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