1 /* Common capabilities, needed by capability.o.
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.
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/lsm_hooks.h>
16 #include <linux/file.h>
18 #include <linux/mman.h>
19 #include <linux/pagemap.h>
20 #include <linux/swap.h>
21 #include <linux/skbuff.h>
22 #include <linux/netlink.h>
23 #include <linux/ptrace.h>
24 #include <linux/xattr.h>
25 #include <linux/hugetlb.h>
26 #include <linux/mount.h>
27 #include <linux/sched.h>
28 #include <linux/prctl.h>
29 #include <linux/securebits.h>
30 #include <linux/user_namespace.h>
31 #include <linux/binfmts.h>
32 #include <linux/personality.h>
34 #ifdef CONFIG_ANDROID_PARANOID_NETWORK
35 #include <linux/android_aid.h>
39 * If a non-root user executes a setuid-root binary in
40 * !secure(SECURE_NOROOT) mode, then we raise capabilities.
41 * However if fE is also set, then the intent is for only
42 * the file capabilities to be applied, and the setuid-root
43 * bit is left on either to change the uid (plausible) or
44 * to get full privilege on a kernel without file capabilities
45 * support. So in that case we do not raise capabilities.
47 * Warn if that happens, once per boot.
49 static void warn_setuid_and_fcaps_mixed(const char *fname)
53 printk(KERN_INFO "warning: `%s' has both setuid-root and"
54 " effective capabilities. Therefore not raising all"
55 " capabilities.\n", fname);
61 * cap_capable - Determine whether a task has a particular effective capability
62 * @cred: The credentials to use
63 * @ns: The user namespace in which we need the capability
64 * @cap: The capability to check for
65 * @audit: Whether to write an audit message or not
67 * Determine whether the nominated task has the specified capability amongst
68 * its effective set, returning 0 if it does, -ve if it does not.
70 * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable()
71 * and has_capability() functions. That is, it has the reverse semantics:
72 * cap_has_capability() returns 0 when a task has a capability, but the
73 * kernel's capable() and has_capability() returns 1 for this case.
75 int cap_capable(const struct cred *cred, struct user_namespace *targ_ns,
78 struct user_namespace *ns = targ_ns;
80 #ifdef CONFIG_ANDROID_PARANOID_NETWORK
81 if (cap == CAP_NET_RAW && in_egroup_p(AID_NET_RAW))
83 if (cap == CAP_NET_ADMIN && in_egroup_p(AID_NET_ADMIN))
87 /* See if cred has the capability in the target user namespace
88 * by examining the target user namespace and all of the target
89 * user namespace's parents.
92 /* Do we have the necessary capabilities? */
93 if (ns == cred->user_ns)
94 return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM;
96 /* Have we tried all of the parent namespaces? */
97 if (ns == &init_user_ns)
101 * The owner of the user namespace in the parent of the
102 * user namespace has all caps.
104 if ((ns->parent == cred->user_ns) && uid_eq(ns->owner, cred->euid))
108 * If you have a capability in a parent user ns, then you have
109 * it over all children user namespaces as well.
114 /* We never get here */
118 * cap_settime - Determine whether the current process may set the system clock
119 * @ts: The time to set
120 * @tz: The timezone to set
122 * Determine whether the current process may set the system clock and timezone
123 * information, returning 0 if permission granted, -ve if denied.
125 int cap_settime(const struct timespec *ts, const struct timezone *tz)
127 if (!capable(CAP_SYS_TIME))
133 * cap_ptrace_access_check - Determine whether the current process may access
135 * @child: The process to be accessed
136 * @mode: The mode of attachment.
138 * If we are in the same or an ancestor user_ns and have all the target
139 * task's capabilities, then ptrace access is allowed.
140 * If we have the ptrace capability to the target user_ns, then ptrace
144 * Determine whether a process may access another, returning 0 if permission
145 * granted, -ve if denied.
147 int cap_ptrace_access_check(struct task_struct *child, unsigned int mode)
150 const struct cred *cred, *child_cred;
151 const kernel_cap_t *caller_caps;
154 cred = current_cred();
155 child_cred = __task_cred(child);
156 if (mode & PTRACE_MODE_FSCREDS)
157 caller_caps = &cred->cap_effective;
159 caller_caps = &cred->cap_permitted;
160 if (cred->user_ns == child_cred->user_ns &&
161 cap_issubset(child_cred->cap_permitted, *caller_caps))
163 if (ns_capable(child_cred->user_ns, CAP_SYS_PTRACE))
172 * cap_ptrace_traceme - Determine whether another process may trace the current
173 * @parent: The task proposed to be the tracer
175 * If parent is in the same or an ancestor user_ns and has all current's
176 * capabilities, then ptrace access is allowed.
177 * If parent has the ptrace capability to current's user_ns, then ptrace
181 * Determine whether the nominated task is permitted to trace the current
182 * process, returning 0 if permission is granted, -ve if denied.
184 int cap_ptrace_traceme(struct task_struct *parent)
187 const struct cred *cred, *child_cred;
190 cred = __task_cred(parent);
191 child_cred = current_cred();
192 if (cred->user_ns == child_cred->user_ns &&
193 cap_issubset(child_cred->cap_permitted, cred->cap_permitted))
195 if (has_ns_capability(parent, child_cred->user_ns, CAP_SYS_PTRACE))
204 * cap_capget - Retrieve a task's capability sets
205 * @target: The task from which to retrieve the capability sets
206 * @effective: The place to record the effective set
207 * @inheritable: The place to record the inheritable set
208 * @permitted: The place to record the permitted set
210 * This function retrieves the capabilities of the nominated task and returns
211 * them to the caller.
213 int cap_capget(struct task_struct *target, kernel_cap_t *effective,
214 kernel_cap_t *inheritable, kernel_cap_t *permitted)
216 const struct cred *cred;
218 /* Derived from kernel/capability.c:sys_capget. */
220 cred = __task_cred(target);
221 *effective = cred->cap_effective;
222 *inheritable = cred->cap_inheritable;
223 *permitted = cred->cap_permitted;
229 * Determine whether the inheritable capabilities are limited to the old
230 * permitted set. Returns 1 if they are limited, 0 if they are not.
232 static inline int cap_inh_is_capped(void)
235 /* they are so limited unless the current task has the CAP_SETPCAP
238 if (cap_capable(current_cred(), current_cred()->user_ns,
239 CAP_SETPCAP, SECURITY_CAP_AUDIT) == 0)
245 * cap_capset - Validate and apply proposed changes to current's capabilities
246 * @new: The proposed new credentials; alterations should be made here
247 * @old: The current task's current credentials
248 * @effective: A pointer to the proposed new effective capabilities set
249 * @inheritable: A pointer to the proposed new inheritable capabilities set
250 * @permitted: A pointer to the proposed new permitted capabilities set
252 * This function validates and applies a proposed mass change to the current
253 * process's capability sets. The changes are made to the proposed new
254 * credentials, and assuming no error, will be committed by the caller of LSM.
256 int cap_capset(struct cred *new,
257 const struct cred *old,
258 const kernel_cap_t *effective,
259 const kernel_cap_t *inheritable,
260 const kernel_cap_t *permitted)
262 if (cap_inh_is_capped() &&
263 !cap_issubset(*inheritable,
264 cap_combine(old->cap_inheritable,
265 old->cap_permitted)))
266 /* incapable of using this inheritable set */
269 if (!cap_issubset(*inheritable,
270 cap_combine(old->cap_inheritable,
272 /* no new pI capabilities outside bounding set */
275 /* verify restrictions on target's new Permitted set */
276 if (!cap_issubset(*permitted, old->cap_permitted))
279 /* verify the _new_Effective_ is a subset of the _new_Permitted_ */
280 if (!cap_issubset(*effective, *permitted))
283 new->cap_effective = *effective;
284 new->cap_inheritable = *inheritable;
285 new->cap_permitted = *permitted;
288 * Mask off ambient bits that are no longer both permitted and
291 new->cap_ambient = cap_intersect(new->cap_ambient,
292 cap_intersect(*permitted,
294 if (WARN_ON(!cap_ambient_invariant_ok(new)))
300 * Clear proposed capability sets for execve().
302 static inline void bprm_clear_caps(struct linux_binprm *bprm)
304 cap_clear(bprm->cred->cap_permitted);
305 bprm->cap_effective = false;
309 * cap_inode_need_killpriv - Determine if inode change affects privileges
310 * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV
312 * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV
313 * affects the security markings on that inode, and if it is, should
314 * inode_killpriv() be invoked or the change rejected?
316 * Returns 0 if granted; +ve if granted, but inode_killpriv() is required; and
317 * -ve to deny the change.
319 int cap_inode_need_killpriv(struct dentry *dentry)
321 struct inode *inode = d_backing_inode(dentry);
324 if (!inode->i_op->getxattr)
327 error = inode->i_op->getxattr(dentry, XATTR_NAME_CAPS, NULL, 0);
334 * cap_inode_killpriv - Erase the security markings on an inode
335 * @dentry: The inode/dentry to alter
337 * Erase the privilege-enhancing security markings on an inode.
339 * Returns 0 if successful, -ve on error.
341 int cap_inode_killpriv(struct dentry *dentry)
343 struct inode *inode = d_backing_inode(dentry);
345 if (!inode->i_op->removexattr)
348 return inode->i_op->removexattr(dentry, XATTR_NAME_CAPS);
352 * Calculate the new process capability sets from the capability sets attached
355 static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps,
356 struct linux_binprm *bprm,
360 struct cred *new = bprm->cred;
364 if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE)
367 if (caps->magic_etc & VFS_CAP_REVISION_MASK)
370 CAP_FOR_EACH_U32(i) {
371 __u32 permitted = caps->permitted.cap[i];
372 __u32 inheritable = caps->inheritable.cap[i];
375 * pP' = (X & fP) | (pI & fI)
376 * The addition of pA' is handled later.
378 new->cap_permitted.cap[i] =
379 (new->cap_bset.cap[i] & permitted) |
380 (new->cap_inheritable.cap[i] & inheritable);
382 if (permitted & ~new->cap_permitted.cap[i])
383 /* insufficient to execute correctly */
388 * For legacy apps, with no internal support for recognizing they
389 * do not have enough capabilities, we return an error if they are
390 * missing some "forced" (aka file-permitted) capabilities.
392 return *effective ? ret : 0;
396 * Extract the on-exec-apply capability sets for an executable file.
398 int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps)
400 struct inode *inode = d_backing_inode(dentry);
404 struct vfs_cap_data caps;
406 memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data));
408 if (!inode || !inode->i_op->getxattr)
411 size = inode->i_op->getxattr((struct dentry *)dentry, XATTR_NAME_CAPS, &caps,
413 if (size == -ENODATA || size == -EOPNOTSUPP)
414 /* no data, that's ok */
419 if (size < sizeof(magic_etc))
422 cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps.magic_etc);
424 switch (magic_etc & VFS_CAP_REVISION_MASK) {
425 case VFS_CAP_REVISION_1:
426 if (size != XATTR_CAPS_SZ_1)
428 tocopy = VFS_CAP_U32_1;
430 case VFS_CAP_REVISION_2:
431 if (size != XATTR_CAPS_SZ_2)
433 tocopy = VFS_CAP_U32_2;
439 CAP_FOR_EACH_U32(i) {
442 cpu_caps->permitted.cap[i] = le32_to_cpu(caps.data[i].permitted);
443 cpu_caps->inheritable.cap[i] = le32_to_cpu(caps.data[i].inheritable);
446 cpu_caps->permitted.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
447 cpu_caps->inheritable.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
453 * Attempt to get the on-exec apply capability sets for an executable file from
454 * its xattrs and, if present, apply them to the proposed credentials being
455 * constructed by execve().
457 static int get_file_caps(struct linux_binprm *bprm, bool *effective, bool *has_cap)
460 struct cpu_vfs_cap_data vcaps;
462 bprm_clear_caps(bprm);
464 if (!file_caps_enabled)
467 if (bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)
470 rc = get_vfs_caps_from_disk(bprm->file->f_path.dentry, &vcaps);
473 printk(KERN_NOTICE "%s: get_vfs_caps_from_disk returned %d for %s\n",
474 __func__, rc, bprm->filename);
475 else if (rc == -ENODATA)
480 rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective, has_cap);
482 printk(KERN_NOTICE "%s: cap_from_disk returned %d for %s\n",
483 __func__, rc, bprm->filename);
487 bprm_clear_caps(bprm);
493 * cap_bprm_set_creds - Set up the proposed credentials for execve().
494 * @bprm: The execution parameters, including the proposed creds
496 * Set up the proposed credentials for a new execution context being
497 * constructed by execve(). The proposed creds in @bprm->cred is altered,
498 * which won't take effect immediately. Returns 0 if successful, -ve on error.
500 int cap_bprm_set_creds(struct linux_binprm *bprm)
502 const struct cred *old = current_cred();
503 struct cred *new = bprm->cred;
504 bool effective, has_cap = false, is_setid;
508 if (WARN_ON(!cap_ambient_invariant_ok(old)))
512 ret = get_file_caps(bprm, &effective, &has_cap);
516 root_uid = make_kuid(new->user_ns, 0);
518 if (!issecure(SECURE_NOROOT)) {
520 * If the legacy file capability is set, then don't set privs
521 * for a setuid root binary run by a non-root user. Do set it
522 * for a root user just to cause least surprise to an admin.
524 if (has_cap && !uid_eq(new->uid, root_uid) && uid_eq(new->euid, root_uid)) {
525 warn_setuid_and_fcaps_mixed(bprm->filename);
529 * To support inheritance of root-permissions and suid-root
530 * executables under compatibility mode, we override the
531 * capability sets for the file.
533 * If only the real uid is 0, we do not set the effective bit.
535 if (uid_eq(new->euid, root_uid) || uid_eq(new->uid, root_uid)) {
536 /* pP' = (cap_bset & ~0) | (pI & ~0) */
537 new->cap_permitted = cap_combine(old->cap_bset,
538 old->cap_inheritable);
540 if (uid_eq(new->euid, root_uid))
545 /* if we have fs caps, clear dangerous personality flags */
546 if (!cap_issubset(new->cap_permitted, old->cap_permitted))
547 bprm->per_clear |= PER_CLEAR_ON_SETID;
550 /* Don't let someone trace a set[ug]id/setpcap binary with the revised
551 * credentials unless they have the appropriate permit.
553 * In addition, if NO_NEW_PRIVS, then ensure we get no new privs.
555 is_setid = !uid_eq(new->euid, old->uid) || !gid_eq(new->egid, old->gid);
558 !cap_issubset(new->cap_permitted, old->cap_permitted)) &&
559 bprm->unsafe & ~LSM_UNSAFE_PTRACE_CAP) {
560 /* downgrade; they get no more than they had, and maybe less */
561 if (!capable(CAP_SETUID) ||
562 (bprm->unsafe & LSM_UNSAFE_NO_NEW_PRIVS)) {
563 new->euid = new->uid;
564 new->egid = new->gid;
566 new->cap_permitted = cap_intersect(new->cap_permitted,
570 new->suid = new->fsuid = new->euid;
571 new->sgid = new->fsgid = new->egid;
573 /* File caps or setid cancels ambient. */
574 if (has_cap || is_setid)
575 cap_clear(new->cap_ambient);
578 * Now that we've computed pA', update pP' to give:
579 * pP' = (X & fP) | (pI & fI) | pA'
581 new->cap_permitted = cap_combine(new->cap_permitted, new->cap_ambient);
584 * Set pE' = (fE ? pP' : pA'). Because pA' is zero if fE is set,
585 * this is the same as pE' = (fE ? pP' : 0) | pA'.
588 new->cap_effective = new->cap_permitted;
590 new->cap_effective = new->cap_ambient;
592 if (WARN_ON(!cap_ambient_invariant_ok(new)))
595 bprm->cap_effective = effective;
598 * Audit candidate if current->cap_effective is set
600 * We do not bother to audit if 3 things are true:
601 * 1) cap_effective has all caps
603 * 3) root is supposed to have all caps (SECURE_NOROOT)
604 * Since this is just a normal root execing a process.
606 * Number 1 above might fail if you don't have a full bset, but I think
607 * that is interesting information to audit.
609 if (!cap_issubset(new->cap_effective, new->cap_ambient)) {
610 if (!cap_issubset(CAP_FULL_SET, new->cap_effective) ||
611 !uid_eq(new->euid, root_uid) || !uid_eq(new->uid, root_uid) ||
612 issecure(SECURE_NOROOT)) {
613 ret = audit_log_bprm_fcaps(bprm, new, old);
619 new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
621 if (WARN_ON(!cap_ambient_invariant_ok(new)))
628 * cap_bprm_secureexec - Determine whether a secure execution is required
629 * @bprm: The execution parameters
631 * Determine whether a secure execution is required, return 1 if it is, and 0
634 * The credentials have been committed by this point, and so are no longer
635 * available through @bprm->cred.
637 int cap_bprm_secureexec(struct linux_binprm *bprm)
639 const struct cred *cred = current_cred();
640 kuid_t root_uid = make_kuid(cred->user_ns, 0);
642 if (!uid_eq(cred->uid, root_uid)) {
643 if (bprm->cap_effective)
645 if (!cap_issubset(cred->cap_permitted, cred->cap_ambient))
649 return (!uid_eq(cred->euid, cred->uid) ||
650 !gid_eq(cred->egid, cred->gid));
654 * cap_inode_setxattr - Determine whether an xattr may be altered
655 * @dentry: The inode/dentry being altered
656 * @name: The name of the xattr to be changed
657 * @value: The value that the xattr will be changed to
658 * @size: The size of value
659 * @flags: The replacement flag
661 * Determine whether an xattr may be altered or set on an inode, returning 0 if
662 * permission is granted, -ve if denied.
664 * This is used to make sure security xattrs don't get updated or set by those
665 * who aren't privileged to do so.
667 int cap_inode_setxattr(struct dentry *dentry, const char *name,
668 const void *value, size_t size, int flags)
670 if (!strcmp(name, XATTR_NAME_CAPS)) {
671 if (!capable(CAP_SETFCAP))
676 if (!strncmp(name, XATTR_SECURITY_PREFIX,
677 sizeof(XATTR_SECURITY_PREFIX) - 1) &&
678 !capable(CAP_SYS_ADMIN))
684 * cap_inode_removexattr - Determine whether an xattr may be removed
685 * @dentry: The inode/dentry being altered
686 * @name: The name of the xattr to be changed
688 * Determine whether an xattr may be removed from an inode, returning 0 if
689 * permission is granted, -ve if denied.
691 * This is used to make sure security xattrs don't get removed by those who
692 * aren't privileged to remove them.
694 int cap_inode_removexattr(struct dentry *dentry, const char *name)
696 if (!strcmp(name, XATTR_NAME_CAPS)) {
697 if (!capable(CAP_SETFCAP))
702 if (!strncmp(name, XATTR_SECURITY_PREFIX,
703 sizeof(XATTR_SECURITY_PREFIX) - 1) &&
704 !capable(CAP_SYS_ADMIN))
710 * cap_emulate_setxuid() fixes the effective / permitted capabilities of
711 * a process after a call to setuid, setreuid, or setresuid.
713 * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
714 * {r,e,s}uid != 0, the permitted and effective capabilities are
717 * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
718 * capabilities of the process are cleared.
720 * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
721 * capabilities are set to the permitted capabilities.
723 * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
728 * cevans - New behaviour, Oct '99
729 * A process may, via prctl(), elect to keep its capabilities when it
730 * calls setuid() and switches away from uid==0. Both permitted and
731 * effective sets will be retained.
732 * Without this change, it was impossible for a daemon to drop only some
733 * of its privilege. The call to setuid(!=0) would drop all privileges!
734 * Keeping uid 0 is not an option because uid 0 owns too many vital
736 * Thanks to Olaf Kirch and Peter Benie for spotting this.
738 static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old)
740 kuid_t root_uid = make_kuid(old->user_ns, 0);
742 if ((uid_eq(old->uid, root_uid) ||
743 uid_eq(old->euid, root_uid) ||
744 uid_eq(old->suid, root_uid)) &&
745 (!uid_eq(new->uid, root_uid) &&
746 !uid_eq(new->euid, root_uid) &&
747 !uid_eq(new->suid, root_uid))) {
748 if (!issecure(SECURE_KEEP_CAPS)) {
749 cap_clear(new->cap_permitted);
750 cap_clear(new->cap_effective);
754 * Pre-ambient programs expect setresuid to nonroot followed
755 * by exec to drop capabilities. We should make sure that
756 * this remains the case.
758 cap_clear(new->cap_ambient);
760 if (uid_eq(old->euid, root_uid) && !uid_eq(new->euid, root_uid))
761 cap_clear(new->cap_effective);
762 if (!uid_eq(old->euid, root_uid) && uid_eq(new->euid, root_uid))
763 new->cap_effective = new->cap_permitted;
767 * cap_task_fix_setuid - Fix up the results of setuid() call
768 * @new: The proposed credentials
769 * @old: The current task's current credentials
770 * @flags: Indications of what has changed
772 * Fix up the results of setuid() call before the credential changes are
773 * actually applied, returning 0 to grant the changes, -ve to deny them.
775 int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags)
781 /* juggle the capabilities to follow [RES]UID changes unless
782 * otherwise suppressed */
783 if (!issecure(SECURE_NO_SETUID_FIXUP))
784 cap_emulate_setxuid(new, old);
788 /* juggle the capabilties to follow FSUID changes, unless
789 * otherwise suppressed
791 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
792 * if not, we might be a bit too harsh here.
794 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
795 kuid_t root_uid = make_kuid(old->user_ns, 0);
796 if (uid_eq(old->fsuid, root_uid) && !uid_eq(new->fsuid, root_uid))
798 cap_drop_fs_set(new->cap_effective);
800 if (!uid_eq(old->fsuid, root_uid) && uid_eq(new->fsuid, root_uid))
802 cap_raise_fs_set(new->cap_effective,
815 * Rationale: code calling task_setscheduler, task_setioprio, and
816 * task_setnice, assumes that
817 * . if capable(cap_sys_nice), then those actions should be allowed
818 * . if not capable(cap_sys_nice), but acting on your own processes,
819 * then those actions should be allowed
820 * This is insufficient now since you can call code without suid, but
821 * yet with increased caps.
822 * So we check for increased caps on the target process.
824 static int cap_safe_nice(struct task_struct *p)
826 int is_subset, ret = 0;
829 is_subset = cap_issubset(__task_cred(p)->cap_permitted,
830 current_cred()->cap_permitted);
831 if (!is_subset && !ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE))
839 * cap_task_setscheduler - Detemine if scheduler policy change is permitted
840 * @p: The task to affect
842 * Detemine if the requested scheduler policy change is permitted for the
843 * specified task, returning 0 if permission is granted, -ve if denied.
845 int cap_task_setscheduler(struct task_struct *p)
847 return cap_safe_nice(p);
851 * cap_task_ioprio - Detemine if I/O priority change is permitted
852 * @p: The task to affect
853 * @ioprio: The I/O priority to set
855 * Detemine if the requested I/O priority change is permitted for the specified
856 * task, returning 0 if permission is granted, -ve if denied.
858 int cap_task_setioprio(struct task_struct *p, int ioprio)
860 return cap_safe_nice(p);
864 * cap_task_ioprio - Detemine if task priority change is permitted
865 * @p: The task to affect
866 * @nice: The nice value to set
868 * Detemine if the requested task priority change is permitted for the
869 * specified task, returning 0 if permission is granted, -ve if denied.
871 int cap_task_setnice(struct task_struct *p, int nice)
873 return cap_safe_nice(p);
877 * Implement PR_CAPBSET_DROP. Attempt to remove the specified capability from
878 * the current task's bounding set. Returns 0 on success, -ve on error.
880 static int cap_prctl_drop(unsigned long cap)
884 if (!ns_capable(current_user_ns(), CAP_SETPCAP))
889 new = prepare_creds();
892 cap_lower(new->cap_bset, cap);
893 return commit_creds(new);
897 * cap_task_prctl - Implement process control functions for this security module
898 * @option: The process control function requested
899 * @arg2, @arg3, @arg4, @arg5: The argument data for this function
901 * Allow process control functions (sys_prctl()) to alter capabilities; may
902 * also deny access to other functions not otherwise implemented here.
904 * Returns 0 or +ve on success, -ENOSYS if this function is not implemented
905 * here, other -ve on error. If -ENOSYS is returned, sys_prctl() and other LSM
906 * modules will consider performing the function.
908 int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,
909 unsigned long arg4, unsigned long arg5)
911 const struct cred *old = current_cred();
915 case PR_CAPBSET_READ:
916 if (!cap_valid(arg2))
918 return !!cap_raised(old->cap_bset, arg2);
920 case PR_CAPBSET_DROP:
921 return cap_prctl_drop(arg2);
924 * The next four prctl's remain to assist with transitioning a
925 * system from legacy UID=0 based privilege (when filesystem
926 * capabilities are not in use) to a system using filesystem
927 * capabilities only - as the POSIX.1e draft intended.
931 * PR_SET_SECUREBITS =
932 * issecure_mask(SECURE_KEEP_CAPS_LOCKED)
933 * | issecure_mask(SECURE_NOROOT)
934 * | issecure_mask(SECURE_NOROOT_LOCKED)
935 * | issecure_mask(SECURE_NO_SETUID_FIXUP)
936 * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
938 * will ensure that the current process and all of its
939 * children will be locked into a pure
940 * capability-based-privilege environment.
942 case PR_SET_SECUREBITS:
943 if ((((old->securebits & SECURE_ALL_LOCKS) >> 1)
944 & (old->securebits ^ arg2)) /*[1]*/
945 || ((old->securebits & SECURE_ALL_LOCKS & ~arg2)) /*[2]*/
946 || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS)) /*[3]*/
947 || (cap_capable(current_cred(),
948 current_cred()->user_ns, CAP_SETPCAP,
949 SECURITY_CAP_AUDIT) != 0) /*[4]*/
951 * [1] no changing of bits that are locked
952 * [2] no unlocking of locks
953 * [3] no setting of unsupported bits
954 * [4] doing anything requires privilege (go read about
955 * the "sendmail capabilities bug")
958 /* cannot change a locked bit */
961 new = prepare_creds();
964 new->securebits = arg2;
965 return commit_creds(new);
967 case PR_GET_SECUREBITS:
968 return old->securebits;
970 case PR_GET_KEEPCAPS:
971 return !!issecure(SECURE_KEEP_CAPS);
973 case PR_SET_KEEPCAPS:
974 if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */
976 if (issecure(SECURE_KEEP_CAPS_LOCKED))
979 new = prepare_creds();
983 new->securebits |= issecure_mask(SECURE_KEEP_CAPS);
985 new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
986 return commit_creds(new);
989 if (arg2 == PR_CAP_AMBIENT_CLEAR_ALL) {
990 if (arg3 | arg4 | arg5)
993 new = prepare_creds();
996 cap_clear(new->cap_ambient);
997 return commit_creds(new);
1000 if (((!cap_valid(arg3)) | arg4 | arg5))
1003 if (arg2 == PR_CAP_AMBIENT_IS_SET) {
1004 return !!cap_raised(current_cred()->cap_ambient, arg3);
1005 } else if (arg2 != PR_CAP_AMBIENT_RAISE &&
1006 arg2 != PR_CAP_AMBIENT_LOWER) {
1009 if (arg2 == PR_CAP_AMBIENT_RAISE &&
1010 (!cap_raised(current_cred()->cap_permitted, arg3) ||
1011 !cap_raised(current_cred()->cap_inheritable,
1013 issecure(SECURE_NO_CAP_AMBIENT_RAISE)))
1016 new = prepare_creds();
1019 if (arg2 == PR_CAP_AMBIENT_RAISE)
1020 cap_raise(new->cap_ambient, arg3);
1022 cap_lower(new->cap_ambient, arg3);
1023 return commit_creds(new);
1027 /* No functionality available - continue with default */
1033 * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted
1034 * @mm: The VM space in which the new mapping is to be made
1035 * @pages: The size of the mapping
1037 * Determine whether the allocation of a new virtual mapping by the current
1038 * task is permitted, returning 1 if permission is granted, 0 if not.
1040 int cap_vm_enough_memory(struct mm_struct *mm, long pages)
1042 int cap_sys_admin = 0;
1044 if (cap_capable(current_cred(), &init_user_ns, CAP_SYS_ADMIN,
1045 SECURITY_CAP_NOAUDIT) == 0)
1047 return cap_sys_admin;
1051 * cap_mmap_addr - check if able to map given addr
1052 * @addr: address attempting to be mapped
1054 * If the process is attempting to map memory below dac_mmap_min_addr they need
1055 * CAP_SYS_RAWIO. The other parameters to this function are unused by the
1056 * capability security module. Returns 0 if this mapping should be allowed
1059 int cap_mmap_addr(unsigned long addr)
1063 if (addr < dac_mmap_min_addr) {
1064 ret = cap_capable(current_cred(), &init_user_ns, CAP_SYS_RAWIO,
1065 SECURITY_CAP_AUDIT);
1066 /* set PF_SUPERPRIV if it turns out we allow the low mmap */
1068 current->flags |= PF_SUPERPRIV;
1073 int cap_mmap_file(struct file *file, unsigned long reqprot,
1074 unsigned long prot, unsigned long flags)
1079 #ifdef CONFIG_SECURITY
1081 struct security_hook_list capability_hooks[] = {
1082 LSM_HOOK_INIT(capable, cap_capable),
1083 LSM_HOOK_INIT(settime, cap_settime),
1084 LSM_HOOK_INIT(ptrace_access_check, cap_ptrace_access_check),
1085 LSM_HOOK_INIT(ptrace_traceme, cap_ptrace_traceme),
1086 LSM_HOOK_INIT(capget, cap_capget),
1087 LSM_HOOK_INIT(capset, cap_capset),
1088 LSM_HOOK_INIT(bprm_set_creds, cap_bprm_set_creds),
1089 LSM_HOOK_INIT(bprm_secureexec, cap_bprm_secureexec),
1090 LSM_HOOK_INIT(inode_need_killpriv, cap_inode_need_killpriv),
1091 LSM_HOOK_INIT(inode_killpriv, cap_inode_killpriv),
1092 LSM_HOOK_INIT(mmap_addr, cap_mmap_addr),
1093 LSM_HOOK_INIT(mmap_file, cap_mmap_file),
1094 LSM_HOOK_INIT(task_fix_setuid, cap_task_fix_setuid),
1095 LSM_HOOK_INIT(task_prctl, cap_task_prctl),
1096 LSM_HOOK_INIT(task_setscheduler, cap_task_setscheduler),
1097 LSM_HOOK_INIT(task_setioprio, cap_task_setioprio),
1098 LSM_HOOK_INIT(task_setnice, cap_task_setnice),
1099 LSM_HOOK_INIT(vm_enough_memory, cap_vm_enough_memory),
1102 void __init capability_add_hooks(void)
1104 security_add_hooks(capability_hooks, ARRAY_SIZE(capability_hooks));
1107 #endif /* CONFIG_SECURITY */