1 /* Common capabilities, needed by capability.o.
2 *
3 * This program is free software; you can redistribute it and/or modify
4 * it under the terms of the GNU General Public License as published by
5 * the Free Software Foundation; either version 2 of the License, or
6 * (at your option) any later version.
7 *
8 */
9
10 #include <linux/capability.h>
11 #include <linux/audit.h>
12 #include <linux/module.h>
13 #include <linux/init.h>
14 #include <linux/kernel.h>
15 #include <linux/security.h>
16 #include <linux/file.h>
17 #include <linux/mm.h>
18 #include <linux/mman.h>
19 #include <linux/pagemap.h>
20 #include <linux/swap.h>
21 #include <linux/skbuff.h>
22 #include <linux/netlink.h>
23 #include <linux/ptrace.h>
24 #include <linux/xattr.h>
25 #include <linux/hugetlb.h>
26 #include <linux/mount.h>
27 #include <linux/sched.h>
28 #include <linux/prctl.h>
29 #include <linux/securebits.h>
30 #include <linux/user_namespace.h>
31 #include <linux/binfmts.h>
32 #include <linux/personality.h>
33
34 /*
35 * If a non-root user executes a setuid-root binary in
36 * !secure(SECURE_NOROOT) mode, then we raise capabilities.
37 * However if fE is also set, then the intent is for only
38 * the file capabilities to be applied, and the setuid-root
39 * bit is left on either to change the uid (plausible) or
40 * to get full privilege on a kernel without file capabilities
41 * support. So in that case we do not raise capabilities.
42 *
43 * Warn if that happens, once per boot.
44 */
warn_setuid_and_fcaps_mixed(const char * fname)45 static void warn_setuid_and_fcaps_mixed(const char *fname)
46 {
47 static int warned;
48 if (!warned) {
49 printk(KERN_INFO "warning: `%s' has both setuid-root and"
50 " effective capabilities. Therefore not raising all"
51 " capabilities.\n", fname);
52 warned = 1;
53 }
54 }
55
cap_netlink_send(struct sock * sk,struct sk_buff * skb)56 int cap_netlink_send(struct sock *sk, struct sk_buff *skb)
57 {
58 return 0;
59 }
60
61 /**
62 * cap_capable - Determine whether a task has a particular effective capability
63 * @cred: The credentials to use
64 * @ns: The user namespace in which we need the capability
65 * @cap: The capability to check for
66 * @audit: Whether to write an audit message or not
67 *
68 * Determine whether the nominated task has the specified capability amongst
69 * its effective set, returning 0 if it does, -ve if it does not.
70 *
71 * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable()
72 * and has_capability() functions. That is, it has the reverse semantics:
73 * cap_has_capability() returns 0 when a task has a capability, but the
74 * kernel's capable() and has_capability() returns 1 for this case.
75 */
cap_capable(const struct cred * cred,struct user_namespace * targ_ns,int cap,int audit)76 int cap_capable(const struct cred *cred, struct user_namespace *targ_ns,
77 int cap, int audit)
78 {
79 struct user_namespace *ns = targ_ns;
80
81 /* See if cred has the capability in the target user namespace
82 * by examining the target user namespace and all of the target
83 * user namespace's parents.
84 */
85 for (;;) {
86 /* Do we have the necessary capabilities? */
87 if (ns == cred->user_ns)
88 return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM;
89
90 /* Have we tried all of the parent namespaces? */
91 if (ns == &init_user_ns)
92 return -EPERM;
93
94 /*
95 * The owner of the user namespace in the parent of the
96 * user namespace has all caps.
97 */
98 if ((ns->parent == cred->user_ns) && uid_eq(ns->owner, cred->euid))
99 return 0;
100
101 /*
102 * If you have a capability in a parent user ns, then you have
103 * it over all children user namespaces as well.
104 */
105 ns = ns->parent;
106 }
107
108 /* We never get here */
109 }
110
111 /**
112 * cap_settime - Determine whether the current process may set the system clock
113 * @ts: The time to set
114 * @tz: The timezone to set
115 *
116 * Determine whether the current process may set the system clock and timezone
117 * information, returning 0 if permission granted, -ve if denied.
118 */
cap_settime(const struct timespec * ts,const struct timezone * tz)119 int cap_settime(const struct timespec *ts, const struct timezone *tz)
120 {
121 if (!capable(CAP_SYS_TIME))
122 return -EPERM;
123 return 0;
124 }
125
126 /**
127 * cap_ptrace_access_check - Determine whether the current process may access
128 * another
129 * @child: The process to be accessed
130 * @mode: The mode of attachment.
131 *
132 * If we are in the same or an ancestor user_ns and have all the target
133 * task's capabilities, then ptrace access is allowed.
134 * If we have the ptrace capability to the target user_ns, then ptrace
135 * access is allowed.
136 * Else denied.
137 *
138 * Determine whether a process may access another, returning 0 if permission
139 * granted, -ve if denied.
140 */
cap_ptrace_access_check(struct task_struct * child,unsigned int mode)141 int cap_ptrace_access_check(struct task_struct *child, unsigned int mode)
142 {
143 int ret = 0;
144 const struct cred *cred, *child_cred;
145 const kernel_cap_t *caller_caps;
146
147 rcu_read_lock();
148 cred = current_cred();
149 child_cred = __task_cred(child);
150 if (mode & PTRACE_MODE_FSCREDS)
151 caller_caps = &cred->cap_effective;
152 else
153 caller_caps = &cred->cap_permitted;
154 if (cred->user_ns == child_cred->user_ns &&
155 cap_issubset(child_cred->cap_permitted, *caller_caps))
156 goto out;
157 if (ns_capable(child_cred->user_ns, CAP_SYS_PTRACE))
158 goto out;
159 ret = -EPERM;
160 out:
161 rcu_read_unlock();
162 return ret;
163 }
164
165 /**
166 * cap_ptrace_traceme - Determine whether another process may trace the current
167 * @parent: The task proposed to be the tracer
168 *
169 * If parent is in the same or an ancestor user_ns and has all current's
170 * capabilities, then ptrace access is allowed.
171 * If parent has the ptrace capability to current's user_ns, then ptrace
172 * access is allowed.
173 * Else denied.
174 *
175 * Determine whether the nominated task is permitted to trace the current
176 * process, returning 0 if permission is granted, -ve if denied.
177 */
cap_ptrace_traceme(struct task_struct * parent)178 int cap_ptrace_traceme(struct task_struct *parent)
179 {
180 int ret = 0;
181 const struct cred *cred, *child_cred;
182
183 rcu_read_lock();
184 cred = __task_cred(parent);
185 child_cred = current_cred();
186 if (cred->user_ns == child_cred->user_ns &&
187 cap_issubset(child_cred->cap_permitted, cred->cap_permitted))
188 goto out;
189 if (has_ns_capability(parent, child_cred->user_ns, CAP_SYS_PTRACE))
190 goto out;
191 ret = -EPERM;
192 out:
193 rcu_read_unlock();
194 return ret;
195 }
196
197 /**
198 * cap_capget - Retrieve a task's capability sets
199 * @target: The task from which to retrieve the capability sets
200 * @effective: The place to record the effective set
201 * @inheritable: The place to record the inheritable set
202 * @permitted: The place to record the permitted set
203 *
204 * This function retrieves the capabilities of the nominated task and returns
205 * them to the caller.
206 */
cap_capget(struct task_struct * target,kernel_cap_t * effective,kernel_cap_t * inheritable,kernel_cap_t * permitted)207 int cap_capget(struct task_struct *target, kernel_cap_t *effective,
208 kernel_cap_t *inheritable, kernel_cap_t *permitted)
209 {
210 const struct cred *cred;
211
212 /* Derived from kernel/capability.c:sys_capget. */
213 rcu_read_lock();
214 cred = __task_cred(target);
215 *effective = cred->cap_effective;
216 *inheritable = cred->cap_inheritable;
217 *permitted = cred->cap_permitted;
218 rcu_read_unlock();
219 return 0;
220 }
221
222 /*
223 * Determine whether the inheritable capabilities are limited to the old
224 * permitted set. Returns 1 if they are limited, 0 if they are not.
225 */
cap_inh_is_capped(void)226 static inline int cap_inh_is_capped(void)
227 {
228
229 /* they are so limited unless the current task has the CAP_SETPCAP
230 * capability
231 */
232 if (cap_capable(current_cred(), current_cred()->user_ns,
233 CAP_SETPCAP, SECURITY_CAP_AUDIT) == 0)
234 return 0;
235 return 1;
236 }
237
238 /**
239 * cap_capset - Validate and apply proposed changes to current's capabilities
240 * @new: The proposed new credentials; alterations should be made here
241 * @old: The current task's current credentials
242 * @effective: A pointer to the proposed new effective capabilities set
243 * @inheritable: A pointer to the proposed new inheritable capabilities set
244 * @permitted: A pointer to the proposed new permitted capabilities set
245 *
246 * This function validates and applies a proposed mass change to the current
247 * process's capability sets. The changes are made to the proposed new
248 * credentials, and assuming no error, will be committed by the caller of LSM.
249 */
cap_capset(struct cred * new,const struct cred * old,const kernel_cap_t * effective,const kernel_cap_t * inheritable,const kernel_cap_t * permitted)250 int cap_capset(struct cred *new,
251 const struct cred *old,
252 const kernel_cap_t *effective,
253 const kernel_cap_t *inheritable,
254 const kernel_cap_t *permitted)
255 {
256 if (cap_inh_is_capped() &&
257 !cap_issubset(*inheritable,
258 cap_combine(old->cap_inheritable,
259 old->cap_permitted)))
260 /* incapable of using this inheritable set */
261 return -EPERM;
262
263 if (!cap_issubset(*inheritable,
264 cap_combine(old->cap_inheritable,
265 old->cap_bset)))
266 /* no new pI capabilities outside bounding set */
267 return -EPERM;
268
269 /* verify restrictions on target's new Permitted set */
270 if (!cap_issubset(*permitted, old->cap_permitted))
271 return -EPERM;
272
273 /* verify the _new_Effective_ is a subset of the _new_Permitted_ */
274 if (!cap_issubset(*effective, *permitted))
275 return -EPERM;
276
277 new->cap_effective = *effective;
278 new->cap_inheritable = *inheritable;
279 new->cap_permitted = *permitted;
280 return 0;
281 }
282
283 /*
284 * Clear proposed capability sets for execve().
285 */
bprm_clear_caps(struct linux_binprm * bprm)286 static inline void bprm_clear_caps(struct linux_binprm *bprm)
287 {
288 cap_clear(bprm->cred->cap_permitted);
289 bprm->cap_effective = false;
290 }
291
292 /**
293 * cap_inode_need_killpriv - Determine if inode change affects privileges
294 * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV
295 *
296 * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV
297 * affects the security markings on that inode, and if it is, should
298 * inode_killpriv() be invoked or the change rejected?
299 *
300 * Returns 0 if granted; +ve if granted, but inode_killpriv() is required; and
301 * -ve to deny the change.
302 */
cap_inode_need_killpriv(struct dentry * dentry)303 int cap_inode_need_killpriv(struct dentry *dentry)
304 {
305 struct inode *inode = d_backing_inode(dentry);
306 int error;
307
308 if (!inode->i_op->getxattr)
309 return 0;
310
311 error = inode->i_op->getxattr(dentry, XATTR_NAME_CAPS, NULL, 0);
312 if (error <= 0)
313 return 0;
314 return 1;
315 }
316
317 /**
318 * cap_inode_killpriv - Erase the security markings on an inode
319 * @dentry: The inode/dentry to alter
320 *
321 * Erase the privilege-enhancing security markings on an inode.
322 *
323 * Returns 0 if successful, -ve on error.
324 */
cap_inode_killpriv(struct dentry * dentry)325 int cap_inode_killpriv(struct dentry *dentry)
326 {
327 struct inode *inode = d_backing_inode(dentry);
328
329 if (!inode->i_op->removexattr)
330 return 0;
331
332 return inode->i_op->removexattr(dentry, XATTR_NAME_CAPS);
333 }
334
335 /*
336 * Calculate the new process capability sets from the capability sets attached
337 * to a file.
338 */
bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data * caps,struct linux_binprm * bprm,bool * effective,bool * has_cap)339 static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps,
340 struct linux_binprm *bprm,
341 bool *effective,
342 bool *has_cap)
343 {
344 struct cred *new = bprm->cred;
345 unsigned i;
346 int ret = 0;
347
348 if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE)
349 *effective = true;
350
351 if (caps->magic_etc & VFS_CAP_REVISION_MASK)
352 *has_cap = true;
353
354 CAP_FOR_EACH_U32(i) {
355 __u32 permitted = caps->permitted.cap[i];
356 __u32 inheritable = caps->inheritable.cap[i];
357
358 /*
359 * pP' = (X & fP) | (pI & fI)
360 */
361 new->cap_permitted.cap[i] =
362 (new->cap_bset.cap[i] & permitted) |
363 (new->cap_inheritable.cap[i] & inheritable);
364
365 if (permitted & ~new->cap_permitted.cap[i])
366 /* insufficient to execute correctly */
367 ret = -EPERM;
368 }
369
370 /*
371 * For legacy apps, with no internal support for recognizing they
372 * do not have enough capabilities, we return an error if they are
373 * missing some "forced" (aka file-permitted) capabilities.
374 */
375 return *effective ? ret : 0;
376 }
377
378 /*
379 * Extract the on-exec-apply capability sets for an executable file.
380 */
get_vfs_caps_from_disk(const struct dentry * dentry,struct cpu_vfs_cap_data * cpu_caps)381 int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps)
382 {
383 struct inode *inode = d_backing_inode(dentry);
384 __u32 magic_etc;
385 unsigned tocopy, i;
386 int size;
387 struct vfs_cap_data caps;
388
389 memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data));
390
391 if (!inode || !inode->i_op->getxattr)
392 return -ENODATA;
393
394 size = inode->i_op->getxattr((struct dentry *)dentry, XATTR_NAME_CAPS, &caps,
395 XATTR_CAPS_SZ);
396 if (size == -ENODATA || size == -EOPNOTSUPP)
397 /* no data, that's ok */
398 return -ENODATA;
399 if (size < 0)
400 return size;
401
402 if (size < sizeof(magic_etc))
403 return -EINVAL;
404
405 cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps.magic_etc);
406
407 switch (magic_etc & VFS_CAP_REVISION_MASK) {
408 case VFS_CAP_REVISION_1:
409 if (size != XATTR_CAPS_SZ_1)
410 return -EINVAL;
411 tocopy = VFS_CAP_U32_1;
412 break;
413 case VFS_CAP_REVISION_2:
414 if (size != XATTR_CAPS_SZ_2)
415 return -EINVAL;
416 tocopy = VFS_CAP_U32_2;
417 break;
418 default:
419 return -EINVAL;
420 }
421
422 CAP_FOR_EACH_U32(i) {
423 if (i >= tocopy)
424 break;
425 cpu_caps->permitted.cap[i] = le32_to_cpu(caps.data[i].permitted);
426 cpu_caps->inheritable.cap[i] = le32_to_cpu(caps.data[i].inheritable);
427 }
428
429 cpu_caps->permitted.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
430 cpu_caps->inheritable.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
431
432 return 0;
433 }
434
435 /*
436 * Attempt to get the on-exec apply capability sets for an executable file from
437 * its xattrs and, if present, apply them to the proposed credentials being
438 * constructed by execve().
439 */
get_file_caps(struct linux_binprm * bprm,bool * effective,bool * has_cap)440 static int get_file_caps(struct linux_binprm *bprm, bool *effective, bool *has_cap)
441 {
442 int rc = 0;
443 struct cpu_vfs_cap_data vcaps;
444
445 bprm_clear_caps(bprm);
446
447 if (!file_caps_enabled)
448 return 0;
449
450 if (bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)
451 return 0;
452
453 rc = get_vfs_caps_from_disk(bprm->file->f_path.dentry, &vcaps);
454 if (rc < 0) {
455 if (rc == -EINVAL)
456 printk(KERN_NOTICE "%s: get_vfs_caps_from_disk returned %d for %s\n",
457 __func__, rc, bprm->filename);
458 else if (rc == -ENODATA)
459 rc = 0;
460 goto out;
461 }
462
463 rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective, has_cap);
464 if (rc == -EINVAL)
465 printk(KERN_NOTICE "%s: cap_from_disk returned %d for %s\n",
466 __func__, rc, bprm->filename);
467
468 out:
469 if (rc)
470 bprm_clear_caps(bprm);
471
472 return rc;
473 }
474
475 /**
476 * cap_bprm_set_creds - Set up the proposed credentials for execve().
477 * @bprm: The execution parameters, including the proposed creds
478 *
479 * Set up the proposed credentials for a new execution context being
480 * constructed by execve(). The proposed creds in @bprm->cred is altered,
481 * which won't take effect immediately. Returns 0 if successful, -ve on error.
482 */
cap_bprm_set_creds(struct linux_binprm * bprm)483 int cap_bprm_set_creds(struct linux_binprm *bprm)
484 {
485 const struct cred *old = current_cred();
486 struct cred *new = bprm->cred;
487 bool effective, has_cap = false;
488 int ret;
489 kuid_t root_uid;
490
491 effective = false;
492 ret = get_file_caps(bprm, &effective, &has_cap);
493 if (ret < 0)
494 return ret;
495
496 root_uid = make_kuid(new->user_ns, 0);
497
498 if (!issecure(SECURE_NOROOT)) {
499 /*
500 * If the legacy file capability is set, then don't set privs
501 * for a setuid root binary run by a non-root user. Do set it
502 * for a root user just to cause least surprise to an admin.
503 */
504 if (has_cap && !uid_eq(new->uid, root_uid) && uid_eq(new->euid, root_uid)) {
505 warn_setuid_and_fcaps_mixed(bprm->filename);
506 goto skip;
507 }
508 /*
509 * To support inheritance of root-permissions and suid-root
510 * executables under compatibility mode, we override the
511 * capability sets for the file.
512 *
513 * If only the real uid is 0, we do not set the effective bit.
514 */
515 if (uid_eq(new->euid, root_uid) || uid_eq(new->uid, root_uid)) {
516 /* pP' = (cap_bset & ~0) | (pI & ~0) */
517 new->cap_permitted = cap_combine(old->cap_bset,
518 old->cap_inheritable);
519 }
520 if (uid_eq(new->euid, root_uid))
521 effective = true;
522 }
523 skip:
524
525 /* if we have fs caps, clear dangerous personality flags */
526 if (!cap_issubset(new->cap_permitted, old->cap_permitted))
527 bprm->per_clear |= PER_CLEAR_ON_SETID;
528
529
530 /* Don't let someone trace a set[ug]id/setpcap binary with the revised
531 * credentials unless they have the appropriate permit.
532 *
533 * In addition, if NO_NEW_PRIVS, then ensure we get no new privs.
534 */
535 if ((!uid_eq(new->euid, old->uid) ||
536 !gid_eq(new->egid, old->gid) ||
537 !cap_issubset(new->cap_permitted, old->cap_permitted)) &&
538 bprm->unsafe & ~LSM_UNSAFE_PTRACE_CAP) {
539 /* downgrade; they get no more than they had, and maybe less */
540 if (!capable(CAP_SETUID) ||
541 (bprm->unsafe & LSM_UNSAFE_NO_NEW_PRIVS)) {
542 new->euid = new->uid;
543 new->egid = new->gid;
544 }
545 new->cap_permitted = cap_intersect(new->cap_permitted,
546 old->cap_permitted);
547 }
548
549 new->suid = new->fsuid = new->euid;
550 new->sgid = new->fsgid = new->egid;
551
552 if (effective)
553 new->cap_effective = new->cap_permitted;
554 else
555 cap_clear(new->cap_effective);
556 bprm->cap_effective = effective;
557
558 /*
559 * Audit candidate if current->cap_effective is set
560 *
561 * We do not bother to audit if 3 things are true:
562 * 1) cap_effective has all caps
563 * 2) we are root
564 * 3) root is supposed to have all caps (SECURE_NOROOT)
565 * Since this is just a normal root execing a process.
566 *
567 * Number 1 above might fail if you don't have a full bset, but I think
568 * that is interesting information to audit.
569 */
570 if (!cap_isclear(new->cap_effective)) {
571 if (!cap_issubset(CAP_FULL_SET, new->cap_effective) ||
572 !uid_eq(new->euid, root_uid) || !uid_eq(new->uid, root_uid) ||
573 issecure(SECURE_NOROOT)) {
574 ret = audit_log_bprm_fcaps(bprm, new, old);
575 if (ret < 0)
576 return ret;
577 }
578 }
579
580 new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
581 return 0;
582 }
583
584 /**
585 * cap_bprm_secureexec - Determine whether a secure execution is required
586 * @bprm: The execution parameters
587 *
588 * Determine whether a secure execution is required, return 1 if it is, and 0
589 * if it is not.
590 *
591 * The credentials have been committed by this point, and so are no longer
592 * available through @bprm->cred.
593 */
cap_bprm_secureexec(struct linux_binprm * bprm)594 int cap_bprm_secureexec(struct linux_binprm *bprm)
595 {
596 const struct cred *cred = current_cred();
597 kuid_t root_uid = make_kuid(cred->user_ns, 0);
598
599 if (!uid_eq(cred->uid, root_uid)) {
600 if (bprm->cap_effective)
601 return 1;
602 if (!cap_isclear(cred->cap_permitted))
603 return 1;
604 }
605
606 return (!uid_eq(cred->euid, cred->uid) ||
607 !gid_eq(cred->egid, cred->gid));
608 }
609
610 /**
611 * cap_inode_setxattr - Determine whether an xattr may be altered
612 * @dentry: The inode/dentry being altered
613 * @name: The name of the xattr to be changed
614 * @value: The value that the xattr will be changed to
615 * @size: The size of value
616 * @flags: The replacement flag
617 *
618 * Determine whether an xattr may be altered or set on an inode, returning 0 if
619 * permission is granted, -ve if denied.
620 *
621 * This is used to make sure security xattrs don't get updated or set by those
622 * who aren't privileged to do so.
623 */
cap_inode_setxattr(struct dentry * dentry,const char * name,const void * value,size_t size,int flags)624 int cap_inode_setxattr(struct dentry *dentry, const char *name,
625 const void *value, size_t size, int flags)
626 {
627 if (!strcmp(name, XATTR_NAME_CAPS)) {
628 if (!capable(CAP_SETFCAP))
629 return -EPERM;
630 return 0;
631 }
632
633 if (!strncmp(name, XATTR_SECURITY_PREFIX,
634 sizeof(XATTR_SECURITY_PREFIX) - 1) &&
635 !capable(CAP_SYS_ADMIN))
636 return -EPERM;
637 return 0;
638 }
639
640 /**
641 * cap_inode_removexattr - Determine whether an xattr may be removed
642 * @dentry: The inode/dentry being altered
643 * @name: The name of the xattr to be changed
644 *
645 * Determine whether an xattr may be removed from an inode, returning 0 if
646 * permission is granted, -ve if denied.
647 *
648 * This is used to make sure security xattrs don't get removed by those who
649 * aren't privileged to remove them.
650 */
cap_inode_removexattr(struct dentry * dentry,const char * name)651 int cap_inode_removexattr(struct dentry *dentry, const char *name)
652 {
653 if (!strcmp(name, XATTR_NAME_CAPS)) {
654 if (!capable(CAP_SETFCAP))
655 return -EPERM;
656 return 0;
657 }
658
659 if (!strncmp(name, XATTR_SECURITY_PREFIX,
660 sizeof(XATTR_SECURITY_PREFIX) - 1) &&
661 !capable(CAP_SYS_ADMIN))
662 return -EPERM;
663 return 0;
664 }
665
666 /*
667 * cap_emulate_setxuid() fixes the effective / permitted capabilities of
668 * a process after a call to setuid, setreuid, or setresuid.
669 *
670 * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
671 * {r,e,s}uid != 0, the permitted and effective capabilities are
672 * cleared.
673 *
674 * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
675 * capabilities of the process are cleared.
676 *
677 * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
678 * capabilities are set to the permitted capabilities.
679 *
680 * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
681 * never happen.
682 *
683 * -astor
684 *
685 * cevans - New behaviour, Oct '99
686 * A process may, via prctl(), elect to keep its capabilities when it
687 * calls setuid() and switches away from uid==0. Both permitted and
688 * effective sets will be retained.
689 * Without this change, it was impossible for a daemon to drop only some
690 * of its privilege. The call to setuid(!=0) would drop all privileges!
691 * Keeping uid 0 is not an option because uid 0 owns too many vital
692 * files..
693 * Thanks to Olaf Kirch and Peter Benie for spotting this.
694 */
cap_emulate_setxuid(struct cred * new,const struct cred * old)695 static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old)
696 {
697 kuid_t root_uid = make_kuid(old->user_ns, 0);
698
699 if ((uid_eq(old->uid, root_uid) ||
700 uid_eq(old->euid, root_uid) ||
701 uid_eq(old->suid, root_uid)) &&
702 (!uid_eq(new->uid, root_uid) &&
703 !uid_eq(new->euid, root_uid) &&
704 !uid_eq(new->suid, root_uid)) &&
705 !issecure(SECURE_KEEP_CAPS)) {
706 cap_clear(new->cap_permitted);
707 cap_clear(new->cap_effective);
708 }
709 if (uid_eq(old->euid, root_uid) && !uid_eq(new->euid, root_uid))
710 cap_clear(new->cap_effective);
711 if (!uid_eq(old->euid, root_uid) && uid_eq(new->euid, root_uid))
712 new->cap_effective = new->cap_permitted;
713 }
714
715 /**
716 * cap_task_fix_setuid - Fix up the results of setuid() call
717 * @new: The proposed credentials
718 * @old: The current task's current credentials
719 * @flags: Indications of what has changed
720 *
721 * Fix up the results of setuid() call before the credential changes are
722 * actually applied, returning 0 to grant the changes, -ve to deny them.
723 */
cap_task_fix_setuid(struct cred * new,const struct cred * old,int flags)724 int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags)
725 {
726 switch (flags) {
727 case LSM_SETID_RE:
728 case LSM_SETID_ID:
729 case LSM_SETID_RES:
730 /* juggle the capabilities to follow [RES]UID changes unless
731 * otherwise suppressed */
732 if (!issecure(SECURE_NO_SETUID_FIXUP))
733 cap_emulate_setxuid(new, old);
734 break;
735
736 case LSM_SETID_FS:
737 /* juggle the capabilties to follow FSUID changes, unless
738 * otherwise suppressed
739 *
740 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
741 * if not, we might be a bit too harsh here.
742 */
743 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
744 kuid_t root_uid = make_kuid(old->user_ns, 0);
745 if (uid_eq(old->fsuid, root_uid) && !uid_eq(new->fsuid, root_uid))
746 new->cap_effective =
747 cap_drop_fs_set(new->cap_effective);
748
749 if (!uid_eq(old->fsuid, root_uid) && uid_eq(new->fsuid, root_uid))
750 new->cap_effective =
751 cap_raise_fs_set(new->cap_effective,
752 new->cap_permitted);
753 }
754 break;
755
756 default:
757 return -EINVAL;
758 }
759
760 return 0;
761 }
762
763 /*
764 * Rationale: code calling task_setscheduler, task_setioprio, and
765 * task_setnice, assumes that
766 * . if capable(cap_sys_nice), then those actions should be allowed
767 * . if not capable(cap_sys_nice), but acting on your own processes,
768 * then those actions should be allowed
769 * This is insufficient now since you can call code without suid, but
770 * yet with increased caps.
771 * So we check for increased caps on the target process.
772 */
cap_safe_nice(struct task_struct * p)773 static int cap_safe_nice(struct task_struct *p)
774 {
775 int is_subset, ret = 0;
776
777 rcu_read_lock();
778 is_subset = cap_issubset(__task_cred(p)->cap_permitted,
779 current_cred()->cap_permitted);
780 if (!is_subset && !ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE))
781 ret = -EPERM;
782 rcu_read_unlock();
783
784 return ret;
785 }
786
787 /**
788 * cap_task_setscheduler - Detemine if scheduler policy change is permitted
789 * @p: The task to affect
790 *
791 * Detemine if the requested scheduler policy change is permitted for the
792 * specified task, returning 0 if permission is granted, -ve if denied.
793 */
cap_task_setscheduler(struct task_struct * p)794 int cap_task_setscheduler(struct task_struct *p)
795 {
796 return cap_safe_nice(p);
797 }
798
799 /**
800 * cap_task_ioprio - Detemine if I/O priority change is permitted
801 * @p: The task to affect
802 * @ioprio: The I/O priority to set
803 *
804 * Detemine if the requested I/O priority change is permitted for the specified
805 * task, returning 0 if permission is granted, -ve if denied.
806 */
cap_task_setioprio(struct task_struct * p,int ioprio)807 int cap_task_setioprio(struct task_struct *p, int ioprio)
808 {
809 return cap_safe_nice(p);
810 }
811
812 /**
813 * cap_task_ioprio - Detemine if task priority change is permitted
814 * @p: The task to affect
815 * @nice: The nice value to set
816 *
817 * Detemine if the requested task priority change is permitted for the
818 * specified task, returning 0 if permission is granted, -ve if denied.
819 */
cap_task_setnice(struct task_struct * p,int nice)820 int cap_task_setnice(struct task_struct *p, int nice)
821 {
822 return cap_safe_nice(p);
823 }
824
825 /*
826 * Implement PR_CAPBSET_DROP. Attempt to remove the specified capability from
827 * the current task's bounding set. Returns 0 on success, -ve on error.
828 */
cap_prctl_drop(unsigned long cap)829 static int cap_prctl_drop(unsigned long cap)
830 {
831 struct cred *new;
832
833 if (!ns_capable(current_user_ns(), CAP_SETPCAP))
834 return -EPERM;
835 if (!cap_valid(cap))
836 return -EINVAL;
837
838 new = prepare_creds();
839 if (!new)
840 return -ENOMEM;
841 cap_lower(new->cap_bset, cap);
842 return commit_creds(new);
843 }
844
845 /**
846 * cap_task_prctl - Implement process control functions for this security module
847 * @option: The process control function requested
848 * @arg2, @arg3, @arg4, @arg5: The argument data for this function
849 *
850 * Allow process control functions (sys_prctl()) to alter capabilities; may
851 * also deny access to other functions not otherwise implemented here.
852 *
853 * Returns 0 or +ve on success, -ENOSYS if this function is not implemented
854 * here, other -ve on error. If -ENOSYS is returned, sys_prctl() and other LSM
855 * modules will consider performing the function.
856 */
cap_task_prctl(int option,unsigned long arg2,unsigned long arg3,unsigned long arg4,unsigned long arg5)857 int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,
858 unsigned long arg4, unsigned long arg5)
859 {
860 const struct cred *old = current_cred();
861 struct cred *new;
862
863 switch (option) {
864 case PR_CAPBSET_READ:
865 if (!cap_valid(arg2))
866 return -EINVAL;
867 return !!cap_raised(old->cap_bset, arg2);
868
869 case PR_CAPBSET_DROP:
870 return cap_prctl_drop(arg2);
871
872 /*
873 * The next four prctl's remain to assist with transitioning a
874 * system from legacy UID=0 based privilege (when filesystem
875 * capabilities are not in use) to a system using filesystem
876 * capabilities only - as the POSIX.1e draft intended.
877 *
878 * Note:
879 *
880 * PR_SET_SECUREBITS =
881 * issecure_mask(SECURE_KEEP_CAPS_LOCKED)
882 * | issecure_mask(SECURE_NOROOT)
883 * | issecure_mask(SECURE_NOROOT_LOCKED)
884 * | issecure_mask(SECURE_NO_SETUID_FIXUP)
885 * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
886 *
887 * will ensure that the current process and all of its
888 * children will be locked into a pure
889 * capability-based-privilege environment.
890 */
891 case PR_SET_SECUREBITS:
892 if ((((old->securebits & SECURE_ALL_LOCKS) >> 1)
893 & (old->securebits ^ arg2)) /*[1]*/
894 || ((old->securebits & SECURE_ALL_LOCKS & ~arg2)) /*[2]*/
895 || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS)) /*[3]*/
896 || (cap_capable(current_cred(),
897 current_cred()->user_ns, CAP_SETPCAP,
898 SECURITY_CAP_AUDIT) != 0) /*[4]*/
899 /*
900 * [1] no changing of bits that are locked
901 * [2] no unlocking of locks
902 * [3] no setting of unsupported bits
903 * [4] doing anything requires privilege (go read about
904 * the "sendmail capabilities bug")
905 */
906 )
907 /* cannot change a locked bit */
908 return -EPERM;
909
910 new = prepare_creds();
911 if (!new)
912 return -ENOMEM;
913 new->securebits = arg2;
914 return commit_creds(new);
915
916 case PR_GET_SECUREBITS:
917 return old->securebits;
918
919 case PR_GET_KEEPCAPS:
920 return !!issecure(SECURE_KEEP_CAPS);
921
922 case PR_SET_KEEPCAPS:
923 if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */
924 return -EINVAL;
925 if (issecure(SECURE_KEEP_CAPS_LOCKED))
926 return -EPERM;
927
928 new = prepare_creds();
929 if (!new)
930 return -ENOMEM;
931 if (arg2)
932 new->securebits |= issecure_mask(SECURE_KEEP_CAPS);
933 else
934 new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
935 return commit_creds(new);
936
937 default:
938 /* No functionality available - continue with default */
939 return -ENOSYS;
940 }
941 }
942
943 /**
944 * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted
945 * @mm: The VM space in which the new mapping is to be made
946 * @pages: The size of the mapping
947 *
948 * Determine whether the allocation of a new virtual mapping by the current
949 * task is permitted, returning 0 if permission is granted, -ve if not.
950 */
cap_vm_enough_memory(struct mm_struct * mm,long pages)951 int cap_vm_enough_memory(struct mm_struct *mm, long pages)
952 {
953 int cap_sys_admin = 0;
954
955 if (cap_capable(current_cred(), &init_user_ns, CAP_SYS_ADMIN,
956 SECURITY_CAP_NOAUDIT) == 0)
957 cap_sys_admin = 1;
958 return __vm_enough_memory(mm, pages, cap_sys_admin);
959 }
960
961 /*
962 * cap_mmap_addr - check if able to map given addr
963 * @addr: address attempting to be mapped
964 *
965 * If the process is attempting to map memory below dac_mmap_min_addr they need
966 * CAP_SYS_RAWIO. The other parameters to this function are unused by the
967 * capability security module. Returns 0 if this mapping should be allowed
968 * -EPERM if not.
969 */
cap_mmap_addr(unsigned long addr)970 int cap_mmap_addr(unsigned long addr)
971 {
972 int ret = 0;
973
974 if (addr < dac_mmap_min_addr) {
975 ret = cap_capable(current_cred(), &init_user_ns, CAP_SYS_RAWIO,
976 SECURITY_CAP_AUDIT);
977 /* set PF_SUPERPRIV if it turns out we allow the low mmap */
978 if (ret == 0)
979 current->flags |= PF_SUPERPRIV;
980 }
981 return ret;
982 }
983
cap_mmap_file(struct file * file,unsigned long reqprot,unsigned long prot,unsigned long flags)984 int cap_mmap_file(struct file *file, unsigned long reqprot,
985 unsigned long prot, unsigned long flags)
986 {
987 return 0;
988 }
989