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