 0; | |
68a9a435 | 661 | for (; pos; pos = next_line) { |
22d917d8 EB |
662 | extent = &new_map.extent[new_map.nr_extents]; |
663 | ||
664 | /* Find the end of line and ensure I don't look past it */ | |
665 | next_line = strchr(pos, '\n'); | |
666 | if (next_line) { | |
667 | *next_line = '\0'; | |
668 | next_line++; | |
669 | if (*next_line == '\0') | |
670 | next_line = NULL; | |
5c1469de | 671 | } |
22d917d8 EB |
672 | |
673 | pos = skip_spaces(pos); | |
674 | extent->first = simple_strtoul(pos, &pos, 10); | |
675 | if (!isspace(*pos)) | |
676 | goto out; | |
677 | ||
678 | pos = skip_spaces(pos); | |
679 | extent->lower_first = simple_strtoul(pos, &pos, 10); | |
680 | if (!isspace(*pos)) | |
681 | goto out; | |
682 | ||
683 | pos = skip_spaces(pos); | |
684 | extent->count = simple_strtoul(pos, &pos, 10); | |
685 | if (*pos && !isspace(*pos)) | |
686 | goto out; | |
687 | ||
688 | /* Verify there is not trailing junk on the line */ | |
689 | pos = skip_spaces(pos); | |
690 | if (*pos != '\0') | |
691 | goto out; | |
692 | ||
693 | /* Verify we have been given valid starting values */ | |
694 | if ((extent->first == (u32) -1) || | |
68a9a435 | 695 | (extent->lower_first == (u32) -1)) |
22d917d8 EB |
696 | goto out; |
697 | ||
68a9a435 FF |
698 | /* Verify count is not zero and does not cause the |
699 | * extent to wrap | |
700 | */ | |
22d917d8 EB |
701 | if ((extent->first + extent->count) <= extent->first) |
702 | goto out; | |
68a9a435 FF |
703 | if ((extent->lower_first + extent->count) <= |
704 | extent->lower_first) | |
22d917d8 EB |
705 | goto out; |
706 | ||
0bd14b4f EB |
707 | /* Do the ranges in extent overlap any previous extents? */ |
708 | if (mappings_overlap(&new_map, extent)) | |
22d917d8 EB |
709 | goto out; |
710 | ||
711 | new_map.nr_extents++; | |
22d917d8 EB |
712 | |
713 | /* Fail if the file contains too many extents */ | |
714 | if ((new_map.nr_extents == UID_GID_MAP_MAX_EXTENTS) && | |
715 | (next_line != NULL)) | |
716 | goto out; | |
5c1469de | 717 | } |
22d917d8 EB |
718 | /* Be very certaint the new map actually exists */ |
719 | if (new_map.nr_extents == 0) | |
720 | goto out; | |
721 | ||
722 | ret = -EPERM; | |
723 | /* Validate the user is allowed to use user id's mapped to. */ | |
6708075f | 724 | if (!new_idmap_permitted(file, ns, cap_setid, &new_map)) |
22d917d8 EB |
725 | goto out; |
726 | ||
727 | /* Map the lower ids from the parent user namespace to the | |
728 | * kernel global id space. | |
729 | */ | |
730 | for (idx = 0; idx < new_map.nr_extents; idx++) { | |
731 | u32 lower_first; | |
732 | extent = &new_map.extent[idx]; | |
733 | ||
734 | lower_first = map_id_range_down(parent_map, | |
735 | extent->lower_first, | |
736 | extent->count); | |
737 | ||
738 | /* Fail if we can not map the specified extent to | |
739 | * the kernel global id space. | |
740 | */ | |
741 | if (lower_first == (u32) -1) | |
742 | goto out; | |
743 | ||
744 | extent->lower_first = lower_first; | |
745 | } | |
746 | ||
747 | /* Install the map */ | |
748 | memcpy(map->extent, new_map.extent, | |
749 | new_map.nr_extents*sizeof(new_map.extent[0])); | |
750 | smp_wmb(); | |
751 | map->nr_extents = new_map.nr_extents; | |
752 | ||
753 | *ppos = count; | |
754 | ret = count; | |
755 | out: | |
f0d62aec | 756 | mutex_unlock(&userns_state_mutex); |
22d917d8 EB |
757 | if (page) |
758 | free_page(page); | |
759 | return ret; | |
760 | } | |
761 | ||
68a9a435 FF |
762 | ssize_t proc_uid_map_write(struct file *file, const char __user *buf, |
763 | size_t size, loff_t *ppos) | |
22d917d8 EB |
764 | { |
765 | struct seq_file *seq = file->private_data; | |
766 | struct user_namespace *ns = seq->private; | |
c450f371 | 767 | struct user_namespace *seq_ns = seq_user_ns(seq); |
22d917d8 EB |
768 | |
769 | if (!ns->parent) | |
770 | return -EPERM; | |
771 | ||
c450f371 EB |
772 | if ((seq_ns != ns) && (seq_ns != ns->parent)) |
773 | return -EPERM; | |
774 | ||
22d917d8 EB |
775 | return map_write(file, buf, size, ppos, CAP_SETUID, |
776 | &ns->uid_map, &ns->parent->uid_map); | |
777 | } | |
778 | ||
68a9a435 FF |
779 | ssize_t proc_gid_map_write(struct file *file, const char __user *buf, |
780 | size_t size, loff_t *ppos) | |
22d917d8 EB |
781 | { |
782 | struct seq_file *seq = file->private_data; | |
783 | struct user_namespace *ns = seq->private; | |
c450f371 | 784 | struct user_namespace *seq_ns = seq_user_ns(seq); |
22d917d8 EB |
785 | |
786 | if (!ns->parent) | |
787 | return -EPERM; | |
788 | ||
c450f371 EB |
789 | if ((seq_ns != ns) && (seq_ns != ns->parent)) |
790 | return -EPERM; | |
791 | ||
22d917d8 EB |
792 | return map_write(file, buf, size, ppos, CAP_SETGID, |
793 | &ns->gid_map, &ns->parent->gid_map); | |
794 | } | |
795 | ||
68a9a435 FF |
796 | ssize_t proc_projid_map_write(struct file *file, const char __user *buf, |
797 | size_t size, loff_t *ppos) | |
f76d207a EB |
798 | { |
799 | struct seq_file *seq = file->private_data; | |
800 | struct user_namespace *ns = seq->private; | |
801 | struct user_namespace *seq_ns = seq_user_ns(seq); | |
802 | ||
803 | if (!ns->parent) | |
804 | return -EPERM; | |
805 | ||
806 | if ((seq_ns != ns) && (seq_ns != ns->parent)) | |
807 | return -EPERM; | |
808 | ||
809 | /* Anyone can set any valid project id no capability needed */ | |
810 | return map_write(file, buf, size, ppos, -1, | |
811 | &ns->projid_map, &ns->parent->projid_map); | |
812 | } | |
813 | ||
68a9a435 | 814 | static bool new_idmap_permitted(const struct file *file, |
6708075f | 815 | struct user_namespace *ns, int cap_setid, |
22d917d8 EB |
816 | struct uid_gid_map *new_map) |
817 | { | |
f95d7918 | 818 | const struct cred *cred = file->f_cred; |
0542f17b EB |
819 | /* Don't allow mappings that would allow anything that wouldn't |
820 | * be allowed without the establishment of unprivileged mappings. | |
821 | */ | |
f95d7918 EB |
822 | if ((new_map->nr_extents == 1) && (new_map->extent[0].count == 1) && |
823 | uid_eq(ns->owner, cred->euid)) { | |
37657da3 EB |
824 | u32 id = new_map->extent[0].lower_first; |
825 | if (cap_setid == CAP_SETUID) { | |
826 | kuid_t uid = make_kuid(ns->parent, id); | |
f95d7918 | 827 | if (uid_eq(uid, cred->euid)) |
37657da3 | 828 | return true; |
37657da3 EB |
829 | } |
830 | } | |
831 | ||
f76d207a EB |
832 | /* Allow anyone to set a mapping that doesn't require privilege */ |
833 | if (!cap_valid(cap_setid)) | |
834 | return true; | |
835 | ||
22d917d8 EB |
836 | /* Allow the specified ids if we have the appropriate capability |
837 | * (CAP_SETUID or CAP_SETGID) over the parent user namespace. | |
6708075f | 838 | * And the opener of the id file also had the approprpiate capability. |
22d917d8 | 839 | */ |
6708075f EB |
840 | if (ns_capable(ns->parent, cap_setid) && |
841 | file_ns_capable(file, ns->parent, cap_setid)) | |
22d917d8 | 842 | return true; |
5c1469de | 843 | |
22d917d8 | 844 | return false; |
5c1469de | 845 | } |
6164281a | 846 | |
9cc46516 EB |
847 | int proc_setgroups_show(struct seq_file *seq, void *v) |
848 | { | |
849 | struct user_namespace *ns = seq->private; | |
850 | unsigned long userns_flags = ACCESS_ONCE(ns->flags); | |
851 | ||
852 | seq_printf(seq, "%s\n", | |
853 | (userns_flags & USERNS_SETGROUPS_ALLOWED) ? | |
854 | "allow" : "deny"); | |
855 | return 0; | |
856 | } | |
857 | ||
858 | ssize_t proc_setgroups_write(struct file *file, const char __user *buf, | |
859 | size_t count, loff_t *ppos) | |
860 | { | |
861 | struct seq_file *seq = file->private_data; | |
862 | struct user_namespace *ns = seq->private; | |
863 | char kbuf[8], *pos; | |
864 | bool setgroups_allowed; | |
865 | ssize_t ret; | |
866 | ||
867 | /* Only allow a very narrow range of strings to be written */ | |
868 | ret = -EINVAL; | |
869 | if ((*ppos != 0) || (count >= sizeof(kbuf))) | |
870 | goto out; | |
871 | ||
872 | /* What was written? */ | |
873 | ret = -EFAULT; | |
874 | if (copy_from_user(kbuf, buf, count)) | |
875 | goto out; | |
876 | kbuf[count] = '\0'; | |
877 | pos = kbuf; | |
878 | ||
879 | /* What is being requested? */ | |
880 | ret = -EINVAL; | |
881 | if (strncmp(pos, "allow", 5) == 0) { | |
882 | pos += 5; | |
883 | setgroups_allowed = true; | |
884 | } | |
885 | else if (strncmp(pos, "deny", 4) == 0) { | |
886 | pos += 4; | |
887 | setgroups_allowed = false; | |
888 | } | |
889 | else | |
890 | goto out; | |
891 | ||
892 | /* Verify there is not trailing junk on the line */ | |
893 | pos = skip_spaces(pos); | |
894 | if (*pos != '\0') | |
895 | goto out; | |
896 | ||
897 | ret = -EPERM; | |
898 | mutex_lock(&userns_state_mutex); | |
899 | if (setgroups_allowed) { | |
900 | /* Enabling setgroups after setgroups has been disabled | |
901 | * is not allowed. | |
902 | */ | |
903 | if (!(ns->flags & USERNS_SETGROUPS_ALLOWED)) | |
904 | goto out_unlock; | |
905 | } else { | |
906 | /* Permanently disabling setgroups after setgroups has | |
907 | * been enabled by writing the gid_map is not allowed. | |
908 | */ | |
909 | if (ns->gid_map.nr_extents != 0) | |
910 | goto out_unlock; | |
911 | ns->flags &= ~USERNS_SETGROUPS_ALLOWED; | |
912 | } | |
913 | mutex_unlock(&userns_state_mutex); | |
914 | ||
915 | /* Report a successful write */ | |
916 | *ppos = count; | |
917 | ret = count; | |
918 | out: | |
919 | return ret; | |
920 | out_unlock: | |
921 | mutex_unlock(&userns_state_mutex); | |
922 | goto out; | |
923 | } | |
924 | ||
273d2c67 EB |
925 | bool userns_may_setgroups(const struct user_namespace *ns) |
926 | { | |
927 | bool allowed; | |
928 | ||
f0d62aec | 929 | mutex_lock(&userns_state_mutex); |
273d2c67 EB |
930 | /* It is not safe to use setgroups until a gid mapping in |
931 | * the user namespace has been established. | |
932 | */ | |
933 | allowed = ns->gid_map.nr_extents != 0; | |
9cc46516 EB |
934 | /* Is setgroups allowed? */ |
935 | allowed = allowed && (ns->flags & USERNS_SETGROUPS_ALLOWED); | |
f0d62aec | 936 | mutex_unlock(&userns_state_mutex); |
273d2c67 EB |
937 | |
938 | return allowed; | |
939 | } | |
940 | ||
cde1975b EB |
941 | static void *userns_get(struct task_struct *task) |
942 | { | |
943 | struct user_namespace *user_ns; | |
944 | ||
945 | rcu_read_lock(); | |
946 | user_ns = get_user_ns(__task_cred(task)->user_ns); | |
947 | rcu_read_unlock(); | |
948 | ||
949 | return user_ns; | |
950 | } | |
951 | ||
952 | static void userns_put(void *ns) | |
953 | { | |
954 | put_user_ns(ns); | |
955 | } | |
956 | ||
957 | static int userns_install(struct nsproxy *nsproxy, void *ns) | |
958 | { | |
959 | struct user_namespace *user_ns = ns; | |
960 | struct cred *cred; | |
961 | ||
962 | /* Don't allow gaining capabilities by reentering | |
963 | * the same user namespace. | |
964 | */ | |
965 | if (user_ns == current_user_ns()) | |
966 | return -EINVAL; | |
967 | ||
5155040e | 968 | /* Threaded processes may not enter a different user namespace */ |
cde1975b EB |
969 | if (atomic_read(¤t->mm->mm_users) > 1) |
970 | return -EINVAL; | |
971 | ||
e66eded8 EB |
972 | if (current->fs->users != 1) |
973 | return -EINVAL; | |
974 | ||
cde1975b EB |
975 | if (!ns_capable(user_ns, CAP_SYS_ADMIN)) |
976 | return -EPERM; | |
977 | ||
978 | cred = prepare_creds(); | |
979 | if (!cred) | |
980 | return -ENOMEM; | |
981 | ||
982 | put_user_ns(cred->user_ns); | |
983 | set_cred_user_ns(cred, get_user_ns(user_ns)); | |
984 | ||
985 | return commit_creds(cred); | |
986 | } | |
987 | ||
98f842e6 EB |
988 | static unsigned int userns_inum(void *ns) |
989 | { | |
990 | struct user_namespace *user_ns = ns; | |
991 | return user_ns->proc_inum; | |
992 | } | |
993 | ||
cde1975b EB |
994 | const struct proc_ns_operations userns_operations = { |
995 | .name = "user", | |
996 | .type = CLONE_NEWUSER, | |
997 | .get = userns_get, | |
998 | .put = userns_put, | |
999 | .install = userns_install, | |
98f842e6 | 1000 | .inum = userns_inum, |
cde1975b EB |
1001 | }; |
1002 | ||
6164281a PE |
1003 | static __init int user_namespaces_init(void) |
1004 | { | |
1005 | user_ns_cachep = KMEM_CACHE(user_namespace, SLAB_PANIC); | |
1006 | return 0; | |
1007 | } | |
c96d6660 | 1008 | subsys_initcall(user_namespaces_init); |