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>
35 * If a non-root user executes a setuid-root binary in
36 * !secure(SECURE_NOROOT) mode, then we raise capabilities.
37 * However if fE is also set, then the intent is for only
38 * the file capabilities to be applied, and the setuid-root
39 * bit is left on either to change the uid (plausible) or
40 * to get full privilege on a kernel without file capabilities
41 * support. So in that case we do not raise capabilities.
43 * Warn if that happens, once per boot.
45 static void warn_setuid_and_fcaps_mixed(const char *fname
)
49 printk(KERN_INFO
"warning: `%s' has both setuid-root and"
50 " effective capabilities. Therefore not raising all"
51 " capabilities.\n", fname
);
57 * cap_capable - Determine whether a task has a particular effective capability
58 * @cred: The credentials to use
59 * @ns: The user namespace in which we need the capability
60 * @cap: The capability to check for
61 * @audit: Whether to write an audit message or not
63 * Determine whether the nominated task has the specified capability amongst
64 * its effective set, returning 0 if it does, -ve if it does not.
66 * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable()
67 * and has_capability() functions. That is, it has the reverse semantics:
68 * cap_has_capability() returns 0 when a task has a capability, but the
69 * kernel's capable() and has_capability() returns 1 for this case.
71 int cap_capable(const struct cred
*cred
, struct user_namespace
*targ_ns
,
74 struct user_namespace
*ns
= targ_ns
;
76 /* See if cred has the capability in the target user namespace
77 * by examining the target user namespace and all of the target
78 * user namespace's parents.
81 /* Do we have the necessary capabilities? */
82 if (ns
== cred
->user_ns
)
83 return cap_raised(cred
->cap_effective
, cap
) ? 0 : -EPERM
;
85 /* Have we tried all of the parent namespaces? */
86 if (ns
== &init_user_ns
)
90 * The owner of the user namespace in the parent of the
91 * user namespace has all caps.
93 if ((ns
->parent
== cred
->user_ns
) && uid_eq(ns
->owner
, cred
->euid
))
97 * If you have a capability in a parent user ns, then you have
98 * it over all children user namespaces as well.
103 /* We never get here */
107 * cap_settime - Determine whether the current process may set the system clock
108 * @ts: The time to set
109 * @tz: The timezone to set
111 * Determine whether the current process may set the system clock and timezone
112 * information, returning 0 if permission granted, -ve if denied.
114 int cap_settime(const struct timespec
*ts
, const struct timezone
*tz
)
116 if (!capable(CAP_SYS_TIME
))
122 * cap_ptrace_access_check - Determine whether the current process may access
124 * @child: The process to be accessed
125 * @mode: The mode of attachment.
127 * If we are in the same or an ancestor user_ns and have all the target
128 * task's capabilities, then ptrace access is allowed.
129 * If we have the ptrace capability to the target user_ns, then ptrace
133 * Determine whether a process may access another, returning 0 if permission
134 * granted, -ve if denied.
136 int cap_ptrace_access_check(struct task_struct
*child
, unsigned int mode
)
139 const struct cred
*cred
, *child_cred
;
140 const kernel_cap_t
*caller_caps
;
143 cred
= current_cred();
144 child_cred
= __task_cred(child
);
145 if (mode
& PTRACE_MODE_FSCREDS
)
146 caller_caps
= &cred
->cap_effective
;
148 caller_caps
= &cred
->cap_permitted
;
149 if (cred
->user_ns
== child_cred
->user_ns
&&
150 cap_issubset(child_cred
->cap_permitted
, *caller_caps
))
152 if (ns_capable(child_cred
->user_ns
, CAP_SYS_PTRACE
))
161 * cap_ptrace_traceme - Determine whether another process may trace the current
162 * @parent: The task proposed to be the tracer
164 * If parent is in the same or an ancestor user_ns and has all current's
165 * capabilities, then ptrace access is allowed.
166 * If parent has the ptrace capability to current's user_ns, then ptrace
170 * Determine whether the nominated task is permitted to trace the current
171 * process, returning 0 if permission is granted, -ve if denied.
173 int cap_ptrace_traceme(struct task_struct
*parent
)
176 const struct cred
*cred
, *child_cred
;
179 cred
= __task_cred(parent
);
180 child_cred
= current_cred();
181 if (cred
->user_ns
== child_cred
->user_ns
&&
182 cap_issubset(child_cred
->cap_permitted
, cred
->cap_permitted
))
184 if (has_ns_capability(parent
, child_cred
->user_ns
, CAP_SYS_PTRACE
))
193 * cap_capget - Retrieve a task's capability sets
194 * @target: The task from which to retrieve the capability sets
195 * @effective: The place to record the effective set
196 * @inheritable: The place to record the inheritable set
197 * @permitted: The place to record the permitted set
199 * This function retrieves the capabilities of the nominated task and returns
200 * them to the caller.
202 int cap_capget(struct task_struct
*target
, kernel_cap_t
*effective
,
203 kernel_cap_t
*inheritable
, kernel_cap_t
*permitted
)
205 const struct cred
*cred
;
207 /* Derived from kernel/capability.c:sys_capget. */
209 cred
= __task_cred(target
);
210 *effective
= cred
->cap_effective
;
211 *inheritable
= cred
->cap_inheritable
;
212 *permitted
= cred
->cap_permitted
;
218 * Determine whether the inheritable capabilities are limited to the old
219 * permitted set. Returns 1 if they are limited, 0 if they are not.
221 static inline int cap_inh_is_capped(void)
224 /* they are so limited unless the current task has the CAP_SETPCAP
227 if (cap_capable(current_cred(), current_cred()->user_ns
,
228 CAP_SETPCAP
, SECURITY_CAP_AUDIT
) == 0)
234 * cap_capset - Validate and apply proposed changes to current's capabilities
235 * @new: The proposed new credentials; alterations should be made here
236 * @old: The current task's current credentials
237 * @effective: A pointer to the proposed new effective capabilities set
238 * @inheritable: A pointer to the proposed new inheritable capabilities set
239 * @permitted: A pointer to the proposed new permitted capabilities set
241 * This function validates and applies a proposed mass change to the current
242 * process's capability sets. The changes are made to the proposed new
243 * credentials, and assuming no error, will be committed by the caller of LSM.
245 int cap_capset(struct cred
*new,
246 const struct cred
*old
,
247 const kernel_cap_t
*effective
,
248 const kernel_cap_t
*inheritable
,
249 const kernel_cap_t
*permitted
)
251 if (cap_inh_is_capped() &&
252 !cap_issubset(*inheritable
,
253 cap_combine(old
->cap_inheritable
,
254 old
->cap_permitted
)))
255 /* incapable of using this inheritable set */
258 if (!cap_issubset(*inheritable
,
259 cap_combine(old
->cap_inheritable
,
261 /* no new pI capabilities outside bounding set */
264 /* verify restrictions on target's new Permitted set */
265 if (!cap_issubset(*permitted
, old
->cap_permitted
))
268 /* verify the _new_Effective_ is a subset of the _new_Permitted_ */
269 if (!cap_issubset(*effective
, *permitted
))
272 new->cap_effective
= *effective
;
273 new->cap_inheritable
= *inheritable
;
274 new->cap_permitted
= *permitted
;
277 * Mask off ambient bits that are no longer both permitted and
280 new->cap_ambient
= cap_intersect(new->cap_ambient
,
281 cap_intersect(*permitted
,
283 if (WARN_ON(!cap_ambient_invariant_ok(new)))
289 * Clear proposed capability sets for execve().
291 static inline void bprm_clear_caps(struct linux_binprm
*bprm
)
293 cap_clear(bprm
->cred
->cap_permitted
);
294 bprm
->cap_effective
= false;
298 * cap_inode_need_killpriv - Determine if inode change affects privileges
299 * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV
301 * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV
302 * affects the security markings on that inode, and if it is, should
303 * inode_killpriv() be invoked or the change rejected?
305 * Returns 0 if granted; +ve if granted, but inode_killpriv() is required; and
306 * -ve to deny the change.
308 int cap_inode_need_killpriv(struct dentry
*dentry
)
310 struct inode
*inode
= d_backing_inode(dentry
);
313 if (!inode
->i_op
->getxattr
)
316 error
= inode
->i_op
->getxattr(dentry
, XATTR_NAME_CAPS
, NULL
, 0);
323 * cap_inode_killpriv - Erase the security markings on an inode
324 * @dentry: The inode/dentry to alter
326 * Erase the privilege-enhancing security markings on an inode.
328 * Returns 0 if successful, -ve on error.
330 int cap_inode_killpriv(struct dentry
*dentry
)
332 struct inode
*inode
= d_backing_inode(dentry
);
334 if (!inode
->i_op
->removexattr
)
337 return inode
->i_op
->removexattr(dentry
, XATTR_NAME_CAPS
);
341 * Calculate the new process capability sets from the capability sets attached
344 static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data
*caps
,
345 struct linux_binprm
*bprm
,
349 struct cred
*new = bprm
->cred
;
353 if (caps
->magic_etc
& VFS_CAP_FLAGS_EFFECTIVE
)
356 if (caps
->magic_etc
& VFS_CAP_REVISION_MASK
)
359 CAP_FOR_EACH_U32(i
) {
360 __u32 permitted
= caps
->permitted
.cap
[i
];
361 __u32 inheritable
= caps
->inheritable
.cap
[i
];
364 * pP' = (X & fP) | (pI & fI)
365 * The addition of pA' is handled later.
367 new->cap_permitted
.cap
[i
] =
368 (new->cap_bset
.cap
[i
] & permitted
) |
369 (new->cap_inheritable
.cap
[i
] & inheritable
);
371 if (permitted
& ~new->cap_permitted
.cap
[i
])
372 /* insufficient to execute correctly */
377 * For legacy apps, with no internal support for recognizing they
378 * do not have enough capabilities, we return an error if they are
379 * missing some "forced" (aka file-permitted) capabilities.
381 return *effective
? ret
: 0;
385 * Extract the on-exec-apply capability sets for an executable file.
387 int get_vfs_caps_from_disk(const struct dentry
*dentry
, struct cpu_vfs_cap_data
*cpu_caps
)
389 struct inode
*inode
= d_backing_inode(dentry
);
393 struct vfs_cap_data caps
;
395 memset(cpu_caps
, 0, sizeof(struct cpu_vfs_cap_data
));
397 if (!inode
|| !inode
->i_op
->getxattr
)
400 size
= inode
->i_op
->getxattr((struct dentry
*)dentry
, XATTR_NAME_CAPS
, &caps
,
402 if (size
== -ENODATA
|| size
== -EOPNOTSUPP
)
403 /* no data, that's ok */
408 if (size
< sizeof(magic_etc
))
411 cpu_caps
->magic_etc
= magic_etc
= le32_to_cpu(caps
.magic_etc
);
413 switch (magic_etc
& VFS_CAP_REVISION_MASK
) {
414 case VFS_CAP_REVISION_1
:
415 if (size
!= XATTR_CAPS_SZ_1
)
417 tocopy
= VFS_CAP_U32_1
;
419 case VFS_CAP_REVISION_2
:
420 if (size
!= XATTR_CAPS_SZ_2
)
422 tocopy
= VFS_CAP_U32_2
;
428 CAP_FOR_EACH_U32(i
) {
431 cpu_caps
->permitted
.cap
[i
] = le32_to_cpu(caps
.data
[i
].permitted
);
432 cpu_caps
->inheritable
.cap
[i
] = le32_to_cpu(caps
.data
[i
].inheritable
);
435 cpu_caps
->permitted
.cap
[CAP_LAST_U32
] &= CAP_LAST_U32_VALID_MASK
;
436 cpu_caps
->inheritable
.cap
[CAP_LAST_U32
] &= CAP_LAST_U32_VALID_MASK
;
442 * Attempt to get the on-exec apply capability sets for an executable file from
443 * its xattrs and, if present, apply them to the proposed credentials being
444 * constructed by execve().
446 static int get_file_caps(struct linux_binprm
*bprm
, bool *effective
, bool *has_cap
)
449 struct cpu_vfs_cap_data vcaps
;
451 bprm_clear_caps(bprm
);
453 if (!file_caps_enabled
)
456 if (path_nosuid(&bprm
->file
->f_path
))
458 if (!current_in_userns(bprm
->file
->f_path
.mnt
->mnt_sb
->s_user_ns
))
461 rc
= get_vfs_caps_from_disk(bprm
->file
->f_path
.dentry
, &vcaps
);
464 printk(KERN_NOTICE
"%s: get_vfs_caps_from_disk returned %d for %s\n",
465 __func__
, rc
, bprm
->filename
);
466 else if (rc
== -ENODATA
)
471 rc
= bprm_caps_from_vfs_caps(&vcaps
, bprm
, effective
, has_cap
);
473 printk(KERN_NOTICE
"%s: cap_from_disk returned %d for %s\n",
474 __func__
, rc
, bprm
->filename
);
478 bprm_clear_caps(bprm
);
484 * cap_bprm_set_creds - Set up the proposed credentials for execve().
485 * @bprm: The execution parameters, including the proposed creds
487 * Set up the proposed credentials for a new execution context being
488 * constructed by execve(). The proposed creds in @bprm->cred is altered,
489 * which won't take effect immediately. Returns 0 if successful, -ve on error.
491 int cap_bprm_set_creds(struct linux_binprm
*bprm
)
493 const struct cred
*old
= current_cred();
494 struct cred
*new = bprm
->cred
;
495 bool effective
, has_cap
= false, is_setid
;
499 if (WARN_ON(!cap_ambient_invariant_ok(old
)))
503 ret
= get_file_caps(bprm
, &effective
, &has_cap
);
507 root_uid
= make_kuid(new->user_ns
, 0);
509 if (!issecure(SECURE_NOROOT
)) {
511 * If the legacy file capability is set, then don't set privs
512 * for a setuid root binary run by a non-root user. Do set it
513 * for a root user just to cause least surprise to an admin.
515 if (has_cap
&& !uid_eq(new->uid
, root_uid
) && uid_eq(new->euid
, root_uid
)) {
516 warn_setuid_and_fcaps_mixed(bprm
->filename
);
520 * To support inheritance of root-permissions and suid-root
521 * executables under compatibility mode, we override the
522 * capability sets for the file.
524 * If only the real uid is 0, we do not set the effective bit.
526 if (uid_eq(new->euid
, root_uid
) || uid_eq(new->uid
, root_uid
)) {
527 /* pP' = (cap_bset & ~0) | (pI & ~0) */
528 new->cap_permitted
= cap_combine(old
->cap_bset
,
529 old
->cap_inheritable
);
531 if (uid_eq(new->euid
, root_uid
))
536 /* if we have fs caps, clear dangerous personality flags */
537 if (!cap_issubset(new->cap_permitted
, old
->cap_permitted
))
538 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
541 /* Don't let someone trace a set[ug]id/setpcap binary with the revised
542 * credentials unless they have the appropriate permit.
544 * In addition, if NO_NEW_PRIVS, then ensure we get no new privs.
546 is_setid
= !uid_eq(new->euid
, old
->uid
) || !gid_eq(new->egid
, old
->gid
);
549 !cap_issubset(new->cap_permitted
, old
->cap_permitted
)) &&
550 bprm
->unsafe
& ~LSM_UNSAFE_PTRACE_CAP
) {
551 /* downgrade; they get no more than they had, and maybe less */
552 if (!capable(CAP_SETUID
) ||
553 (bprm
->unsafe
& LSM_UNSAFE_NO_NEW_PRIVS
)) {
554 new->euid
= new->uid
;
555 new->egid
= new->gid
;
557 new->cap_permitted
= cap_intersect(new->cap_permitted
,
561 new->suid
= new->fsuid
= new->euid
;
562 new->sgid
= new->fsgid
= new->egid
;
564 /* File caps or setid cancels ambient. */
565 if (has_cap
|| is_setid
)
566 cap_clear(new->cap_ambient
);
569 * Now that we've computed pA', update pP' to give:
570 * pP' = (X & fP) | (pI & fI) | pA'
572 new->cap_permitted
= cap_combine(new->cap_permitted
, new->cap_ambient
);
575 * Set pE' = (fE ? pP' : pA'). Because pA' is zero if fE is set,
576 * this is the same as pE' = (fE ? pP' : 0) | pA'.
579 new->cap_effective
= new->cap_permitted
;
581 new->cap_effective
= new->cap_ambient
;
583 if (WARN_ON(!cap_ambient_invariant_ok(new)))
586 bprm
->cap_effective
= effective
;
589 * Audit candidate if current->cap_effective is set
591 * We do not bother to audit if 3 things are true:
592 * 1) cap_effective has all caps
594 * 3) root is supposed to have all caps (SECURE_NOROOT)
595 * Since this is just a normal root execing a process.
597 * Number 1 above might fail if you don't have a full bset, but I think
598 * that is interesting information to audit.
600 if (!cap_issubset(new->cap_effective
, new->cap_ambient
)) {
601 if (!cap_issubset(CAP_FULL_SET
, new->cap_effective
) ||
602 !uid_eq(new->euid
, root_uid
) || !uid_eq(new->uid
, root_uid
) ||
603 issecure(SECURE_NOROOT
)) {
604 ret
= audit_log_bprm_fcaps(bprm
, new, old
);
610 new->securebits
&= ~issecure_mask(SECURE_KEEP_CAPS
);
612 if (WARN_ON(!cap_ambient_invariant_ok(new)))
619 * cap_bprm_secureexec - Determine whether a secure execution is required
620 * @bprm: The execution parameters
622 * Determine whether a secure execution is required, return 1 if it is, and 0
625 * The credentials have been committed by this point, and so are no longer
626 * available through @bprm->cred.
628 int cap_bprm_secureexec(struct linux_binprm
*bprm
)
630 const struct cred
*cred
= current_cred();
631 kuid_t root_uid
= make_kuid(cred
->user_ns
, 0);
633 if (!uid_eq(cred
->uid
, root_uid
)) {
634 if (bprm
->cap_effective
)
636 if (!cap_issubset(cred
->cap_permitted
, cred
->cap_ambient
))
640 return (!uid_eq(cred
->euid
, cred
->uid
) ||
641 !gid_eq(cred
->egid
, cred
->gid
));
645 * cap_inode_setxattr - Determine whether an xattr may be altered
646 * @dentry: The inode/dentry being altered
647 * @name: The name of the xattr to be changed
648 * @value: The value that the xattr will be changed to
649 * @size: The size of value
650 * @flags: The replacement flag
652 * Determine whether an xattr may be altered or set on an inode, returning 0 if
653 * permission is granted, -ve if denied.
655 * This is used to make sure security xattrs don't get updated or set by those
656 * who aren't privileged to do so.
658 int cap_inode_setxattr(struct dentry
*dentry
, const char *name
,
659 const void *value
, size_t size
, int flags
)
661 struct user_namespace
*user_ns
= dentry
->d_sb
->s_user_ns
;
663 if (!strcmp(name
, XATTR_NAME_CAPS
)) {
664 if (!ns_capable(user_ns
, CAP_SETFCAP
))
669 if (!strncmp(name
, XATTR_SECURITY_PREFIX
,
670 sizeof(XATTR_SECURITY_PREFIX
) - 1) &&
671 !ns_capable(user_ns
, CAP_SYS_ADMIN
))
677 * cap_inode_removexattr - Determine whether an xattr may be removed
678 * @dentry: The inode/dentry being altered
679 * @name: The name of the xattr to be changed
681 * Determine whether an xattr may be removed from an inode, returning 0 if
682 * permission is granted, -ve if denied.
684 * This is used to make sure security xattrs don't get removed by those who
685 * aren't privileged to remove them.
687 int cap_inode_removexattr(struct dentry
*dentry
, const char *name
)
689 struct user_namespace
*user_ns
= dentry
->d_sb
->s_user_ns
;
691 if (!strcmp(name
, XATTR_NAME_CAPS
)) {
692 if (!ns_capable(user_ns
, CAP_SETFCAP
))
697 if (!strncmp(name
, XATTR_SECURITY_PREFIX
,
698 sizeof(XATTR_SECURITY_PREFIX
) - 1) &&
699 !ns_capable(user_ns
, CAP_SYS_ADMIN
))
705 * cap_emulate_setxuid() fixes the effective / permitted capabilities of
706 * a process after a call to setuid, setreuid, or setresuid.
708 * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
709 * {r,e,s}uid != 0, the permitted and effective capabilities are
712 * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
713 * capabilities of the process are cleared.
715 * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
716 * capabilities are set to the permitted capabilities.
718 * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
723 * cevans - New behaviour, Oct '99
724 * A process may, via prctl(), elect to keep its capabilities when it
725 * calls setuid() and switches away from uid==0. Both permitted and
726 * effective sets will be retained.
727 * Without this change, it was impossible for a daemon to drop only some
728 * of its privilege. The call to setuid(!=0) would drop all privileges!
729 * Keeping uid 0 is not an option because uid 0 owns too many vital
731 * Thanks to Olaf Kirch and Peter Benie for spotting this.
733 static inline void cap_emulate_setxuid(struct cred
*new, const struct cred
*old
)
735 kuid_t root_uid
= make_kuid(old
->user_ns
, 0);
737 if ((uid_eq(old
->uid
, root_uid
) ||
738 uid_eq(old
->euid
, root_uid
) ||
739 uid_eq(old
->suid
, root_uid
)) &&
740 (!uid_eq(new->uid
, root_uid
) &&
741 !uid_eq(new->euid
, root_uid
) &&
742 !uid_eq(new->suid
, root_uid
))) {
743 if (!issecure(SECURE_KEEP_CAPS
)) {
744 cap_clear(new->cap_permitted
);
745 cap_clear(new->cap_effective
);
749 * Pre-ambient programs expect setresuid to nonroot followed
750 * by exec to drop capabilities. We should make sure that
751 * this remains the case.
753 cap_clear(new->cap_ambient
);
755 if (uid_eq(old
->euid
, root_uid
) && !uid_eq(new->euid
, root_uid
))
756 cap_clear(new->cap_effective
);
757 if (!uid_eq(old
->euid
, root_uid
) && uid_eq(new->euid
, root_uid
))
758 new->cap_effective
= new->cap_permitted
;
762 * cap_task_fix_setuid - Fix up the results of setuid() call
763 * @new: The proposed credentials
764 * @old: The current task's current credentials
765 * @flags: Indications of what has changed
767 * Fix up the results of setuid() call before the credential changes are
768 * actually applied, returning 0 to grant the changes, -ve to deny them.
770 int cap_task_fix_setuid(struct cred
*new, const struct cred
*old
, int flags
)
776 /* juggle the capabilities to follow [RES]UID changes unless
777 * otherwise suppressed */
778 if (!issecure(SECURE_NO_SETUID_FIXUP
))
779 cap_emulate_setxuid(new, old
);
783 /* juggle the capabilties to follow FSUID changes, unless
784 * otherwise suppressed
786 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
787 * if not, we might be a bit too harsh here.
789 if (!issecure(SECURE_NO_SETUID_FIXUP
)) {
790 kuid_t root_uid
= make_kuid(old
->user_ns
, 0);
791 if (uid_eq(old
->fsuid
, root_uid
) && !uid_eq(new->fsuid
, root_uid
))
793 cap_drop_fs_set(new->cap_effective
);
795 if (!uid_eq(old
->fsuid
, root_uid
) && uid_eq(new->fsuid
, root_uid
))
797 cap_raise_fs_set(new->cap_effective
,
810 * Rationale: code calling task_setscheduler, task_setioprio, and
811 * task_setnice, assumes that
812 * . if capable(cap_sys_nice), then those actions should be allowed
813 * . if not capable(cap_sys_nice), but acting on your own processes,
814 * then those actions should be allowed
815 * This is insufficient now since you can call code without suid, but
816 * yet with increased caps.
817 * So we check for increased caps on the target process.
819 static int cap_safe_nice(struct task_struct
*p
)
821 int is_subset
, ret
= 0;
824 is_subset
= cap_issubset(__task_cred(p
)->cap_permitted
,
825 current_cred()->cap_permitted
);
826 if (!is_subset
&& !ns_capable(__task_cred(p
)->user_ns
, CAP_SYS_NICE
))
834 * cap_task_setscheduler - Detemine if scheduler policy change is permitted
835 * @p: The task to affect
837 * Detemine if the requested scheduler policy change is permitted for the
838 * specified task, returning 0 if permission is granted, -ve if denied.
840 int cap_task_setscheduler(struct task_struct
*p
)
842 return cap_safe_nice(p
);
846 * cap_task_ioprio - Detemine if I/O priority change is permitted
847 * @p: The task to affect
848 * @ioprio: The I/O priority to set
850 * Detemine if the requested I/O priority change is permitted for the specified
851 * task, returning 0 if permission is granted, -ve if denied.
853 int cap_task_setioprio(struct task_struct
*p
, int ioprio
)
855 return cap_safe_nice(p
);
859 * cap_task_ioprio - Detemine if task priority change is permitted
860 * @p: The task to affect
861 * @nice: The nice value to set
863 * Detemine if the requested task priority change is permitted for the
864 * specified task, returning 0 if permission is granted, -ve if denied.
866 int cap_task_setnice(struct task_struct
*p
, int nice
)
868 return cap_safe_nice(p
);
872 * Implement PR_CAPBSET_DROP. Attempt to remove the specified capability from
873 * the current task's bounding set. Returns 0 on success, -ve on error.
875 static int cap_prctl_drop(unsigned long cap
)
879 if (!ns_capable(current_user_ns(), CAP_SETPCAP
))
884 new = prepare_creds();
887 cap_lower(new->cap_bset
, cap
);
888 return commit_creds(new);
892 * cap_task_prctl - Implement process control functions for this security module
893 * @option: The process control function requested
894 * @arg2, @arg3, @arg4, @arg5: The argument data for this function
896 * Allow process control functions (sys_prctl()) to alter capabilities; may
897 * also deny access to other functions not otherwise implemented here.
899 * Returns 0 or +ve on success, -ENOSYS if this function is not implemented
900 * here, other -ve on error. If -ENOSYS is returned, sys_prctl() and other LSM
901 * modules will consider performing the function.
903 int cap_task_prctl(int option
, unsigned long arg2
, unsigned long arg3
,
904 unsigned long arg4
, unsigned long arg5
)
906 const struct cred
*old
= current_cred();
910 case PR_CAPBSET_READ
:
911 if (!cap_valid(arg2
))
913 return !!cap_raised(old
->cap_bset
, arg2
);
915 case PR_CAPBSET_DROP
:
916 return cap_prctl_drop(arg2
);
919 * The next four prctl's remain to assist with transitioning a
920 * system from legacy UID=0 based privilege (when filesystem
921 * capabilities are not in use) to a system using filesystem
922 * capabilities only - as the POSIX.1e draft intended.
926 * PR_SET_SECUREBITS =
927 * issecure_mask(SECURE_KEEP_CAPS_LOCKED)
928 * | issecure_mask(SECURE_NOROOT)
929 * | issecure_mask(SECURE_NOROOT_LOCKED)
930 * | issecure_mask(SECURE_NO_SETUID_FIXUP)
931 * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
933 * will ensure that the current process and all of its
934 * children will be locked into a pure
935 * capability-based-privilege environment.
937 case PR_SET_SECUREBITS
:
938 if ((((old
->securebits
& SECURE_ALL_LOCKS
) >> 1)
939 & (old
->securebits
^ arg2
)) /*[1]*/
940 || ((old
->securebits
& SECURE_ALL_LOCKS
& ~arg2
)) /*[2]*/
941 || (arg2
& ~(SECURE_ALL_LOCKS
| SECURE_ALL_BITS
)) /*[3]*/
942 || (cap_capable(current_cred(),
943 current_cred()->user_ns
, CAP_SETPCAP
,
944 SECURITY_CAP_AUDIT
) != 0) /*[4]*/
946 * [1] no changing of bits that are locked
947 * [2] no unlocking of locks
948 * [3] no setting of unsupported bits
949 * [4] doing anything requires privilege (go read about
950 * the "sendmail capabilities bug")
953 /* cannot change a locked bit */
956 new = prepare_creds();
959 new->securebits
= arg2
;
960 return commit_creds(new);
962 case PR_GET_SECUREBITS
:
963 return old
->securebits
;
965 case PR_GET_KEEPCAPS
:
966 return !!issecure(SECURE_KEEP_CAPS
);
968 case PR_SET_KEEPCAPS
:
969 if (arg2
> 1) /* Note, we rely on arg2 being unsigned here */
971 if (issecure(SECURE_KEEP_CAPS_LOCKED
))
974 new = prepare_creds();
978 new->securebits
|= issecure_mask(SECURE_KEEP_CAPS
);
980 new->securebits
&= ~issecure_mask(SECURE_KEEP_CAPS
);
981 return commit_creds(new);
984 if (arg2
== PR_CAP_AMBIENT_CLEAR_ALL
) {
985 if (arg3
| arg4
| arg5
)
988 new = prepare_creds();
991 cap_clear(new->cap_ambient
);
992 return commit_creds(new);
995 if (((!cap_valid(arg3
)) | arg4
| arg5
))
998 if (arg2
== PR_CAP_AMBIENT_IS_SET
) {
999 return !!cap_raised(current_cred()->cap_ambient
, arg3
);
1000 } else if (arg2
!= PR_CAP_AMBIENT_RAISE
&&
1001 arg2
!= PR_CAP_AMBIENT_LOWER
) {
1004 if (arg2
== PR_CAP_AMBIENT_RAISE
&&
1005 (!cap_raised(current_cred()->cap_permitted
, arg3
) ||
1006 !cap_raised(current_cred()->cap_inheritable
,
1008 issecure(SECURE_NO_CAP_AMBIENT_RAISE
)))
1011 new = prepare_creds();
1014 if (arg2
== PR_CAP_AMBIENT_RAISE
)
1015 cap_raise(new->cap_ambient
, arg3
);
1017 cap_lower(new->cap_ambient
, arg3
);
1018 return commit_creds(new);
1022 /* No functionality available - continue with default */
1028 * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted
1029 * @mm: The VM space in which the new mapping is to be made
1030 * @pages: The size of the mapping
1032 * Determine whether the allocation of a new virtual mapping by the current
1033 * task is permitted, returning 1 if permission is granted, 0 if not.
1035 int cap_vm_enough_memory(struct mm_struct
*mm
, long pages
)
1037 int cap_sys_admin
= 0;
1039 if (cap_capable(current_cred(), &init_user_ns
, CAP_SYS_ADMIN
,
1040 SECURITY_CAP_NOAUDIT
) == 0)
1042 return cap_sys_admin
;
1046 * cap_mmap_addr - check if able to map given addr
1047 * @addr: address attempting to be mapped
1049 * If the process is attempting to map memory below dac_mmap_min_addr they need
1050 * CAP_SYS_RAWIO. The other parameters to this function are unused by the
1051 * capability security module. Returns 0 if this mapping should be allowed
1054 int cap_mmap_addr(unsigned long addr
)
1058 if (addr
< dac_mmap_min_addr
) {
1059 ret
= cap_capable(current_cred(), &init_user_ns
, CAP_SYS_RAWIO
,
1060 SECURITY_CAP_AUDIT
);
1061 /* set PF_SUPERPRIV if it turns out we allow the low mmap */
1063 current
->flags
|= PF_SUPERPRIV
;
1067 EXPORT_SYMBOL(cap_mmap_addr
);
1069 int cap_mmap_file(struct file
*file
, unsigned long reqprot
,
1070 unsigned long prot
, unsigned long flags
)
1074 EXPORT_SYMBOL(cap_mmap_file
);
1076 #ifdef CONFIG_SECURITY
1078 struct security_hook_list capability_hooks
[] = {
1079 LSM_HOOK_INIT(capable
, cap_capable
),
1080 LSM_HOOK_INIT(settime
, cap_settime
),
1081 LSM_HOOK_INIT(ptrace_access_check
, cap_ptrace_access_check
),
1082 LSM_HOOK_INIT(ptrace_traceme
, cap_ptrace_traceme
),
1083 LSM_HOOK_INIT(capget
, cap_capget
),
1084 LSM_HOOK_INIT(capset
, cap_capset
),
1085 LSM_HOOK_INIT(bprm_set_creds
, cap_bprm_set_creds
),
1086 LSM_HOOK_INIT(bprm_secureexec
, cap_bprm_secureexec
),
1087 LSM_HOOK_INIT(inode_need_killpriv
, cap_inode_need_killpriv
),
1088 LSM_HOOK_INIT(inode_killpriv
, cap_inode_killpriv
),
1089 LSM_HOOK_INIT(mmap_addr
, cap_mmap_addr
),
1090 LSM_HOOK_INIT(mmap_file
, cap_mmap_file
),
1091 LSM_HOOK_INIT(task_fix_setuid
, cap_task_fix_setuid
),
1092 LSM_HOOK_INIT(task_prctl
, cap_task_prctl
),
1093 LSM_HOOK_INIT(task_setscheduler
, cap_task_setscheduler
),
1094 LSM_HOOK_INIT(task_setioprio
, cap_task_setioprio
),
1095 LSM_HOOK_INIT(task_setnice
, cap_task_setnice
),
1096 LSM_HOOK_INIT(vm_enough_memory
, cap_vm_enough_memory
),
1099 void __init
capability_add_hooks(void)
1101 security_add_hooks(capability_hooks
, ARRAY_SIZE(capability_hooks
));
1104 #endif /* CONFIG_SECURITY */