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 timespec64
*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 error
= __vfs_getxattr(dentry
, inode
, XATTR_NAME_CAPS
, NULL
, 0);
318 * cap_inode_killpriv - Erase the security markings on an inode
319 * @dentry: The inode/dentry to alter
321 * Erase the privilege-enhancing security markings on an inode.
323 * Returns 0 if successful, -ve on error.
325 int cap_inode_killpriv(struct dentry
*dentry
)
329 error
= __vfs_removexattr(dentry
, XATTR_NAME_CAPS
);
330 if (error
== -EOPNOTSUPP
)
335 static bool rootid_owns_currentns(kuid_t kroot
)
337 struct user_namespace
*ns
;
339 if (!uid_valid(kroot
))
342 for (ns
= current_user_ns(); ; ns
= ns
->parent
) {
343 if (from_kuid(ns
, kroot
) == 0)
345 if (ns
== &init_user_ns
)
352 static __u32
sansflags(__u32 m
)
354 return m
& ~VFS_CAP_FLAGS_EFFECTIVE
;
357 static bool is_v2header(size_t size
, __le32 magic
)
359 __u32 m
= le32_to_cpu(magic
);
360 if (size
!= XATTR_CAPS_SZ_2
)
362 return sansflags(m
) == VFS_CAP_REVISION_2
;
365 static bool is_v3header(size_t size
, __le32 magic
)
367 __u32 m
= le32_to_cpu(magic
);
369 if (size
!= XATTR_CAPS_SZ_3
)
371 return sansflags(m
) == VFS_CAP_REVISION_3
;
375 * getsecurity: We are called for security.* before any attempt to read the
376 * xattr from the inode itself.
378 * This gives us a chance to read the on-disk value and convert it. If we
379 * return -EOPNOTSUPP, then vfs_getxattr() will call the i_op handler.
381 * Note we are not called by vfs_getxattr_alloc(), but that is only called
382 * by the integrity subsystem, which really wants the unconverted values -
385 int cap_inode_getsecurity(struct inode
*inode
, const char *name
, void **buffer
,
390 uid_t root
, mappedroot
;
392 struct vfs_cap_data
*cap
;
393 struct vfs_ns_cap_data
*nscap
;
394 struct dentry
*dentry
;
395 struct user_namespace
*fs_ns
;
397 if (strcmp(name
, "capability") != 0)
400 dentry
= d_find_alias(inode
);
404 size
= sizeof(struct vfs_ns_cap_data
);
405 ret
= (int) vfs_getxattr_alloc(dentry
, XATTR_NAME_CAPS
,
406 &tmpbuf
, size
, GFP_NOFS
);
412 fs_ns
= inode
->i_sb
->s_user_ns
;
413 cap
= (struct vfs_cap_data
*) tmpbuf
;
414 if (is_v2header((size_t) ret
, cap
->magic_etc
)) {
415 /* If this is sizeof(vfs_cap_data) then we're ok with the
416 * on-disk value, so return that. */
422 } else if (!is_v3header((size_t) ret
, cap
->magic_etc
)) {
427 nscap
= (struct vfs_ns_cap_data
*) tmpbuf
;
428 root
= le32_to_cpu(nscap
->rootid
);
429 kroot
= make_kuid(fs_ns
, root
);
431 /* If the root kuid maps to a valid uid in current ns, then return
432 * this as a nscap. */
433 mappedroot
= from_kuid(current_user_ns(), kroot
);
434 if (mappedroot
!= (uid_t
)-1 && mappedroot
!= (uid_t
)0) {
437 nscap
->rootid
= cpu_to_le32(mappedroot
);
443 if (!rootid_owns_currentns(kroot
)) {
448 /* This comes from a parent namespace. Return as a v2 capability */
449 size
= sizeof(struct vfs_cap_data
);
451 *buffer
= kmalloc(size
, GFP_ATOMIC
);
453 struct vfs_cap_data
*cap
= *buffer
;
454 __le32 nsmagic
, magic
;
455 magic
= VFS_CAP_REVISION_2
;
456 nsmagic
= le32_to_cpu(nscap
->magic_etc
);
457 if (nsmagic
& VFS_CAP_FLAGS_EFFECTIVE
)
458 magic
|= VFS_CAP_FLAGS_EFFECTIVE
;
459 memcpy(&cap
->data
, &nscap
->data
, sizeof(__le32
) * 2 * VFS_CAP_U32
);
460 cap
->magic_etc
= cpu_to_le32(magic
);
467 static kuid_t
rootid_from_xattr(const void *value
, size_t size
,
468 struct user_namespace
*task_ns
)
470 const struct vfs_ns_cap_data
*nscap
= value
;
473 if (size
== XATTR_CAPS_SZ_3
)
474 rootid
= le32_to_cpu(nscap
->rootid
);
476 return make_kuid(task_ns
, rootid
);
479 static bool validheader(size_t size
, __le32 magic
)
481 return is_v2header(size
, magic
) || is_v3header(size
, magic
);
485 * User requested a write of security.capability. If needed, update the
486 * xattr to change from v2 to v3, or to fixup the v3 rootid.
488 * If all is ok, we return the new size, on error return < 0.
490 int cap_convert_nscap(struct dentry
*dentry
, void **ivalue
, size_t size
)
492 struct vfs_ns_cap_data
*nscap
;
494 const struct vfs_cap_data
*cap
= *ivalue
;
495 __u32 magic
, nsmagic
;
496 struct inode
*inode
= d_backing_inode(dentry
);
497 struct user_namespace
*task_ns
= current_user_ns(),
498 *fs_ns
= inode
->i_sb
->s_user_ns
;
504 if (!validheader(size
, cap
->magic_etc
))
506 if (!capable_wrt_inode_uidgid(inode
, CAP_SETFCAP
))
508 if (size
== XATTR_CAPS_SZ_2
)
509 if (ns_capable(inode
->i_sb
->s_user_ns
, CAP_SETFCAP
))
510 /* user is privileged, just write the v2 */
513 rootid
= rootid_from_xattr(*ivalue
, size
, task_ns
);
514 if (!uid_valid(rootid
))
517 nsrootid
= from_kuid(fs_ns
, rootid
);
521 newsize
= sizeof(struct vfs_ns_cap_data
);
522 nscap
= kmalloc(newsize
, GFP_ATOMIC
);
525 nscap
->rootid
= cpu_to_le32(nsrootid
);
526 nsmagic
= VFS_CAP_REVISION_3
;
527 magic
= le32_to_cpu(cap
->magic_etc
);
528 if (magic
& VFS_CAP_FLAGS_EFFECTIVE
)
529 nsmagic
|= VFS_CAP_FLAGS_EFFECTIVE
;
530 nscap
->magic_etc
= cpu_to_le32(nsmagic
);
531 memcpy(&nscap
->data
, &cap
->data
, sizeof(__le32
) * 2 * VFS_CAP_U32
);
539 * Calculate the new process capability sets from the capability sets attached
542 static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data
*caps
,
543 struct linux_binprm
*bprm
,
547 struct cred
*new = bprm
->cred
;
551 if (caps
->magic_etc
& VFS_CAP_FLAGS_EFFECTIVE
)
554 if (caps
->magic_etc
& VFS_CAP_REVISION_MASK
)
557 CAP_FOR_EACH_U32(i
) {
558 __u32 permitted
= caps
->permitted
.cap
[i
];
559 __u32 inheritable
= caps
->inheritable
.cap
[i
];
562 * pP' = (X & fP) | (pI & fI)
563 * The addition of pA' is handled later.
565 new->cap_permitted
.cap
[i
] =
566 (new->cap_bset
.cap
[i
] & permitted
) |
567 (new->cap_inheritable
.cap
[i
] & inheritable
);
569 if (permitted
& ~new->cap_permitted
.cap
[i
])
570 /* insufficient to execute correctly */
575 * For legacy apps, with no internal support for recognizing they
576 * do not have enough capabilities, we return an error if they are
577 * missing some "forced" (aka file-permitted) capabilities.
579 return *effective
? ret
: 0;
583 * Extract the on-exec-apply capability sets for an executable file.
585 int get_vfs_caps_from_disk(const struct dentry
*dentry
, struct cpu_vfs_cap_data
*cpu_caps
)
587 struct inode
*inode
= d_backing_inode(dentry
);
591 struct vfs_ns_cap_data data
, *nscaps
= &data
;
592 struct vfs_cap_data
*caps
= (struct vfs_cap_data
*) &data
;
594 struct user_namespace
*fs_ns
= inode
->i_sb
->s_user_ns
;
596 memset(cpu_caps
, 0, sizeof(struct cpu_vfs_cap_data
));
601 size
= __vfs_getxattr((struct dentry
*)dentry
, inode
,
602 XATTR_NAME_CAPS
, &data
, XATTR_CAPS_SZ
);
603 if (size
== -ENODATA
|| size
== -EOPNOTSUPP
)
604 /* no data, that's ok */
610 if (size
< sizeof(magic_etc
))
613 cpu_caps
->magic_etc
= magic_etc
= le32_to_cpu(caps
->magic_etc
);
615 rootkuid
= make_kuid(fs_ns
, 0);
616 switch (magic_etc
& VFS_CAP_REVISION_MASK
) {
617 case VFS_CAP_REVISION_1
:
618 if (size
!= XATTR_CAPS_SZ_1
)
620 tocopy
= VFS_CAP_U32_1
;
622 case VFS_CAP_REVISION_2
:
623 if (size
!= XATTR_CAPS_SZ_2
)
625 tocopy
= VFS_CAP_U32_2
;
627 case VFS_CAP_REVISION_3
:
628 if (size
!= XATTR_CAPS_SZ_3
)
630 tocopy
= VFS_CAP_U32_3
;
631 rootkuid
= make_kuid(fs_ns
, le32_to_cpu(nscaps
->rootid
));
637 /* Limit the caps to the mounter of the filesystem
638 * or the more limited uid specified in the xattr.
640 if (!rootid_owns_currentns(rootkuid
))
643 CAP_FOR_EACH_U32(i
) {
646 cpu_caps
->permitted
.cap
[i
] = le32_to_cpu(caps
->data
[i
].permitted
);
647 cpu_caps
->inheritable
.cap
[i
] = le32_to_cpu(caps
->data
[i
].inheritable
);
650 cpu_caps
->permitted
.cap
[CAP_LAST_U32
] &= CAP_LAST_U32_VALID_MASK
;
651 cpu_caps
->inheritable
.cap
[CAP_LAST_U32
] &= CAP_LAST_U32_VALID_MASK
;
657 * Attempt to get the on-exec apply capability sets for an executable file from
658 * its xattrs and, if present, apply them to the proposed credentials being
659 * constructed by execve().
661 static int get_file_caps(struct linux_binprm
*bprm
, bool *effective
, bool *has_cap
)
664 struct cpu_vfs_cap_data vcaps
;
666 bprm_clear_caps(bprm
);
668 if (!file_caps_enabled
)
671 if (path_nosuid(&bprm
->file
->f_path
))
675 * This check is redundant with mnt_may_suid() but is kept to make
676 * explicit that capability bits are limited to s_user_ns and its
679 if (!current_in_userns(bprm
->file
->f_path
.mnt
->mnt_sb
->s_user_ns
))
682 rc
= get_vfs_caps_from_disk(bprm
->file
->f_path
.dentry
, &vcaps
);
685 printk(KERN_NOTICE
"Invalid argument reading file caps for %s\n",
687 else if (rc
== -ENODATA
)
692 rc
= bprm_caps_from_vfs_caps(&vcaps
, bprm
, effective
, has_cap
);
694 printk(KERN_NOTICE
"%s: cap_from_disk returned %d for %s\n",
695 __func__
, rc
, bprm
->filename
);
699 bprm_clear_caps(bprm
);
705 * cap_bprm_set_creds - Set up the proposed credentials for execve().
706 * @bprm: The execution parameters, including the proposed creds
708 * Set up the proposed credentials for a new execution context being
709 * constructed by execve(). The proposed creds in @bprm->cred is altered,
710 * which won't take effect immediately. Returns 0 if successful, -ve on error.
712 int cap_bprm_set_creds(struct linux_binprm
*bprm
)
714 const struct cred
*old
= current_cred();
715 struct cred
*new = bprm
->cred
;
716 bool effective
, has_cap
= false, is_setid
;
720 if (WARN_ON(!cap_ambient_invariant_ok(old
)))
724 ret
= get_file_caps(bprm
, &effective
, &has_cap
);
728 root_uid
= make_kuid(new->user_ns
, 0);
730 if (!issecure(SECURE_NOROOT
)) {
732 * If the legacy file capability is set, then don't set privs
733 * for a setuid root binary run by a non-root user. Do set it
734 * for a root user just to cause least surprise to an admin.
736 if (has_cap
&& !uid_eq(new->uid
, root_uid
) && uid_eq(new->euid
, root_uid
)) {
737 warn_setuid_and_fcaps_mixed(bprm
->filename
);
741 * To support inheritance of root-permissions and suid-root
742 * executables under compatibility mode, we override the
743 * capability sets for the file.
745 * If only the real uid is 0, we do not set the effective bit.
747 if (uid_eq(new->euid
, root_uid
) || uid_eq(new->uid
, root_uid
)) {
748 /* pP' = (cap_bset & ~0) | (pI & ~0) */
749 new->cap_permitted
= cap_combine(old
->cap_bset
,
750 old
->cap_inheritable
);
752 if (uid_eq(new->euid
, root_uid
))
757 /* if we have fs caps, clear dangerous personality flags */
758 if (!cap_issubset(new->cap_permitted
, old
->cap_permitted
))
759 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
762 /* Don't let someone trace a set[ug]id/setpcap binary with the revised
763 * credentials unless they have the appropriate permit.
765 * In addition, if NO_NEW_PRIVS, then ensure we get no new privs.
767 is_setid
= !uid_eq(new->euid
, old
->uid
) || !gid_eq(new->egid
, old
->gid
);
770 !cap_issubset(new->cap_permitted
, old
->cap_permitted
)) &&
771 ((bprm
->unsafe
& ~LSM_UNSAFE_PTRACE
) ||
772 !ptracer_capable(current
, new->user_ns
))) {
773 /* downgrade; they get no more than they had, and maybe less */
774 if (!ns_capable(new->user_ns
, CAP_SETUID
) ||
775 (bprm
->unsafe
& LSM_UNSAFE_NO_NEW_PRIVS
)) {
776 new->euid
= new->uid
;
777 new->egid
= new->gid
;
779 new->cap_permitted
= cap_intersect(new->cap_permitted
,
783 new->suid
= new->fsuid
= new->euid
;
784 new->sgid
= new->fsgid
= new->egid
;
786 /* File caps or setid cancels ambient. */
787 if (has_cap
|| is_setid
)
788 cap_clear(new->cap_ambient
);
791 * Now that we've computed pA', update pP' to give:
792 * pP' = (X & fP) | (pI & fI) | pA'
794 new->cap_permitted
= cap_combine(new->cap_permitted
, new->cap_ambient
);
797 * Set pE' = (fE ? pP' : pA'). Because pA' is zero if fE is set,
798 * this is the same as pE' = (fE ? pP' : 0) | pA'.
801 new->cap_effective
= new->cap_permitted
;
803 new->cap_effective
= new->cap_ambient
;
805 if (WARN_ON(!cap_ambient_invariant_ok(new)))
808 bprm
->cap_effective
= effective
;
811 * Audit candidate if current->cap_effective is set
813 * We do not bother to audit if 3 things are true:
814 * 1) cap_effective has all caps
816 * 3) root is supposed to have all caps (SECURE_NOROOT)
817 * Since this is just a normal root execing a process.
819 * Number 1 above might fail if you don't have a full bset, but I think
820 * that is interesting information to audit.
822 if (!cap_issubset(new->cap_effective
, new->cap_ambient
)) {
823 if (!cap_issubset(CAP_FULL_SET
, new->cap_effective
) ||
824 !uid_eq(new->euid
, root_uid
) || !uid_eq(new->uid
, root_uid
) ||
825 issecure(SECURE_NOROOT
)) {
826 ret
= audit_log_bprm_fcaps(bprm
, new, old
);
832 new->securebits
&= ~issecure_mask(SECURE_KEEP_CAPS
);
834 if (WARN_ON(!cap_ambient_invariant_ok(new)))
841 * cap_bprm_secureexec - Determine whether a secure execution is required
842 * @bprm: The execution parameters
844 * Determine whether a secure execution is required, return 1 if it is, and 0
847 * The credentials have been committed by this point, and so are no longer
848 * available through @bprm->cred.
850 int cap_bprm_secureexec(struct linux_binprm
*bprm
)
852 const struct cred
*cred
= current_cred();
853 kuid_t root_uid
= make_kuid(cred
->user_ns
, 0);
855 if (!uid_eq(cred
->uid
, root_uid
)) {
856 if (bprm
->cap_effective
)
858 if (!cap_issubset(cred
->cap_permitted
, cred
->cap_ambient
))
862 return (!uid_eq(cred
->euid
, cred
->uid
) ||
863 !gid_eq(cred
->egid
, cred
->gid
));
867 * cap_inode_setxattr - Determine whether an xattr may be altered
868 * @dentry: The inode/dentry being altered
869 * @name: The name of the xattr to be changed
870 * @value: The value that the xattr will be changed to
871 * @size: The size of value
872 * @flags: The replacement flag
874 * Determine whether an xattr may be altered or set on an inode, returning 0 if
875 * permission is granted, -ve if denied.
877 * This is used to make sure security xattrs don't get updated or set by those
878 * who aren't privileged to do so.
880 int cap_inode_setxattr(struct dentry
*dentry
, const char *name
,
881 const void *value
, size_t size
, int flags
)
883 /* Ignore non-security xattrs */
884 if (strncmp(name
, XATTR_SECURITY_PREFIX
,
885 sizeof(XATTR_SECURITY_PREFIX
) - 1) != 0)
889 * For XATTR_NAME_CAPS the check will be done in
890 * cap_convert_nscap(), called by setxattr()
892 if (strcmp(name
, XATTR_NAME_CAPS
) == 0)
895 if (!ns_capable(dentry
->d_sb
->s_user_ns
, CAP_SYS_ADMIN
))
901 * cap_inode_removexattr - Determine whether an xattr may be removed
902 * @dentry: The inode/dentry being altered
903 * @name: The name of the xattr to be changed
905 * Determine whether an xattr may be removed from an inode, returning 0 if
906 * permission is granted, -ve if denied.
908 * This is used to make sure security xattrs don't get removed by those who
909 * aren't privileged to remove them.
911 int cap_inode_removexattr(struct dentry
*dentry
, const char *name
)
913 /* Ignore non-security xattrs */
914 if (strncmp(name
, XATTR_SECURITY_PREFIX
,
915 sizeof(XATTR_SECURITY_PREFIX
) - 1) != 0)
918 if (strcmp(name
, XATTR_NAME_CAPS
) == 0) {
919 /* security.capability gets namespaced */
920 struct inode
*inode
= d_backing_inode(dentry
);
923 if (!capable_wrt_inode_uidgid(inode
, CAP_SETFCAP
))
928 if (!ns_capable(dentry
->d_sb
->s_user_ns
, CAP_SYS_ADMIN
))
934 * cap_emulate_setxuid() fixes the effective / permitted capabilities of
935 * a process after a call to setuid, setreuid, or setresuid.
937 * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
938 * {r,e,s}uid != 0, the permitted and effective capabilities are
941 * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
942 * capabilities of the process are cleared.
944 * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
945 * capabilities are set to the permitted capabilities.
947 * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
952 * cevans - New behaviour, Oct '99
953 * A process may, via prctl(), elect to keep its capabilities when it
954 * calls setuid() and switches away from uid==0. Both permitted and
955 * effective sets will be retained.
956 * Without this change, it was impossible for a daemon to drop only some
957 * of its privilege. The call to setuid(!=0) would drop all privileges!
958 * Keeping uid 0 is not an option because uid 0 owns too many vital
960 * Thanks to Olaf Kirch and Peter Benie for spotting this.
962 static inline void cap_emulate_setxuid(struct cred
*new, const struct cred
*old
)
964 kuid_t root_uid
= make_kuid(old
->user_ns
, 0);
966 if ((uid_eq(old
->uid
, root_uid
) ||
967 uid_eq(old
->euid
, root_uid
) ||
968 uid_eq(old
->suid
, root_uid
)) &&
969 (!uid_eq(new->uid
, root_uid
) &&
970 !uid_eq(new->euid
, root_uid
) &&
971 !uid_eq(new->suid
, root_uid
))) {
972 if (!issecure(SECURE_KEEP_CAPS
)) {
973 cap_clear(new->cap_permitted
);
974 cap_clear(new->cap_effective
);
978 * Pre-ambient programs expect setresuid to nonroot followed
979 * by exec to drop capabilities. We should make sure that
980 * this remains the case.
982 cap_clear(new->cap_ambient
);
984 if (uid_eq(old
->euid
, root_uid
) && !uid_eq(new->euid
, root_uid
))
985 cap_clear(new->cap_effective
);
986 if (!uid_eq(old
->euid
, root_uid
) && uid_eq(new->euid
, root_uid
))
987 new->cap_effective
= new->cap_permitted
;
991 * cap_task_fix_setuid - Fix up the results of setuid() call
992 * @new: The proposed credentials
993 * @old: The current task's current credentials
994 * @flags: Indications of what has changed
996 * Fix up the results of setuid() call before the credential changes are
997 * actually applied, returning 0 to grant the changes, -ve to deny them.
999 int cap_task_fix_setuid(struct cred
*new, const struct cred
*old
, int flags
)
1005 /* juggle the capabilities to follow [RES]UID changes unless
1006 * otherwise suppressed */
1007 if (!issecure(SECURE_NO_SETUID_FIXUP
))
1008 cap_emulate_setxuid(new, old
);
1012 /* juggle the capabilties to follow FSUID changes, unless
1013 * otherwise suppressed
1015 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
1016 * if not, we might be a bit too harsh here.
1018 if (!issecure(SECURE_NO_SETUID_FIXUP
)) {
1019 kuid_t root_uid
= make_kuid(old
->user_ns
, 0);
1020 if (uid_eq(old
->fsuid
, root_uid
) && !uid_eq(new->fsuid
, root_uid
))
1021 new->cap_effective
=
1022 cap_drop_fs_set(new->cap_effective
);
1024 if (!uid_eq(old
->fsuid
, root_uid
) && uid_eq(new->fsuid
, root_uid
))
1025 new->cap_effective
=
1026 cap_raise_fs_set(new->cap_effective
,
1027 new->cap_permitted
);
1039 * Rationale: code calling task_setscheduler, task_setioprio, and
1040 * task_setnice, assumes that
1041 * . if capable(cap_sys_nice), then those actions should be allowed
1042 * . if not capable(cap_sys_nice), but acting on your own processes,
1043 * then those actions should be allowed
1044 * This is insufficient now since you can call code without suid, but
1045 * yet with increased caps.
1046 * So we check for increased caps on the target process.
1048 static int cap_safe_nice(struct task_struct
*p
)
1050 int is_subset
, ret
= 0;
1053 is_subset
= cap_issubset(__task_cred(p
)->cap_permitted
,
1054 current_cred()->cap_permitted
);
1055 if (!is_subset
&& !ns_capable(__task_cred(p
)->user_ns
, CAP_SYS_NICE
))
1063 * cap_task_setscheduler - Detemine if scheduler policy change is permitted
1064 * @p: The task to affect
1066 * Detemine if the requested scheduler policy change is permitted for the
1067 * specified task, returning 0 if permission is granted, -ve if denied.
1069 int cap_task_setscheduler(struct task_struct
*p
)
1071 return cap_safe_nice(p
);
1075 * cap_task_ioprio - Detemine if I/O priority change is permitted
1076 * @p: The task to affect
1077 * @ioprio: The I/O priority to set
1079 * Detemine if the requested I/O priority change is permitted for the specified
1080 * task, returning 0 if permission is granted, -ve if denied.
1082 int cap_task_setioprio(struct task_struct
*p
, int ioprio
)
1084 return cap_safe_nice(p
);
1088 * cap_task_ioprio - Detemine if task priority change is permitted
1089 * @p: The task to affect
1090 * @nice: The nice value to set
1092 * Detemine if the requested task priority change is permitted for the
1093 * specified task, returning 0 if permission is granted, -ve if denied.
1095 int cap_task_setnice(struct task_struct
*p
, int nice
)
1097 return cap_safe_nice(p
);
1101 * Implement PR_CAPBSET_DROP. Attempt to remove the specified capability from
1102 * the current task's bounding set. Returns 0 on success, -ve on error.
1104 static int cap_prctl_drop(unsigned long cap
)
1108 if (!ns_capable(current_user_ns(), CAP_SETPCAP
))
1110 if (!cap_valid(cap
))
1113 new = prepare_creds();
1116 cap_lower(new->cap_bset
, cap
);
1117 return commit_creds(new);
1121 * cap_task_prctl - Implement process control functions for this security module
1122 * @option: The process control function requested
1123 * @arg2, @arg3, @arg4, @arg5: The argument data for this function
1125 * Allow process control functions (sys_prctl()) to alter capabilities; may
1126 * also deny access to other functions not otherwise implemented here.
1128 * Returns 0 or +ve on success, -ENOSYS if this function is not implemented
1129 * here, other -ve on error. If -ENOSYS is returned, sys_prctl() and other LSM
1130 * modules will consider performing the function.
1132 int cap_task_prctl(int option
, unsigned long arg2
, unsigned long arg3
,
1133 unsigned long arg4
, unsigned long arg5
)
1135 const struct cred
*old
= current_cred();
1139 case PR_CAPBSET_READ
:
1140 if (!cap_valid(arg2
))
1142 return !!cap_raised(old
->cap_bset
, arg2
);
1144 case PR_CAPBSET_DROP
:
1145 return cap_prctl_drop(arg2
);
1148 * The next four prctl's remain to assist with transitioning a
1149 * system from legacy UID=0 based privilege (when filesystem
1150 * capabilities are not in use) to a system using filesystem
1151 * capabilities only - as the POSIX.1e draft intended.
1155 * PR_SET_SECUREBITS =
1156 * issecure_mask(SECURE_KEEP_CAPS_LOCKED)
1157 * | issecure_mask(SECURE_NOROOT)
1158 * | issecure_mask(SECURE_NOROOT_LOCKED)
1159 * | issecure_mask(SECURE_NO_SETUID_FIXUP)
1160 * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
1162 * will ensure that the current process and all of its
1163 * children will be locked into a pure
1164 * capability-based-privilege environment.
1166 case PR_SET_SECUREBITS
:
1167 if ((((old
->securebits
& SECURE_ALL_LOCKS
) >> 1)
1168 & (old
->securebits
^ arg2
)) /*[1]*/
1169 || ((old
->securebits
& SECURE_ALL_LOCKS
& ~arg2
)) /*[2]*/
1170 || (arg2
& ~(SECURE_ALL_LOCKS
| SECURE_ALL_BITS
)) /*[3]*/
1171 || (cap_capable(current_cred(),
1172 current_cred()->user_ns
, CAP_SETPCAP
,
1173 SECURITY_CAP_AUDIT
) != 0) /*[4]*/
1175 * [1] no changing of bits that are locked
1176 * [2] no unlocking of locks
1177 * [3] no setting of unsupported bits
1178 * [4] doing anything requires privilege (go read about
1179 * the "sendmail capabilities bug")
1182 /* cannot change a locked bit */
1185 new = prepare_creds();
1188 new->securebits
= arg2
;
1189 return commit_creds(new);
1191 case PR_GET_SECUREBITS
:
1192 return old
->securebits
;
1194 case PR_GET_KEEPCAPS
:
1195 return !!issecure(SECURE_KEEP_CAPS
);
1197 case PR_SET_KEEPCAPS
:
1198 if (arg2
> 1) /* Note, we rely on arg2 being unsigned here */
1200 if (issecure(SECURE_KEEP_CAPS_LOCKED
))
1203 new = prepare_creds();
1207 new->securebits
|= issecure_mask(SECURE_KEEP_CAPS
);
1209 new->securebits
&= ~issecure_mask(SECURE_KEEP_CAPS
);
1210 return commit_creds(new);
1212 case PR_CAP_AMBIENT
:
1213 if (arg2
== PR_CAP_AMBIENT_CLEAR_ALL
) {
1214 if (arg3
| arg4
| arg5
)
1217 new = prepare_creds();
1220 cap_clear(new->cap_ambient
);
1221 return commit_creds(new);
1224 if (((!cap_valid(arg3
)) | arg4
| arg5
))
1227 if (arg2
== PR_CAP_AMBIENT_IS_SET
) {
1228 return !!cap_raised(current_cred()->cap_ambient
, arg3
);
1229 } else if (arg2
!= PR_CAP_AMBIENT_RAISE
&&
1230 arg2
!= PR_CAP_AMBIENT_LOWER
) {
1233 if (arg2
== PR_CAP_AMBIENT_RAISE
&&
1234 (!cap_raised(current_cred()->cap_permitted
, arg3
) ||
1235 !cap_raised(current_cred()->cap_inheritable
,
1237 issecure(SECURE_NO_CAP_AMBIENT_RAISE
)))
1240 new = prepare_creds();
1243 if (arg2
== PR_CAP_AMBIENT_RAISE
)
1244 cap_raise(new->cap_ambient
, arg3
);
1246 cap_lower(new->cap_ambient
, arg3
);
1247 return commit_creds(new);
1251 /* No functionality available - continue with default */
1257 * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted
1258 * @mm: The VM space in which the new mapping is to be made
1259 * @pages: The size of the mapping
1261 * Determine whether the allocation of a new virtual mapping by the current
1262 * task is permitted, returning 1 if permission is granted, 0 if not.
1264 int cap_vm_enough_memory(struct mm_struct
*mm
, long pages
)
1266 int cap_sys_admin
= 0;
1268 if (cap_capable(current_cred(), &init_user_ns
, CAP_SYS_ADMIN
,
1269 SECURITY_CAP_NOAUDIT
) == 0)
1271 return cap_sys_admin
;
1275 * cap_mmap_addr - check if able to map given addr
1276 * @addr: address attempting to be mapped
1278 * If the process is attempting to map memory below dac_mmap_min_addr they need
1279 * CAP_SYS_RAWIO. The other parameters to this function are unused by the
1280 * capability security module. Returns 0 if this mapping should be allowed
1283 int cap_mmap_addr(unsigned long addr
)
1287 if (addr
< dac_mmap_min_addr
) {
1288 ret
= cap_capable(current_cred(), &init_user_ns
, CAP_SYS_RAWIO
,
1289 SECURITY_CAP_AUDIT
);
1290 /* set PF_SUPERPRIV if it turns out we allow the low mmap */
1292 current
->flags
|= PF_SUPERPRIV
;
1296 EXPORT_SYMBOL_GPL(cap_mmap_addr
);
1298 int cap_mmap_file(struct file
*file
, unsigned long reqprot
,
1299 unsigned long prot
, unsigned long flags
)
1303 EXPORT_SYMBOL_GPL(cap_mmap_file
);
1305 #ifdef CONFIG_SECURITY
1307 struct security_hook_list capability_hooks
[] __lsm_ro_after_init
= {
1308 LSM_HOOK_INIT(capable
, cap_capable
),
1309 LSM_HOOK_INIT(settime
, cap_settime
),
1310 LSM_HOOK_INIT(ptrace_access_check
, cap_ptrace_access_check
),
1311 LSM_HOOK_INIT(ptrace_traceme
, cap_ptrace_traceme
),
1312 LSM_HOOK_INIT(capget
, cap_capget
),
1313 LSM_HOOK_INIT(capset
, cap_capset
),
1314 LSM_HOOK_INIT(bprm_set_creds
, cap_bprm_set_creds
),
1315 LSM_HOOK_INIT(bprm_secureexec
, cap_bprm_secureexec
),
1316 LSM_HOOK_INIT(inode_need_killpriv
, cap_inode_need_killpriv
),
1317 LSM_HOOK_INIT(inode_killpriv
, cap_inode_killpriv
),
1318 LSM_HOOK_INIT(inode_getsecurity
, cap_inode_getsecurity
),
1319 LSM_HOOK_INIT(mmap_addr
, cap_mmap_addr
),
1320 LSM_HOOK_INIT(mmap_file
, cap_mmap_file
),
1321 LSM_HOOK_INIT(task_fix_setuid
, cap_task_fix_setuid
),
1322 LSM_HOOK_INIT(task_prctl
, cap_task_prctl
),
1323 LSM_HOOK_INIT(task_setscheduler
, cap_task_setscheduler
),
1324 LSM_HOOK_INIT(task_setioprio
, cap_task_setioprio
),
1325 LSM_HOOK_INIT(task_setnice
, cap_task_setnice
),
1326 LSM_HOOK_INIT(vm_enough_memory
, cap_vm_enough_memory
),
1329 void __init
capability_add_hooks(void)
1331 security_add_hooks(capability_hooks
, ARRAY_SIZE(capability_hooks
),
1335 #endif /* CONFIG_SECURITY */