1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Common capabilities, needed by capability.o.
3 */
4
5 #include <linux/capability.h>
6 #include <linux/audit.h>
7 #include <linux/init.h>
8 #include <linux/kernel.h>
9 #include <linux/lsm_hooks.h>
10 #include <linux/file.h>
11 #include <linux/mm.h>
12 #include <linux/mman.h>
13 #include <linux/pagemap.h>
14 #include <linux/swap.h>
15 #include <linux/skbuff.h>
16 #include <linux/netlink.h>
17 #include <linux/ptrace.h>
18 #include <linux/xattr.h>
19 #include <linux/hugetlb.h>
20 #include <linux/mount.h>
21 #include <linux/sched.h>
22 #include <linux/prctl.h>
23 #include <linux/securebits.h>
24 #include <linux/user_namespace.h>
25 #include <linux/binfmts.h>
26 #include <linux/personality.h>
27
28 /*
29 * If a non-root user executes a setuid-root binary in
30 * !secure(SECURE_NOROOT) mode, then we raise capabilities.
31 * However if fE is also set, then the intent is for only
32 * the file capabilities to be applied, and the setuid-root
33 * bit is left on either to change the uid (plausible) or
34 * to get full privilege on a kernel without file capabilities
35 * support. So in that case we do not raise capabilities.
36 *
37 * Warn if that happens, once per boot.
38 */
warn_setuid_and_fcaps_mixed(const char * fname)39 static void warn_setuid_and_fcaps_mixed(const char *fname)
40 {
41 static int warned;
42 if (!warned) {
43 printk(KERN_INFO "warning: `%s' has both setuid-root and"
44 " effective capabilities. Therefore not raising all"
45 " capabilities.\n", fname);
46 warned = 1;
47 }
48 }
49
50 /**
51 * cap_capable - Determine whether a task has a particular effective capability
52 * @cred: The credentials to use
53 * @ns: The user namespace in which we need the capability
54 * @cap: The capability to check for
55 * @opts: Bitmask of options defined in include/linux/security.h
56 *
57 * Determine whether the nominated task has the specified capability amongst
58 * its effective set, returning 0 if it does, -ve if it does not.
59 *
60 * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable()
61 * and has_capability() functions. That is, it has the reverse semantics:
62 * cap_has_capability() returns 0 when a task has a capability, but the
63 * kernel's capable() and has_capability() returns 1 for this case.
64 */
cap_capable(const struct cred * cred,struct user_namespace * targ_ns,int cap,unsigned int opts)65 int cap_capable(const struct cred *cred, struct user_namespace *targ_ns,
66 int cap, unsigned int opts)
67 {
68 struct user_namespace *ns = targ_ns;
69
70 /* See if cred has the capability in the target user namespace
71 * by examining the target user namespace and all of the target
72 * user namespace's parents.
73 */
74 for (;;) {
75 /* Do we have the necessary capabilities? */
76 if (ns == cred->user_ns)
77 return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM;
78
79 /*
80 * If we're already at a lower level than we're looking for,
81 * we're done searching.
82 */
83 if (ns->level <= cred->user_ns->level)
84 return -EPERM;
85
86 /*
87 * The owner of the user namespace in the parent of the
88 * user namespace has all caps.
89 */
90 if ((ns->parent == cred->user_ns) && uid_eq(ns->owner, cred->euid))
91 return 0;
92
93 /*
94 * If you have a capability in a parent user ns, then you have
95 * it over all children user namespaces as well.
96 */
97 ns = ns->parent;
98 }
99
100 /* We never get here */
101 }
102
103 /**
104 * cap_settime - Determine whether the current process may set the system clock
105 * @ts: The time to set
106 * @tz: The timezone to set
107 *
108 * Determine whether the current process may set the system clock and timezone
109 * information, returning 0 if permission granted, -ve if denied.
110 */
cap_settime(const struct timespec64 * ts,const struct timezone * tz)111 int cap_settime(const struct timespec64 *ts, const struct timezone *tz)
112 {
113 if (!capable(CAP_SYS_TIME))
114 return -EPERM;
115 return 0;
116 }
117
118 /**
119 * cap_ptrace_access_check - Determine whether the current process may access
120 * another
121 * @child: The process to be accessed
122 * @mode: The mode of attachment.
123 *
124 * If we are in the same or an ancestor user_ns and have all the target
125 * task's capabilities, then ptrace access is allowed.
126 * If we have the ptrace capability to the target user_ns, then ptrace
127 * access is allowed.
128 * Else denied.
129 *
130 * Determine whether a process may access another, returning 0 if permission
131 * granted, -ve if denied.
132 */
cap_ptrace_access_check(struct task_struct * child,unsigned int mode)133 int cap_ptrace_access_check(struct task_struct *child, unsigned int mode)
134 {
135 int ret = 0;
136 const struct cred *cred, *child_cred;
137 const kernel_cap_t *caller_caps;
138
139 rcu_read_lock();
140 cred = current_cred();
141 child_cred = __task_cred(child);
142 if (mode & PTRACE_MODE_FSCREDS)
143 caller_caps = &cred->cap_effective;
144 else
145 caller_caps = &cred->cap_permitted;
146 if (cred->user_ns == child_cred->user_ns &&
147 cap_issubset(child_cred->cap_permitted, *caller_caps))
148 goto out;
149 if (ns_capable(child_cred->user_ns, CAP_SYS_PTRACE))
150 goto out;
151 ret = -EPERM;
152 out:
153 rcu_read_unlock();
154 return ret;
155 }
156
157 /**
158 * cap_ptrace_traceme - Determine whether another process may trace the current
159 * @parent: The task proposed to be the tracer
160 *
161 * If parent is in the same or an ancestor user_ns and has all current's
162 * capabilities, then ptrace access is allowed.
163 * If parent has the ptrace capability to current's user_ns, then ptrace
164 * access is allowed.
165 * Else denied.
166 *
167 * Determine whether the nominated task is permitted to trace the current
168 * process, returning 0 if permission is granted, -ve if denied.
169 */
cap_ptrace_traceme(struct task_struct * parent)170 int cap_ptrace_traceme(struct task_struct *parent)
171 {
172 int ret = 0;
173 const struct cred *cred, *child_cred;
174
175 rcu_read_lock();
176 cred = __task_cred(parent);
177 child_cred = current_cred();
178 if (cred->user_ns == child_cred->user_ns &&
179 cap_issubset(child_cred->cap_permitted, cred->cap_permitted))
180 goto out;
181 if (has_ns_capability(parent, child_cred->user_ns, CAP_SYS_PTRACE))
182 goto out;
183 ret = -EPERM;
184 out:
185 rcu_read_unlock();
186 return ret;
187 }
188
189 /**
190 * cap_capget - Retrieve a task's capability sets
191 * @target: The task from which to retrieve the capability sets
192 * @effective: The place to record the effective set
193 * @inheritable: The place to record the inheritable set
194 * @permitted: The place to record the permitted set
195 *
196 * This function retrieves the capabilities of the nominated task and returns
197 * them to the caller.
198 */
cap_capget(struct task_struct * target,kernel_cap_t * effective,kernel_cap_t * inheritable,kernel_cap_t * permitted)199 int cap_capget(struct task_struct *target, kernel_cap_t *effective,
200 kernel_cap_t *inheritable, kernel_cap_t *permitted)
201 {
202 const struct cred *cred;
203
204 /* Derived from kernel/capability.c:sys_capget. */
205 rcu_read_lock();
206 cred = __task_cred(target);
207 *effective = cred->cap_effective;
208 *inheritable = cred->cap_inheritable;
209 *permitted = cred->cap_permitted;
210 rcu_read_unlock();
211 return 0;
212 }
213
214 /*
215 * Determine whether the inheritable capabilities are limited to the old
216 * permitted set. Returns 1 if they are limited, 0 if they are not.
217 */
cap_inh_is_capped(void)218 static inline int cap_inh_is_capped(void)
219 {
220 /* they are so limited unless the current task has the CAP_SETPCAP
221 * capability
222 */
223 if (cap_capable(current_cred(), current_cred()->user_ns,
224 CAP_SETPCAP, CAP_OPT_NONE) == 0)
225 return 0;
226 return 1;
227 }
228
229 /**
230 * cap_capset - Validate and apply proposed changes to current's capabilities
231 * @new: The proposed new credentials; alterations should be made here
232 * @old: The current task's current credentials
233 * @effective: A pointer to the proposed new effective capabilities set
234 * @inheritable: A pointer to the proposed new inheritable capabilities set
235 * @permitted: A pointer to the proposed new permitted capabilities set
236 *
237 * This function validates and applies a proposed mass change to the current
238 * process's capability sets. The changes are made to the proposed new
239 * credentials, and assuming no error, will be committed by the caller of LSM.
240 */
cap_capset(struct cred * new,const struct cred * old,const kernel_cap_t * effective,const kernel_cap_t * inheritable,const kernel_cap_t * permitted)241 int cap_capset(struct cred *new,
242 const struct cred *old,
243 const kernel_cap_t *effective,
244 const kernel_cap_t *inheritable,
245 const kernel_cap_t *permitted)
246 {
247 if (cap_inh_is_capped() &&
248 !cap_issubset(*inheritable,
249 cap_combine(old->cap_inheritable,
250 old->cap_permitted)))
251 /* incapable of using this inheritable set */
252 return -EPERM;
253
254 if (!cap_issubset(*inheritable,
255 cap_combine(old->cap_inheritable,
256 old->cap_bset)))
257 /* no new pI capabilities outside bounding set */
258 return -EPERM;
259
260 /* verify restrictions on target's new Permitted set */
261 if (!cap_issubset(*permitted, old->cap_permitted))
262 return -EPERM;
263
264 /* verify the _new_Effective_ is a subset of the _new_Permitted_ */
265 if (!cap_issubset(*effective, *permitted))
266 return -EPERM;
267
268 new->cap_effective = *effective;
269 new->cap_inheritable = *inheritable;
270 new->cap_permitted = *permitted;
271
272 /*
273 * Mask off ambient bits that are no longer both permitted and
274 * inheritable.
275 */
276 new->cap_ambient = cap_intersect(new->cap_ambient,
277 cap_intersect(*permitted,
278 *inheritable));
279 if (WARN_ON(!cap_ambient_invariant_ok(new)))
280 return -EINVAL;
281 return 0;
282 }
283
284 /**
285 * cap_inode_need_killpriv - Determine if inode change affects privileges
286 * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV
287 *
288 * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV
289 * affects the security markings on that inode, and if it is, should
290 * inode_killpriv() be invoked or the change rejected.
291 *
292 * Returns 1 if security.capability has a value, meaning inode_killpriv()
293 * is required, 0 otherwise, meaning inode_killpriv() is not required.
294 */
cap_inode_need_killpriv(struct dentry * dentry)295 int cap_inode_need_killpriv(struct dentry *dentry)
296 {
297 struct inode *inode = d_backing_inode(dentry);
298 int error;
299
300 error = __vfs_getxattr(dentry, inode, XATTR_NAME_CAPS, NULL, 0,
301 XATTR_NOSECURITY);
302 return error > 0;
303 }
304
305 /**
306 * cap_inode_killpriv - Erase the security markings on an inode
307 * @dentry: The inode/dentry to alter
308 *
309 * Erase the privilege-enhancing security markings on an inode.
310 *
311 * Returns 0 if successful, -ve on error.
312 */
cap_inode_killpriv(struct dentry * dentry)313 int cap_inode_killpriv(struct dentry *dentry)
314 {
315 int error;
316
317 error = __vfs_removexattr(dentry, XATTR_NAME_CAPS);
318 if (error == -EOPNOTSUPP)
319 error = 0;
320 return error;
321 }
322
rootid_owns_currentns(kuid_t kroot)323 static bool rootid_owns_currentns(kuid_t kroot)
324 {
325 struct user_namespace *ns;
326
327 if (!uid_valid(kroot))
328 return false;
329
330 for (ns = current_user_ns(); ; ns = ns->parent) {
331 if (from_kuid(ns, kroot) == 0)
332 return true;
333 if (ns == &init_user_ns)
334 break;
335 }
336
337 return false;
338 }
339
sansflags(__u32 m)340 static __u32 sansflags(__u32 m)
341 {
342 return m & ~VFS_CAP_FLAGS_EFFECTIVE;
343 }
344
is_v2header(size_t size,const struct vfs_cap_data * cap)345 static bool is_v2header(size_t size, const struct vfs_cap_data *cap)
346 {
347 if (size != XATTR_CAPS_SZ_2)
348 return false;
349 return sansflags(le32_to_cpu(cap->magic_etc)) == VFS_CAP_REVISION_2;
350 }
351
is_v3header(size_t size,const struct vfs_cap_data * cap)352 static bool is_v3header(size_t size, const struct vfs_cap_data *cap)
353 {
354 if (size != XATTR_CAPS_SZ_3)
355 return false;
356 return sansflags(le32_to_cpu(cap->magic_etc)) == VFS_CAP_REVISION_3;
357 }
358
359 /*
360 * getsecurity: We are called for security.* before any attempt to read the
361 * xattr from the inode itself.
362 *
363 * This gives us a chance to read the on-disk value and convert it. If we
364 * return -EOPNOTSUPP, then vfs_getxattr() will call the i_op handler.
365 *
366 * Note we are not called by vfs_getxattr_alloc(), but that is only called
367 * by the integrity subsystem, which really wants the unconverted values -
368 * so that's good.
369 */
cap_inode_getsecurity(struct inode * inode,const char * name,void ** buffer,bool alloc)370 int cap_inode_getsecurity(struct inode *inode, const char *name, void **buffer,
371 bool alloc)
372 {
373 int size, ret;
374 kuid_t kroot;
375 u32 nsmagic, magic;
376 uid_t root, mappedroot;
377 char *tmpbuf = NULL;
378 struct vfs_cap_data *cap;
379 struct vfs_ns_cap_data *nscap = NULL;
380 struct dentry *dentry;
381 struct user_namespace *fs_ns;
382
383 if (strcmp(name, "capability") != 0)
384 return -EOPNOTSUPP;
385
386 dentry = d_find_any_alias(inode);
387 if (!dentry)
388 return -EINVAL;
389
390 size = sizeof(struct vfs_ns_cap_data);
391 ret = (int) vfs_getxattr_alloc(dentry, XATTR_NAME_CAPS,
392 &tmpbuf, size, GFP_NOFS);
393 dput(dentry);
394
395 if (ret < 0 || !tmpbuf) {
396 size = ret;
397 goto out_free;
398 }
399
400 fs_ns = inode->i_sb->s_user_ns;
401 cap = (struct vfs_cap_data *) tmpbuf;
402 if (is_v2header((size_t) ret, cap)) {
403 root = 0;
404 } else if (is_v3header((size_t) ret, cap)) {
405 nscap = (struct vfs_ns_cap_data *) tmpbuf;
406 root = le32_to_cpu(nscap->rootid);
407 } else {
408 size = -EINVAL;
409 goto out_free;
410 }
411
412 kroot = make_kuid(fs_ns, root);
413
414 /* If the root kuid maps to a valid uid in current ns, then return
415 * this as a nscap. */
416 mappedroot = from_kuid(current_user_ns(), kroot);
417 if (mappedroot != (uid_t)-1 && mappedroot != (uid_t)0) {
418 size = sizeof(struct vfs_ns_cap_data);
419 if (alloc) {
420 if (!nscap) {
421 /* v2 -> v3 conversion */
422 nscap = kzalloc(size, GFP_ATOMIC);
423 if (!nscap) {
424 size = -ENOMEM;
425 goto out_free;
426 }
427 nsmagic = VFS_CAP_REVISION_3;
428 magic = le32_to_cpu(cap->magic_etc);
429 if (magic & VFS_CAP_FLAGS_EFFECTIVE)
430 nsmagic |= VFS_CAP_FLAGS_EFFECTIVE;
431 memcpy(&nscap->data, &cap->data, sizeof(__le32) * 2 * VFS_CAP_U32);
432 nscap->magic_etc = cpu_to_le32(nsmagic);
433 } else {
434 /* use allocated v3 buffer */
435 tmpbuf = NULL;
436 }
437 nscap->rootid = cpu_to_le32(mappedroot);
438 *buffer = nscap;
439 }
440 goto out_free;
441 }
442
443 if (!rootid_owns_currentns(kroot)) {
444 size = -EOVERFLOW;
445 goto out_free;
446 }
447
448 /* This comes from a parent namespace. Return as a v2 capability */
449 size = sizeof(struct vfs_cap_data);
450 if (alloc) {
451 if (nscap) {
452 /* v3 -> v2 conversion */
453 cap = kzalloc(size, GFP_ATOMIC);
454 if (!cap) {
455 size = -ENOMEM;
456 goto out_free;
457 }
458 magic = VFS_CAP_REVISION_2;
459 nsmagic = le32_to_cpu(nscap->magic_etc);
460 if (nsmagic & VFS_CAP_FLAGS_EFFECTIVE)
461 magic |= VFS_CAP_FLAGS_EFFECTIVE;
462 memcpy(&cap->data, &nscap->data, sizeof(__le32) * 2 * VFS_CAP_U32);
463 cap->magic_etc = cpu_to_le32(magic);
464 } else {
465 /* use unconverted v2 */
466 tmpbuf = NULL;
467 }
468 *buffer = cap;
469 }
470 out_free:
471 kfree(tmpbuf);
472 return size;
473 }
474
rootid_from_xattr(const void * value,size_t size,struct user_namespace * task_ns)475 static kuid_t rootid_from_xattr(const void *value, size_t size,
476 struct user_namespace *task_ns)
477 {
478 const struct vfs_ns_cap_data *nscap = value;
479 uid_t rootid = 0;
480
481 if (size == XATTR_CAPS_SZ_3)
482 rootid = le32_to_cpu(nscap->rootid);
483
484 return make_kuid(task_ns, rootid);
485 }
486
validheader(size_t size,const struct vfs_cap_data * cap)487 static bool validheader(size_t size, const struct vfs_cap_data *cap)
488 {
489 return is_v2header(size, cap) || is_v3header(size, cap);
490 }
491
492 /*
493 * User requested a write of security.capability. If needed, update the
494 * xattr to change from v2 to v3, or to fixup the v3 rootid.
495 *
496 * If all is ok, we return the new size, on error return < 0.
497 */
cap_convert_nscap(struct dentry * dentry,void ** ivalue,size_t size)498 int cap_convert_nscap(struct dentry *dentry, void **ivalue, size_t size)
499 {
500 struct vfs_ns_cap_data *nscap;
501 uid_t nsrootid;
502 const struct vfs_cap_data *cap = *ivalue;
503 __u32 magic, nsmagic;
504 struct inode *inode = d_backing_inode(dentry);
505 struct user_namespace *task_ns = current_user_ns(),
506 *fs_ns = inode->i_sb->s_user_ns;
507 kuid_t rootid;
508 size_t newsize;
509
510 if (!*ivalue)
511 return -EINVAL;
512 if (!validheader(size, cap))
513 return -EINVAL;
514 if (!capable_wrt_inode_uidgid(inode, CAP_SETFCAP))
515 return -EPERM;
516 if (size == XATTR_CAPS_SZ_2)
517 if (ns_capable(inode->i_sb->s_user_ns, CAP_SETFCAP))
518 /* user is privileged, just write the v2 */
519 return size;
520
521 rootid = rootid_from_xattr(*ivalue, size, task_ns);
522 if (!uid_valid(rootid))
523 return -EINVAL;
524
525 nsrootid = from_kuid(fs_ns, rootid);
526 if (nsrootid == -1)
527 return -EINVAL;
528
529 newsize = sizeof(struct vfs_ns_cap_data);
530 nscap = kmalloc(newsize, GFP_ATOMIC);
531 if (!nscap)
532 return -ENOMEM;
533 nscap->rootid = cpu_to_le32(nsrootid);
534 nsmagic = VFS_CAP_REVISION_3;
535 magic = le32_to_cpu(cap->magic_etc);
536 if (magic & VFS_CAP_FLAGS_EFFECTIVE)
537 nsmagic |= VFS_CAP_FLAGS_EFFECTIVE;
538 nscap->magic_etc = cpu_to_le32(nsmagic);
539 memcpy(&nscap->data, &cap->data, sizeof(__le32) * 2 * VFS_CAP_U32);
540
541 kvfree(*ivalue);
542 *ivalue = nscap;
543 return newsize;
544 }
545
546 /*
547 * Calculate the new process capability sets from the capability sets attached
548 * to a file.
549 */
bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data * caps,struct linux_binprm * bprm,bool * effective,bool * has_fcap)550 static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps,
551 struct linux_binprm *bprm,
552 bool *effective,
553 bool *has_fcap)
554 {
555 struct cred *new = bprm->cred;
556 unsigned i;
557 int ret = 0;
558
559 if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE)
560 *effective = true;
561
562 if (caps->magic_etc & VFS_CAP_REVISION_MASK)
563 *has_fcap = true;
564
565 CAP_FOR_EACH_U32(i) {
566 __u32 permitted = caps->permitted.cap[i];
567 __u32 inheritable = caps->inheritable.cap[i];
568
569 /*
570 * pP' = (X & fP) | (pI & fI)
571 * The addition of pA' is handled later.
572 */
573 new->cap_permitted.cap[i] =
574 (new->cap_bset.cap[i] & permitted) |
575 (new->cap_inheritable.cap[i] & inheritable);
576
577 if (permitted & ~new->cap_permitted.cap[i])
578 /* insufficient to execute correctly */
579 ret = -EPERM;
580 }
581
582 /*
583 * For legacy apps, with no internal support for recognizing they
584 * do not have enough capabilities, we return an error if they are
585 * missing some "forced" (aka file-permitted) capabilities.
586 */
587 return *effective ? ret : 0;
588 }
589
590 /*
591 * Extract the on-exec-apply capability sets for an executable file.
592 */
get_vfs_caps_from_disk(const struct dentry * dentry,struct cpu_vfs_cap_data * cpu_caps)593 int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps)
594 {
595 struct inode *inode = d_backing_inode(dentry);
596 __u32 magic_etc;
597 unsigned tocopy, i;
598 int size;
599 struct vfs_ns_cap_data data, *nscaps = &data;
600 struct vfs_cap_data *caps = (struct vfs_cap_data *) &data;
601 kuid_t rootkuid;
602 struct user_namespace *fs_ns;
603
604 memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data));
605
606 if (!inode)
607 return -ENODATA;
608
609 fs_ns = inode->i_sb->s_user_ns;
610 size = __vfs_getxattr((struct dentry *)dentry, inode,
611 XATTR_NAME_CAPS, &data, XATTR_CAPS_SZ,
612 XATTR_NOSECURITY);
613 if (size == -ENODATA || size == -EOPNOTSUPP)
614 /* no data, that's ok */
615 return -ENODATA;
616
617 if (size < 0)
618 return size;
619
620 if (size < sizeof(magic_etc))
621 return -EINVAL;
622
623 cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps->magic_etc);
624
625 rootkuid = make_kuid(fs_ns, 0);
626 switch (magic_etc & VFS_CAP_REVISION_MASK) {
627 case VFS_CAP_REVISION_1:
628 if (size != XATTR_CAPS_SZ_1)
629 return -EINVAL;
630 tocopy = VFS_CAP_U32_1;
631 break;
632 case VFS_CAP_REVISION_2:
633 if (size != XATTR_CAPS_SZ_2)
634 return -EINVAL;
635 tocopy = VFS_CAP_U32_2;
636 break;
637 case VFS_CAP_REVISION_3:
638 if (size != XATTR_CAPS_SZ_3)
639 return -EINVAL;
640 tocopy = VFS_CAP_U32_3;
641 rootkuid = make_kuid(fs_ns, le32_to_cpu(nscaps->rootid));
642 break;
643
644 default:
645 return -EINVAL;
646 }
647 /* Limit the caps to the mounter of the filesystem
648 * or the more limited uid specified in the xattr.
649 */
650 if (!rootid_owns_currentns(rootkuid))
651 return -ENODATA;
652
653 CAP_FOR_EACH_U32(i) {
654 if (i >= tocopy)
655 break;
656 cpu_caps->permitted.cap[i] = le32_to_cpu(caps->data[i].permitted);
657 cpu_caps->inheritable.cap[i] = le32_to_cpu(caps->data[i].inheritable);
658 }
659
660 cpu_caps->permitted.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
661 cpu_caps->inheritable.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
662
663 cpu_caps->rootid = rootkuid;
664
665 return 0;
666 }
667
668 /*
669 * Attempt to get the on-exec apply capability sets for an executable file from
670 * its xattrs and, if present, apply them to the proposed credentials being
671 * constructed by execve().
672 */
get_file_caps(struct linux_binprm * bprm,struct file * file,bool * effective,bool * has_fcap)673 static int get_file_caps(struct linux_binprm *bprm, struct file *file,
674 bool *effective, bool *has_fcap)
675 {
676 int rc = 0;
677 struct cpu_vfs_cap_data vcaps;
678
679 cap_clear(bprm->cred->cap_permitted);
680
681 if (!file_caps_enabled)
682 return 0;
683
684 if (!mnt_may_suid(file->f_path.mnt))
685 return 0;
686
687 /*
688 * This check is redundant with mnt_may_suid() but is kept to make
689 * explicit that capability bits are limited to s_user_ns and its
690 * descendants.
691 */
692 if (!current_in_userns(file->f_path.mnt->mnt_sb->s_user_ns))
693 return 0;
694
695 rc = get_vfs_caps_from_disk(file->f_path.dentry, &vcaps);
696 if (rc < 0) {
697 if (rc == -EINVAL)
698 printk(KERN_NOTICE "Invalid argument reading file caps for %s\n",
699 bprm->filename);
700 else if (rc == -ENODATA)
701 rc = 0;
702 goto out;
703 }
704
705 rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective, has_fcap);
706
707 out:
708 if (rc)
709 cap_clear(bprm->cred->cap_permitted);
710
711 return rc;
712 }
713
root_privileged(void)714 static inline bool root_privileged(void) { return !issecure(SECURE_NOROOT); }
715
__is_real(kuid_t uid,struct cred * cred)716 static inline bool __is_real(kuid_t uid, struct cred *cred)
717 { return uid_eq(cred->uid, uid); }
718
__is_eff(kuid_t uid,struct cred * cred)719 static inline bool __is_eff(kuid_t uid, struct cred *cred)
720 { return uid_eq(cred->euid, uid); }
721
__is_suid(kuid_t uid,struct cred * cred)722 static inline bool __is_suid(kuid_t uid, struct cred *cred)
723 { return !__is_real(uid, cred) && __is_eff(uid, cred); }
724
725 /*
726 * handle_privileged_root - Handle case of privileged root
727 * @bprm: The execution parameters, including the proposed creds
728 * @has_fcap: Are any file capabilities set?
729 * @effective: Do we have effective root privilege?
730 * @root_uid: This namespace' root UID WRT initial USER namespace
731 *
732 * Handle the case where root is privileged and hasn't been neutered by
733 * SECURE_NOROOT. If file capabilities are set, they won't be combined with
734 * set UID root and nothing is changed. If we are root, cap_permitted is
735 * updated. If we have become set UID root, the effective bit is set.
736 */
handle_privileged_root(struct linux_binprm * bprm,bool has_fcap,bool * effective,kuid_t root_uid)737 static void handle_privileged_root(struct linux_binprm *bprm, bool has_fcap,
738 bool *effective, kuid_t root_uid)
739 {
740 const struct cred *old = current_cred();
741 struct cred *new = bprm->cred;
742
743 if (!root_privileged())
744 return;
745 /*
746 * If the legacy file capability is set, then don't set privs
747 * for a setuid root binary run by a non-root user. Do set it
748 * for a root user just to cause least surprise to an admin.
749 */
750 if (has_fcap && __is_suid(root_uid, new)) {
751 warn_setuid_and_fcaps_mixed(bprm->filename);
752 return;
753 }
754 /*
755 * To support inheritance of root-permissions and suid-root
756 * executables under compatibility mode, we override the
757 * capability sets for the file.
758 */
759 if (__is_eff(root_uid, new) || __is_real(root_uid, new)) {
760 /* pP' = (cap_bset & ~0) | (pI & ~0) */
761 new->cap_permitted = cap_combine(old->cap_bset,
762 old->cap_inheritable);
763 }
764 /*
765 * If only the real uid is 0, we do not set the effective bit.
766 */
767 if (__is_eff(root_uid, new))
768 *effective = true;
769 }
770
771 #define __cap_gained(field, target, source) \
772 !cap_issubset(target->cap_##field, source->cap_##field)
773 #define __cap_grew(target, source, cred) \
774 !cap_issubset(cred->cap_##target, cred->cap_##source)
775 #define __cap_full(field, cred) \
776 cap_issubset(CAP_FULL_SET, cred->cap_##field)
777
__is_setuid(struct cred * new,const struct cred * old)778 static inline bool __is_setuid(struct cred *new, const struct cred *old)
779 { return !uid_eq(new->euid, old->uid); }
780
__is_setgid(struct cred * new,const struct cred * old)781 static inline bool __is_setgid(struct cred *new, const struct cred *old)
782 { return !gid_eq(new->egid, old->gid); }
783
784 /*
785 * 1) Audit candidate if current->cap_effective is set
786 *
787 * We do not bother to audit if 3 things are true:
788 * 1) cap_effective has all caps
789 * 2) we became root *OR* are were already root
790 * 3) root is supposed to have all caps (SECURE_NOROOT)
791 * Since this is just a normal root execing a process.
792 *
793 * Number 1 above might fail if you don't have a full bset, but I think
794 * that is interesting information to audit.
795 *
796 * A number of other conditions require logging:
797 * 2) something prevented setuid root getting all caps
798 * 3) non-setuid root gets fcaps
799 * 4) non-setuid root gets ambient
800 */
nonroot_raised_pE(struct cred * new,const struct cred * old,kuid_t root,bool has_fcap)801 static inline bool nonroot_raised_pE(struct cred *new, const struct cred *old,
802 kuid_t root, bool has_fcap)
803 {
804 bool ret = false;
805
806 if ((__cap_grew(effective, ambient, new) &&
807 !(__cap_full(effective, new) &&
808 (__is_eff(root, new) || __is_real(root, new)) &&
809 root_privileged())) ||
810 (root_privileged() &&
811 __is_suid(root, new) &&
812 !__cap_full(effective, new)) ||
813 (!__is_setuid(new, old) &&
814 ((has_fcap &&
815 __cap_gained(permitted, new, old)) ||
816 __cap_gained(ambient, new, old))))
817
818 ret = true;
819
820 return ret;
821 }
822
823 /**
824 * cap_bprm_creds_from_file - Set up the proposed credentials for execve().
825 * @bprm: The execution parameters, including the proposed creds
826 * @file: The file to pull the credentials from
827 *
828 * Set up the proposed credentials for a new execution context being
829 * constructed by execve(). The proposed creds in @bprm->cred is altered,
830 * which won't take effect immediately. Returns 0 if successful, -ve on error.
831 */
cap_bprm_creds_from_file(struct linux_binprm * bprm,struct file * file)832 int cap_bprm_creds_from_file(struct linux_binprm *bprm, struct file *file)
833 {
834 /* Process setpcap binaries and capabilities for uid 0 */
835 const struct cred *old = current_cred();
836 struct cred *new = bprm->cred;
837 bool effective = false, has_fcap = false, is_setid;
838 int ret;
839 kuid_t root_uid;
840
841 if (WARN_ON(!cap_ambient_invariant_ok(old)))
842 return -EPERM;
843
844 ret = get_file_caps(bprm, file, &effective, &has_fcap);
845 if (ret < 0)
846 return ret;
847
848 root_uid = make_kuid(new->user_ns, 0);
849
850 handle_privileged_root(bprm, has_fcap, &effective, root_uid);
851
852 /* if we have fs caps, clear dangerous personality flags */
853 if (__cap_gained(permitted, new, old))
854 bprm->per_clear |= PER_CLEAR_ON_SETID;
855
856 /* Don't let someone trace a set[ug]id/setpcap binary with the revised
857 * credentials unless they have the appropriate permit.
858 *
859 * In addition, if NO_NEW_PRIVS, then ensure we get no new privs.
860 */
861 is_setid = __is_setuid(new, old) || __is_setgid(new, old);
862
863 if ((is_setid || __cap_gained(permitted, new, old)) &&
864 ((bprm->unsafe & ~LSM_UNSAFE_PTRACE) ||
865 !ptracer_capable(current, new->user_ns))) {
866 /* downgrade; they get no more than they had, and maybe less */
867 if (!ns_capable(new->user_ns, CAP_SETUID) ||
868 (bprm->unsafe & LSM_UNSAFE_NO_NEW_PRIVS)) {
869 new->euid = new->uid;
870 new->egid = new->gid;
871 }
872 new->cap_permitted = cap_intersect(new->cap_permitted,
873 old->cap_permitted);
874 }
875
876 new->suid = new->fsuid = new->euid;
877 new->sgid = new->fsgid = new->egid;
878
879 /* File caps or setid cancels ambient. */
880 if (has_fcap || is_setid)
881 cap_clear(new->cap_ambient);
882
883 /*
884 * Now that we've computed pA', update pP' to give:
885 * pP' = (X & fP) | (pI & fI) | pA'
886 */
887 new->cap_permitted = cap_combine(new->cap_permitted, new->cap_ambient);
888
889 /*
890 * Set pE' = (fE ? pP' : pA'). Because pA' is zero if fE is set,
891 * this is the same as pE' = (fE ? pP' : 0) | pA'.
892 */
893 if (effective)
894 new->cap_effective = new->cap_permitted;
895 else
896 new->cap_effective = new->cap_ambient;
897
898 if (WARN_ON(!cap_ambient_invariant_ok(new)))
899 return -EPERM;
900
901 if (nonroot_raised_pE(new, old, root_uid, has_fcap)) {
902 ret = audit_log_bprm_fcaps(bprm, new, old);
903 if (ret < 0)
904 return ret;
905 }
906
907 new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
908
909 if (WARN_ON(!cap_ambient_invariant_ok(new)))
910 return -EPERM;
911
912 /* Check for privilege-elevated exec. */
913 if (is_setid ||
914 (!__is_real(root_uid, new) &&
915 (effective ||
916 __cap_grew(permitted, ambient, new))))
917 bprm->secureexec = 1;
918
919 return 0;
920 }
921
922 /**
923 * cap_inode_setxattr - Determine whether an xattr may be altered
924 * @dentry: The inode/dentry being altered
925 * @name: The name of the xattr to be changed
926 * @value: The value that the xattr will be changed to
927 * @size: The size of value
928 * @flags: The replacement flag
929 *
930 * Determine whether an xattr may be altered or set on an inode, returning 0 if
931 * permission is granted, -ve if denied.
932 *
933 * This is used to make sure security xattrs don't get updated or set by those
934 * who aren't privileged to do so.
935 */
cap_inode_setxattr(struct dentry * dentry,const char * name,const void * value,size_t size,int flags)936 int cap_inode_setxattr(struct dentry *dentry, const char *name,
937 const void *value, size_t size, int flags)
938 {
939 struct user_namespace *user_ns = dentry->d_sb->s_user_ns;
940
941 /* Ignore non-security xattrs */
942 if (strncmp(name, XATTR_SECURITY_PREFIX,
943 XATTR_SECURITY_PREFIX_LEN) != 0)
944 return 0;
945
946 /*
947 * For XATTR_NAME_CAPS the check will be done in
948 * cap_convert_nscap(), called by setxattr()
949 */
950 if (strcmp(name, XATTR_NAME_CAPS) == 0)
951 return 0;
952
953 if (!ns_capable(user_ns, CAP_SYS_ADMIN))
954 return -EPERM;
955 return 0;
956 }
957
958 /**
959 * cap_inode_removexattr - Determine whether an xattr may be removed
960 * @dentry: The inode/dentry being altered
961 * @name: The name of the xattr to be changed
962 *
963 * Determine whether an xattr may be removed from an inode, returning 0 if
964 * permission is granted, -ve if denied.
965 *
966 * This is used to make sure security xattrs don't get removed by those who
967 * aren't privileged to remove them.
968 */
cap_inode_removexattr(struct dentry * dentry,const char * name)969 int cap_inode_removexattr(struct dentry *dentry, const char *name)
970 {
971 struct user_namespace *user_ns = dentry->d_sb->s_user_ns;
972
973 /* Ignore non-security xattrs */
974 if (strncmp(name, XATTR_SECURITY_PREFIX,
975 XATTR_SECURITY_PREFIX_LEN) != 0)
976 return 0;
977
978 if (strcmp(name, XATTR_NAME_CAPS) == 0) {
979 /* security.capability gets namespaced */
980 struct inode *inode = d_backing_inode(dentry);
981 if (!inode)
982 return -EINVAL;
983 if (!capable_wrt_inode_uidgid(inode, CAP_SETFCAP))
984 return -EPERM;
985 return 0;
986 }
987
988 if (!ns_capable(user_ns, CAP_SYS_ADMIN))
989 return -EPERM;
990 return 0;
991 }
992
993 /*
994 * cap_emulate_setxuid() fixes the effective / permitted capabilities of
995 * a process after a call to setuid, setreuid, or setresuid.
996 *
997 * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
998 * {r,e,s}uid != 0, the permitted and effective capabilities are
999 * cleared.
1000 *
1001 * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
1002 * capabilities of the process are cleared.
1003 *
1004 * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
1005 * capabilities are set to the permitted capabilities.
1006 *
1007 * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
1008 * never happen.
1009 *
1010 * -astor
1011 *
1012 * cevans - New behaviour, Oct '99
1013 * A process may, via prctl(), elect to keep its capabilities when it
1014 * calls setuid() and switches away from uid==0. Both permitted and
1015 * effective sets will be retained.
1016 * Without this change, it was impossible for a daemon to drop only some
1017 * of its privilege. The call to setuid(!=0) would drop all privileges!
1018 * Keeping uid 0 is not an option because uid 0 owns too many vital
1019 * files..
1020 * Thanks to Olaf Kirch and Peter Benie for spotting this.
1021 */
cap_emulate_setxuid(struct cred * new,const struct cred * old)1022 static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old)
1023 {
1024 kuid_t root_uid = make_kuid(old->user_ns, 0);
1025
1026 if ((uid_eq(old->uid, root_uid) ||
1027 uid_eq(old->euid, root_uid) ||
1028 uid_eq(old->suid, root_uid)) &&
1029 (!uid_eq(new->uid, root_uid) &&
1030 !uid_eq(new->euid, root_uid) &&
1031 !uid_eq(new->suid, root_uid))) {
1032 if (!issecure(SECURE_KEEP_CAPS)) {
1033 cap_clear(new->cap_permitted);
1034 cap_clear(new->cap_effective);
1035 }
1036
1037 /*
1038 * Pre-ambient programs expect setresuid to nonroot followed
1039 * by exec to drop capabilities. We should make sure that
1040 * this remains the case.
1041 */
1042 cap_clear(new->cap_ambient);
1043 }
1044 if (uid_eq(old->euid, root_uid) && !uid_eq(new->euid, root_uid))
1045 cap_clear(new->cap_effective);
1046 if (!uid_eq(old->euid, root_uid) && uid_eq(new->euid, root_uid))
1047 new->cap_effective = new->cap_permitted;
1048 }
1049
1050 /**
1051 * cap_task_fix_setuid - Fix up the results of setuid() call
1052 * @new: The proposed credentials
1053 * @old: The current task's current credentials
1054 * @flags: Indications of what has changed
1055 *
1056 * Fix up the results of setuid() call before the credential changes are
1057 * actually applied, returning 0 to grant the changes, -ve to deny them.
1058 */
cap_task_fix_setuid(struct cred * new,const struct cred * old,int flags)1059 int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags)
1060 {
1061 switch (flags) {
1062 case LSM_SETID_RE:
1063 case LSM_SETID_ID:
1064 case LSM_SETID_RES:
1065 /* juggle the capabilities to follow [RES]UID changes unless
1066 * otherwise suppressed */
1067 if (!issecure(SECURE_NO_SETUID_FIXUP))
1068 cap_emulate_setxuid(new, old);
1069 break;
1070
1071 case LSM_SETID_FS:
1072 /* juggle the capabilties to follow FSUID changes, unless
1073 * otherwise suppressed
1074 *
1075 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
1076 * if not, we might be a bit too harsh here.
1077 */
1078 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
1079 kuid_t root_uid = make_kuid(old->user_ns, 0);
1080 if (uid_eq(old->fsuid, root_uid) && !uid_eq(new->fsuid, root_uid))
1081 new->cap_effective =
1082 cap_drop_fs_set(new->cap_effective);
1083
1084 if (!uid_eq(old->fsuid, root_uid) && uid_eq(new->fsuid, root_uid))
1085 new->cap_effective =
1086 cap_raise_fs_set(new->cap_effective,
1087 new->cap_permitted);
1088 }
1089 break;
1090
1091 default:
1092 return -EINVAL;
1093 }
1094
1095 return 0;
1096 }
1097
1098 /*
1099 * Rationale: code calling task_setscheduler, task_setioprio, and
1100 * task_setnice, assumes that
1101 * . if capable(cap_sys_nice), then those actions should be allowed
1102 * . if not capable(cap_sys_nice), but acting on your own processes,
1103 * then those actions should be allowed
1104 * This is insufficient now since you can call code without suid, but
1105 * yet with increased caps.
1106 * So we check for increased caps on the target process.
1107 */
cap_safe_nice(struct task_struct * p)1108 static int cap_safe_nice(struct task_struct *p)
1109 {
1110 int is_subset, ret = 0;
1111
1112 rcu_read_lock();
1113 is_subset = cap_issubset(__task_cred(p)->cap_permitted,
1114 current_cred()->cap_permitted);
1115 if (!is_subset && !ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE))
1116 ret = -EPERM;
1117 rcu_read_unlock();
1118
1119 return ret;
1120 }
1121
1122 /**
1123 * cap_task_setscheduler - Detemine if scheduler policy change is permitted
1124 * @p: The task to affect
1125 *
1126 * Detemine if the requested scheduler policy change is permitted for the
1127 * specified task, returning 0 if permission is granted, -ve if denied.
1128 */
cap_task_setscheduler(struct task_struct * p)1129 int cap_task_setscheduler(struct task_struct *p)
1130 {
1131 return cap_safe_nice(p);
1132 }
1133
1134 /**
1135 * cap_task_ioprio - Detemine if I/O priority change is permitted
1136 * @p: The task to affect
1137 * @ioprio: The I/O priority to set
1138 *
1139 * Detemine if the requested I/O priority change is permitted for the specified
1140 * task, returning 0 if permission is granted, -ve if denied.
1141 */
cap_task_setioprio(struct task_struct * p,int ioprio)1142 int cap_task_setioprio(struct task_struct *p, int ioprio)
1143 {
1144 return cap_safe_nice(p);
1145 }
1146
1147 /**
1148 * cap_task_ioprio - Detemine if task priority change is permitted
1149 * @p: The task to affect
1150 * @nice: The nice value to set
1151 *
1152 * Detemine if the requested task priority change is permitted for the
1153 * specified task, returning 0 if permission is granted, -ve if denied.
1154 */
cap_task_setnice(struct task_struct * p,int nice)1155 int cap_task_setnice(struct task_struct *p, int nice)
1156 {
1157 return cap_safe_nice(p);
1158 }
1159
1160 /*
1161 * Implement PR_CAPBSET_DROP. Attempt to remove the specified capability from
1162 * the current task's bounding set. Returns 0 on success, -ve on error.
1163 */
cap_prctl_drop(unsigned long cap)1164 static int cap_prctl_drop(unsigned long cap)
1165 {
1166 struct cred *new;
1167
1168 if (!ns_capable(current_user_ns(), CAP_SETPCAP))
1169 return -EPERM;
1170 if (!cap_valid(cap))
1171 return -EINVAL;
1172
1173 new = prepare_creds();
1174 if (!new)
1175 return -ENOMEM;
1176 cap_lower(new->cap_bset, cap);
1177 return commit_creds(new);
1178 }
1179
1180 /**
1181 * cap_task_prctl - Implement process control functions for this security module
1182 * @option: The process control function requested
1183 * @arg2, @arg3, @arg4, @arg5: The argument data for this function
1184 *
1185 * Allow process control functions (sys_prctl()) to alter capabilities; may
1186 * also deny access to other functions not otherwise implemented here.
1187 *
1188 * Returns 0 or +ve on success, -ENOSYS if this function is not implemented
1189 * here, other -ve on error. If -ENOSYS is returned, sys_prctl() and other LSM
1190 * modules will consider performing the function.
1191 */
cap_task_prctl(int option,unsigned long arg2,unsigned long arg3,unsigned long arg4,unsigned long arg5)1192 int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,
1193 unsigned long arg4, unsigned long arg5)
1194 {
1195 const struct cred *old = current_cred();
1196 struct cred *new;
1197
1198 switch (option) {
1199 case PR_CAPBSET_READ:
1200 if (!cap_valid(arg2))
1201 return -EINVAL;
1202 return !!cap_raised(old->cap_bset, arg2);
1203
1204 case PR_CAPBSET_DROP:
1205 return cap_prctl_drop(arg2);
1206
1207 /*
1208 * The next four prctl's remain to assist with transitioning a
1209 * system from legacy UID=0 based privilege (when filesystem
1210 * capabilities are not in use) to a system using filesystem
1211 * capabilities only - as the POSIX.1e draft intended.
1212 *
1213 * Note:
1214 *
1215 * PR_SET_SECUREBITS =
1216 * issecure_mask(SECURE_KEEP_CAPS_LOCKED)
1217 * | issecure_mask(SECURE_NOROOT)
1218 * | issecure_mask(SECURE_NOROOT_LOCKED)
1219 * | issecure_mask(SECURE_NO_SETUID_FIXUP)
1220 * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
1221 *
1222 * will ensure that the current process and all of its
1223 * children will be locked into a pure
1224 * capability-based-privilege environment.
1225 */
1226 case PR_SET_SECUREBITS:
1227 if ((((old->securebits & SECURE_ALL_LOCKS) >> 1)
1228 & (old->securebits ^ arg2)) /*[1]*/
1229 || ((old->securebits & SECURE_ALL_LOCKS & ~arg2)) /*[2]*/
1230 || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS)) /*[3]*/
1231 || (cap_capable(current_cred(),
1232 current_cred()->user_ns,
1233 CAP_SETPCAP,
1234 CAP_OPT_NONE) != 0) /*[4]*/
1235 /*
1236 * [1] no changing of bits that are locked
1237 * [2] no unlocking of locks
1238 * [3] no setting of unsupported bits
1239 * [4] doing anything requires privilege (go read about
1240 * the "sendmail capabilities bug")
1241 */
1242 )
1243 /* cannot change a locked bit */
1244 return -EPERM;
1245
1246 new = prepare_creds();
1247 if (!new)
1248 return -ENOMEM;
1249 new->securebits = arg2;
1250 return commit_creds(new);
1251
1252 case PR_GET_SECUREBITS:
1253 return old->securebits;
1254
1255 case PR_GET_KEEPCAPS:
1256 return !!issecure(SECURE_KEEP_CAPS);
1257
1258 case PR_SET_KEEPCAPS:
1259 if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */
1260 return -EINVAL;
1261 if (issecure(SECURE_KEEP_CAPS_LOCKED))
1262 return -EPERM;
1263
1264 new = prepare_creds();
1265 if (!new)
1266 return -ENOMEM;
1267 if (arg2)
1268 new->securebits |= issecure_mask(SECURE_KEEP_CAPS);
1269 else
1270 new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
1271 return commit_creds(new);
1272
1273 case PR_CAP_AMBIENT:
1274 if (arg2 == PR_CAP_AMBIENT_CLEAR_ALL) {
1275 if (arg3 | arg4 | arg5)
1276 return -EINVAL;
1277
1278 new = prepare_creds();
1279 if (!new)
1280 return -ENOMEM;
1281 cap_clear(new->cap_ambient);
1282 return commit_creds(new);
1283 }
1284
1285 if (((!cap_valid(arg3)) | arg4 | arg5))
1286 return -EINVAL;
1287
1288 if (arg2 == PR_CAP_AMBIENT_IS_SET) {
1289 return !!cap_raised(current_cred()->cap_ambient, arg3);
1290 } else if (arg2 != PR_CAP_AMBIENT_RAISE &&
1291 arg2 != PR_CAP_AMBIENT_LOWER) {
1292 return -EINVAL;
1293 } else {
1294 if (arg2 == PR_CAP_AMBIENT_RAISE &&
1295 (!cap_raised(current_cred()->cap_permitted, arg3) ||
1296 !cap_raised(current_cred()->cap_inheritable,
1297 arg3) ||
1298 issecure(SECURE_NO_CAP_AMBIENT_RAISE)))
1299 return -EPERM;
1300
1301 new = prepare_creds();
1302 if (!new)
1303 return -ENOMEM;
1304 if (arg2 == PR_CAP_AMBIENT_RAISE)
1305 cap_raise(new->cap_ambient, arg3);
1306 else
1307 cap_lower(new->cap_ambient, arg3);
1308 return commit_creds(new);
1309 }
1310
1311 default:
1312 /* No functionality available - continue with default */
1313 return -ENOSYS;
1314 }
1315 }
1316
1317 /**
1318 * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted
1319 * @mm: The VM space in which the new mapping is to be made
1320 * @pages: The size of the mapping
1321 *
1322 * Determine whether the allocation of a new virtual mapping by the current
1323 * task is permitted, returning 1 if permission is granted, 0 if not.
1324 */
cap_vm_enough_memory(struct mm_struct * mm,long pages)1325 int cap_vm_enough_memory(struct mm_struct *mm, long pages)
1326 {
1327 int cap_sys_admin = 0;
1328
1329 if (cap_capable(current_cred(), &init_user_ns,
1330 CAP_SYS_ADMIN, CAP_OPT_NOAUDIT) == 0)
1331 cap_sys_admin = 1;
1332
1333 return cap_sys_admin;
1334 }
1335
1336 /*
1337 * cap_mmap_addr - check if able to map given addr
1338 * @addr: address attempting to be mapped
1339 *
1340 * If the process is attempting to map memory below dac_mmap_min_addr they need
1341 * CAP_SYS_RAWIO. The other parameters to this function are unused by the
1342 * capability security module. Returns 0 if this mapping should be allowed
1343 * -EPERM if not.
1344 */
cap_mmap_addr(unsigned long addr)1345 int cap_mmap_addr(unsigned long addr)
1346 {
1347 int ret = 0;
1348
1349 if (addr < dac_mmap_min_addr) {
1350 ret = cap_capable(current_cred(), &init_user_ns, CAP_SYS_RAWIO,
1351 CAP_OPT_NONE);
1352 /* set PF_SUPERPRIV if it turns out we allow the low mmap */
1353 if (ret == 0)
1354 current->flags |= PF_SUPERPRIV;
1355 }
1356 return ret;
1357 }
1358
cap_mmap_file(struct file * file,unsigned long reqprot,unsigned long prot,unsigned long flags)1359 int cap_mmap_file(struct file *file, unsigned long reqprot,
1360 unsigned long prot, unsigned long flags)
1361 {
1362 return 0;
1363 }
1364
1365 #ifdef CONFIG_SECURITY
1366
1367 static struct security_hook_list capability_hooks[] __lsm_ro_after_init = {
1368 LSM_HOOK_INIT(capable, cap_capable),
1369 LSM_HOOK_INIT(settime, cap_settime),
1370 LSM_HOOK_INIT(ptrace_access_check, cap_ptrace_access_check),
1371 LSM_HOOK_INIT(ptrace_traceme, cap_ptrace_traceme),
1372 LSM_HOOK_INIT(capget, cap_capget),
1373 LSM_HOOK_INIT(capset, cap_capset),
1374 LSM_HOOK_INIT(bprm_creds_from_file, cap_bprm_creds_from_file),
1375 LSM_HOOK_INIT(inode_need_killpriv, cap_inode_need_killpriv),
1376 LSM_HOOK_INIT(inode_killpriv, cap_inode_killpriv),
1377 LSM_HOOK_INIT(inode_getsecurity, cap_inode_getsecurity),
1378 LSM_HOOK_INIT(mmap_addr, cap_mmap_addr),
1379 LSM_HOOK_INIT(mmap_file, cap_mmap_file),
1380 LSM_HOOK_INIT(task_fix_setuid, cap_task_fix_setuid),
1381 LSM_HOOK_INIT(task_prctl, cap_task_prctl),
1382 LSM_HOOK_INIT(task_setscheduler, cap_task_setscheduler),
1383 LSM_HOOK_INIT(task_setioprio, cap_task_setioprio),
1384 LSM_HOOK_INIT(task_setnice, cap_task_setnice),
1385 LSM_HOOK_INIT(vm_enough_memory, cap_vm_enough_memory),
1386 };
1387
capability_init(void)1388 static int __init capability_init(void)
1389 {
1390 security_add_hooks(capability_hooks, ARRAY_SIZE(capability_hooks),
1391 "capability");
1392 return 0;
1393 }
1394
1395 DEFINE_LSM(capability) = {
1396 .name = "capability",
1397 .order = LSM_ORDER_FIRST,
1398 .init = capability_init,
1399 };
1400
1401 #endif /* CONFIG_SECURITY */
1402