xref: /linux/security/commoncap.c (revision c54ea4918c2b7722d7242ea53271356501988a9b)
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 
31 /*
32  * If a non-root user executes a setuid-root binary in
33  * !secure(SECURE_NOROOT) mode, then we raise capabilities.
34  * However if fE is also set, then the intent is for only
35  * the file capabilities to be applied, and the setuid-root
36  * bit is left on either to change the uid (plausible) or
37  * to get full privilege on a kernel without file capabilities
38  * support.  So in that case we do not raise capabilities.
39  *
40  * Warn if that happens, once per boot.
41  */
42 static void warn_setuid_and_fcaps_mixed(const char *fname)
43 {
44 	static int warned;
45 	if (!warned) {
46 		printk(KERN_INFO "warning: `%s' has both setuid-root and"
47 			" effective capabilities. Therefore not raising all"
48 			" capabilities.\n", fname);
49 		warned = 1;
50 	}
51 }
52 
53 int cap_netlink_send(struct sock *sk, struct sk_buff *skb)
54 {
55 	return 0;
56 }
57 
58 int cap_netlink_recv(struct sk_buff *skb, int cap)
59 {
60 	if (!cap_raised(current_cap(), cap))
61 		return -EPERM;
62 	return 0;
63 }
64 EXPORT_SYMBOL(cap_netlink_recv);
65 
66 /**
67  * cap_capable - Determine whether a task has a particular effective capability
68  * @tsk: The task to query
69  * @cred: The credentials to use
70  * @cap: The capability to check for
71  * @audit: Whether to write an audit message or not
72  *
73  * Determine whether the nominated task has the specified capability amongst
74  * its effective set, returning 0 if it does, -ve if it does not.
75  *
76  * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable()
77  * and has_capability() functions.  That is, it has the reverse semantics:
78  * cap_has_capability() returns 0 when a task has a capability, but the
79  * kernel's capable() and has_capability() returns 1 for this case.
80  */
81 int cap_capable(struct task_struct *tsk, const struct cred *cred, int cap,
82 		int audit)
83 {
84 	return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM;
85 }
86 
87 /**
88  * cap_settime - Determine whether the current process may set the system clock
89  * @ts: The time to set
90  * @tz: The timezone to set
91  *
92  * Determine whether the current process may set the system clock and timezone
93  * information, returning 0 if permission granted, -ve if denied.
94  */
95 int cap_settime(const struct timespec *ts, const struct timezone *tz)
96 {
97 	if (!capable(CAP_SYS_TIME))
98 		return -EPERM;
99 	return 0;
100 }
101 
102 /**
103  * cap_ptrace_access_check - Determine whether the current process may access
104  *			   another
105  * @child: The process to be accessed
106  * @mode: The mode of attachment.
107  *
108  * Determine whether a process may access another, returning 0 if permission
109  * granted, -ve if denied.
110  */
111 int cap_ptrace_access_check(struct task_struct *child, unsigned int mode)
112 {
113 	int ret = 0;
114 
115 	rcu_read_lock();
116 	if (!cap_issubset(__task_cred(child)->cap_permitted,
117 			  current_cred()->cap_permitted) &&
118 	    !capable(CAP_SYS_PTRACE))
119 		ret = -EPERM;
120 	rcu_read_unlock();
121 	return ret;
122 }
123 
124 /**
125  * cap_ptrace_traceme - Determine whether another process may trace the current
126  * @parent: The task proposed to be the tracer
127  *
128  * Determine whether the nominated task is permitted to trace the current
129  * process, returning 0 if permission is granted, -ve if denied.
130  */
131 int cap_ptrace_traceme(struct task_struct *parent)
132 {
133 	int ret = 0;
134 
135 	rcu_read_lock();
136 	if (!cap_issubset(current_cred()->cap_permitted,
137 			  __task_cred(parent)->cap_permitted) &&
138 	    !has_capability(parent, CAP_SYS_PTRACE))
139 		ret = -EPERM;
140 	rcu_read_unlock();
141 	return ret;
142 }
143 
144 /**
145  * cap_capget - Retrieve a task's capability sets
146  * @target: The task from which to retrieve the capability sets
147  * @effective: The place to record the effective set
148  * @inheritable: The place to record the inheritable set
149  * @permitted: The place to record the permitted set
150  *
151  * This function retrieves the capabilities of the nominated task and returns
152  * them to the caller.
153  */
154 int cap_capget(struct task_struct *target, kernel_cap_t *effective,
155 	       kernel_cap_t *inheritable, kernel_cap_t *permitted)
156 {
157 	const struct cred *cred;
158 
159 	/* Derived from kernel/capability.c:sys_capget. */
160 	rcu_read_lock();
161 	cred = __task_cred(target);
162 	*effective   = cred->cap_effective;
163 	*inheritable = cred->cap_inheritable;
164 	*permitted   = cred->cap_permitted;
165 	rcu_read_unlock();
166 	return 0;
167 }
168 
169 /*
170  * Determine whether the inheritable capabilities are limited to the old
171  * permitted set.  Returns 1 if they are limited, 0 if they are not.
172  */
173 static inline int cap_inh_is_capped(void)
174 {
175 
176 	/* they are so limited unless the current task has the CAP_SETPCAP
177 	 * capability
178 	 */
179 	if (cap_capable(current, current_cred(), CAP_SETPCAP,
180 			SECURITY_CAP_AUDIT) == 0)
181 		return 0;
182 	return 1;
183 }
184 
185 /**
186  * cap_capset - Validate and apply proposed changes to current's capabilities
187  * @new: The proposed new credentials; alterations should be made here
188  * @old: The current task's current credentials
189  * @effective: A pointer to the proposed new effective capabilities set
190  * @inheritable: A pointer to the proposed new inheritable capabilities set
191  * @permitted: A pointer to the proposed new permitted capabilities set
192  *
193  * This function validates and applies a proposed mass change to the current
194  * process's capability sets.  The changes are made to the proposed new
195  * credentials, and assuming no error, will be committed by the caller of LSM.
196  */
197 int cap_capset(struct cred *new,
198 	       const struct cred *old,
199 	       const kernel_cap_t *effective,
200 	       const kernel_cap_t *inheritable,
201 	       const kernel_cap_t *permitted)
202 {
203 	if (cap_inh_is_capped() &&
204 	    !cap_issubset(*inheritable,
205 			  cap_combine(old->cap_inheritable,
206 				      old->cap_permitted)))
207 		/* incapable of using this inheritable set */
208 		return -EPERM;
209 
210 	if (!cap_issubset(*inheritable,
211 			  cap_combine(old->cap_inheritable,
212 				      old->cap_bset)))
213 		/* no new pI capabilities outside bounding set */
214 		return -EPERM;
215 
216 	/* verify restrictions on target's new Permitted set */
217 	if (!cap_issubset(*permitted, old->cap_permitted))
218 		return -EPERM;
219 
220 	/* verify the _new_Effective_ is a subset of the _new_Permitted_ */
221 	if (!cap_issubset(*effective, *permitted))
222 		return -EPERM;
223 
224 	new->cap_effective   = *effective;
225 	new->cap_inheritable = *inheritable;
226 	new->cap_permitted   = *permitted;
227 	return 0;
228 }
229 
230 /*
231  * Clear proposed capability sets for execve().
232  */
233 static inline void bprm_clear_caps(struct linux_binprm *bprm)
234 {
235 	cap_clear(bprm->cred->cap_permitted);
236 	bprm->cap_effective = false;
237 }
238 
239 /**
240  * cap_inode_need_killpriv - Determine if inode change affects privileges
241  * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV
242  *
243  * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV
244  * affects the security markings on that inode, and if it is, should
245  * inode_killpriv() be invoked or the change rejected?
246  *
247  * Returns 0 if granted; +ve if granted, but inode_killpriv() is required; and
248  * -ve to deny the change.
249  */
250 int cap_inode_need_killpriv(struct dentry *dentry)
251 {
252 	struct inode *inode = dentry->d_inode;
253 	int error;
254 
255 	if (!inode->i_op->getxattr)
256 	       return 0;
257 
258 	error = inode->i_op->getxattr(dentry, XATTR_NAME_CAPS, NULL, 0);
259 	if (error <= 0)
260 		return 0;
261 	return 1;
262 }
263 
264 /**
265  * cap_inode_killpriv - Erase the security markings on an inode
266  * @dentry: The inode/dentry to alter
267  *
268  * Erase the privilege-enhancing security markings on an inode.
269  *
270  * Returns 0 if successful, -ve on error.
271  */
272 int cap_inode_killpriv(struct dentry *dentry)
273 {
274 	struct inode *inode = dentry->d_inode;
275 
276 	if (!inode->i_op->removexattr)
277 	       return 0;
278 
279 	return inode->i_op->removexattr(dentry, XATTR_NAME_CAPS);
280 }
281 
282 /*
283  * Calculate the new process capability sets from the capability sets attached
284  * to a file.
285  */
286 static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps,
287 					  struct linux_binprm *bprm,
288 					  bool *effective)
289 {
290 	struct cred *new = bprm->cred;
291 	unsigned i;
292 	int ret = 0;
293 
294 	if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE)
295 		*effective = true;
296 
297 	CAP_FOR_EACH_U32(i) {
298 		__u32 permitted = caps->permitted.cap[i];
299 		__u32 inheritable = caps->inheritable.cap[i];
300 
301 		/*
302 		 * pP' = (X & fP) | (pI & fI)
303 		 */
304 		new->cap_permitted.cap[i] =
305 			(new->cap_bset.cap[i] & permitted) |
306 			(new->cap_inheritable.cap[i] & inheritable);
307 
308 		if (permitted & ~new->cap_permitted.cap[i])
309 			/* insufficient to execute correctly */
310 			ret = -EPERM;
311 	}
312 
313 	/*
314 	 * For legacy apps, with no internal support for recognizing they
315 	 * do not have enough capabilities, we return an error if they are
316 	 * missing some "forced" (aka file-permitted) capabilities.
317 	 */
318 	return *effective ? ret : 0;
319 }
320 
321 /*
322  * Extract the on-exec-apply capability sets for an executable file.
323  */
324 int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps)
325 {
326 	struct inode *inode = dentry->d_inode;
327 	__u32 magic_etc;
328 	unsigned tocopy, i;
329 	int size;
330 	struct vfs_cap_data caps;
331 
332 	memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data));
333 
334 	if (!inode || !inode->i_op->getxattr)
335 		return -ENODATA;
336 
337 	size = inode->i_op->getxattr((struct dentry *)dentry, XATTR_NAME_CAPS, &caps,
338 				   XATTR_CAPS_SZ);
339 	if (size == -ENODATA || size == -EOPNOTSUPP)
340 		/* no data, that's ok */
341 		return -ENODATA;
342 	if (size < 0)
343 		return size;
344 
345 	if (size < sizeof(magic_etc))
346 		return -EINVAL;
347 
348 	cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps.magic_etc);
349 
350 	switch (magic_etc & VFS_CAP_REVISION_MASK) {
351 	case VFS_CAP_REVISION_1:
352 		if (size != XATTR_CAPS_SZ_1)
353 			return -EINVAL;
354 		tocopy = VFS_CAP_U32_1;
355 		break;
356 	case VFS_CAP_REVISION_2:
357 		if (size != XATTR_CAPS_SZ_2)
358 			return -EINVAL;
359 		tocopy = VFS_CAP_U32_2;
360 		break;
361 	default:
362 		return -EINVAL;
363 	}
364 
365 	CAP_FOR_EACH_U32(i) {
366 		if (i >= tocopy)
367 			break;
368 		cpu_caps->permitted.cap[i] = le32_to_cpu(caps.data[i].permitted);
369 		cpu_caps->inheritable.cap[i] = le32_to_cpu(caps.data[i].inheritable);
370 	}
371 
372 	return 0;
373 }
374 
375 /*
376  * Attempt to get the on-exec apply capability sets for an executable file from
377  * its xattrs and, if present, apply them to the proposed credentials being
378  * constructed by execve().
379  */
380 static int get_file_caps(struct linux_binprm *bprm, bool *effective)
381 {
382 	struct dentry *dentry;
383 	int rc = 0;
384 	struct cpu_vfs_cap_data vcaps;
385 
386 	bprm_clear_caps(bprm);
387 
388 	if (!file_caps_enabled)
389 		return 0;
390 
391 	if (bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID)
392 		return 0;
393 
394 	dentry = dget(bprm->file->f_dentry);
395 
396 	rc = get_vfs_caps_from_disk(dentry, &vcaps);
397 	if (rc < 0) {
398 		if (rc == -EINVAL)
399 			printk(KERN_NOTICE "%s: get_vfs_caps_from_disk returned %d for %s\n",
400 				__func__, rc, bprm->filename);
401 		else if (rc == -ENODATA)
402 			rc = 0;
403 		goto out;
404 	}
405 
406 	rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective);
407 	if (rc == -EINVAL)
408 		printk(KERN_NOTICE "%s: cap_from_disk returned %d for %s\n",
409 		       __func__, rc, bprm->filename);
410 
411 out:
412 	dput(dentry);
413 	if (rc)
414 		bprm_clear_caps(bprm);
415 
416 	return rc;
417 }
418 
419 /**
420  * cap_bprm_set_creds - Set up the proposed credentials for execve().
421  * @bprm: The execution parameters, including the proposed creds
422  *
423  * Set up the proposed credentials for a new execution context being
424  * constructed by execve().  The proposed creds in @bprm->cred is altered,
425  * which won't take effect immediately.  Returns 0 if successful, -ve on error.
426  */
427 int cap_bprm_set_creds(struct linux_binprm *bprm)
428 {
429 	const struct cred *old = current_cred();
430 	struct cred *new = bprm->cred;
431 	bool effective;
432 	int ret;
433 
434 	effective = false;
435 	ret = get_file_caps(bprm, &effective);
436 	if (ret < 0)
437 		return ret;
438 
439 	if (!issecure(SECURE_NOROOT)) {
440 		/*
441 		 * If the legacy file capability is set, then don't set privs
442 		 * for a setuid root binary run by a non-root user.  Do set it
443 		 * for a root user just to cause least surprise to an admin.
444 		 */
445 		if (effective && new->uid != 0 && new->euid == 0) {
446 			warn_setuid_and_fcaps_mixed(bprm->filename);
447 			goto skip;
448 		}
449 		/*
450 		 * To support inheritance of root-permissions and suid-root
451 		 * executables under compatibility mode, we override the
452 		 * capability sets for the file.
453 		 *
454 		 * If only the real uid is 0, we do not set the effective bit.
455 		 */
456 		if (new->euid == 0 || new->uid == 0) {
457 			/* pP' = (cap_bset & ~0) | (pI & ~0) */
458 			new->cap_permitted = cap_combine(old->cap_bset,
459 							 old->cap_inheritable);
460 		}
461 		if (new->euid == 0)
462 			effective = true;
463 	}
464 skip:
465 
466 	/* Don't let someone trace a set[ug]id/setpcap binary with the revised
467 	 * credentials unless they have the appropriate permit
468 	 */
469 	if ((new->euid != old->uid ||
470 	     new->egid != old->gid ||
471 	     !cap_issubset(new->cap_permitted, old->cap_permitted)) &&
472 	    bprm->unsafe & ~LSM_UNSAFE_PTRACE_CAP) {
473 		/* downgrade; they get no more than they had, and maybe less */
474 		if (!capable(CAP_SETUID)) {
475 			new->euid = new->uid;
476 			new->egid = new->gid;
477 		}
478 		new->cap_permitted = cap_intersect(new->cap_permitted,
479 						   old->cap_permitted);
480 	}
481 
482 	new->suid = new->fsuid = new->euid;
483 	new->sgid = new->fsgid = new->egid;
484 
485 	/* For init, we want to retain the capabilities set in the initial
486 	 * task.  Thus we skip the usual capability rules
487 	 */
488 	if (!is_global_init(current)) {
489 		if (effective)
490 			new->cap_effective = new->cap_permitted;
491 		else
492 			cap_clear(new->cap_effective);
493 	}
494 	bprm->cap_effective = effective;
495 
496 	/*
497 	 * Audit candidate if current->cap_effective is set
498 	 *
499 	 * We do not bother to audit if 3 things are true:
500 	 *   1) cap_effective has all caps
501 	 *   2) we are root
502 	 *   3) root is supposed to have all caps (SECURE_NOROOT)
503 	 * Since this is just a normal root execing a process.
504 	 *
505 	 * Number 1 above might fail if you don't have a full bset, but I think
506 	 * that is interesting information to audit.
507 	 */
508 	if (!cap_isclear(new->cap_effective)) {
509 		if (!cap_issubset(CAP_FULL_SET, new->cap_effective) ||
510 		    new->euid != 0 || new->uid != 0 ||
511 		    issecure(SECURE_NOROOT)) {
512 			ret = audit_log_bprm_fcaps(bprm, new, old);
513 			if (ret < 0)
514 				return ret;
515 		}
516 	}
517 
518 	new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
519 	return 0;
520 }
521 
522 /**
523  * cap_bprm_secureexec - Determine whether a secure execution is required
524  * @bprm: The execution parameters
525  *
526  * Determine whether a secure execution is required, return 1 if it is, and 0
527  * if it is not.
528  *
529  * The credentials have been committed by this point, and so are no longer
530  * available through @bprm->cred.
531  */
532 int cap_bprm_secureexec(struct linux_binprm *bprm)
533 {
534 	const struct cred *cred = current_cred();
535 
536 	if (cred->uid != 0) {
537 		if (bprm->cap_effective)
538 			return 1;
539 		if (!cap_isclear(cred->cap_permitted))
540 			return 1;
541 	}
542 
543 	return (cred->euid != cred->uid ||
544 		cred->egid != cred->gid);
545 }
546 
547 /**
548  * cap_inode_setxattr - Determine whether an xattr may be altered
549  * @dentry: The inode/dentry being altered
550  * @name: The name of the xattr to be changed
551  * @value: The value that the xattr will be changed to
552  * @size: The size of value
553  * @flags: The replacement flag
554  *
555  * Determine whether an xattr may be altered or set on an inode, returning 0 if
556  * permission is granted, -ve if denied.
557  *
558  * This is used to make sure security xattrs don't get updated or set by those
559  * who aren't privileged to do so.
560  */
561 int cap_inode_setxattr(struct dentry *dentry, const char *name,
562 		       const void *value, size_t size, int flags)
563 {
564 	if (!strcmp(name, XATTR_NAME_CAPS)) {
565 		if (!capable(CAP_SETFCAP))
566 			return -EPERM;
567 		return 0;
568 	}
569 
570 	if (!strncmp(name, XATTR_SECURITY_PREFIX,
571 		     sizeof(XATTR_SECURITY_PREFIX) - 1) &&
572 	    !capable(CAP_SYS_ADMIN))
573 		return -EPERM;
574 	return 0;
575 }
576 
577 /**
578  * cap_inode_removexattr - Determine whether an xattr may be removed
579  * @dentry: The inode/dentry being altered
580  * @name: The name of the xattr to be changed
581  *
582  * Determine whether an xattr may be removed from an inode, returning 0 if
583  * permission is granted, -ve if denied.
584  *
585  * This is used to make sure security xattrs don't get removed by those who
586  * aren't privileged to remove them.
587  */
588 int cap_inode_removexattr(struct dentry *dentry, const char *name)
589 {
590 	if (!strcmp(name, XATTR_NAME_CAPS)) {
591 		if (!capable(CAP_SETFCAP))
592 			return -EPERM;
593 		return 0;
594 	}
595 
596 	if (!strncmp(name, XATTR_SECURITY_PREFIX,
597 		     sizeof(XATTR_SECURITY_PREFIX) - 1) &&
598 	    !capable(CAP_SYS_ADMIN))
599 		return -EPERM;
600 	return 0;
601 }
602 
603 /*
604  * cap_emulate_setxuid() fixes the effective / permitted capabilities of
605  * a process after a call to setuid, setreuid, or setresuid.
606  *
607  *  1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
608  *  {r,e,s}uid != 0, the permitted and effective capabilities are
609  *  cleared.
610  *
611  *  2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
612  *  capabilities of the process are cleared.
613  *
614  *  3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
615  *  capabilities are set to the permitted capabilities.
616  *
617  *  fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
618  *  never happen.
619  *
620  *  -astor
621  *
622  * cevans - New behaviour, Oct '99
623  * A process may, via prctl(), elect to keep its capabilities when it
624  * calls setuid() and switches away from uid==0. Both permitted and
625  * effective sets will be retained.
626  * Without this change, it was impossible for a daemon to drop only some
627  * of its privilege. The call to setuid(!=0) would drop all privileges!
628  * Keeping uid 0 is not an option because uid 0 owns too many vital
629  * files..
630  * Thanks to Olaf Kirch and Peter Benie for spotting this.
631  */
632 static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old)
633 {
634 	if ((old->uid == 0 || old->euid == 0 || old->suid == 0) &&
635 	    (new->uid != 0 && new->euid != 0 && new->suid != 0) &&
636 	    !issecure(SECURE_KEEP_CAPS)) {
637 		cap_clear(new->cap_permitted);
638 		cap_clear(new->cap_effective);
639 	}
640 	if (old->euid == 0 && new->euid != 0)
641 		cap_clear(new->cap_effective);
642 	if (old->euid != 0 && new->euid == 0)
643 		new->cap_effective = new->cap_permitted;
644 }
645 
646 /**
647  * cap_task_fix_setuid - Fix up the results of setuid() call
648  * @new: The proposed credentials
649  * @old: The current task's current credentials
650  * @flags: Indications of what has changed
651  *
652  * Fix up the results of setuid() call before the credential changes are
653  * actually applied, returning 0 to grant the changes, -ve to deny them.
654  */
655 int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags)
656 {
657 	switch (flags) {
658 	case LSM_SETID_RE:
659 	case LSM_SETID_ID:
660 	case LSM_SETID_RES:
661 		/* juggle the capabilities to follow [RES]UID changes unless
662 		 * otherwise suppressed */
663 		if (!issecure(SECURE_NO_SETUID_FIXUP))
664 			cap_emulate_setxuid(new, old);
665 		break;
666 
667 	case LSM_SETID_FS:
668 		/* juggle the capabilties to follow FSUID changes, unless
669 		 * otherwise suppressed
670 		 *
671 		 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
672 		 *          if not, we might be a bit too harsh here.
673 		 */
674 		if (!issecure(SECURE_NO_SETUID_FIXUP)) {
675 			if (old->fsuid == 0 && new->fsuid != 0)
676 				new->cap_effective =
677 					cap_drop_fs_set(new->cap_effective);
678 
679 			if (old->fsuid != 0 && new->fsuid == 0)
680 				new->cap_effective =
681 					cap_raise_fs_set(new->cap_effective,
682 							 new->cap_permitted);
683 		}
684 		break;
685 
686 	default:
687 		return -EINVAL;
688 	}
689 
690 	return 0;
691 }
692 
693 /*
694  * Rationale: code calling task_setscheduler, task_setioprio, and
695  * task_setnice, assumes that
696  *   . if capable(cap_sys_nice), then those actions should be allowed
697  *   . if not capable(cap_sys_nice), but acting on your own processes,
698  *   	then those actions should be allowed
699  * This is insufficient now since you can call code without suid, but
700  * yet with increased caps.
701  * So we check for increased caps on the target process.
702  */
703 static int cap_safe_nice(struct task_struct *p)
704 {
705 	int is_subset;
706 
707 	rcu_read_lock();
708 	is_subset = cap_issubset(__task_cred(p)->cap_permitted,
709 				 current_cred()->cap_permitted);
710 	rcu_read_unlock();
711 
712 	if (!is_subset && !capable(CAP_SYS_NICE))
713 		return -EPERM;
714 	return 0;
715 }
716 
717 /**
718  * cap_task_setscheduler - Detemine if scheduler policy change is permitted
719  * @p: The task to affect
720  *
721  * Detemine if the requested scheduler policy change is permitted for the
722  * specified task, returning 0 if permission is granted, -ve if denied.
723  */
724 int cap_task_setscheduler(struct task_struct *p)
725 {
726 	return cap_safe_nice(p);
727 }
728 
729 /**
730  * cap_task_ioprio - Detemine if I/O priority change is permitted
731  * @p: The task to affect
732  * @ioprio: The I/O priority to set
733  *
734  * Detemine if the requested I/O priority change is permitted for the specified
735  * task, returning 0 if permission is granted, -ve if denied.
736  */
737 int cap_task_setioprio(struct task_struct *p, int ioprio)
738 {
739 	return cap_safe_nice(p);
740 }
741 
742 /**
743  * cap_task_ioprio - Detemine if task priority change is permitted
744  * @p: The task to affect
745  * @nice: The nice value to set
746  *
747  * Detemine if the requested task priority change is permitted for the
748  * specified task, returning 0 if permission is granted, -ve if denied.
749  */
750 int cap_task_setnice(struct task_struct *p, int nice)
751 {
752 	return cap_safe_nice(p);
753 }
754 
755 /*
756  * Implement PR_CAPBSET_DROP.  Attempt to remove the specified capability from
757  * the current task's bounding set.  Returns 0 on success, -ve on error.
758  */
759 static long cap_prctl_drop(struct cred *new, unsigned long cap)
760 {
761 	if (!capable(CAP_SETPCAP))
762 		return -EPERM;
763 	if (!cap_valid(cap))
764 		return -EINVAL;
765 
766 	cap_lower(new->cap_bset, cap);
767 	return 0;
768 }
769 
770 /**
771  * cap_task_prctl - Implement process control functions for this security module
772  * @option: The process control function requested
773  * @arg2, @arg3, @arg4, @arg5: The argument data for this function
774  *
775  * Allow process control functions (sys_prctl()) to alter capabilities; may
776  * also deny access to other functions not otherwise implemented here.
777  *
778  * Returns 0 or +ve on success, -ENOSYS if this function is not implemented
779  * here, other -ve on error.  If -ENOSYS is returned, sys_prctl() and other LSM
780  * modules will consider performing the function.
781  */
782 int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,
783 		   unsigned long arg4, unsigned long arg5)
784 {
785 	struct cred *new;
786 	long error = 0;
787 
788 	new = prepare_creds();
789 	if (!new)
790 		return -ENOMEM;
791 
792 	switch (option) {
793 	case PR_CAPBSET_READ:
794 		error = -EINVAL;
795 		if (!cap_valid(arg2))
796 			goto error;
797 		error = !!cap_raised(new->cap_bset, arg2);
798 		goto no_change;
799 
800 	case PR_CAPBSET_DROP:
801 		error = cap_prctl_drop(new, arg2);
802 		if (error < 0)
803 			goto error;
804 		goto changed;
805 
806 	/*
807 	 * The next four prctl's remain to assist with transitioning a
808 	 * system from legacy UID=0 based privilege (when filesystem
809 	 * capabilities are not in use) to a system using filesystem
810 	 * capabilities only - as the POSIX.1e draft intended.
811 	 *
812 	 * Note:
813 	 *
814 	 *  PR_SET_SECUREBITS =
815 	 *      issecure_mask(SECURE_KEEP_CAPS_LOCKED)
816 	 *    | issecure_mask(SECURE_NOROOT)
817 	 *    | issecure_mask(SECURE_NOROOT_LOCKED)
818 	 *    | issecure_mask(SECURE_NO_SETUID_FIXUP)
819 	 *    | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
820 	 *
821 	 * will ensure that the current process and all of its
822 	 * children will be locked into a pure
823 	 * capability-based-privilege environment.
824 	 */
825 	case PR_SET_SECUREBITS:
826 		error = -EPERM;
827 		if ((((new->securebits & SECURE_ALL_LOCKS) >> 1)
828 		     & (new->securebits ^ arg2))			/*[1]*/
829 		    || ((new->securebits & SECURE_ALL_LOCKS & ~arg2))	/*[2]*/
830 		    || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS))	/*[3]*/
831 		    || (cap_capable(current, current_cred(), CAP_SETPCAP,
832 				    SECURITY_CAP_AUDIT) != 0)		/*[4]*/
833 			/*
834 			 * [1] no changing of bits that are locked
835 			 * [2] no unlocking of locks
836 			 * [3] no setting of unsupported bits
837 			 * [4] doing anything requires privilege (go read about
838 			 *     the "sendmail capabilities bug")
839 			 */
840 		    )
841 			/* cannot change a locked bit */
842 			goto error;
843 		new->securebits = arg2;
844 		goto changed;
845 
846 	case PR_GET_SECUREBITS:
847 		error = new->securebits;
848 		goto no_change;
849 
850 	case PR_GET_KEEPCAPS:
851 		if (issecure(SECURE_KEEP_CAPS))
852 			error = 1;
853 		goto no_change;
854 
855 	case PR_SET_KEEPCAPS:
856 		error = -EINVAL;
857 		if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */
858 			goto error;
859 		error = -EPERM;
860 		if (issecure(SECURE_KEEP_CAPS_LOCKED))
861 			goto error;
862 		if (arg2)
863 			new->securebits |= issecure_mask(SECURE_KEEP_CAPS);
864 		else
865 			new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
866 		goto changed;
867 
868 	default:
869 		/* No functionality available - continue with default */
870 		error = -ENOSYS;
871 		goto error;
872 	}
873 
874 	/* Functionality provided */
875 changed:
876 	return commit_creds(new);
877 
878 no_change:
879 error:
880 	abort_creds(new);
881 	return error;
882 }
883 
884 /**
885  * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted
886  * @mm: The VM space in which the new mapping is to be made
887  * @pages: The size of the mapping
888  *
889  * Determine whether the allocation of a new virtual mapping by the current
890  * task is permitted, returning 0 if permission is granted, -ve if not.
891  */
892 int cap_vm_enough_memory(struct mm_struct *mm, long pages)
893 {
894 	int cap_sys_admin = 0;
895 
896 	if (cap_capable(current, current_cred(), CAP_SYS_ADMIN,
897 			SECURITY_CAP_NOAUDIT) == 0)
898 		cap_sys_admin = 1;
899 	return __vm_enough_memory(mm, pages, cap_sys_admin);
900 }
901 
902 /*
903  * cap_file_mmap - check if able to map given addr
904  * @file: unused
905  * @reqprot: unused
906  * @prot: unused
907  * @flags: unused
908  * @addr: address attempting to be mapped
909  * @addr_only: unused
910  *
911  * If the process is attempting to map memory below dac_mmap_min_addr they need
912  * CAP_SYS_RAWIO.  The other parameters to this function are unused by the
913  * capability security module.  Returns 0 if this mapping should be allowed
914  * -EPERM if not.
915  */
916 int cap_file_mmap(struct file *file, unsigned long reqprot,
917 		  unsigned long prot, unsigned long flags,
918 		  unsigned long addr, unsigned long addr_only)
919 {
920 	int ret = 0;
921 
922 	if (addr < dac_mmap_min_addr) {
923 		ret = cap_capable(current, current_cred(), CAP_SYS_RAWIO,
924 				  SECURITY_CAP_AUDIT);
925 		/* set PF_SUPERPRIV if it turns out we allow the low mmap */
926 		if (ret == 0)
927 			current->flags |= PF_SUPERPRIV;
928 	}
929 	return ret;
930 }
931