xref: /openbmc/linux/security/commoncap.c (revision b627b4ed)
1 /* Common capabilities, needed by capability.o and root_plug.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 int cap_netlink_send(struct sock *sk, struct sk_buff *skb)
32 {
33 	NETLINK_CB(skb).eff_cap = current_cap();
34 	return 0;
35 }
36 
37 int cap_netlink_recv(struct sk_buff *skb, int cap)
38 {
39 	if (!cap_raised(NETLINK_CB(skb).eff_cap, cap))
40 		return -EPERM;
41 	return 0;
42 }
43 EXPORT_SYMBOL(cap_netlink_recv);
44 
45 /**
46  * cap_capable - Determine whether a task has a particular effective capability
47  * @tsk: The task to query
48  * @cred: The credentials to use
49  * @cap: The capability to check for
50  * @audit: Whether to write an audit message or not
51  *
52  * Determine whether the nominated task has the specified capability amongst
53  * its effective set, returning 0 if it does, -ve if it does not.
54  *
55  * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable()
56  * and has_capability() functions.  That is, it has the reverse semantics:
57  * cap_has_capability() returns 0 when a task has a capability, but the
58  * kernel's capable() and has_capability() returns 1 for this case.
59  */
60 int cap_capable(struct task_struct *tsk, const struct cred *cred, int cap,
61 		int audit)
62 {
63 	return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM;
64 }
65 
66 /**
67  * cap_settime - Determine whether the current process may set the system clock
68  * @ts: The time to set
69  * @tz: The timezone to set
70  *
71  * Determine whether the current process may set the system clock and timezone
72  * information, returning 0 if permission granted, -ve if denied.
73  */
74 int cap_settime(struct timespec *ts, struct timezone *tz)
75 {
76 	if (!capable(CAP_SYS_TIME))
77 		return -EPERM;
78 	return 0;
79 }
80 
81 /**
82  * cap_ptrace_may_access - Determine whether the current process may access
83  *			   another
84  * @child: The process to be accessed
85  * @mode: The mode of attachment.
86  *
87  * Determine whether a process may access another, returning 0 if permission
88  * granted, -ve if denied.
89  */
90 int cap_ptrace_may_access(struct task_struct *child, unsigned int mode)
91 {
92 	int ret = 0;
93 
94 	rcu_read_lock();
95 	if (!cap_issubset(__task_cred(child)->cap_permitted,
96 			  current_cred()->cap_permitted) &&
97 	    !capable(CAP_SYS_PTRACE))
98 		ret = -EPERM;
99 	rcu_read_unlock();
100 	return ret;
101 }
102 
103 /**
104  * cap_ptrace_traceme - Determine whether another process may trace the current
105  * @parent: The task proposed to be the tracer
106  *
107  * Determine whether the nominated task is permitted to trace the current
108  * process, returning 0 if permission is granted, -ve if denied.
109  */
110 int cap_ptrace_traceme(struct task_struct *parent)
111 {
112 	int ret = 0;
113 
114 	rcu_read_lock();
115 	if (!cap_issubset(current_cred()->cap_permitted,
116 			  __task_cred(parent)->cap_permitted) &&
117 	    !has_capability(parent, CAP_SYS_PTRACE))
118 		ret = -EPERM;
119 	rcu_read_unlock();
120 	return ret;
121 }
122 
123 /**
124  * cap_capget - Retrieve a task's capability sets
125  * @target: The task from which to retrieve the capability sets
126  * @effective: The place to record the effective set
127  * @inheritable: The place to record the inheritable set
128  * @permitted: The place to record the permitted set
129  *
130  * This function retrieves the capabilities of the nominated task and returns
131  * them to the caller.
132  */
133 int cap_capget(struct task_struct *target, kernel_cap_t *effective,
134 	       kernel_cap_t *inheritable, kernel_cap_t *permitted)
135 {
136 	const struct cred *cred;
137 
138 	/* Derived from kernel/capability.c:sys_capget. */
139 	rcu_read_lock();
140 	cred = __task_cred(target);
141 	*effective   = cred->cap_effective;
142 	*inheritable = cred->cap_inheritable;
143 	*permitted   = cred->cap_permitted;
144 	rcu_read_unlock();
145 	return 0;
146 }
147 
148 /*
149  * Determine whether the inheritable capabilities are limited to the old
150  * permitted set.  Returns 1 if they are limited, 0 if they are not.
151  */
152 static inline int cap_inh_is_capped(void)
153 {
154 #ifdef CONFIG_SECURITY_FILE_CAPABILITIES
155 
156 	/* they are so limited unless the current task has the CAP_SETPCAP
157 	 * capability
158 	 */
159 	if (cap_capable(current, current_cred(), CAP_SETPCAP,
160 			SECURITY_CAP_AUDIT) == 0)
161 		return 0;
162 #endif
163 	return 1;
164 }
165 
166 /**
167  * cap_capset - Validate and apply proposed changes to current's capabilities
168  * @new: The proposed new credentials; alterations should be made here
169  * @old: The current task's current credentials
170  * @effective: A pointer to the proposed new effective capabilities set
171  * @inheritable: A pointer to the proposed new inheritable capabilities set
172  * @permitted: A pointer to the proposed new permitted capabilities set
173  *
174  * This function validates and applies a proposed mass change to the current
175  * process's capability sets.  The changes are made to the proposed new
176  * credentials, and assuming no error, will be committed by the caller of LSM.
177  */
178 int cap_capset(struct cred *new,
179 	       const struct cred *old,
180 	       const kernel_cap_t *effective,
181 	       const kernel_cap_t *inheritable,
182 	       const kernel_cap_t *permitted)
183 {
184 	if (cap_inh_is_capped() &&
185 	    !cap_issubset(*inheritable,
186 			  cap_combine(old->cap_inheritable,
187 				      old->cap_permitted)))
188 		/* incapable of using this inheritable set */
189 		return -EPERM;
190 
191 	if (!cap_issubset(*inheritable,
192 			  cap_combine(old->cap_inheritable,
193 				      old->cap_bset)))
194 		/* no new pI capabilities outside bounding set */
195 		return -EPERM;
196 
197 	/* verify restrictions on target's new Permitted set */
198 	if (!cap_issubset(*permitted, old->cap_permitted))
199 		return -EPERM;
200 
201 	/* verify the _new_Effective_ is a subset of the _new_Permitted_ */
202 	if (!cap_issubset(*effective, *permitted))
203 		return -EPERM;
204 
205 	new->cap_effective   = *effective;
206 	new->cap_inheritable = *inheritable;
207 	new->cap_permitted   = *permitted;
208 	return 0;
209 }
210 
211 /*
212  * Clear proposed capability sets for execve().
213  */
214 static inline void bprm_clear_caps(struct linux_binprm *bprm)
215 {
216 	cap_clear(bprm->cred->cap_permitted);
217 	bprm->cap_effective = false;
218 }
219 
220 #ifdef CONFIG_SECURITY_FILE_CAPABILITIES
221 
222 /**
223  * cap_inode_need_killpriv - Determine if inode change affects privileges
224  * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV
225  *
226  * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV
227  * affects the security markings on that inode, and if it is, should
228  * inode_killpriv() be invoked or the change rejected?
229  *
230  * Returns 0 if granted; +ve if granted, but inode_killpriv() is required; and
231  * -ve to deny the change.
232  */
233 int cap_inode_need_killpriv(struct dentry *dentry)
234 {
235 	struct inode *inode = dentry->d_inode;
236 	int error;
237 
238 	if (!inode->i_op->getxattr)
239 	       return 0;
240 
241 	error = inode->i_op->getxattr(dentry, XATTR_NAME_CAPS, NULL, 0);
242 	if (error <= 0)
243 		return 0;
244 	return 1;
245 }
246 
247 /**
248  * cap_inode_killpriv - Erase the security markings on an inode
249  * @dentry: The inode/dentry to alter
250  *
251  * Erase the privilege-enhancing security markings on an inode.
252  *
253  * Returns 0 if successful, -ve on error.
254  */
255 int cap_inode_killpriv(struct dentry *dentry)
256 {
257 	struct inode *inode = dentry->d_inode;
258 
259 	if (!inode->i_op->removexattr)
260 	       return 0;
261 
262 	return inode->i_op->removexattr(dentry, XATTR_NAME_CAPS);
263 }
264 
265 /*
266  * Calculate the new process capability sets from the capability sets attached
267  * to a file.
268  */
269 static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps,
270 					  struct linux_binprm *bprm,
271 					  bool *effective)
272 {
273 	struct cred *new = bprm->cred;
274 	unsigned i;
275 	int ret = 0;
276 
277 	if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE)
278 		*effective = true;
279 
280 	CAP_FOR_EACH_U32(i) {
281 		__u32 permitted = caps->permitted.cap[i];
282 		__u32 inheritable = caps->inheritable.cap[i];
283 
284 		/*
285 		 * pP' = (X & fP) | (pI & fI)
286 		 */
287 		new->cap_permitted.cap[i] =
288 			(new->cap_bset.cap[i] & permitted) |
289 			(new->cap_inheritable.cap[i] & inheritable);
290 
291 		if (permitted & ~new->cap_permitted.cap[i])
292 			/* insufficient to execute correctly */
293 			ret = -EPERM;
294 	}
295 
296 	/*
297 	 * For legacy apps, with no internal support for recognizing they
298 	 * do not have enough capabilities, we return an error if they are
299 	 * missing some "forced" (aka file-permitted) capabilities.
300 	 */
301 	return *effective ? ret : 0;
302 }
303 
304 /*
305  * Extract the on-exec-apply capability sets for an executable file.
306  */
307 int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps)
308 {
309 	struct inode *inode = dentry->d_inode;
310 	__u32 magic_etc;
311 	unsigned tocopy, i;
312 	int size;
313 	struct vfs_cap_data caps;
314 
315 	memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data));
316 
317 	if (!inode || !inode->i_op->getxattr)
318 		return -ENODATA;
319 
320 	size = inode->i_op->getxattr((struct dentry *)dentry, XATTR_NAME_CAPS, &caps,
321 				   XATTR_CAPS_SZ);
322 	if (size == -ENODATA || size == -EOPNOTSUPP)
323 		/* no data, that's ok */
324 		return -ENODATA;
325 	if (size < 0)
326 		return size;
327 
328 	if (size < sizeof(magic_etc))
329 		return -EINVAL;
330 
331 	cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps.magic_etc);
332 
333 	switch (magic_etc & VFS_CAP_REVISION_MASK) {
334 	case VFS_CAP_REVISION_1:
335 		if (size != XATTR_CAPS_SZ_1)
336 			return -EINVAL;
337 		tocopy = VFS_CAP_U32_1;
338 		break;
339 	case VFS_CAP_REVISION_2:
340 		if (size != XATTR_CAPS_SZ_2)
341 			return -EINVAL;
342 		tocopy = VFS_CAP_U32_2;
343 		break;
344 	default:
345 		return -EINVAL;
346 	}
347 
348 	CAP_FOR_EACH_U32(i) {
349 		if (i >= tocopy)
350 			break;
351 		cpu_caps->permitted.cap[i] = le32_to_cpu(caps.data[i].permitted);
352 		cpu_caps->inheritable.cap[i] = le32_to_cpu(caps.data[i].inheritable);
353 	}
354 
355 	return 0;
356 }
357 
358 /*
359  * Attempt to get the on-exec apply capability sets for an executable file from
360  * its xattrs and, if present, apply them to the proposed credentials being
361  * constructed by execve().
362  */
363 static int get_file_caps(struct linux_binprm *bprm, bool *effective)
364 {
365 	struct dentry *dentry;
366 	int rc = 0;
367 	struct cpu_vfs_cap_data vcaps;
368 
369 	bprm_clear_caps(bprm);
370 
371 	if (!file_caps_enabled)
372 		return 0;
373 
374 	if (bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID)
375 		return 0;
376 
377 	dentry = dget(bprm->file->f_dentry);
378 
379 	rc = get_vfs_caps_from_disk(dentry, &vcaps);
380 	if (rc < 0) {
381 		if (rc == -EINVAL)
382 			printk(KERN_NOTICE "%s: get_vfs_caps_from_disk returned %d for %s\n",
383 				__func__, rc, bprm->filename);
384 		else if (rc == -ENODATA)
385 			rc = 0;
386 		goto out;
387 	}
388 
389 	rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective);
390 	if (rc == -EINVAL)
391 		printk(KERN_NOTICE "%s: cap_from_disk returned %d for %s\n",
392 		       __func__, rc, bprm->filename);
393 
394 out:
395 	dput(dentry);
396 	if (rc)
397 		bprm_clear_caps(bprm);
398 
399 	return rc;
400 }
401 
402 #else
403 int cap_inode_need_killpriv(struct dentry *dentry)
404 {
405 	return 0;
406 }
407 
408 int cap_inode_killpriv(struct dentry *dentry)
409 {
410 	return 0;
411 }
412 
413 int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps)
414 {
415 	memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data));
416  	return -ENODATA;
417 }
418 
419 static inline int get_file_caps(struct linux_binprm *bprm, bool *effective)
420 {
421 	bprm_clear_caps(bprm);
422 	return 0;
423 }
424 #endif
425 
426 /*
427  * Determine whether a exec'ing process's new permitted capabilities should be
428  * limited to just what it already has.
429  *
430  * This prevents processes that are being ptraced from gaining access to
431  * CAP_SETPCAP, unless the process they're tracing already has it, and the
432  * binary they're executing has filecaps that elevate it.
433  *
434  *  Returns 1 if they should be limited, 0 if they are not.
435  */
436 static inline int cap_limit_ptraced_target(void)
437 {
438 #ifndef CONFIG_SECURITY_FILE_CAPABILITIES
439 	if (capable(CAP_SETPCAP))
440 		return 0;
441 #endif
442 	return 1;
443 }
444 
445 /**
446  * cap_bprm_set_creds - Set up the proposed credentials for execve().
447  * @bprm: The execution parameters, including the proposed creds
448  *
449  * Set up the proposed credentials for a new execution context being
450  * constructed by execve().  The proposed creds in @bprm->cred is altered,
451  * which won't take effect immediately.  Returns 0 if successful, -ve on error.
452  */
453 int cap_bprm_set_creds(struct linux_binprm *bprm)
454 {
455 	const struct cred *old = current_cred();
456 	struct cred *new = bprm->cred;
457 	bool effective;
458 	int ret;
459 
460 	effective = false;
461 	ret = get_file_caps(bprm, &effective);
462 	if (ret < 0)
463 		return ret;
464 
465 	if (!issecure(SECURE_NOROOT)) {
466 		/*
467 		 * To support inheritance of root-permissions and suid-root
468 		 * executables under compatibility mode, we override the
469 		 * capability sets for the file.
470 		 *
471 		 * If only the real uid is 0, we do not set the effective bit.
472 		 */
473 		if (new->euid == 0 || new->uid == 0) {
474 			/* pP' = (cap_bset & ~0) | (pI & ~0) */
475 			new->cap_permitted = cap_combine(old->cap_bset,
476 							 old->cap_inheritable);
477 		}
478 		if (new->euid == 0)
479 			effective = true;
480 	}
481 
482 	/* Don't let someone trace a set[ug]id/setpcap binary with the revised
483 	 * credentials unless they have the appropriate permit
484 	 */
485 	if ((new->euid != old->uid ||
486 	     new->egid != old->gid ||
487 	     !cap_issubset(new->cap_permitted, old->cap_permitted)) &&
488 	    bprm->unsafe & ~LSM_UNSAFE_PTRACE_CAP) {
489 		/* downgrade; they get no more than they had, and maybe less */
490 		if (!capable(CAP_SETUID)) {
491 			new->euid = new->uid;
492 			new->egid = new->gid;
493 		}
494 		if (cap_limit_ptraced_target())
495 			new->cap_permitted = cap_intersect(new->cap_permitted,
496 							   old->cap_permitted);
497 	}
498 
499 	new->suid = new->fsuid = new->euid;
500 	new->sgid = new->fsgid = new->egid;
501 
502 	/* For init, we want to retain the capabilities set in the initial
503 	 * task.  Thus we skip the usual capability rules
504 	 */
505 	if (!is_global_init(current)) {
506 		if (effective)
507 			new->cap_effective = new->cap_permitted;
508 		else
509 			cap_clear(new->cap_effective);
510 	}
511 	bprm->cap_effective = effective;
512 
513 	/*
514 	 * Audit candidate if current->cap_effective is set
515 	 *
516 	 * We do not bother to audit if 3 things are true:
517 	 *   1) cap_effective has all caps
518 	 *   2) we are root
519 	 *   3) root is supposed to have all caps (SECURE_NOROOT)
520 	 * Since this is just a normal root execing a process.
521 	 *
522 	 * Number 1 above might fail if you don't have a full bset, but I think
523 	 * that is interesting information to audit.
524 	 */
525 	if (!cap_isclear(new->cap_effective)) {
526 		if (!cap_issubset(CAP_FULL_SET, new->cap_effective) ||
527 		    new->euid != 0 || new->uid != 0 ||
528 		    issecure(SECURE_NOROOT)) {
529 			ret = audit_log_bprm_fcaps(bprm, new, old);
530 			if (ret < 0)
531 				return ret;
532 		}
533 	}
534 
535 	new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
536 	return 0;
537 }
538 
539 /**
540  * cap_bprm_secureexec - Determine whether a secure execution is required
541  * @bprm: The execution parameters
542  *
543  * Determine whether a secure execution is required, return 1 if it is, and 0
544  * if it is not.
545  *
546  * The credentials have been committed by this point, and so are no longer
547  * available through @bprm->cred.
548  */
549 int cap_bprm_secureexec(struct linux_binprm *bprm)
550 {
551 	const struct cred *cred = current_cred();
552 
553 	if (cred->uid != 0) {
554 		if (bprm->cap_effective)
555 			return 1;
556 		if (!cap_isclear(cred->cap_permitted))
557 			return 1;
558 	}
559 
560 	return (cred->euid != cred->uid ||
561 		cred->egid != cred->gid);
562 }
563 
564 /**
565  * cap_inode_setxattr - Determine whether an xattr may be altered
566  * @dentry: The inode/dentry being altered
567  * @name: The name of the xattr to be changed
568  * @value: The value that the xattr will be changed to
569  * @size: The size of value
570  * @flags: The replacement flag
571  *
572  * Determine whether an xattr may be altered or set on an inode, returning 0 if
573  * permission is granted, -ve if denied.
574  *
575  * This is used to make sure security xattrs don't get updated or set by those
576  * who aren't privileged to do so.
577  */
578 int cap_inode_setxattr(struct dentry *dentry, const char *name,
579 		       const void *value, size_t size, int flags)
580 {
581 	if (!strcmp(name, XATTR_NAME_CAPS)) {
582 		if (!capable(CAP_SETFCAP))
583 			return -EPERM;
584 		return 0;
585 	}
586 
587 	if (!strncmp(name, XATTR_SECURITY_PREFIX,
588 		     sizeof(XATTR_SECURITY_PREFIX) - 1)  &&
589 	    !capable(CAP_SYS_ADMIN))
590 		return -EPERM;
591 	return 0;
592 }
593 
594 /**
595  * cap_inode_removexattr - Determine whether an xattr may be removed
596  * @dentry: The inode/dentry being altered
597  * @name: The name of the xattr to be changed
598  *
599  * Determine whether an xattr may be removed from an inode, returning 0 if
600  * permission is granted, -ve if denied.
601  *
602  * This is used to make sure security xattrs don't get removed by those who
603  * aren't privileged to remove them.
604  */
605 int cap_inode_removexattr(struct dentry *dentry, const char *name)
606 {
607 	if (!strcmp(name, XATTR_NAME_CAPS)) {
608 		if (!capable(CAP_SETFCAP))
609 			return -EPERM;
610 		return 0;
611 	}
612 
613 	if (!strncmp(name, XATTR_SECURITY_PREFIX,
614 		     sizeof(XATTR_SECURITY_PREFIX) - 1)  &&
615 	    !capable(CAP_SYS_ADMIN))
616 		return -EPERM;
617 	return 0;
618 }
619 
620 /*
621  * cap_emulate_setxuid() fixes the effective / permitted capabilities of
622  * a process after a call to setuid, setreuid, or setresuid.
623  *
624  *  1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
625  *  {r,e,s}uid != 0, the permitted and effective capabilities are
626  *  cleared.
627  *
628  *  2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
629  *  capabilities of the process are cleared.
630  *
631  *  3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
632  *  capabilities are set to the permitted capabilities.
633  *
634  *  fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
635  *  never happen.
636  *
637  *  -astor
638  *
639  * cevans - New behaviour, Oct '99
640  * A process may, via prctl(), elect to keep its capabilities when it
641  * calls setuid() and switches away from uid==0. Both permitted and
642  * effective sets will be retained.
643  * Without this change, it was impossible for a daemon to drop only some
644  * of its privilege. The call to setuid(!=0) would drop all privileges!
645  * Keeping uid 0 is not an option because uid 0 owns too many vital
646  * files..
647  * Thanks to Olaf Kirch and Peter Benie for spotting this.
648  */
649 static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old)
650 {
651 	if ((old->uid == 0 || old->euid == 0 || old->suid == 0) &&
652 	    (new->uid != 0 && new->euid != 0 && new->suid != 0) &&
653 	    !issecure(SECURE_KEEP_CAPS)) {
654 		cap_clear(new->cap_permitted);
655 		cap_clear(new->cap_effective);
656 	}
657 	if (old->euid == 0 && new->euid != 0)
658 		cap_clear(new->cap_effective);
659 	if (old->euid != 0 && new->euid == 0)
660 		new->cap_effective = new->cap_permitted;
661 }
662 
663 /**
664  * cap_task_fix_setuid - Fix up the results of setuid() call
665  * @new: The proposed credentials
666  * @old: The current task's current credentials
667  * @flags: Indications of what has changed
668  *
669  * Fix up the results of setuid() call before the credential changes are
670  * actually applied, returning 0 to grant the changes, -ve to deny them.
671  */
672 int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags)
673 {
674 	switch (flags) {
675 	case LSM_SETID_RE:
676 	case LSM_SETID_ID:
677 	case LSM_SETID_RES:
678 		/* juggle the capabilities to follow [RES]UID changes unless
679 		 * otherwise suppressed */
680 		if (!issecure(SECURE_NO_SETUID_FIXUP))
681 			cap_emulate_setxuid(new, old);
682 		break;
683 
684 	case LSM_SETID_FS:
685 		/* juggle the capabilties to follow FSUID changes, unless
686 		 * otherwise suppressed
687 		 *
688 		 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
689 		 *          if not, we might be a bit too harsh here.
690 		 */
691 		if (!issecure(SECURE_NO_SETUID_FIXUP)) {
692 			if (old->fsuid == 0 && new->fsuid != 0)
693 				new->cap_effective =
694 					cap_drop_fs_set(new->cap_effective);
695 
696 			if (old->fsuid != 0 && new->fsuid == 0)
697 				new->cap_effective =
698 					cap_raise_fs_set(new->cap_effective,
699 							 new->cap_permitted);
700 		}
701 		break;
702 
703 	default:
704 		return -EINVAL;
705 	}
706 
707 	return 0;
708 }
709 
710 #ifdef CONFIG_SECURITY_FILE_CAPABILITIES
711 /*
712  * Rationale: code calling task_setscheduler, task_setioprio, and
713  * task_setnice, assumes that
714  *   . if capable(cap_sys_nice), then those actions should be allowed
715  *   . if not capable(cap_sys_nice), but acting on your own processes,
716  *   	then those actions should be allowed
717  * This is insufficient now since you can call code without suid, but
718  * yet with increased caps.
719  * So we check for increased caps on the target process.
720  */
721 static int cap_safe_nice(struct task_struct *p)
722 {
723 	int is_subset;
724 
725 	rcu_read_lock();
726 	is_subset = cap_issubset(__task_cred(p)->cap_permitted,
727 				 current_cred()->cap_permitted);
728 	rcu_read_unlock();
729 
730 	if (!is_subset && !capable(CAP_SYS_NICE))
731 		return -EPERM;
732 	return 0;
733 }
734 
735 /**
736  * cap_task_setscheduler - Detemine if scheduler policy change is permitted
737  * @p: The task to affect
738  * @policy: The policy to effect
739  * @lp: The parameters to the scheduling policy
740  *
741  * Detemine if the requested scheduler policy change is permitted for the
742  * specified task, returning 0 if permission is granted, -ve if denied.
743  */
744 int cap_task_setscheduler(struct task_struct *p, int policy,
745 			   struct sched_param *lp)
746 {
747 	return cap_safe_nice(p);
748 }
749 
750 /**
751  * cap_task_ioprio - Detemine if I/O priority change is permitted
752  * @p: The task to affect
753  * @ioprio: The I/O priority to set
754  *
755  * Detemine if the requested I/O priority change is permitted for the specified
756  * task, returning 0 if permission is granted, -ve if denied.
757  */
758 int cap_task_setioprio(struct task_struct *p, int ioprio)
759 {
760 	return cap_safe_nice(p);
761 }
762 
763 /**
764  * cap_task_ioprio - Detemine if task priority change is permitted
765  * @p: The task to affect
766  * @nice: The nice value to set
767  *
768  * Detemine if the requested task priority change is permitted for the
769  * specified task, returning 0 if permission is granted, -ve if denied.
770  */
771 int cap_task_setnice(struct task_struct *p, int nice)
772 {
773 	return cap_safe_nice(p);
774 }
775 
776 /*
777  * Implement PR_CAPBSET_DROP.  Attempt to remove the specified capability from
778  * the current task's bounding set.  Returns 0 on success, -ve on error.
779  */
780 static long cap_prctl_drop(struct cred *new, unsigned long cap)
781 {
782 	if (!capable(CAP_SETPCAP))
783 		return -EPERM;
784 	if (!cap_valid(cap))
785 		return -EINVAL;
786 
787 	cap_lower(new->cap_bset, cap);
788 	return 0;
789 }
790 
791 #else
792 int cap_task_setscheduler (struct task_struct *p, int policy,
793 			   struct sched_param *lp)
794 {
795 	return 0;
796 }
797 int cap_task_setioprio (struct task_struct *p, int ioprio)
798 {
799 	return 0;
800 }
801 int cap_task_setnice (struct task_struct *p, int nice)
802 {
803 	return 0;
804 }
805 #endif
806 
807 /**
808  * cap_task_prctl - Implement process control functions for this security module
809  * @option: The process control function requested
810  * @arg2, @arg3, @arg4, @arg5: The argument data for this function
811  *
812  * Allow process control functions (sys_prctl()) to alter capabilities; may
813  * also deny access to other functions not otherwise implemented here.
814  *
815  * Returns 0 or +ve on success, -ENOSYS if this function is not implemented
816  * here, other -ve on error.  If -ENOSYS is returned, sys_prctl() and other LSM
817  * modules will consider performing the function.
818  */
819 int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,
820 		   unsigned long arg4, unsigned long arg5)
821 {
822 	struct cred *new;
823 	long error = 0;
824 
825 	new = prepare_creds();
826 	if (!new)
827 		return -ENOMEM;
828 
829 	switch (option) {
830 	case PR_CAPBSET_READ:
831 		error = -EINVAL;
832 		if (!cap_valid(arg2))
833 			goto error;
834 		error = !!cap_raised(new->cap_bset, arg2);
835 		goto no_change;
836 
837 #ifdef CONFIG_SECURITY_FILE_CAPABILITIES
838 	case PR_CAPBSET_DROP:
839 		error = cap_prctl_drop(new, arg2);
840 		if (error < 0)
841 			goto error;
842 		goto changed;
843 
844 	/*
845 	 * The next four prctl's remain to assist with transitioning a
846 	 * system from legacy UID=0 based privilege (when filesystem
847 	 * capabilities are not in use) to a system using filesystem
848 	 * capabilities only - as the POSIX.1e draft intended.
849 	 *
850 	 * Note:
851 	 *
852 	 *  PR_SET_SECUREBITS =
853 	 *      issecure_mask(SECURE_KEEP_CAPS_LOCKED)
854 	 *    | issecure_mask(SECURE_NOROOT)
855 	 *    | issecure_mask(SECURE_NOROOT_LOCKED)
856 	 *    | issecure_mask(SECURE_NO_SETUID_FIXUP)
857 	 *    | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
858 	 *
859 	 * will ensure that the current process and all of its
860 	 * children will be locked into a pure
861 	 * capability-based-privilege environment.
862 	 */
863 	case PR_SET_SECUREBITS:
864 		error = -EPERM;
865 		if ((((new->securebits & SECURE_ALL_LOCKS) >> 1)
866 		     & (new->securebits ^ arg2))			/*[1]*/
867 		    || ((new->securebits & SECURE_ALL_LOCKS & ~arg2))	/*[2]*/
868 		    || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS))	/*[3]*/
869 		    || (cap_capable(current, current_cred(), CAP_SETPCAP,
870 				    SECURITY_CAP_AUDIT) != 0)		/*[4]*/
871 			/*
872 			 * [1] no changing of bits that are locked
873 			 * [2] no unlocking of locks
874 			 * [3] no setting of unsupported bits
875 			 * [4] doing anything requires privilege (go read about
876 			 *     the "sendmail capabilities bug")
877 			 */
878 		    )
879 			/* cannot change a locked bit */
880 			goto error;
881 		new->securebits = arg2;
882 		goto changed;
883 
884 	case PR_GET_SECUREBITS:
885 		error = new->securebits;
886 		goto no_change;
887 
888 #endif /* def CONFIG_SECURITY_FILE_CAPABILITIES */
889 
890 	case PR_GET_KEEPCAPS:
891 		if (issecure(SECURE_KEEP_CAPS))
892 			error = 1;
893 		goto no_change;
894 
895 	case PR_SET_KEEPCAPS:
896 		error = -EINVAL;
897 		if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */
898 			goto error;
899 		error = -EPERM;
900 		if (issecure(SECURE_KEEP_CAPS_LOCKED))
901 			goto error;
902 		if (arg2)
903 			new->securebits |= issecure_mask(SECURE_KEEP_CAPS);
904 		else
905 			new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
906 		goto changed;
907 
908 	default:
909 		/* No functionality available - continue with default */
910 		error = -ENOSYS;
911 		goto error;
912 	}
913 
914 	/* Functionality provided */
915 changed:
916 	return commit_creds(new);
917 
918 no_change:
919 error:
920 	abort_creds(new);
921 	return error;
922 }
923 
924 /**
925  * cap_syslog - Determine whether syslog function is permitted
926  * @type: Function requested
927  *
928  * Determine whether the current process is permitted to use a particular
929  * syslog function, returning 0 if permission is granted, -ve if not.
930  */
931 int cap_syslog(int type)
932 {
933 	if ((type != 3 && type != 10) && !capable(CAP_SYS_ADMIN))
934 		return -EPERM;
935 	return 0;
936 }
937 
938 /**
939  * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted
940  * @mm: The VM space in which the new mapping is to be made
941  * @pages: The size of the mapping
942  *
943  * Determine whether the allocation of a new virtual mapping by the current
944  * task is permitted, returning 0 if permission is granted, -ve if not.
945  */
946 int cap_vm_enough_memory(struct mm_struct *mm, long pages)
947 {
948 	int cap_sys_admin = 0;
949 
950 	if (cap_capable(current, current_cred(), CAP_SYS_ADMIN,
951 			SECURITY_CAP_NOAUDIT) == 0)
952 		cap_sys_admin = 1;
953 	return __vm_enough_memory(mm, pages, cap_sys_admin);
954 }
955