~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/kernel/user_namespace.c

Version: ~ [ linux-5.8 ] ~ [ linux-5.7.14 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.57 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.138 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.193 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.232 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.232 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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 
  8 #include <linux/export.h>
  9 #include <linux/nsproxy.h>
 10 #include <linux/slab.h>
 11 #include <linux/user_namespace.h>
 12 #include <linux/proc_ns.h>
 13 #include <linux/highuid.h>
 14 #include <linux/cred.h>
 15 #include <linux/securebits.h>
 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>
 23 #include <linux/projid.h>
 24 #include <linux/fs_struct.h>
 25 
 26 static struct kmem_cache *user_ns_cachep __read_mostly;
 27 static DEFINE_MUTEX(userns_state_mutex);
 28 
 29 static bool new_idmap_permitted(const struct file *file,
 30                                 struct user_namespace *ns, int cap_setid,
 31                                 struct uid_gid_map *map);
 32 
 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 
 51 /*
 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.
 58  */
 59 int create_user_ns(struct cred *new)
 60 {
 61         struct user_namespace *ns, *parent_ns = new->user_ns;
 62         kuid_t owner = new->euid;
 63         kgid_t group = new->egid;
 64         int ret;
 65 
 66         if (parent_ns->level > 32)
 67                 return -EUSERS;
 68 
 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 
 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;
 85 
 86         ns = kmem_cache_zalloc(user_ns_cachep, GFP_KERNEL);
 87         if (!ns)
 88                 return -ENOMEM;
 89 
 90         ret = proc_alloc_inum(&ns->proc_inum);
 91         if (ret) {
 92                 kmem_cache_free(user_ns_cachep, ns);
 93                 return ret;
 94         }
 95 
 96         atomic_set(&ns->count, 1);
 97         /* Leave the new->user_ns reference with the new user namespace. */
 98         ns->parent = parent_ns;
 99         ns->level = parent_ns->level + 1;
100         ns->owner = owner;
101         ns->group = group;
102 
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 
108         set_cred_user_ns(new, ns);
109 
110 #ifdef CONFIG_PERSISTENT_KEYRINGS
111         init_rwsem(&ns->persistent_keyring_register_sem);
112 #endif
113         return 0;
114 }
115 
116 int unshare_userns(unsigned long unshare_flags, struct cred **new_cred)
117 {
118         struct cred *cred;
119         int err = -ENOMEM;
120 
121         if (!(unshare_flags & CLONE_NEWUSER))
122                 return 0;
123 
124         cred = prepare_creds();
125         if (cred) {
126                 err = create_user_ns(cred);
127                 if (err)
128                         put_cred(cred);
129                 else
130                         *new_cred = cred;
131         }
132 
133         return err;
134 }
135 
136 void free_user_ns(struct user_namespace *ns)
137 {
138         struct user_namespace *parent;
139 
140         do {
141                 parent = ns->parent;
142 #ifdef CONFIG_PERSISTENT_KEYRINGS
143                 key_put(ns->persistent_keyring_register);
144 #endif
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));
149 }
150 EXPORT_SYMBOL(free_user_ns);
151 
152 static u32 map_id_range_down(struct uid_gid_map *map, u32 id, u32 count)
153 {
154         unsigned idx, extents;
155         u32 first, last, id2;
156 
157         id2 = id + count - 1;
158 
159         /* Find the matching extent */
160         extents = map->nr_extents;
161         smp_rmb();
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;
185         smp_rmb();
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;
208         smp_rmb();
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;
214         }
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
234  *      for and handle INVALID_UID being returned.  INVALID_UID
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
295  *      @gid: group identifier
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 
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 
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;
434 
435         lower_ns = seq_user_ns(seq);
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;
447 }
448 
449 static int gid_m_show(struct seq_file *seq, void *v)
450 {
451         struct user_namespace *ns = seq->private;
452         struct uid_gid_extent *extent = v;
453         struct user_namespace *lower_ns;
454         gid_t lower;
455 
456         lower_ns = seq_user_ns(seq);
457         if ((lower_ns == ns) && lower_ns->parent)
458                 lower_ns = lower_ns->parent;
459 
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 
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 
491 static void *m_start(struct seq_file *seq, loff_t *ppos,
492                      struct uid_gid_map *map)
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 
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 
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 
535 const struct seq_operations proc_uid_seq_operations = {
536         .start = uid_m_start,
537         .stop = m_stop,
538         .next = m_next,
539         .show = uid_m_show,
540 };
541 
542 const struct seq_operations proc_gid_seq_operations = {
543         .start = gid_m_start,
544         .stop = m_stop,
545         .next = m_next,
546         .show = gid_m_show,
547 };
548 
549 const struct seq_operations proc_projid_seq_operations = {
550         .start = projid_m_start,
551         .stop = m_stop,
552         .next = m_next,
553         .show = projid_m_show,
554 };
555 
556 static bool mappings_overlap(struct uid_gid_map *new_map,
557                              struct uid_gid_extent *extent)
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 
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;
602         struct uid_gid_extent *extent = NULL;
603         unsigned long page;
604         char *kbuf, *pos, *next_line;
605         ssize_t ret;
606 
607         /* Only allow < page size writes at the beginning of the file */
608         if ((*ppos != 0) || (count >= PAGE_SIZE))
609                 return -EINVAL;
610 
611         /* Get a buffer */
612         page = __get_free_page(GFP_TEMPORARY);
613         kbuf = (char *) page;
614         if (!page)
615                 return -ENOMEM;
616 
617         /* Slurp in the user data */
618         if (copy_from_user(kbuf, buf, count)) {
619                 free_page(page);
620                 return -EFAULT;
621         }
622         kbuf[count] = '\0';
623 
624         /*
625          * The userns_state_mutex serializes all writes to any given map.
626          *
627          * Any map is only ever written once.
628          *
629          * An id map fits within 1 cache line on most architectures.
630          *
631          * On read nothing needs to be done unless you are on an
632          * architecture with a crazy cache coherency model like alpha.
633          *
634          * There is a one time data dependency between reading the
635          * count of the extents and the values of the extents.  The
636          * desired behavior is to see the values of the extents that
637          * were written before the count of the extents.
638          *
639          * To achieve this smp_wmb() is used on guarantee the write
640          * order and smp_rmb() is guaranteed that we don't have crazy
641          * architectures returning stale data.
642          */
643         mutex_lock(&userns_state_mutex);
644 
645         ret = -EPERM;
646         /* Only allow one successful write to the map */
647         if (map->nr_extents != 0)
648                 goto out;
649 
650         /*
651          * Adjusting namespace settings requires capabilities on the target.
652          */
653         if (cap_valid(cap_setid) && !file_ns_capable(file, ns, CAP_SYS_ADMIN))
654                 goto out;
655 
656         /* Parse the user data */
657         ret = -EINVAL;
658         pos = kbuf;
659         new_map.nr_extents = 0;
660         for (; pos; pos = next_line) {
661                 extent = &new_map.extent[new_map.nr_extents];
662 
663                 /* Find the end of line and ensure I don't look past it */
664                 next_line = strchr(pos, '\n');
665                 if (next_line) {
666                         *next_line = '\0';
667                         next_line++;
668                         if (*next_line == '\0')
669                                 next_line = NULL;
670                 }
671 
672                 pos = skip_spaces(pos);
673                 extent->first = simple_strtoul(pos, &pos, 10);
674                 if (!isspace(*pos))
675                         goto out;
676 
677                 pos = skip_spaces(pos);
678                 extent->lower_first = simple_strtoul(pos, &pos, 10);
679                 if (!isspace(*pos))
680                         goto out;
681 
682                 pos = skip_spaces(pos);
683                 extent->count = simple_strtoul(pos, &pos, 10);
684                 if (*pos && !isspace(*pos))
685                         goto out;
686 
687                 /* Verify there is not trailing junk on the line */
688                 pos = skip_spaces(pos);
689                 if (*pos != '\0')
690                         goto out;
691 
692                 /* Verify we have been given valid starting values */
693                 if ((extent->first == (u32) -1) ||
694                     (extent->lower_first == (u32) -1))
695                         goto out;
696 
697                 /* Verify count is not zero and does not cause the
698                  * extent to wrap
699                  */
700                 if ((extent->first + extent->count) <= extent->first)
701                         goto out;
702                 if ((extent->lower_first + extent->count) <=
703                      extent->lower_first)
704                         goto out;
705 
706                 /* Do the ranges in extent overlap any previous extents? */
707                 if (mappings_overlap(&new_map, extent))
708                         goto out;
709 
710                 new_map.nr_extents++;
711 
712                 /* Fail if the file contains too many extents */
713                 if ((new_map.nr_extents == UID_GID_MAP_MAX_EXTENTS) &&
714                     (next_line != NULL))
715                         goto out;
716         }
717         /* Be very certaint the new map actually exists */
718         if (new_map.nr_extents == 0)
719                 goto out;
720 
721         ret = -EPERM;
722         /* Validate the user is allowed to use user id's mapped to. */
723         if (!new_idmap_permitted(file, ns, cap_setid, &new_map))
724                 goto out;
725 
726         /* Map the lower ids from the parent user namespace to the
727          * kernel global id space.
728          */
729         for (idx = 0; idx < new_map.nr_extents; idx++) {
730                 u32 lower_first;
731                 extent = &new_map.extent[idx];
732 
733                 lower_first = map_id_range_down(parent_map,
734                                                 extent->lower_first,
735                                                 extent->count);
736 
737                 /* Fail if we can not map the specified extent to
738                  * the kernel global id space.
739                  */
740                 if (lower_first == (u32) -1)
741                         goto out;
742 
743                 extent->lower_first = lower_first;
744         }
745 
746         /* Install the map */
747         memcpy(map->extent, new_map.extent,
748                 new_map.nr_extents*sizeof(new_map.extent[0]));
749         smp_wmb();
750         map->nr_extents = new_map.nr_extents;
751 
752         *ppos = count;
753         ret = count;
754 out:
755         mutex_unlock(&userns_state_mutex);
756         if (page)
757                 free_page(page);
758         return ret;
759 }
760 
761 ssize_t proc_uid_map_write(struct file *file, const char __user *buf,
762                            size_t size, loff_t *ppos)
763 {
764         struct seq_file *seq = file->private_data;
765         struct user_namespace *ns = seq->private;
766         struct user_namespace *seq_ns = seq_user_ns(seq);
767 
768         if (!ns->parent)
769                 return -EPERM;
770 
771         if ((seq_ns != ns) && (seq_ns != ns->parent))
772                 return -EPERM;
773 
774         return map_write(file, buf, size, ppos, CAP_SETUID,
775                          &ns->uid_map, &ns->parent->uid_map);
776 }
777 
778 ssize_t proc_gid_map_write(struct file *file, const char __user *buf,
779                            size_t size, loff_t *ppos)
780 {
781         struct seq_file *seq = file->private_data;
782         struct user_namespace *ns = seq->private;
783         struct user_namespace *seq_ns = seq_user_ns(seq);
784 
785         if (!ns->parent)
786                 return -EPERM;
787 
788         if ((seq_ns != ns) && (seq_ns != ns->parent))
789                 return -EPERM;
790 
791         return map_write(file, buf, size, ppos, CAP_SETGID,
792                          &ns->gid_map, &ns->parent->gid_map);
793 }
794 
795 ssize_t proc_projid_map_write(struct file *file, const char __user *buf,
796                               size_t size, loff_t *ppos)
797 {
798         struct seq_file *seq = file->private_data;
799         struct user_namespace *ns = seq->private;
800         struct user_namespace *seq_ns = seq_user_ns(seq);
801 
802         if (!ns->parent)
803                 return -EPERM;
804 
805         if ((seq_ns != ns) && (seq_ns != ns->parent))
806                 return -EPERM;
807 
808         /* Anyone can set any valid project id no capability needed */
809         return map_write(file, buf, size, ppos, -1,
810                          &ns->projid_map, &ns->parent->projid_map);
811 }
812 
813 static bool new_idmap_permitted(const struct file *file,
814                                 struct user_namespace *ns, int cap_setid,
815                                 struct uid_gid_map *new_map)
816 {
817         const struct cred *cred = file->f_cred;
818         /* Don't allow mappings that would allow anything that wouldn't
819          * be allowed without the establishment of unprivileged mappings.
820          */
821         if ((new_map->nr_extents == 1) && (new_map->extent[0].count == 1) &&
822             uid_eq(ns->owner, cred->euid)) {
823                 u32 id = new_map->extent[0].lower_first;
824                 if (cap_setid == CAP_SETUID) {
825                         kuid_t uid = make_kuid(ns->parent, id);
826                         if (uid_eq(uid, cred->euid))
827                                 return true;
828                 } else if (cap_setid == CAP_SETGID) {
829                         kgid_t gid = make_kgid(ns->parent, id);
830                         if (!(ns->flags & USERNS_SETGROUPS_ALLOWED) &&
831                             gid_eq(gid, cred->egid))
832                                 return true;
833                 }
834         }
835 
836         /* Allow anyone to set a mapping that doesn't require privilege */
837         if (!cap_valid(cap_setid))
838                 return true;
839 
840         /* Allow the specified ids if we have the appropriate capability
841          * (CAP_SETUID or CAP_SETGID) over the parent user namespace.
842          * And the opener of the id file also had the approprpiate capability.
843          */
844         if (ns_capable(ns->parent, cap_setid) &&
845             file_ns_capable(file, ns->parent, cap_setid))
846                 return true;
847 
848         return false;
849 }
850 
851 int proc_setgroups_show(struct seq_file *seq, void *v)
852 {
853         struct user_namespace *ns = seq->private;
854         unsigned long userns_flags = ACCESS_ONCE(ns->flags);
855 
856         seq_printf(seq, "%s\n",
857                    (userns_flags & USERNS_SETGROUPS_ALLOWED) ?
858                    "allow" : "deny");
859         return 0;
860 }
861 
862 ssize_t proc_setgroups_write(struct file *file, const char __user *buf,
863                              size_t count, loff_t *ppos)
864 {
865         struct seq_file *seq = file->private_data;
866         struct user_namespace *ns = seq->private;
867         char kbuf[8], *pos;
868         bool setgroups_allowed;
869         ssize_t ret;
870 
871         /* Only allow a very narrow range of strings to be written */
872         ret = -EINVAL;
873         if ((*ppos != 0) || (count >= sizeof(kbuf)))
874                 goto out;
875 
876         /* What was written? */
877         ret = -EFAULT;
878         if (copy_from_user(kbuf, buf, count))
879                 goto out;
880         kbuf[count] = '\0';
881         pos = kbuf;
882 
883         /* What is being requested? */
884         ret = -EINVAL;
885         if (strncmp(pos, "allow", 5) == 0) {
886                 pos += 5;
887                 setgroups_allowed = true;
888         }
889         else if (strncmp(pos, "deny", 4) == 0) {
890                 pos += 4;
891                 setgroups_allowed = false;
892         }
893         else
894                 goto out;
895 
896         /* Verify there is not trailing junk on the line */
897         pos = skip_spaces(pos);
898         if (*pos != '\0')
899                 goto out;
900 
901         ret = -EPERM;
902         mutex_lock(&userns_state_mutex);
903         if (setgroups_allowed) {
904                 /* Enabling setgroups after setgroups has been disabled
905                  * is not allowed.
906                  */
907                 if (!(ns->flags & USERNS_SETGROUPS_ALLOWED))
908                         goto out_unlock;
909         } else {
910                 /* Permanently disabling setgroups after setgroups has
911                  * been enabled by writing the gid_map is not allowed.
912                  */
913                 if (ns->gid_map.nr_extents != 0)
914                         goto out_unlock;
915                 ns->flags &= ~USERNS_SETGROUPS_ALLOWED;
916         }
917         mutex_unlock(&userns_state_mutex);
918 
919         /* Report a successful write */
920         *ppos = count;
921         ret = count;
922 out:
923         return ret;
924 out_unlock:
925         mutex_unlock(&userns_state_mutex);
926         goto out;
927 }
928 
929 bool userns_may_setgroups(const struct user_namespace *ns)
930 {
931         bool allowed;
932 
933         mutex_lock(&userns_state_mutex);
934         /* It is not safe to use setgroups until a gid mapping in
935          * the user namespace has been established.
936          */
937         allowed = ns->gid_map.nr_extents != 0;
938         /* Is setgroups allowed? */
939         allowed = allowed && (ns->flags & USERNS_SETGROUPS_ALLOWED);
940         mutex_unlock(&userns_state_mutex);
941 
942         return allowed;
943 }
944 
945 static void *userns_get(struct task_struct *task)
946 {
947         struct user_namespace *user_ns;
948 
949         rcu_read_lock();
950         user_ns = get_user_ns(__task_cred(task)->user_ns);
951         rcu_read_unlock();
952 
953         return user_ns;
954 }
955 
956 static void userns_put(void *ns)
957 {
958         put_user_ns(ns);
959 }
960 
961 static int userns_install(struct nsproxy *nsproxy, void *ns)
962 {
963         struct user_namespace *user_ns = ns;
964         struct cred *cred;
965 
966         /* Don't allow gaining capabilities by reentering
967          * the same user namespace.
968          */
969         if (user_ns == current_user_ns())
970                 return -EINVAL;
971 
972         /* Threaded processes may not enter a different user namespace */
973         if (atomic_read(&current->mm->mm_users) > 1)
974                 return -EINVAL;
975 
976         if (current->fs->users != 1)
977                 return -EINVAL;
978 
979         if (!ns_capable(user_ns, CAP_SYS_ADMIN))
980                 return -EPERM;
981 
982         cred = prepare_creds();
983         if (!cred)
984                 return -ENOMEM;
985 
986         put_user_ns(cred->user_ns);
987         set_cred_user_ns(cred, get_user_ns(user_ns));
988 
989         return commit_creds(cred);
990 }
991 
992 static unsigned int userns_inum(void *ns)
993 {
994         struct user_namespace *user_ns = ns;
995         return user_ns->proc_inum;
996 }
997 
998 const struct proc_ns_operations userns_operations = {
999         .name           = "user",
1000         .type           = CLONE_NEWUSER,
1001         .get            = userns_get,
1002         .put            = userns_put,
1003         .install        = userns_install,
1004         .inum           = userns_inum,
1005 };
1006 
1007 static __init int user_namespaces_init(void)
1008 {
1009         user_ns_cachep = KMEM_CACHE(user_namespace, SLAB_PANIC);
1010         return 0;
1011 }
1012 subsys_initcall(user_namespaces_init);
1013 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp