xref: /linux/security/commoncap.c (revision 69dbdd819599e2f3b77c172e83af512845bca5ad)
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/syslog.h>
31 
32 /*
33  * If a non-root user executes a setuid-root binary in
34  * !secure(SECURE_NOROOT) mode, then we raise capabilities.
35  * However if fE is also set, then the intent is for only
36  * the file capabilities to be applied, and the setuid-root
37  * bit is left on either to change the uid (plausible) or
38  * to get full privilege on a kernel without file capabilities
39  * support.  So in that case we do not raise capabilities.
40  *
41  * Warn if that happens, once per boot.
42  */
43 static void warn_setuid_and_fcaps_mixed(const char *fname)
44 {
45 	static int warned;
46 	if (!warned) {
47 		printk(KERN_INFO "warning: `%s' has both setuid-root and"
48 			" effective capabilities. Therefore not raising all"
49 			" capabilities.\n", fname);
50 		warned = 1;
51 	}
52 }
53 
54 int cap_netlink_send(struct sock *sk, struct sk_buff *skb)
55 {
56 	NETLINK_CB(skb).eff_cap = current_cap();
57 	return 0;
58 }
59 
60 int cap_netlink_recv(struct sk_buff *skb, int cap)
61 {
62 	if (!cap_raised(NETLINK_CB(skb).eff_cap, cap))
63 		return -EPERM;
64 	return 0;
65 }
66 EXPORT_SYMBOL(cap_netlink_recv);
67 
68 /**
69  * cap_capable - Determine whether a task has a particular effective capability
70  * @tsk: The task to query
71  * @cred: The credentials to use
72  * @cap: The capability to check for
73  * @audit: Whether to write an audit message or not
74  *
75  * Determine whether the nominated task has the specified capability amongst
76  * its effective set, returning 0 if it does, -ve if it does not.
77  *
78  * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable()
79  * and has_capability() functions.  That is, it has the reverse semantics:
80  * cap_has_capability() returns 0 when a task has a capability, but the
81  * kernel's capable() and has_capability() returns 1 for this case.
82  */
83 int cap_capable(struct task_struct *tsk, const struct cred *cred, int cap,
84 		int audit)
85 {
86 	return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM;
87 }
88 
89 /**
90  * cap_settime - Determine whether the current process may set the system clock
91  * @ts: The time to set
92  * @tz: The timezone to set
93  *
94  * Determine whether the current process may set the system clock and timezone
95  * information, returning 0 if permission granted, -ve if denied.
96  */
97 int cap_settime(struct timespec *ts, struct timezone *tz)
98 {
99 	if (!capable(CAP_SYS_TIME))
100 		return -EPERM;
101 	return 0;
102 }
103 
104 /**
105  * cap_ptrace_access_check - Determine whether the current process may access
106  *			   another
107  * @child: The process to be accessed
108  * @mode: The mode of attachment.
109  *
110  * Determine whether a process may access another, returning 0 if permission
111  * granted, -ve if denied.
112  */
113 int cap_ptrace_access_check(struct task_struct *child, unsigned int mode)
114 {
115 	int ret = 0;
116 
117 	rcu_read_lock();
118 	if (!cap_issubset(__task_cred(child)->cap_permitted,
119 			  current_cred()->cap_permitted) &&
120 	    !capable(CAP_SYS_PTRACE))
121 		ret = -EPERM;
122 	rcu_read_unlock();
123 	return ret;
124 }
125 
126 /**
127  * cap_ptrace_traceme - Determine whether another process may trace the current
128  * @parent: The task proposed to be the tracer
129  *
130  * Determine whether the nominated task is permitted to trace the current
131  * process, returning 0 if permission is granted, -ve if denied.
132  */
133 int cap_ptrace_traceme(struct task_struct *parent)
134 {
135 	int ret = 0;
136 
137 	rcu_read_lock();
138 	if (!cap_issubset(current_cred()->cap_permitted,
139 			  __task_cred(parent)->cap_permitted) &&
140 	    !has_capability(parent, CAP_SYS_PTRACE))
141 		ret = -EPERM;
142 	rcu_read_unlock();
143 	return ret;
144 }
145 
146 /**
147  * cap_capget - Retrieve a task's capability sets
148  * @target: The task from which to retrieve the capability sets
149  * @effective: The place to record the effective set
150  * @inheritable: The place to record the inheritable set
151  * @permitted: The place to record the permitted set
152  *
153  * This function retrieves the capabilities of the nominated task and returns
154  * them to the caller.
155  */
156 int cap_capget(struct task_struct *target, kernel_cap_t *effective,
157 	       kernel_cap_t *inheritable, kernel_cap_t *permitted)
158 {
159 	const struct cred *cred;
160 
161 	/* Derived from kernel/capability.c:sys_capget. */
162 	rcu_read_lock();
163 	cred = __task_cred(target);
164 	*effective   = cred->cap_effective;
165 	*inheritable = cred->cap_inheritable;
166 	*permitted   = cred->cap_permitted;
167 	rcu_read_unlock();
168 	return 0;
169 }
170 
171 /*
172  * Determine whether the inheritable capabilities are limited to the old
173  * permitted set.  Returns 1 if they are limited, 0 if they are not.
174  */
175 static inline int cap_inh_is_capped(void)
176 {
177 
178 	/* they are so limited unless the current task has the CAP_SETPCAP
179 	 * capability
180 	 */
181 	if (cap_capable(current, current_cred(), CAP_SETPCAP,
182 			SECURITY_CAP_AUDIT) == 0)
183 		return 0;
184 	return 1;
185 }
186 
187 /**
188  * cap_capset - Validate and apply proposed changes to current's capabilities
189  * @new: The proposed new credentials; alterations should be made here
190  * @old: The current task's current credentials
191  * @effective: A pointer to the proposed new effective capabilities set
192  * @inheritable: A pointer to the proposed new inheritable capabilities set
193  * @permitted: A pointer to the proposed new permitted capabilities set
194  *
195  * This function validates and applies a proposed mass change to the current
196  * process's capability sets.  The changes are made to the proposed new
197  * credentials, and assuming no error, will be committed by the caller of LSM.
198  */
199 int cap_capset(struct cred *new,
200 	       const struct cred *old,
201 	       const kernel_cap_t *effective,
202 	       const kernel_cap_t *inheritable,
203 	       const kernel_cap_t *permitted)
204 {
205 	if (cap_inh_is_capped() &&
206 	    !cap_issubset(*inheritable,
207 			  cap_combine(old->cap_inheritable,
208 				      old->cap_permitted)))
209 		/* incapable of using this inheritable set */
210 		return -EPERM;
211 
212 	if (!cap_issubset(*inheritable,
213 			  cap_combine(old->cap_inheritable,
214 				      old->cap_bset)))
215 		/* no new pI capabilities outside bounding set */
216 		return -EPERM;
217 
218 	/* verify restrictions on target's new Permitted set */
219 	if (!cap_issubset(*permitted, old->cap_permitted))
220 		return -EPERM;
221 
222 	/* verify the _new_Effective_ is a subset of the _new_Permitted_ */
223 	if (!cap_issubset(*effective, *permitted))
224 		return -EPERM;
225 
226 	new->cap_effective   = *effective;
227 	new->cap_inheritable = *inheritable;
228 	new->cap_permitted   = *permitted;
229 	return 0;
230 }
231 
232 /*
233  * Clear proposed capability sets for execve().
234  */
235 static inline void bprm_clear_caps(struct linux_binprm *bprm)
236 {
237 	cap_clear(bprm->cred->cap_permitted);
238 	bprm->cap_effective = false;
239 }
240 
241 /**
242  * cap_inode_need_killpriv - Determine if inode change affects privileges
243  * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV
244  *
245  * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV
246  * affects the security markings on that inode, and if it is, should
247  * inode_killpriv() be invoked or the change rejected?
248  *
249  * Returns 0 if granted; +ve if granted, but inode_killpriv() is required; and
250  * -ve to deny the change.
251  */
252 int cap_inode_need_killpriv(struct dentry *dentry)
253 {
254 	struct inode *inode = dentry->d_inode;
255 	int error;
256 
257 	if (!inode->i_op->getxattr)
258 	       return 0;
259 
260 	error = inode->i_op->getxattr(dentry, XATTR_NAME_CAPS, NULL, 0);
261 	if (error <= 0)
262 		return 0;
263 	return 1;
264 }
265 
266 /**
267  * cap_inode_killpriv - Erase the security markings on an inode
268  * @dentry: The inode/dentry to alter
269  *
270  * Erase the privilege-enhancing security markings on an inode.
271  *
272  * Returns 0 if successful, -ve on error.
273  */
274 int cap_inode_killpriv(struct dentry *dentry)
275 {
276 	struct inode *inode = dentry->d_inode;
277 
278 	if (!inode->i_op->removexattr)
279 	       return 0;
280 
281 	return inode->i_op->removexattr(dentry, XATTR_NAME_CAPS);
282 }
283 
284 /*
285  * Calculate the new process capability sets from the capability sets attached
286  * to a file.
287  */
288 static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps,
289 					  struct linux_binprm *bprm,
290 					  bool *effective)
291 {
292 	struct cred *new = bprm->cred;
293 	unsigned i;
294 	int ret = 0;
295 
296 	if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE)
297 		*effective = true;
298 
299 	CAP_FOR_EACH_U32(i) {
300 		__u32 permitted = caps->permitted.cap[i];
301 		__u32 inheritable = caps->inheritable.cap[i];
302 
303 		/*
304 		 * pP' = (X & fP) | (pI & fI)
305 		 */
306 		new->cap_permitted.cap[i] =
307 			(new->cap_bset.cap[i] & permitted) |
308 			(new->cap_inheritable.cap[i] & inheritable);
309 
310 		if (permitted & ~new->cap_permitted.cap[i])
311 			/* insufficient to execute correctly */
312 			ret = -EPERM;
313 	}
314 
315 	/*
316 	 * For legacy apps, with no internal support for recognizing they
317 	 * do not have enough capabilities, we return an error if they are
318 	 * missing some "forced" (aka file-permitted) capabilities.
319 	 */
320 	return *effective ? ret : 0;
321 }
322 
323 /*
324  * Extract the on-exec-apply capability sets for an executable file.
325  */
326 int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps)
327 {
328 	struct inode *inode = dentry->d_inode;
329 	__u32 magic_etc;
330 	unsigned tocopy, i;
331 	int size;
332 	struct vfs_cap_data caps;
333 
334 	memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data));
335 
336 	if (!inode || !inode->i_op->getxattr)
337 		return -ENODATA;
338 
339 	size = inode->i_op->getxattr((struct dentry *)dentry, XATTR_NAME_CAPS, &caps,
340 				   XATTR_CAPS_SZ);
341 	if (size == -ENODATA || size == -EOPNOTSUPP)
342 		/* no data, that's ok */
343 		return -ENODATA;
344 	if (size < 0)
345 		return size;
346 
347 	if (size < sizeof(magic_etc))
348 		return -EINVAL;
349 
350 	cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps.magic_etc);
351 
352 	switch (magic_etc & VFS_CAP_REVISION_MASK) {
353 	case VFS_CAP_REVISION_1:
354 		if (size != XATTR_CAPS_SZ_1)
355 			return -EINVAL;
356 		tocopy = VFS_CAP_U32_1;
357 		break;
358 	case VFS_CAP_REVISION_2:
359 		if (size != XATTR_CAPS_SZ_2)
360 			return -EINVAL;
361 		tocopy = VFS_CAP_U32_2;
362 		break;
363 	default:
364 		return -EINVAL;
365 	}
366 
367 	CAP_FOR_EACH_U32(i) {
368 		if (i >= tocopy)
369 			break;
370 		cpu_caps->permitted.cap[i] = le32_to_cpu(caps.data[i].permitted);
371 		cpu_caps->inheritable.cap[i] = le32_to_cpu(caps.data[i].inheritable);
372 	}
373 
374 	return 0;
375 }
376 
377 /*
378  * Attempt to get the on-exec apply capability sets for an executable file from
379  * its xattrs and, if present, apply them to the proposed credentials being
380  * constructed by execve().
381  */
382 static int get_file_caps(struct linux_binprm *bprm, bool *effective)
383 {
384 	struct dentry *dentry;
385 	int rc = 0;
386 	struct cpu_vfs_cap_data vcaps;
387 
388 	bprm_clear_caps(bprm);
389 
390 	if (!file_caps_enabled)
391 		return 0;
392 
393 	if (bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID)
394 		return 0;
395 
396 	dentry = dget(bprm->file->f_dentry);
397 
398 	rc = get_vfs_caps_from_disk(dentry, &vcaps);
399 	if (rc < 0) {
400 		if (rc == -EINVAL)
401 			printk(KERN_NOTICE "%s: get_vfs_caps_from_disk returned %d for %s\n",
402 				__func__, rc, bprm->filename);
403 		else if (rc == -ENODATA)
404 			rc = 0;
405 		goto out;
406 	}
407 
408 	rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective);
409 	if (rc == -EINVAL)
410 		printk(KERN_NOTICE "%s: cap_from_disk returned %d for %s\n",
411 		       __func__, rc, bprm->filename);
412 
413 out:
414 	dput(dentry);
415 	if (rc)
416 		bprm_clear_caps(bprm);
417 
418 	return rc;
419 }
420 
421 /**
422  * cap_bprm_set_creds - Set up the proposed credentials for execve().
423  * @bprm: The execution parameters, including the proposed creds
424  *
425  * Set up the proposed credentials for a new execution context being
426  * constructed by execve().  The proposed creds in @bprm->cred is altered,
427  * which won't take effect immediately.  Returns 0 if successful, -ve on error.
428  */
429 int cap_bprm_set_creds(struct linux_binprm *bprm)
430 {
431 	const struct cred *old = current_cred();
432 	struct cred *new = bprm->cred;
433 	bool effective;
434 	int ret;
435 
436 	effective = false;
437 	ret = get_file_caps(bprm, &effective);
438 	if (ret < 0)
439 		return ret;
440 
441 	if (!issecure(SECURE_NOROOT)) {
442 		/*
443 		 * If the legacy file capability is set, then don't set privs
444 		 * for a setuid root binary run by a non-root user.  Do set it
445 		 * for a root user just to cause least surprise to an admin.
446 		 */
447 		if (effective && new->uid != 0 && new->euid == 0) {
448 			warn_setuid_and_fcaps_mixed(bprm->filename);
449 			goto skip;
450 		}
451 		/*
452 		 * To support inheritance of root-permissions and suid-root
453 		 * executables under compatibility mode, we override the
454 		 * capability sets for the file.
455 		 *
456 		 * If only the real uid is 0, we do not set the effective bit.
457 		 */
458 		if (new->euid == 0 || new->uid == 0) {
459 			/* pP' = (cap_bset & ~0) | (pI & ~0) */
460 			new->cap_permitted = cap_combine(old->cap_bset,
461 							 old->cap_inheritable);
462 		}
463 		if (new->euid == 0)
464 			effective = true;
465 	}
466 skip:
467 
468 	/* Don't let someone trace a set[ug]id/setpcap binary with the revised
469 	 * credentials unless they have the appropriate permit
470 	 */
471 	if ((new->euid != old->uid ||
472 	     new->egid != old->gid ||
473 	     !cap_issubset(new->cap_permitted, old->cap_permitted)) &&
474 	    bprm->unsafe & ~LSM_UNSAFE_PTRACE_CAP) {
475 		/* downgrade; they get no more than they had, and maybe less */
476 		if (!capable(CAP_SETUID)) {
477 			new->euid = new->uid;
478 			new->egid = new->gid;
479 		}
480 		new->cap_permitted = cap_intersect(new->cap_permitted,
481 						   old->cap_permitted);
482 	}
483 
484 	new->suid = new->fsuid = new->euid;
485 	new->sgid = new->fsgid = new->egid;
486 
487 	/* For init, we want to retain the capabilities set in the initial
488 	 * task.  Thus we skip the usual capability rules
489 	 */
490 	if (!is_global_init(current)) {
491 		if (effective)
492 			new->cap_effective = new->cap_permitted;
493 		else
494 			cap_clear(new->cap_effective);
495 	}
496 	bprm->cap_effective = effective;
497 
498 	/*
499 	 * Audit candidate if current->cap_effective is set
500 	 *
501 	 * We do not bother to audit if 3 things are true:
502 	 *   1) cap_effective has all caps
503 	 *   2) we are root
504 	 *   3) root is supposed to have all caps (SECURE_NOROOT)
505 	 * Since this is just a normal root execing a process.
506 	 *
507 	 * Number 1 above might fail if you don't have a full bset, but I think
508 	 * that is interesting information to audit.
509 	 */
510 	if (!cap_isclear(new->cap_effective)) {
511 		if (!cap_issubset(CAP_FULL_SET, new->cap_effective) ||
512 		    new->euid != 0 || new->uid != 0 ||
513 		    issecure(SECURE_NOROOT)) {
514 			ret = audit_log_bprm_fcaps(bprm, new, old);
515 			if (ret < 0)
516 				return ret;
517 		}
518 	}
519 
520 	new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
521 	return 0;
522 }
523 
524 /**
525  * cap_bprm_secureexec - Determine whether a secure execution is required
526  * @bprm: The execution parameters
527  *
528  * Determine whether a secure execution is required, return 1 if it is, and 0
529  * if it is not.
530  *
531  * The credentials have been committed by this point, and so are no longer
532  * available through @bprm->cred.
533  */
534 int cap_bprm_secureexec(struct linux_binprm *bprm)
535 {
536 	const struct cred *cred = current_cred();
537 
538 	if (cred->uid != 0) {
539 		if (bprm->cap_effective)
540 			return 1;
541 		if (!cap_isclear(cred->cap_permitted))
542 			return 1;
543 	}
544 
545 	return (cred->euid != cred->uid ||
546 		cred->egid != cred->gid);
547 }
548 
549 /**
550  * cap_inode_setxattr - Determine whether an xattr may be altered
551  * @dentry: The inode/dentry being altered
552  * @name: The name of the xattr to be changed
553  * @value: The value that the xattr will be changed to
554  * @size: The size of value
555  * @flags: The replacement flag
556  *
557  * Determine whether an xattr may be altered or set on an inode, returning 0 if
558  * permission is granted, -ve if denied.
559  *
560  * This is used to make sure security xattrs don't get updated or set by those
561  * who aren't privileged to do so.
562  */
563 int cap_inode_setxattr(struct dentry *dentry, const char *name,
564 		       const void *value, size_t size, int flags)
565 {
566 	if (!strcmp(name, XATTR_NAME_CAPS)) {
567 		if (!capable(CAP_SETFCAP))
568 			return -EPERM;
569 		return 0;
570 	}
571 
572 	if (!strncmp(name, XATTR_SECURITY_PREFIX,
573 		     sizeof(XATTR_SECURITY_PREFIX) - 1) &&
574 	    !capable(CAP_SYS_ADMIN))
575 		return -EPERM;
576 	return 0;
577 }
578 
579 /**
580  * cap_inode_removexattr - Determine whether an xattr may be removed
581  * @dentry: The inode/dentry being altered
582  * @name: The name of the xattr to be changed
583  *
584  * Determine whether an xattr may be removed from an inode, returning 0 if
585  * permission is granted, -ve if denied.
586  *
587  * This is used to make sure security xattrs don't get removed by those who
588  * aren't privileged to remove them.
589  */
590 int cap_inode_removexattr(struct dentry *dentry, const char *name)
591 {
592 	if (!strcmp(name, XATTR_NAME_CAPS)) {
593 		if (!capable(CAP_SETFCAP))
594 			return -EPERM;
595 		return 0;
596 	}
597 
598 	if (!strncmp(name, XATTR_SECURITY_PREFIX,
599 		     sizeof(XATTR_SECURITY_PREFIX) - 1) &&
600 	    !capable(CAP_SYS_ADMIN))
601 		return -EPERM;
602 	return 0;
603 }
604 
605 /*
606  * cap_emulate_setxuid() fixes the effective / permitted capabilities of
607  * a process after a call to setuid, setreuid, or setresuid.
608  *
609  *  1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
610  *  {r,e,s}uid != 0, the permitted and effective capabilities are
611  *  cleared.
612  *
613  *  2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
614  *  capabilities of the process are cleared.
615  *
616  *  3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
617  *  capabilities are set to the permitted capabilities.
618  *
619  *  fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
620  *  never happen.
621  *
622  *  -astor
623  *
624  * cevans - New behaviour, Oct '99
625  * A process may, via prctl(), elect to keep its capabilities when it
626  * calls setuid() and switches away from uid==0. Both permitted and
627  * effective sets will be retained.
628  * Without this change, it was impossible for a daemon to drop only some
629  * of its privilege. The call to setuid(!=0) would drop all privileges!
630  * Keeping uid 0 is not an option because uid 0 owns too many vital
631  * files..
632  * Thanks to Olaf Kirch and Peter Benie for spotting this.
633  */
634 static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old)
635 {
636 	if ((old->uid == 0 || old->euid == 0 || old->suid == 0) &&
637 	    (new->uid != 0 && new->euid != 0 && new->suid != 0) &&
638 	    !issecure(SECURE_KEEP_CAPS)) {
639 		cap_clear(new->cap_permitted);
640 		cap_clear(new->cap_effective);
641 	}
642 	if (old->euid == 0 && new->euid != 0)
643 		cap_clear(new->cap_effective);
644 	if (old->euid != 0 && new->euid == 0)
645 		new->cap_effective = new->cap_permitted;
646 }
647 
648 /**
649  * cap_task_fix_setuid - Fix up the results of setuid() call
650  * @new: The proposed credentials
651  * @old: The current task's current credentials
652  * @flags: Indications of what has changed
653  *
654  * Fix up the results of setuid() call before the credential changes are
655  * actually applied, returning 0 to grant the changes, -ve to deny them.
656  */
657 int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags)
658 {
659 	switch (flags) {
660 	case LSM_SETID_RE:
661 	case LSM_SETID_ID:
662 	case LSM_SETID_RES:
663 		/* juggle the capabilities to follow [RES]UID changes unless
664 		 * otherwise suppressed */
665 		if (!issecure(SECURE_NO_SETUID_FIXUP))
666 			cap_emulate_setxuid(new, old);
667 		break;
668 
669 	case LSM_SETID_FS:
670 		/* juggle the capabilties to follow FSUID changes, unless
671 		 * otherwise suppressed
672 		 *
673 		 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
674 		 *          if not, we might be a bit too harsh here.
675 		 */
676 		if (!issecure(SECURE_NO_SETUID_FIXUP)) {
677 			if (old->fsuid == 0 && new->fsuid != 0)
678 				new->cap_effective =
679 					cap_drop_fs_set(new->cap_effective);
680 
681 			if (old->fsuid != 0 && new->fsuid == 0)
682 				new->cap_effective =
683 					cap_raise_fs_set(new->cap_effective,
684 							 new->cap_permitted);
685 		}
686 		break;
687 
688 	default:
689 		return -EINVAL;
690 	}
691 
692 	return 0;
693 }
694 
695 /*
696  * Rationale: code calling task_setscheduler, task_setioprio, and
697  * task_setnice, assumes that
698  *   . if capable(cap_sys_nice), then those actions should be allowed
699  *   . if not capable(cap_sys_nice), but acting on your own processes,
700  *   	then those actions should be allowed
701  * This is insufficient now since you can call code without suid, but
702  * yet with increased caps.
703  * So we check for increased caps on the target process.
704  */
705 static int cap_safe_nice(struct task_struct *p)
706 {
707 	int is_subset;
708 
709 	rcu_read_lock();
710 	is_subset = cap_issubset(__task_cred(p)->cap_permitted,
711 				 current_cred()->cap_permitted);
712 	rcu_read_unlock();
713 
714 	if (!is_subset && !capable(CAP_SYS_NICE))
715 		return -EPERM;
716 	return 0;
717 }
718 
719 /**
720  * cap_task_setscheduler - Detemine if scheduler policy change is permitted
721  * @p: The task to affect
722  *
723  * Detemine if the requested scheduler policy change is permitted for the
724  * specified task, returning 0 if permission is granted, -ve if denied.
725  */
726 int cap_task_setscheduler(struct task_struct *p)
727 {
728 	return cap_safe_nice(p);
729 }
730 
731 /**
732  * cap_task_ioprio - Detemine if I/O priority change is permitted
733  * @p: The task to affect
734  * @ioprio: The I/O priority to set
735  *
736  * Detemine if the requested I/O priority change is permitted for the specified
737  * task, returning 0 if permission is granted, -ve if denied.
738  */
739 int cap_task_setioprio(struct task_struct *p, int ioprio)
740 {
741 	return cap_safe_nice(p);
742 }
743 
744 /**
745  * cap_task_ioprio - Detemine if task priority change is permitted
746  * @p: The task to affect
747  * @nice: The nice value to set
748  *
749  * Detemine if the requested task priority change is permitted for the
750  * specified task, returning 0 if permission is granted, -ve if denied.
751  */
752 int cap_task_setnice(struct task_struct *p, int nice)
753 {
754 	return cap_safe_nice(p);
755 }
756 
757 /*
758  * Implement PR_CAPBSET_DROP.  Attempt to remove the specified capability from
759  * the current task's bounding set.  Returns 0 on success, -ve on error.
760  */
761 static long cap_prctl_drop(struct cred *new, unsigned long cap)
762 {
763 	if (!capable(CAP_SETPCAP))
764 		return -EPERM;
765 	if (!cap_valid(cap))
766 		return -EINVAL;
767 
768 	cap_lower(new->cap_bset, cap);
769 	return 0;
770 }
771 
772 /**
773  * cap_task_prctl - Implement process control functions for this security module
774  * @option: The process control function requested
775  * @arg2, @arg3, @arg4, @arg5: The argument data for this function
776  *
777  * Allow process control functions (sys_prctl()) to alter capabilities; may
778  * also deny access to other functions not otherwise implemented here.
779  *
780  * Returns 0 or +ve on success, -ENOSYS if this function is not implemented
781  * here, other -ve on error.  If -ENOSYS is returned, sys_prctl() and other LSM
782  * modules will consider performing the function.
783  */
784 int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,
785 		   unsigned long arg4, unsigned long arg5)
786 {
787 	struct cred *new;
788 	long error = 0;
789 
790 	new = prepare_creds();
791 	if (!new)
792 		return -ENOMEM;
793 
794 	switch (option) {
795 	case PR_CAPBSET_READ:
796 		error = -EINVAL;
797 		if (!cap_valid(arg2))
798 			goto error;
799 		error = !!cap_raised(new->cap_bset, arg2);
800 		goto no_change;
801 
802 	case PR_CAPBSET_DROP:
803 		error = cap_prctl_drop(new, arg2);
804 		if (error < 0)
805 			goto error;
806 		goto changed;
807 
808 	/*
809 	 * The next four prctl's remain to assist with transitioning a
810 	 * system from legacy UID=0 based privilege (when filesystem
811 	 * capabilities are not in use) to a system using filesystem
812 	 * capabilities only - as the POSIX.1e draft intended.
813 	 *
814 	 * Note:
815 	 *
816 	 *  PR_SET_SECUREBITS =
817 	 *      issecure_mask(SECURE_KEEP_CAPS_LOCKED)
818 	 *    | issecure_mask(SECURE_NOROOT)
819 	 *    | issecure_mask(SECURE_NOROOT_LOCKED)
820 	 *    | issecure_mask(SECURE_NO_SETUID_FIXUP)
821 	 *    | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
822 	 *
823 	 * will ensure that the current process and all of its
824 	 * children will be locked into a pure
825 	 * capability-based-privilege environment.
826 	 */
827 	case PR_SET_SECUREBITS:
828 		error = -EPERM;
829 		if ((((new->securebits & SECURE_ALL_LOCKS) >> 1)
830 		     & (new->securebits ^ arg2))			/*[1]*/
831 		    || ((new->securebits & SECURE_ALL_LOCKS & ~arg2))	/*[2]*/
832 		    || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS))	/*[3]*/
833 		    || (cap_capable(current, current_cred(), CAP_SETPCAP,
834 				    SECURITY_CAP_AUDIT) != 0)		/*[4]*/
835 			/*
836 			 * [1] no changing of bits that are locked
837 			 * [2] no unlocking of locks
838 			 * [3] no setting of unsupported bits
839 			 * [4] doing anything requires privilege (go read about
840 			 *     the "sendmail capabilities bug")
841 			 */
842 		    )
843 			/* cannot change a locked bit */
844 			goto error;
845 		new->securebits = arg2;
846 		goto changed;
847 
848 	case PR_GET_SECUREBITS:
849 		error = new->securebits;
850 		goto no_change;
851 
852 	case PR_GET_KEEPCAPS:
853 		if (issecure(SECURE_KEEP_CAPS))
854 			error = 1;
855 		goto no_change;
856 
857 	case PR_SET_KEEPCAPS:
858 		error = -EINVAL;
859 		if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */
860 			goto error;
861 		error = -EPERM;
862 		if (issecure(SECURE_KEEP_CAPS_LOCKED))
863 			goto error;
864 		if (arg2)
865 			new->securebits |= issecure_mask(SECURE_KEEP_CAPS);
866 		else
867 			new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
868 		goto changed;
869 
870 	default:
871 		/* No functionality available - continue with default */
872 		error = -ENOSYS;
873 		goto error;
874 	}
875 
876 	/* Functionality provided */
877 changed:
878 	return commit_creds(new);
879 
880 no_change:
881 error:
882 	abort_creds(new);
883 	return error;
884 }
885 
886 /**
887  * cap_syslog - Determine whether syslog function is permitted
888  * @type: Function requested
889  * @from_file: Whether this request came from an open file (i.e. /proc)
890  *
891  * Determine whether the current process is permitted to use a particular
892  * syslog function, returning 0 if permission is granted, -ve if not.
893  */
894 int cap_syslog(int type, bool from_file)
895 {
896 	if (type != SYSLOG_ACTION_OPEN && from_file)
897 		return 0;
898 	if ((type != SYSLOG_ACTION_READ_ALL &&
899 	     type != SYSLOG_ACTION_SIZE_BUFFER) && !capable(CAP_SYS_ADMIN))
900 		return -EPERM;
901 	return 0;
902 }
903 
904 /**
905  * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted
906  * @mm: The VM space in which the new mapping is to be made
907  * @pages: The size of the mapping
908  *
909  * Determine whether the allocation of a new virtual mapping by the current
910  * task is permitted, returning 0 if permission is granted, -ve if not.
911  */
912 int cap_vm_enough_memory(struct mm_struct *mm, long pages)
913 {
914 	int cap_sys_admin = 0;
915 
916 	if (cap_capable(current, current_cred(), CAP_SYS_ADMIN,
917 			SECURITY_CAP_NOAUDIT) == 0)
918 		cap_sys_admin = 1;
919 	return __vm_enough_memory(mm, pages, cap_sys_admin);
920 }
921 
922 /*
923  * cap_file_mmap - check if able to map given addr
924  * @file: unused
925  * @reqprot: unused
926  * @prot: unused
927  * @flags: unused
928  * @addr: address attempting to be mapped
929  * @addr_only: unused
930  *
931  * If the process is attempting to map memory below dac_mmap_min_addr they need
932  * CAP_SYS_RAWIO.  The other parameters to this function are unused by the
933  * capability security module.  Returns 0 if this mapping should be allowed
934  * -EPERM if not.
935  */
936 int cap_file_mmap(struct file *file, unsigned long reqprot,
937 		  unsigned long prot, unsigned long flags,
938 		  unsigned long addr, unsigned long addr_only)
939 {
940 	int ret = 0;
941 
942 	if (addr < dac_mmap_min_addr) {
943 		ret = cap_capable(current, current_cred(), CAP_SYS_RAWIO,
944 				  SECURITY_CAP_AUDIT);
945 		/* set PF_SUPERPRIV if it turns out we allow the low mmap */
946 		if (ret == 0)
947 			current->flags |= PF_SUPERPRIV;
948 	}
949 	return ret;
950 }
951