hwmon: DA9055 HWMON driver
[linux/fpc-iii.git] / kernel / user_namespace.c
blob456a6b9fba34f3104bc9db64119acd9d3fdfa8df
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 */
8 #include <linux/export.h>
9 #include <linux/nsproxy.h>
10 #include <linux/slab.h>
11 #include <linux/user_namespace.h>
12 #include <linux/highuid.h>
13 #include <linux/cred.h>
14 #include <linux/securebits.h>
15 #include <linux/keyctl.h>
16 #include <linux/key-type.h>
17 #include <keys/user-type.h>
18 #include <linux/seq_file.h>
19 #include <linux/fs.h>
20 #include <linux/uaccess.h>
21 #include <linux/ctype.h>
22 #include <linux/projid.h>
24 static struct kmem_cache *user_ns_cachep __read_mostly;
26 static bool new_idmap_permitted(struct user_namespace *ns, int cap_setid,
27 struct uid_gid_map *map);
30 * Create a new user namespace, deriving the creator from the user in the
31 * passed credentials, and replacing that user with the new root user for the
32 * new namespace.
34 * This is called by copy_creds(), which will finish setting the target task's
35 * credentials.
37 int create_user_ns(struct cred *new)
39 struct user_namespace *ns, *parent_ns = new->user_ns;
40 kuid_t owner = new->euid;
41 kgid_t group = new->egid;
43 /* The creator needs a mapping in the parent user namespace
44 * or else we won't be able to reasonably tell userspace who
45 * created a user_namespace.
47 if (!kuid_has_mapping(parent_ns, owner) ||
48 !kgid_has_mapping(parent_ns, group))
49 return -EPERM;
51 ns = kmem_cache_zalloc(user_ns_cachep, GFP_KERNEL);
52 if (!ns)
53 return -ENOMEM;
55 kref_init(&ns->kref);
56 ns->parent = parent_ns;
57 ns->owner = owner;
58 ns->group = group;
60 /* Start with the same capabilities as init but useless for doing
61 * anything as the capabilities are bound to the new user namespace.
63 new->securebits = SECUREBITS_DEFAULT;
64 new->cap_inheritable = CAP_EMPTY_SET;
65 new->cap_permitted = CAP_FULL_SET;
66 new->cap_effective = CAP_FULL_SET;
67 new->cap_bset = CAP_FULL_SET;
68 #ifdef CONFIG_KEYS
69 key_put(new->request_key_auth);
70 new->request_key_auth = NULL;
71 #endif
72 /* tgcred will be cleared in our caller bc CLONE_THREAD won't be set */
74 /* Leave the new->user_ns reference with the new user namespace. */
75 /* Leave the reference to our user_ns with the new cred. */
76 new->user_ns = ns;
78 return 0;
81 void free_user_ns(struct kref *kref)
83 struct user_namespace *parent, *ns =
84 container_of(kref, struct user_namespace, kref);
86 parent = ns->parent;
87 kmem_cache_free(user_ns_cachep, ns);
88 put_user_ns(parent);
90 EXPORT_SYMBOL(free_user_ns);
92 static u32 map_id_range_down(struct uid_gid_map *map, u32 id, u32 count)
94 unsigned idx, extents;
95 u32 first, last, id2;
97 id2 = id + count - 1;
99 /* Find the matching extent */
100 extents = map->nr_extents;
101 smp_read_barrier_depends();
102 for (idx = 0; idx < extents; idx++) {
103 first = map->extent[idx].first;
104 last = first + map->extent[idx].count - 1;
105 if (id >= first && id <= last &&
106 (id2 >= first && id2 <= last))
107 break;
109 /* Map the id or note failure */
110 if (idx < extents)
111 id = (id - first) + map->extent[idx].lower_first;
112 else
113 id = (u32) -1;
115 return id;
118 static u32 map_id_down(struct uid_gid_map *map, u32 id)
120 unsigned idx, extents;
121 u32 first, last;
123 /* Find the matching extent */
124 extents = map->nr_extents;
125 smp_read_barrier_depends();
126 for (idx = 0; idx < extents; idx++) {
127 first = map->extent[idx].first;
128 last = first + map->extent[idx].count - 1;
129 if (id >= first && id <= last)
130 break;
132 /* Map the id or note failure */
133 if (idx < extents)
134 id = (id - first) + map->extent[idx].lower_first;
135 else
136 id = (u32) -1;
138 return id;
141 static u32 map_id_up(struct uid_gid_map *map, u32 id)
143 unsigned idx, extents;
144 u32 first, last;
146 /* Find the matching extent */
147 extents = map->nr_extents;
148 smp_read_barrier_depends();
149 for (idx = 0; idx < extents; idx++) {
150 first = map->extent[idx].lower_first;
151 last = first + map->extent[idx].count - 1;
152 if (id >= first && id <= last)
153 break;
155 /* Map the id or note failure */
156 if (idx < extents)
157 id = (id - first) + map->extent[idx].first;
158 else
159 id = (u32) -1;
161 return id;
165 * make_kuid - Map a user-namespace uid pair into a kuid.
166 * @ns: User namespace that the uid is in
167 * @uid: User identifier
169 * Maps a user-namespace uid pair into a kernel internal kuid,
170 * and returns that kuid.
172 * When there is no mapping defined for the user-namespace uid
173 * pair INVALID_UID is returned. Callers are expected to test
174 * for and handle handle INVALID_UID being returned. INVALID_UID
175 * may be tested for using uid_valid().
177 kuid_t make_kuid(struct user_namespace *ns, uid_t uid)
179 /* Map the uid to a global kernel uid */
180 return KUIDT_INIT(map_id_down(&ns->uid_map, uid));
182 EXPORT_SYMBOL(make_kuid);
185 * from_kuid - Create a uid from a kuid user-namespace pair.
186 * @targ: The user namespace we want a uid in.
187 * @kuid: The kernel internal uid to start with.
189 * Map @kuid into the user-namespace specified by @targ and
190 * return the resulting uid.
192 * There is always a mapping into the initial user_namespace.
194 * If @kuid has no mapping in @targ (uid_t)-1 is returned.
196 uid_t from_kuid(struct user_namespace *targ, kuid_t kuid)
198 /* Map the uid from a global kernel uid */
199 return map_id_up(&targ->uid_map, __kuid_val(kuid));
201 EXPORT_SYMBOL(from_kuid);
204 * from_kuid_munged - Create a uid from a kuid user-namespace pair.
205 * @targ: The user namespace we want a uid in.
206 * @kuid: The kernel internal uid to start with.
208 * Map @kuid into the user-namespace specified by @targ and
209 * return the resulting uid.
211 * There is always a mapping into the initial user_namespace.
213 * Unlike from_kuid from_kuid_munged never fails and always
214 * returns a valid uid. This makes from_kuid_munged appropriate
215 * for use in syscalls like stat and getuid where failing the
216 * system call and failing to provide a valid uid are not an
217 * options.
219 * If @kuid has no mapping in @targ overflowuid is returned.
221 uid_t from_kuid_munged(struct user_namespace *targ, kuid_t kuid)
223 uid_t uid;
224 uid = from_kuid(targ, kuid);
226 if (uid == (uid_t) -1)
227 uid = overflowuid;
228 return uid;
230 EXPORT_SYMBOL(from_kuid_munged);
233 * make_kgid - Map a user-namespace gid pair into a kgid.
234 * @ns: User namespace that the gid is in
235 * @uid: group identifier
237 * Maps a user-namespace gid pair into a kernel internal kgid,
238 * and returns that kgid.
240 * When there is no mapping defined for the user-namespace gid
241 * pair INVALID_GID is returned. Callers are expected to test
242 * for and handle INVALID_GID being returned. INVALID_GID may be
243 * tested for using gid_valid().
245 kgid_t make_kgid(struct user_namespace *ns, gid_t gid)
247 /* Map the gid to a global kernel gid */
248 return KGIDT_INIT(map_id_down(&ns->gid_map, gid));
250 EXPORT_SYMBOL(make_kgid);
253 * from_kgid - Create a gid from a kgid user-namespace pair.
254 * @targ: The user namespace we want a gid in.
255 * @kgid: The kernel internal gid to start with.
257 * Map @kgid into the user-namespace specified by @targ and
258 * return the resulting gid.
260 * There is always a mapping into the initial user_namespace.
262 * If @kgid has no mapping in @targ (gid_t)-1 is returned.
264 gid_t from_kgid(struct user_namespace *targ, kgid_t kgid)
266 /* Map the gid from a global kernel gid */
267 return map_id_up(&targ->gid_map, __kgid_val(kgid));
269 EXPORT_SYMBOL(from_kgid);
272 * from_kgid_munged - Create a gid from a kgid user-namespace pair.
273 * @targ: The user namespace we want a gid in.
274 * @kgid: The kernel internal gid to start with.
276 * Map @kgid into the user-namespace specified by @targ and
277 * return the resulting gid.
279 * There is always a mapping into the initial user_namespace.
281 * Unlike from_kgid from_kgid_munged never fails and always
282 * returns a valid gid. This makes from_kgid_munged appropriate
283 * for use in syscalls like stat and getgid where failing the
284 * system call and failing to provide a valid gid are not options.
286 * If @kgid has no mapping in @targ overflowgid is returned.
288 gid_t from_kgid_munged(struct user_namespace *targ, kgid_t kgid)
290 gid_t gid;
291 gid = from_kgid(targ, kgid);
293 if (gid == (gid_t) -1)
294 gid = overflowgid;
295 return gid;
297 EXPORT_SYMBOL(from_kgid_munged);
300 * make_kprojid - Map a user-namespace projid pair into a kprojid.
301 * @ns: User namespace that the projid is in
302 * @projid: Project identifier
304 * Maps a user-namespace uid pair into a kernel internal kuid,
305 * and returns that kuid.
307 * When there is no mapping defined for the user-namespace projid
308 * pair INVALID_PROJID is returned. Callers are expected to test
309 * for and handle handle INVALID_PROJID being returned. INVALID_PROJID
310 * may be tested for using projid_valid().
312 kprojid_t make_kprojid(struct user_namespace *ns, projid_t projid)
314 /* Map the uid to a global kernel uid */
315 return KPROJIDT_INIT(map_id_down(&ns->projid_map, projid));
317 EXPORT_SYMBOL(make_kprojid);
320 * from_kprojid - Create a projid from a kprojid user-namespace pair.
321 * @targ: The user namespace we want a projid in.
322 * @kprojid: The kernel internal project identifier to start with.
324 * Map @kprojid into the user-namespace specified by @targ and
325 * return the resulting projid.
327 * There is always a mapping into the initial user_namespace.
329 * If @kprojid has no mapping in @targ (projid_t)-1 is returned.
331 projid_t from_kprojid(struct user_namespace *targ, kprojid_t kprojid)
333 /* Map the uid from a global kernel uid */
334 return map_id_up(&targ->projid_map, __kprojid_val(kprojid));
336 EXPORT_SYMBOL(from_kprojid);
339 * from_kprojid_munged - Create a projiid from a kprojid user-namespace pair.
340 * @targ: The user namespace we want a projid in.
341 * @kprojid: The kernel internal projid to start with.
343 * Map @kprojid into the user-namespace specified by @targ and
344 * return the resulting projid.
346 * There is always a mapping into the initial user_namespace.
348 * Unlike from_kprojid from_kprojid_munged never fails and always
349 * returns a valid projid. This makes from_kprojid_munged
350 * appropriate for use in syscalls like stat and where
351 * failing the system call and failing to provide a valid projid are
352 * not an options.
354 * If @kprojid has no mapping in @targ OVERFLOW_PROJID is returned.
356 projid_t from_kprojid_munged(struct user_namespace *targ, kprojid_t kprojid)
358 projid_t projid;
359 projid = from_kprojid(targ, kprojid);
361 if (projid == (projid_t) -1)
362 projid = OVERFLOW_PROJID;
363 return projid;
365 EXPORT_SYMBOL(from_kprojid_munged);
368 static int uid_m_show(struct seq_file *seq, void *v)
370 struct user_namespace *ns = seq->private;
371 struct uid_gid_extent *extent = v;
372 struct user_namespace *lower_ns;
373 uid_t lower;
375 lower_ns = current_user_ns();
376 if ((lower_ns == ns) && lower_ns->parent)
377 lower_ns = lower_ns->parent;
379 lower = from_kuid(lower_ns, KUIDT_INIT(extent->lower_first));
381 seq_printf(seq, "%10u %10u %10u\n",
382 extent->first,
383 lower,
384 extent->count);
386 return 0;
389 static int gid_m_show(struct seq_file *seq, void *v)
391 struct user_namespace *ns = seq->private;
392 struct uid_gid_extent *extent = v;
393 struct user_namespace *lower_ns;
394 gid_t lower;
396 lower_ns = current_user_ns();
397 if ((lower_ns == ns) && lower_ns->parent)
398 lower_ns = lower_ns->parent;
400 lower = from_kgid(lower_ns, KGIDT_INIT(extent->lower_first));
402 seq_printf(seq, "%10u %10u %10u\n",
403 extent->first,
404 lower,
405 extent->count);
407 return 0;
410 static int projid_m_show(struct seq_file *seq, void *v)
412 struct user_namespace *ns = seq->private;
413 struct uid_gid_extent *extent = v;
414 struct user_namespace *lower_ns;
415 projid_t lower;
417 lower_ns = seq_user_ns(seq);
418 if ((lower_ns == ns) && lower_ns->parent)
419 lower_ns = lower_ns->parent;
421 lower = from_kprojid(lower_ns, KPROJIDT_INIT(extent->lower_first));
423 seq_printf(seq, "%10u %10u %10u\n",
424 extent->first,
425 lower,
426 extent->count);
428 return 0;
431 static void *m_start(struct seq_file *seq, loff_t *ppos, struct uid_gid_map *map)
433 struct uid_gid_extent *extent = NULL;
434 loff_t pos = *ppos;
436 if (pos < map->nr_extents)
437 extent = &map->extent[pos];
439 return extent;
442 static void *uid_m_start(struct seq_file *seq, loff_t *ppos)
444 struct user_namespace *ns = seq->private;
446 return m_start(seq, ppos, &ns->uid_map);
449 static void *gid_m_start(struct seq_file *seq, loff_t *ppos)
451 struct user_namespace *ns = seq->private;
453 return m_start(seq, ppos, &ns->gid_map);
456 static void *projid_m_start(struct seq_file *seq, loff_t *ppos)
458 struct user_namespace *ns = seq->private;
460 return m_start(seq, ppos, &ns->projid_map);
463 static void *m_next(struct seq_file *seq, void *v, loff_t *pos)
465 (*pos)++;
466 return seq->op->start(seq, pos);
469 static void m_stop(struct seq_file *seq, void *v)
471 return;
474 struct seq_operations proc_uid_seq_operations = {
475 .start = uid_m_start,
476 .stop = m_stop,
477 .next = m_next,
478 .show = uid_m_show,
481 struct seq_operations proc_gid_seq_operations = {
482 .start = gid_m_start,
483 .stop = m_stop,
484 .next = m_next,
485 .show = gid_m_show,
488 struct seq_operations proc_projid_seq_operations = {
489 .start = projid_m_start,
490 .stop = m_stop,
491 .next = m_next,
492 .show = projid_m_show,
495 static DEFINE_MUTEX(id_map_mutex);
497 static ssize_t map_write(struct file *file, const char __user *buf,
498 size_t count, loff_t *ppos,
499 int cap_setid,
500 struct uid_gid_map *map,
501 struct uid_gid_map *parent_map)
503 struct seq_file *seq = file->private_data;
504 struct user_namespace *ns = seq->private;
505 struct uid_gid_map new_map;
506 unsigned idx;
507 struct uid_gid_extent *extent, *last = NULL;
508 unsigned long page = 0;
509 char *kbuf, *pos, *next_line;
510 ssize_t ret = -EINVAL;
513 * The id_map_mutex serializes all writes to any given map.
515 * Any map is only ever written once.
517 * An id map fits within 1 cache line on most architectures.
519 * On read nothing needs to be done unless you are on an
520 * architecture with a crazy cache coherency model like alpha.
522 * There is a one time data dependency between reading the
523 * count of the extents and the values of the extents. The
524 * desired behavior is to see the values of the extents that
525 * were written before the count of the extents.
527 * To achieve this smp_wmb() is used on guarantee the write
528 * order and smp_read_barrier_depends() is guaranteed that we
529 * don't have crazy architectures returning stale data.
532 mutex_lock(&id_map_mutex);
534 ret = -EPERM;
535 /* Only allow one successful write to the map */
536 if (map->nr_extents != 0)
537 goto out;
539 /* Require the appropriate privilege CAP_SETUID or CAP_SETGID
540 * over the user namespace in order to set the id mapping.
542 if (cap_valid(cap_setid) && !ns_capable(ns, cap_setid))
543 goto out;
545 /* Get a buffer */
546 ret = -ENOMEM;
547 page = __get_free_page(GFP_TEMPORARY);
548 kbuf = (char *) page;
549 if (!page)
550 goto out;
552 /* Only allow <= page size writes at the beginning of the file */
553 ret = -EINVAL;
554 if ((*ppos != 0) || (count >= PAGE_SIZE))
555 goto out;
557 /* Slurp in the user data */
558 ret = -EFAULT;
559 if (copy_from_user(kbuf, buf, count))
560 goto out;
561 kbuf[count] = '\0';
563 /* Parse the user data */
564 ret = -EINVAL;
565 pos = kbuf;
566 new_map.nr_extents = 0;
567 for (;pos; pos = next_line) {
568 extent = &new_map.extent[new_map.nr_extents];
570 /* Find the end of line and ensure I don't look past it */
571 next_line = strchr(pos, '\n');
572 if (next_line) {
573 *next_line = '\0';
574 next_line++;
575 if (*next_line == '\0')
576 next_line = NULL;
579 pos = skip_spaces(pos);
580 extent->first = simple_strtoul(pos, &pos, 10);
581 if (!isspace(*pos))
582 goto out;
584 pos = skip_spaces(pos);
585 extent->lower_first = simple_strtoul(pos, &pos, 10);
586 if (!isspace(*pos))
587 goto out;
589 pos = skip_spaces(pos);
590 extent->count = simple_strtoul(pos, &pos, 10);
591 if (*pos && !isspace(*pos))
592 goto out;
594 /* Verify there is not trailing junk on the line */
595 pos = skip_spaces(pos);
596 if (*pos != '\0')
597 goto out;
599 /* Verify we have been given valid starting values */
600 if ((extent->first == (u32) -1) ||
601 (extent->lower_first == (u32) -1 ))
602 goto out;
604 /* Verify count is not zero and does not cause the extent to wrap */
605 if ((extent->first + extent->count) <= extent->first)
606 goto out;
607 if ((extent->lower_first + extent->count) <= extent->lower_first)
608 goto out;
610 /* For now only accept extents that are strictly in order */
611 if (last &&
612 (((last->first + last->count) > extent->first) ||
613 ((last->lower_first + last->count) > extent->lower_first)))
614 goto out;
616 new_map.nr_extents++;
617 last = extent;
619 /* Fail if the file contains too many extents */
620 if ((new_map.nr_extents == UID_GID_MAP_MAX_EXTENTS) &&
621 (next_line != NULL))
622 goto out;
624 /* Be very certaint the new map actually exists */
625 if (new_map.nr_extents == 0)
626 goto out;
628 ret = -EPERM;
629 /* Validate the user is allowed to use user id's mapped to. */
630 if (!new_idmap_permitted(ns, cap_setid, &new_map))
631 goto out;
633 /* Map the lower ids from the parent user namespace to the
634 * kernel global id space.
636 for (idx = 0; idx < new_map.nr_extents; idx++) {
637 u32 lower_first;
638 extent = &new_map.extent[idx];
640 lower_first = map_id_range_down(parent_map,
641 extent->lower_first,
642 extent->count);
644 /* Fail if we can not map the specified extent to
645 * the kernel global id space.
647 if (lower_first == (u32) -1)
648 goto out;
650 extent->lower_first = lower_first;
653 /* Install the map */
654 memcpy(map->extent, new_map.extent,
655 new_map.nr_extents*sizeof(new_map.extent[0]));
656 smp_wmb();
657 map->nr_extents = new_map.nr_extents;
659 *ppos = count;
660 ret = count;
661 out:
662 mutex_unlock(&id_map_mutex);
663 if (page)
664 free_page(page);
665 return ret;
668 ssize_t proc_uid_map_write(struct file *file, const char __user *buf, size_t size, loff_t *ppos)
670 struct seq_file *seq = file->private_data;
671 struct user_namespace *ns = seq->private;
673 if (!ns->parent)
674 return -EPERM;
676 return map_write(file, buf, size, ppos, CAP_SETUID,
677 &ns->uid_map, &ns->parent->uid_map);
680 ssize_t proc_gid_map_write(struct file *file, const char __user *buf, size_t size, loff_t *ppos)
682 struct seq_file *seq = file->private_data;
683 struct user_namespace *ns = seq->private;
685 if (!ns->parent)
686 return -EPERM;
688 return map_write(file, buf, size, ppos, CAP_SETGID,
689 &ns->gid_map, &ns->parent->gid_map);
692 ssize_t proc_projid_map_write(struct file *file, const char __user *buf, size_t size, loff_t *ppos)
694 struct seq_file *seq = file->private_data;
695 struct user_namespace *ns = seq->private;
696 struct user_namespace *seq_ns = seq_user_ns(seq);
698 if (!ns->parent)
699 return -EPERM;
701 if ((seq_ns != ns) && (seq_ns != ns->parent))
702 return -EPERM;
704 /* Anyone can set any valid project id no capability needed */
705 return map_write(file, buf, size, ppos, -1,
706 &ns->projid_map, &ns->parent->projid_map);
709 static bool new_idmap_permitted(struct user_namespace *ns, int cap_setid,
710 struct uid_gid_map *new_map)
712 /* Allow anyone to set a mapping that doesn't require privilege */
713 if (!cap_valid(cap_setid))
714 return true;
716 /* Allow the specified ids if we have the appropriate capability
717 * (CAP_SETUID or CAP_SETGID) over the parent user namespace.
719 if (ns_capable(ns->parent, cap_setid))
720 return true;
722 return false;
725 static __init int user_namespaces_init(void)
727 user_ns_cachep = KMEM_CACHE(user_namespace, SLAB_PANIC);
728 return 0;
730 module_init(user_namespaces_init);