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.
11 * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved.
14 #include <linux/capability.h>
15 #include <linux/audit.h>
16 #include <linux/module.h>
17 #include <linux/init.h>
18 #include <linux/kernel.h>
19 #include <linux/lsm_hooks.h>
20 #include <linux/file.h>
22 #include <linux/mman.h>
23 #include <linux/pagemap.h>
24 #include <linux/swap.h>
25 #include <linux/skbuff.h>
26 #include <linux/netlink.h>
27 #include <linux/ptrace.h>
28 #include <linux/xattr.h>
29 #include <linux/hugetlb.h>
30 #include <linux/mount.h>
31 #include <linux/sched.h>
32 #include <linux/prctl.h>
33 #include <linux/securebits.h>
34 #include <linux/user_namespace.h>
35 #include <linux/binfmts.h>
36 #include <linux/personality.h>
38 #ifdef CONFIG_ANDROID_PARANOID_NETWORK
39 #include <linux/android_aid.h>
42 u8 disable_android_paranoid_network;
43 EXPORT_SYMBOL(disable_android_paranoid_network);
45 static int __init disable_android_paranoid_network_feature(char *s)
47 disable_android_paranoid_network = 1;
52 __setup("disable_android_paranoid_network",
53 disable_android_paranoid_network_feature);
56 * If a non-root user executes a setuid-root binary in
57 * !secure(SECURE_NOROOT) mode, then we raise capabilities.
58 * However if fE is also set, then the intent is for only
59 * the file capabilities to be applied, and the setuid-root
60 * bit is left on either to change the uid (plausible) or
61 * to get full privilege on a kernel without file capabilities
62 * support. So in that case we do not raise capabilities.
64 * Warn if that happens, once per boot.
66 static void warn_setuid_and_fcaps_mixed(const char *fname)
70 printk(KERN_INFO "warning: `%s' has both setuid-root and"
71 " effective capabilities. Therefore not raising all"
72 " capabilities.\n", fname);
78 * cap_capable - Determine whether a task has a particular effective capability
79 * @cred: The credentials to use
80 * @ns: The user namespace in which we need the capability
81 * @cap: The capability to check for
82 * @audit: Whether to write an audit message or not
84 * Determine whether the nominated task has the specified capability amongst
85 * its effective set, returning 0 if it does, -ve if it does not.
87 * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable()
88 * and has_capability() functions. That is, it has the reverse semantics:
89 * cap_has_capability() returns 0 when a task has a capability, but the
90 * kernel's capable() and has_capability() returns 1 for this case.
92 int cap_capable(const struct cred *cred, struct user_namespace *targ_ns,
95 struct user_namespace *ns = targ_ns;
97 #ifdef CONFIG_ANDROID_PARANOID_NETWORK
98 if (!disable_android_paranoid_network) {
99 if (cap == CAP_NET_RAW && in_egroup_p(AID_NET_RAW))
101 if (cap == CAP_NET_ADMIN && in_egroup_p(AID_NET_ADMIN))
106 /* See if cred has the capability in the target user namespace
107 * by examining the target user namespace and all of the target
108 * user namespace's parents.
111 /* Do we have the necessary capabilities? */
112 if (ns == cred->user_ns)
113 return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM;
115 /* Have we tried all of the parent namespaces? */
116 if (ns == &init_user_ns)
120 * The owner of the user namespace in the parent of the
121 * user namespace has all caps.
123 if ((ns->parent == cred->user_ns) && uid_eq(ns->owner, cred->euid))
127 * If you have a capability in a parent user ns, then you have
128 * it over all children user namespaces as well.
133 /* We never get here */
137 * cap_settime - Determine whether the current process may set the system clock
138 * @ts: The time to set
139 * @tz: The timezone to set
141 * Determine whether the current process may set the system clock and timezone
142 * information, returning 0 if permission granted, -ve if denied.
144 int cap_settime(const struct timespec *ts, const struct timezone *tz)
146 if (!capable(CAP_SYS_TIME))
152 * cap_ptrace_access_check - Determine whether the current process may access
154 * @child: The process to be accessed
155 * @mode: The mode of attachment.
157 * If we are in the same or an ancestor user_ns and have all the target
158 * task's capabilities, then ptrace access is allowed.
159 * If we have the ptrace capability to the target user_ns, then ptrace
163 * Determine whether a process may access another, returning 0 if permission
164 * granted, -ve if denied.
166 int cap_ptrace_access_check(struct task_struct *child, unsigned int mode)
169 const struct cred *cred, *child_cred;
170 const kernel_cap_t *caller_caps;
173 cred = current_cred();
174 child_cred = __task_cred(child);
175 if (mode & PTRACE_MODE_FSCREDS)
176 caller_caps = &cred->cap_effective;
178 caller_caps = &cred->cap_permitted;
179 if (cred->user_ns == child_cred->user_ns &&
180 cap_issubset(child_cred->cap_permitted, *caller_caps))
182 if (ns_capable(child_cred->user_ns, CAP_SYS_PTRACE))
191 * cap_ptrace_traceme - Determine whether another process may trace the current
192 * @parent: The task proposed to be the tracer
194 * If parent is in the same or an ancestor user_ns and has all current's
195 * capabilities, then ptrace access is allowed.
196 * If parent has the ptrace capability to current's user_ns, then ptrace
200 * Determine whether the nominated task is permitted to trace the current
201 * process, returning 0 if permission is granted, -ve if denied.
203 int cap_ptrace_traceme(struct task_struct *parent)
206 const struct cred *cred, *child_cred;
209 cred = __task_cred(parent);
210 child_cred = current_cred();
211 if (cred->user_ns == child_cred->user_ns &&
212 cap_issubset(child_cred->cap_permitted, cred->cap_permitted))
214 if (has_ns_capability(parent, child_cred->user_ns, CAP_SYS_PTRACE))
223 * cap_capget - Retrieve a task's capability sets
224 * @target: The task from which to retrieve the capability sets
225 * @effective: The place to record the effective set
226 * @inheritable: The place to record the inheritable set
227 * @permitted: The place to record the permitted set
229 * This function retrieves the capabilities of the nominated task and returns
230 * them to the caller.
232 int cap_capget(struct task_struct *target, kernel_cap_t *effective,
233 kernel_cap_t *inheritable, kernel_cap_t *permitted)
235 const struct cred *cred;
237 /* Derived from kernel/capability.c:sys_capget. */
239 cred = __task_cred(target);
240 *effective = cred->cap_effective;
241 *inheritable = cred->cap_inheritable;
242 *permitted = cred->cap_permitted;
248 * Determine whether the inheritable capabilities are limited to the old
249 * permitted set. Returns 1 if they are limited, 0 if they are not.
251 static inline int cap_inh_is_capped(void)
254 /* they are so limited unless the current task has the CAP_SETPCAP
257 if (cap_capable(current_cred(), current_cred()->user_ns,
258 CAP_SETPCAP, SECURITY_CAP_AUDIT) == 0)
264 * cap_capset - Validate and apply proposed changes to current's capabilities
265 * @new: The proposed new credentials; alterations should be made here
266 * @old: The current task's current credentials
267 * @effective: A pointer to the proposed new effective capabilities set
268 * @inheritable: A pointer to the proposed new inheritable capabilities set
269 * @permitted: A pointer to the proposed new permitted capabilities set
271 * This function validates and applies a proposed mass change to the current
272 * process's capability sets. The changes are made to the proposed new
273 * credentials, and assuming no error, will be committed by the caller of LSM.
275 int cap_capset(struct cred *new,
276 const struct cred *old,
277 const kernel_cap_t *effective,
278 const kernel_cap_t *inheritable,
279 const kernel_cap_t *permitted)
281 if (cap_inh_is_capped() &&
282 !cap_issubset(*inheritable,
283 cap_combine(old->cap_inheritable,
284 old->cap_permitted)))
285 /* incapable of using this inheritable set */
288 if (!cap_issubset(*inheritable,
289 cap_combine(old->cap_inheritable,
291 /* no new pI capabilities outside bounding set */
294 /* verify restrictions on target's new Permitted set */
295 if (!cap_issubset(*permitted, old->cap_permitted))
298 /* verify the _new_Effective_ is a subset of the _new_Permitted_ */
299 if (!cap_issubset(*effective, *permitted))
302 new->cap_effective = *effective;
303 new->cap_inheritable = *inheritable;
304 new->cap_permitted = *permitted;
307 * Mask off ambient bits that are no longer both permitted and
310 new->cap_ambient = cap_intersect(new->cap_ambient,
311 cap_intersect(*permitted,
313 if (WARN_ON(!cap_ambient_invariant_ok(new)))
319 * Clear proposed capability sets for execve().
321 static inline void bprm_clear_caps(struct linux_binprm *bprm)
323 cap_clear(bprm->cred->cap_permitted);
324 bprm->cap_effective = false;
328 * cap_inode_need_killpriv - Determine if inode change affects privileges
329 * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV
331 * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV
332 * affects the security markings on that inode, and if it is, should
333 * inode_killpriv() be invoked or the change rejected?
335 * Returns 0 if granted; +ve if granted, but inode_killpriv() is required; and
336 * -ve to deny the change.
338 int cap_inode_need_killpriv(struct dentry *dentry)
340 struct inode *inode = d_backing_inode(dentry);
343 if (!inode->i_op->getxattr)
346 error = inode->i_op->getxattr(dentry, XATTR_NAME_CAPS, NULL, 0);
353 * cap_inode_killpriv - Erase the security markings on an inode
354 * @dentry: The inode/dentry to alter
356 * Erase the privilege-enhancing security markings on an inode.
358 * Returns 0 if successful, -ve on error.
360 int cap_inode_killpriv(struct dentry *dentry)
362 struct inode *inode = d_backing_inode(dentry);
364 if (!inode->i_op->removexattr)
367 return inode->i_op->removexattr(dentry, XATTR_NAME_CAPS);
371 * Calculate the new process capability sets from the capability sets attached
374 static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps,
375 struct linux_binprm *bprm,
379 struct cred *new = bprm->cred;
383 if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE)
386 if (caps->magic_etc & VFS_CAP_REVISION_MASK)
389 CAP_FOR_EACH_U32(i) {
390 __u32 permitted = caps->permitted.cap[i];
391 __u32 inheritable = caps->inheritable.cap[i];
394 * pP' = (X & fP) | (pI & fI)
395 * The addition of pA' is handled later.
397 new->cap_permitted.cap[i] =
398 (new->cap_bset.cap[i] & permitted) |
399 (new->cap_inheritable.cap[i] & inheritable);
401 if (permitted & ~new->cap_permitted.cap[i])
402 /* insufficient to execute correctly */
407 * For legacy apps, with no internal support for recognizing they
408 * do not have enough capabilities, we return an error if they are
409 * missing some "forced" (aka file-permitted) capabilities.
411 return *effective ? ret : 0;
415 * Extract the on-exec-apply capability sets for an executable file.
417 int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps)
419 struct inode *inode = d_backing_inode(dentry);
423 struct vfs_cap_data caps;
425 memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data));
427 if (!inode || !inode->i_op->getxattr)
430 size = inode->i_op->getxattr((struct dentry *)dentry, XATTR_NAME_CAPS, &caps,
432 if (size == -ENODATA || size == -EOPNOTSUPP)
433 /* no data, that's ok */
438 if (size < sizeof(magic_etc))
441 cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps.magic_etc);
443 switch (magic_etc & VFS_CAP_REVISION_MASK) {
444 case VFS_CAP_REVISION_1:
445 if (size != XATTR_CAPS_SZ_1)
447 tocopy = VFS_CAP_U32_1;
449 case VFS_CAP_REVISION_2:
450 if (size != XATTR_CAPS_SZ_2)
452 tocopy = VFS_CAP_U32_2;
458 CAP_FOR_EACH_U32(i) {
461 cpu_caps->permitted.cap[i] = le32_to_cpu(caps.data[i].permitted);
462 cpu_caps->inheritable.cap[i] = le32_to_cpu(caps.data[i].inheritable);
465 cpu_caps->permitted.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
466 cpu_caps->inheritable.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
472 * Attempt to get the on-exec apply capability sets for an executable file from
473 * its xattrs and, if present, apply them to the proposed credentials being
474 * constructed by execve().
476 static int get_file_caps(struct linux_binprm *bprm, bool *effective, bool *has_cap)
479 struct cpu_vfs_cap_data vcaps;
481 bprm_clear_caps(bprm);
483 if (!file_caps_enabled)
486 if (bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)
489 rc = get_vfs_caps_from_disk(bprm->file->f_path.dentry, &vcaps);
492 printk(KERN_NOTICE "%s: get_vfs_caps_from_disk returned %d for %s\n",
493 __func__, rc, bprm->filename);
494 else if (rc == -ENODATA)
499 rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective, has_cap);
501 printk(KERN_NOTICE "%s: cap_from_disk returned %d for %s\n",
502 __func__, rc, bprm->filename);
506 bprm_clear_caps(bprm);
512 * cap_bprm_set_creds - Set up the proposed credentials for execve().
513 * @bprm: The execution parameters, including the proposed creds
515 * Set up the proposed credentials for a new execution context being
516 * constructed by execve(). The proposed creds in @bprm->cred is altered,
517 * which won't take effect immediately. Returns 0 if successful, -ve on error.
519 int cap_bprm_set_creds(struct linux_binprm *bprm)
521 const struct cred *old = current_cred();
522 struct cred *new = bprm->cred;
523 bool effective, has_cap = false, is_setid;
527 if (WARN_ON(!cap_ambient_invariant_ok(old)))
531 ret = get_file_caps(bprm, &effective, &has_cap);
535 root_uid = make_kuid(new->user_ns, 0);
537 if (!issecure(SECURE_NOROOT)) {
539 * If the legacy file capability is set, then don't set privs
540 * for a setuid root binary run by a non-root user. Do set it
541 * for a root user just to cause least surprise to an admin.
543 if (has_cap && !uid_eq(new->uid, root_uid) && uid_eq(new->euid, root_uid)) {
544 warn_setuid_and_fcaps_mixed(bprm->filename);
548 * To support inheritance of root-permissions and suid-root
549 * executables under compatibility mode, we override the
550 * capability sets for the file.
552 * If only the real uid is 0, we do not set the effective bit.
554 if (uid_eq(new->euid, root_uid) || uid_eq(new->uid, root_uid)) {
555 /* pP' = (cap_bset & ~0) | (pI & ~0) */
556 new->cap_permitted = cap_combine(old->cap_bset,
557 old->cap_inheritable);
559 if (uid_eq(new->euid, root_uid))
564 /* if we have fs caps, clear dangerous personality flags */
565 if (!cap_issubset(new->cap_permitted, old->cap_permitted))
566 bprm->per_clear |= PER_CLEAR_ON_SETID;
569 /* Don't let someone trace a set[ug]id/setpcap binary with the revised
570 * credentials unless they have the appropriate permit.
572 * In addition, if NO_NEW_PRIVS, then ensure we get no new privs.
574 is_setid = !uid_eq(new->euid, old->uid) || !gid_eq(new->egid, old->gid);
577 !cap_issubset(new->cap_permitted, old->cap_permitted)) &&
578 bprm->unsafe & ~LSM_UNSAFE_PTRACE_CAP) {
579 /* downgrade; they get no more than they had, and maybe less */
580 if (!capable(CAP_SETUID) ||
581 (bprm->unsafe & LSM_UNSAFE_NO_NEW_PRIVS)) {
582 new->euid = new->uid;
583 new->egid = new->gid;
585 new->cap_permitted = cap_intersect(new->cap_permitted,
589 new->suid = new->fsuid = new->euid;
590 new->sgid = new->fsgid = new->egid;
592 /* File caps or setid cancels ambient. */
593 if (has_cap || is_setid)
594 cap_clear(new->cap_ambient);
597 * Now that we've computed pA', update pP' to give:
598 * pP' = (X & fP) | (pI & fI) | pA'
600 new->cap_permitted = cap_combine(new->cap_permitted, new->cap_ambient);
603 * Set pE' = (fE ? pP' : pA'). Because pA' is zero if fE is set,
604 * this is the same as pE' = (fE ? pP' : 0) | pA'.
607 new->cap_effective = new->cap_permitted;
609 new->cap_effective = new->cap_ambient;
611 if (WARN_ON(!cap_ambient_invariant_ok(new)))
614 bprm->cap_effective = effective;
617 * Audit candidate if current->cap_effective is set
619 * We do not bother to audit if 3 things are true:
620 * 1) cap_effective has all caps
622 * 3) root is supposed to have all caps (SECURE_NOROOT)
623 * Since this is just a normal root execing a process.
625 * Number 1 above might fail if you don't have a full bset, but I think
626 * that is interesting information to audit.
628 if (!cap_issubset(new->cap_effective, new->cap_ambient)) {
629 if (!cap_issubset(CAP_FULL_SET, new->cap_effective) ||
630 !uid_eq(new->euid, root_uid) || !uid_eq(new->uid, root_uid) ||
631 issecure(SECURE_NOROOT)) {
632 ret = audit_log_bprm_fcaps(bprm, new, old);
638 new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
640 if (WARN_ON(!cap_ambient_invariant_ok(new)))
647 * cap_bprm_secureexec - Determine whether a secure execution is required
648 * @bprm: The execution parameters
650 * Determine whether a secure execution is required, return 1 if it is, and 0
653 * The credentials have been committed by this point, and so are no longer
654 * available through @bprm->cred.
656 int cap_bprm_secureexec(struct linux_binprm *bprm)
658 const struct cred *cred = current_cred();
659 kuid_t root_uid = make_kuid(cred->user_ns, 0);
661 if (!uid_eq(cred->uid, root_uid)) {
662 if (bprm->cap_effective)
664 if (!cap_issubset(cred->cap_permitted, cred->cap_ambient))
668 return (!uid_eq(cred->euid, cred->uid) ||
669 !gid_eq(cred->egid, cred->gid));
673 * cap_inode_setxattr - Determine whether an xattr may be altered
674 * @dentry: The inode/dentry being altered
675 * @name: The name of the xattr to be changed
676 * @value: The value that the xattr will be changed to
677 * @size: The size of value
678 * @flags: The replacement flag
680 * Determine whether an xattr may be altered or set on an inode, returning 0 if
681 * permission is granted, -ve if denied.
683 * This is used to make sure security xattrs don't get updated or set by those
684 * who aren't privileged to do so.
686 int cap_inode_setxattr(struct dentry *dentry, const char *name,
687 const void *value, size_t size, int flags)
689 if (!strcmp(name, XATTR_NAME_CAPS)) {
690 if (!capable(CAP_SETFCAP))
695 if (!strncmp(name, XATTR_SECURITY_PREFIX,
696 sizeof(XATTR_SECURITY_PREFIX) - 1) &&
697 !capable(CAP_SYS_ADMIN))
703 * cap_inode_removexattr - Determine whether an xattr may be removed
704 * @dentry: The inode/dentry being altered
705 * @name: The name of the xattr to be changed
707 * Determine whether an xattr may be removed from an inode, returning 0 if
708 * permission is granted, -ve if denied.
710 * This is used to make sure security xattrs don't get removed by those who
711 * aren't privileged to remove them.
713 int cap_inode_removexattr(struct dentry *dentry, const char *name)
715 if (!strcmp(name, XATTR_NAME_CAPS)) {
716 if (!capable(CAP_SETFCAP))
721 if (!strncmp(name, XATTR_SECURITY_PREFIX,
722 sizeof(XATTR_SECURITY_PREFIX) - 1) &&
723 !capable(CAP_SYS_ADMIN))
729 * cap_emulate_setxuid() fixes the effective / permitted capabilities of
730 * a process after a call to setuid, setreuid, or setresuid.
732 * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
733 * {r,e,s}uid != 0, the permitted and effective capabilities are
736 * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
737 * capabilities of the process are cleared.
739 * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
740 * capabilities are set to the permitted capabilities.
742 * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
747 * cevans - New behaviour, Oct '99
748 * A process may, via prctl(), elect to keep its capabilities when it
749 * calls setuid() and switches away from uid==0. Both permitted and
750 * effective sets will be retained.
751 * Without this change, it was impossible for a daemon to drop only some
752 * of its privilege. The call to setuid(!=0) would drop all privileges!
753 * Keeping uid 0 is not an option because uid 0 owns too many vital
755 * Thanks to Olaf Kirch and Peter Benie for spotting this.
757 static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old)
759 kuid_t root_uid = make_kuid(old->user_ns, 0);
761 if ((uid_eq(old->uid, root_uid) ||
762 uid_eq(old->euid, root_uid) ||
763 uid_eq(old->suid, root_uid)) &&
764 (!uid_eq(new->uid, root_uid) &&
765 !uid_eq(new->euid, root_uid) &&
766 !uid_eq(new->suid, root_uid))) {
767 if (!issecure(SECURE_KEEP_CAPS)) {
768 cap_clear(new->cap_permitted);
769 cap_clear(new->cap_effective);
773 * Pre-ambient programs expect setresuid to nonroot followed
774 * by exec to drop capabilities. We should make sure that
775 * this remains the case.
777 cap_clear(new->cap_ambient);
779 if (uid_eq(old->euid, root_uid) && !uid_eq(new->euid, root_uid))
780 cap_clear(new->cap_effective);
781 if (!uid_eq(old->euid, root_uid) && uid_eq(new->euid, root_uid))
782 new->cap_effective = new->cap_permitted;
786 * cap_task_fix_setuid - Fix up the results of setuid() call
787 * @new: The proposed credentials
788 * @old: The current task's current credentials
789 * @flags: Indications of what has changed
791 * Fix up the results of setuid() call before the credential changes are
792 * actually applied, returning 0 to grant the changes, -ve to deny them.
794 int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags)
800 /* juggle the capabilities to follow [RES]UID changes unless
801 * otherwise suppressed */
802 if (!issecure(SECURE_NO_SETUID_FIXUP))
803 cap_emulate_setxuid(new, old);
807 /* juggle the capabilties to follow FSUID changes, unless
808 * otherwise suppressed
810 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
811 * if not, we might be a bit too harsh here.
813 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
814 kuid_t root_uid = make_kuid(old->user_ns, 0);
815 if (uid_eq(old->fsuid, root_uid) && !uid_eq(new->fsuid, root_uid))
817 cap_drop_fs_set(new->cap_effective);
819 if (!uid_eq(old->fsuid, root_uid) && uid_eq(new->fsuid, root_uid))
821 cap_raise_fs_set(new->cap_effective,
834 * Rationale: code calling task_setscheduler, task_setioprio, and
835 * task_setnice, assumes that
836 * . if capable(cap_sys_nice), then those actions should be allowed
837 * . if not capable(cap_sys_nice), but acting on your own processes,
838 * then those actions should be allowed
839 * This is insufficient now since you can call code without suid, but
840 * yet with increased caps.
841 * So we check for increased caps on the target process.
843 static int cap_safe_nice(struct task_struct *p)
845 int is_subset, ret = 0;
848 is_subset = cap_issubset(__task_cred(p)->cap_permitted,
849 current_cred()->cap_permitted);
850 if (!is_subset && !ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE))
858 * cap_task_setscheduler - Detemine if scheduler policy change is permitted
859 * @p: The task to affect
861 * Detemine if the requested scheduler policy change is permitted for the
862 * specified task, returning 0 if permission is granted, -ve if denied.
864 int cap_task_setscheduler(struct task_struct *p)
866 return cap_safe_nice(p);
870 * cap_task_ioprio - Detemine if I/O priority change is permitted
871 * @p: The task to affect
872 * @ioprio: The I/O priority to set
874 * Detemine if the requested I/O priority change is permitted for the specified
875 * task, returning 0 if permission is granted, -ve if denied.
877 int cap_task_setioprio(struct task_struct *p, int ioprio)
879 return cap_safe_nice(p);
883 * cap_task_ioprio - Detemine if task priority change is permitted
884 * @p: The task to affect
885 * @nice: The nice value to set
887 * Detemine if the requested task priority change is permitted for the
888 * specified task, returning 0 if permission is granted, -ve if denied.
890 int cap_task_setnice(struct task_struct *p, int nice)
892 return cap_safe_nice(p);
896 * Implement PR_CAPBSET_DROP. Attempt to remove the specified capability from
897 * the current task's bounding set. Returns 0 on success, -ve on error.
899 static int cap_prctl_drop(unsigned long cap)
903 if (!ns_capable(current_user_ns(), CAP_SETPCAP))
908 new = prepare_creds();
911 cap_lower(new->cap_bset, cap);
912 return commit_creds(new);
916 * cap_task_prctl - Implement process control functions for this security module
917 * @option: The process control function requested
918 * @arg2, @arg3, @arg4, @arg5: The argument data for this function
920 * Allow process control functions (sys_prctl()) to alter capabilities; may
921 * also deny access to other functions not otherwise implemented here.
923 * Returns 0 or +ve on success, -ENOSYS if this function is not implemented
924 * here, other -ve on error. If -ENOSYS is returned, sys_prctl() and other LSM
925 * modules will consider performing the function.
927 int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,
928 unsigned long arg4, unsigned long arg5)
930 const struct cred *old = current_cred();
934 case PR_CAPBSET_READ:
935 if (!cap_valid(arg2))
937 return !!cap_raised(old->cap_bset, arg2);
939 case PR_CAPBSET_DROP:
940 return cap_prctl_drop(arg2);
943 * The next four prctl's remain to assist with transitioning a
944 * system from legacy UID=0 based privilege (when filesystem
945 * capabilities are not in use) to a system using filesystem
946 * capabilities only - as the POSIX.1e draft intended.
950 * PR_SET_SECUREBITS =
951 * issecure_mask(SECURE_KEEP_CAPS_LOCKED)
952 * | issecure_mask(SECURE_NOROOT)
953 * | issecure_mask(SECURE_NOROOT_LOCKED)
954 * | issecure_mask(SECURE_NO_SETUID_FIXUP)
955 * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
957 * will ensure that the current process and all of its
958 * children will be locked into a pure
959 * capability-based-privilege environment.
961 case PR_SET_SECUREBITS:
962 if ((((old->securebits & SECURE_ALL_LOCKS) >> 1)
963 & (old->securebits ^ arg2)) /*[1]*/
964 || ((old->securebits & SECURE_ALL_LOCKS & ~arg2)) /*[2]*/
965 || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS)) /*[3]*/
966 || (cap_capable(current_cred(),
967 current_cred()->user_ns, CAP_SETPCAP,
968 SECURITY_CAP_AUDIT) != 0) /*[4]*/
970 * [1] no changing of bits that are locked
971 * [2] no unlocking of locks
972 * [3] no setting of unsupported bits
973 * [4] doing anything requires privilege (go read about
974 * the "sendmail capabilities bug")
977 /* cannot change a locked bit */
980 new = prepare_creds();
983 new->securebits = arg2;
984 return commit_creds(new);
986 case PR_GET_SECUREBITS:
987 return old->securebits;
989 case PR_GET_KEEPCAPS:
990 return !!issecure(SECURE_KEEP_CAPS);
992 case PR_SET_KEEPCAPS:
993 if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */
995 if (issecure(SECURE_KEEP_CAPS_LOCKED))
998 new = prepare_creds();
1002 new->securebits |= issecure_mask(SECURE_KEEP_CAPS);
1004 new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
1005 return commit_creds(new);
1007 case PR_CAP_AMBIENT:
1008 if (arg2 == PR_CAP_AMBIENT_CLEAR_ALL) {
1009 if (arg3 | arg4 | arg5)
1012 new = prepare_creds();
1015 cap_clear(new->cap_ambient);
1016 return commit_creds(new);
1019 if (((!cap_valid(arg3)) | arg4 | arg5))
1022 if (arg2 == PR_CAP_AMBIENT_IS_SET) {
1023 return !!cap_raised(current_cred()->cap_ambient, arg3);
1024 } else if (arg2 != PR_CAP_AMBIENT_RAISE &&
1025 arg2 != PR_CAP_AMBIENT_LOWER) {
1028 if (arg2 == PR_CAP_AMBIENT_RAISE &&
1029 (!cap_raised(current_cred()->cap_permitted, arg3) ||
1030 !cap_raised(current_cred()->cap_inheritable,
1032 issecure(SECURE_NO_CAP_AMBIENT_RAISE)))
1035 new = prepare_creds();
1038 if (arg2 == PR_CAP_AMBIENT_RAISE)
1039 cap_raise(new->cap_ambient, arg3);
1041 cap_lower(new->cap_ambient, arg3);
1042 return commit_creds(new);
1046 /* No functionality available - continue with default */
1052 * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted
1053 * @mm: The VM space in which the new mapping is to be made
1054 * @pages: The size of the mapping
1056 * Determine whether the allocation of a new virtual mapping by the current
1057 * task is permitted, returning 1 if permission is granted, 0 if not.
1059 int cap_vm_enough_memory(struct mm_struct *mm, long pages)
1061 int cap_sys_admin = 0;
1063 if (cap_capable(current_cred(), &init_user_ns, CAP_SYS_ADMIN,
1064 SECURITY_CAP_NOAUDIT) == 0)
1066 return cap_sys_admin;
1070 * cap_mmap_addr - check if able to map given addr
1071 * @addr: address attempting to be mapped
1073 * If the process is attempting to map memory below dac_mmap_min_addr they need
1074 * CAP_SYS_RAWIO. The other parameters to this function are unused by the
1075 * capability security module. Returns 0 if this mapping should be allowed
1078 int cap_mmap_addr(unsigned long addr)
1082 if (addr < dac_mmap_min_addr) {
1083 ret = cap_capable(current_cred(), &init_user_ns, CAP_SYS_RAWIO,
1084 SECURITY_CAP_AUDIT);
1085 /* set PF_SUPERPRIV if it turns out we allow the low mmap */
1087 current->flags |= PF_SUPERPRIV;
1092 int cap_mmap_file(struct file *file, unsigned long reqprot,
1093 unsigned long prot, unsigned long flags)
1098 #ifdef CONFIG_SECURITY
1100 struct security_hook_list capability_hooks[] = {
1101 LSM_HOOK_INIT(capable, cap_capable),
1102 LSM_HOOK_INIT(settime, cap_settime),
1103 LSM_HOOK_INIT(ptrace_access_check, cap_ptrace_access_check),
1104 LSM_HOOK_INIT(ptrace_traceme, cap_ptrace_traceme),
1105 LSM_HOOK_INIT(capget, cap_capget),
1106 LSM_HOOK_INIT(capset, cap_capset),
1107 LSM_HOOK_INIT(bprm_set_creds, cap_bprm_set_creds),
1108 LSM_HOOK_INIT(bprm_secureexec, cap_bprm_secureexec),
1109 LSM_HOOK_INIT(inode_need_killpriv, cap_inode_need_killpriv),
1110 LSM_HOOK_INIT(inode_killpriv, cap_inode_killpriv),
1111 LSM_HOOK_INIT(mmap_addr, cap_mmap_addr),
1112 LSM_HOOK_INIT(mmap_file, cap_mmap_file),
1113 LSM_HOOK_INIT(task_fix_setuid, cap_task_fix_setuid),
1114 LSM_HOOK_INIT(task_prctl, cap_task_prctl),
1115 LSM_HOOK_INIT(task_setscheduler, cap_task_setscheduler),
1116 LSM_HOOK_INIT(task_setioprio, cap_task_setioprio),
1117 LSM_HOOK_INIT(task_setnice, cap_task_setnice),
1118 LSM_HOOK_INIT(vm_enough_memory, cap_vm_enough_memory),
1121 void __init capability_add_hooks(void)
1123 security_add_hooks(capability_hooks, ARRAY_SIZE(capability_hooks));
1126 #endif /* CONFIG_SECURITY */