1 /* Common capabilities, needed by capability.o and root_plug.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/module.h>
12 #include <linux/init.h>
13 #include <linux/kernel.h>
14 #include <linux/security.h>
15 #include <linux/file.h>
17 #include <linux/mman.h>
18 #include <linux/pagemap.h>
19 #include <linux/swap.h>
20 #include <linux/skbuff.h>
21 #include <linux/netlink.h>
22 #include <linux/ptrace.h>
23 #include <linux/xattr.h>
24 #include <linux/hugetlb.h>
25 #include <linux/mount.h>
26 #include <linux/sched.h>
27 #include <linux/prctl.h>
28 #include <linux/securebits.h>
30 int cap_netlink_send(struct sock
*sk
, struct sk_buff
*skb
)
32 NETLINK_CB(skb
).eff_cap
= current
->cap_effective
;
36 int cap_netlink_recv(struct sk_buff
*skb
, int cap
)
38 if (!cap_raised(NETLINK_CB(skb
).eff_cap
, cap
))
43 EXPORT_SYMBOL(cap_netlink_recv
);
46 * NOTE WELL: cap_capable() cannot be used like the kernel's capable()
47 * function. That is, it has the reverse semantics: cap_capable()
48 * returns 0 when a task has a capability, but the kernel's capable()
49 * returns 1 for this case.
51 int cap_capable (struct task_struct
*tsk
, int cap
)
53 /* Derived from include/linux/sched.h:capable. */
54 if (cap_raised(tsk
->cap_effective
, cap
))
59 int cap_settime(struct timespec
*ts
, struct timezone
*tz
)
61 if (!capable(CAP_SYS_TIME
))
66 int cap_ptrace_may_access(struct task_struct
*child
, unsigned int mode
)
68 /* Derived from arch/i386/kernel/ptrace.c:sys_ptrace. */
69 if (cap_issubset(child
->cap_permitted
, current
->cap_permitted
))
71 if (capable(CAP_SYS_PTRACE
))
76 int cap_ptrace_traceme(struct task_struct
*parent
)
78 /* Derived from arch/i386/kernel/ptrace.c:sys_ptrace. */
79 if (cap_issubset(current
->cap_permitted
, parent
->cap_permitted
))
81 if (has_capability(parent
, CAP_SYS_PTRACE
))
86 int cap_capget (struct task_struct
*target
, kernel_cap_t
*effective
,
87 kernel_cap_t
*inheritable
, kernel_cap_t
*permitted
)
89 /* Derived from kernel/capability.c:sys_capget. */
90 *effective
= target
->cap_effective
;
91 *inheritable
= target
->cap_inheritable
;
92 *permitted
= target
->cap_permitted
;
96 #ifdef CONFIG_SECURITY_FILE_CAPABILITIES
98 static inline int cap_block_setpcap(struct task_struct
*target
)
101 * No support for remote process capability manipulation with
102 * filesystem capability support.
104 return (target
!= current
);
107 static inline int cap_inh_is_capped(void)
110 * Return 1 if changes to the inheritable set are limited
111 * to the old permitted set. That is, if the current task
112 * does *not* possess the CAP_SETPCAP capability.
114 return (cap_capable(current
, CAP_SETPCAP
) != 0);
117 static inline int cap_limit_ptraced_target(void) { return 1; }
119 #else /* ie., ndef CONFIG_SECURITY_FILE_CAPABILITIES */
121 static inline int cap_block_setpcap(struct task_struct
*t
) { return 0; }
122 static inline int cap_inh_is_capped(void) { return 1; }
123 static inline int cap_limit_ptraced_target(void)
125 return !capable(CAP_SETPCAP
);
128 #endif /* def CONFIG_SECURITY_FILE_CAPABILITIES */
130 int cap_capset_check (struct task_struct
*target
, kernel_cap_t
*effective
,
131 kernel_cap_t
*inheritable
, kernel_cap_t
*permitted
)
133 if (cap_block_setpcap(target
)) {
136 if (cap_inh_is_capped()
137 && !cap_issubset(*inheritable
,
138 cap_combine(target
->cap_inheritable
,
139 current
->cap_permitted
))) {
140 /* incapable of using this inheritable set */
143 if (!cap_issubset(*inheritable
,
144 cap_combine(target
->cap_inheritable
,
145 current
->cap_bset
))) {
146 /* no new pI capabilities outside bounding set */
150 /* verify restrictions on target's new Permitted set */
151 if (!cap_issubset (*permitted
,
152 cap_combine (target
->cap_permitted
,
153 current
->cap_permitted
))) {
157 /* verify the _new_Effective_ is a subset of the _new_Permitted_ */
158 if (!cap_issubset (*effective
, *permitted
)) {
165 void cap_capset_set (struct task_struct
*target
, kernel_cap_t
*effective
,
166 kernel_cap_t
*inheritable
, kernel_cap_t
*permitted
)
168 target
->cap_effective
= *effective
;
169 target
->cap_inheritable
= *inheritable
;
170 target
->cap_permitted
= *permitted
;
173 static inline void bprm_clear_caps(struct linux_binprm
*bprm
)
175 cap_clear(bprm
->cap_post_exec_permitted
);
176 bprm
->cap_effective
= false;
179 #ifdef CONFIG_SECURITY_FILE_CAPABILITIES
181 int cap_inode_need_killpriv(struct dentry
*dentry
)
183 struct inode
*inode
= dentry
->d_inode
;
186 if (!inode
->i_op
|| !inode
->i_op
->getxattr
)
189 error
= inode
->i_op
->getxattr(dentry
, XATTR_NAME_CAPS
, NULL
, 0);
195 int cap_inode_killpriv(struct dentry
*dentry
)
197 struct inode
*inode
= dentry
->d_inode
;
199 if (!inode
->i_op
|| !inode
->i_op
->removexattr
)
202 return inode
->i_op
->removexattr(dentry
, XATTR_NAME_CAPS
);
205 static inline int cap_from_disk(struct vfs_cap_data
*caps
,
206 struct linux_binprm
*bprm
, unsigned size
)
212 if (size
< sizeof(magic_etc
))
215 magic_etc
= le32_to_cpu(caps
->magic_etc
);
217 switch ((magic_etc
& VFS_CAP_REVISION_MASK
)) {
218 case VFS_CAP_REVISION_1
:
219 if (size
!= XATTR_CAPS_SZ_1
)
221 tocopy
= VFS_CAP_U32_1
;
223 case VFS_CAP_REVISION_2
:
224 if (size
!= XATTR_CAPS_SZ_2
)
226 tocopy
= VFS_CAP_U32_2
;
232 if (magic_etc
& VFS_CAP_FLAGS_EFFECTIVE
) {
233 bprm
->cap_effective
= true;
235 bprm
->cap_effective
= false;
240 CAP_FOR_EACH_U32(i
) {
245 * Legacy capability sets have no upper bits
247 bprm
->cap_post_exec_permitted
.cap
[i
] = 0;
251 * pP' = (X & fP) | (pI & fI)
253 value_cpu
= le32_to_cpu(caps
->data
[i
].permitted
);
254 bprm
->cap_post_exec_permitted
.cap
[i
] =
255 (current
->cap_bset
.cap
[i
] & value_cpu
) |
256 (current
->cap_inheritable
.cap
[i
] &
257 le32_to_cpu(caps
->data
[i
].inheritable
));
258 if (value_cpu
& ~bprm
->cap_post_exec_permitted
.cap
[i
]) {
260 * insufficient to execute correctly
267 * For legacy apps, with no internal support for recognizing they
268 * do not have enough capabilities, we return an error if they are
269 * missing some "forced" (aka file-permitted) capabilities.
271 return bprm
->cap_effective
? ret
: 0;
274 /* Locate any VFS capabilities: */
275 static int get_file_caps(struct linux_binprm
*bprm
)
277 struct dentry
*dentry
;
279 struct vfs_cap_data vcaps
;
282 bprm_clear_caps(bprm
);
284 if (!file_caps_enabled
)
287 if (bprm
->file
->f_vfsmnt
->mnt_flags
& MNT_NOSUID
)
290 dentry
= dget(bprm
->file
->f_dentry
);
291 inode
= dentry
->d_inode
;
292 if (!inode
->i_op
|| !inode
->i_op
->getxattr
)
295 rc
= inode
->i_op
->getxattr(dentry
, XATTR_NAME_CAPS
, &vcaps
,
297 if (rc
== -ENODATA
|| rc
== -EOPNOTSUPP
) {
298 /* no data, that's ok */
305 rc
= cap_from_disk(&vcaps
, bprm
, rc
);
307 printk(KERN_NOTICE
"%s: cap_from_disk returned %d for %s\n",
308 __func__
, rc
, bprm
->filename
);
313 bprm_clear_caps(bprm
);
319 int cap_inode_need_killpriv(struct dentry
*dentry
)
324 int cap_inode_killpriv(struct dentry
*dentry
)
329 static inline int get_file_caps(struct linux_binprm
*bprm
)
331 bprm_clear_caps(bprm
);
336 int cap_bprm_set_security (struct linux_binprm
*bprm
)
340 ret
= get_file_caps(bprm
);
342 if (!issecure(SECURE_NOROOT
)) {
344 * To support inheritance of root-permissions and suid-root
345 * executables under compatibility mode, we override the
346 * capability sets for the file.
348 * If only the real uid is 0, we do not set the effective
351 if (bprm
->e_uid
== 0 || current
->uid
== 0) {
352 /* pP' = (cap_bset & ~0) | (pI & ~0) */
353 bprm
->cap_post_exec_permitted
= cap_combine(
354 current
->cap_bset
, current
->cap_inheritable
356 bprm
->cap_effective
= (bprm
->e_uid
== 0);
364 void cap_bprm_apply_creds (struct linux_binprm
*bprm
, int unsafe
)
366 if (bprm
->e_uid
!= current
->uid
|| bprm
->e_gid
!= current
->gid
||
367 !cap_issubset(bprm
->cap_post_exec_permitted
,
368 current
->cap_permitted
)) {
369 set_dumpable(current
->mm
, suid_dumpable
);
370 current
->pdeath_signal
= 0;
372 if (unsafe
& ~LSM_UNSAFE_PTRACE_CAP
) {
373 if (!capable(CAP_SETUID
)) {
374 bprm
->e_uid
= current
->uid
;
375 bprm
->e_gid
= current
->gid
;
377 if (cap_limit_ptraced_target()) {
378 bprm
->cap_post_exec_permitted
= cap_intersect(
379 bprm
->cap_post_exec_permitted
,
380 current
->cap_permitted
);
385 current
->suid
= current
->euid
= current
->fsuid
= bprm
->e_uid
;
386 current
->sgid
= current
->egid
= current
->fsgid
= bprm
->e_gid
;
388 /* For init, we want to retain the capabilities set
389 * in the init_task struct. Thus we skip the usual
390 * capability rules */
391 if (!is_global_init(current
)) {
392 current
->cap_permitted
= bprm
->cap_post_exec_permitted
;
393 if (bprm
->cap_effective
)
394 current
->cap_effective
= bprm
->cap_post_exec_permitted
;
396 cap_clear(current
->cap_effective
);
399 /* AUD: Audit candidate if current->cap_effective is set */
401 current
->securebits
&= ~issecure_mask(SECURE_KEEP_CAPS
);
404 int cap_bprm_secureexec (struct linux_binprm
*bprm
)
406 if (current
->uid
!= 0) {
407 if (bprm
->cap_effective
)
409 if (!cap_isclear(bprm
->cap_post_exec_permitted
))
413 return (current
->euid
!= current
->uid
||
414 current
->egid
!= current
->gid
);
417 int cap_inode_setxattr(struct dentry
*dentry
, const char *name
,
418 const void *value
, size_t size
, int flags
)
420 if (!strcmp(name
, XATTR_NAME_CAPS
)) {
421 if (!capable(CAP_SETFCAP
))
424 } else if (!strncmp(name
, XATTR_SECURITY_PREFIX
,
425 sizeof(XATTR_SECURITY_PREFIX
) - 1) &&
426 !capable(CAP_SYS_ADMIN
))
431 int cap_inode_removexattr(struct dentry
*dentry
, const char *name
)
433 if (!strcmp(name
, XATTR_NAME_CAPS
)) {
434 if (!capable(CAP_SETFCAP
))
437 } else if (!strncmp(name
, XATTR_SECURITY_PREFIX
,
438 sizeof(XATTR_SECURITY_PREFIX
) - 1) &&
439 !capable(CAP_SYS_ADMIN
))
444 /* moved from kernel/sys.c. */
446 * cap_emulate_setxuid() fixes the effective / permitted capabilities of
447 * a process after a call to setuid, setreuid, or setresuid.
449 * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
450 * {r,e,s}uid != 0, the permitted and effective capabilities are
453 * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
454 * capabilities of the process are cleared.
456 * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
457 * capabilities are set to the permitted capabilities.
459 * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
464 * cevans - New behaviour, Oct '99
465 * A process may, via prctl(), elect to keep its capabilities when it
466 * calls setuid() and switches away from uid==0. Both permitted and
467 * effective sets will be retained.
468 * Without this change, it was impossible for a daemon to drop only some
469 * of its privilege. The call to setuid(!=0) would drop all privileges!
470 * Keeping uid 0 is not an option because uid 0 owns too many vital
472 * Thanks to Olaf Kirch and Peter Benie for spotting this.
474 static inline void cap_emulate_setxuid (int old_ruid
, int old_euid
,
477 if ((old_ruid
== 0 || old_euid
== 0 || old_suid
== 0) &&
478 (current
->uid
!= 0 && current
->euid
!= 0 && current
->suid
!= 0) &&
479 !issecure(SECURE_KEEP_CAPS
)) {
480 cap_clear (current
->cap_permitted
);
481 cap_clear (current
->cap_effective
);
483 if (old_euid
== 0 && current
->euid
!= 0) {
484 cap_clear (current
->cap_effective
);
486 if (old_euid
!= 0 && current
->euid
== 0) {
487 current
->cap_effective
= current
->cap_permitted
;
491 int cap_task_post_setuid (uid_t old_ruid
, uid_t old_euid
, uid_t old_suid
,
498 /* Copied from kernel/sys.c:setreuid/setuid/setresuid. */
499 if (!issecure (SECURE_NO_SETUID_FIXUP
)) {
500 cap_emulate_setxuid (old_ruid
, old_euid
, old_suid
);
505 uid_t old_fsuid
= old_ruid
;
507 /* Copied from kernel/sys.c:setfsuid. */
510 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
511 * if not, we might be a bit too harsh here.
514 if (!issecure (SECURE_NO_SETUID_FIXUP
)) {
515 if (old_fsuid
== 0 && current
->fsuid
!= 0) {
516 current
->cap_effective
=
518 current
->cap_effective
);
520 if (old_fsuid
!= 0 && current
->fsuid
== 0) {
521 current
->cap_effective
=
523 current
->cap_effective
,
524 current
->cap_permitted
);
536 #ifdef CONFIG_SECURITY_FILE_CAPABILITIES
538 * Rationale: code calling task_setscheduler, task_setioprio, and
539 * task_setnice, assumes that
540 * . if capable(cap_sys_nice), then those actions should be allowed
541 * . if not capable(cap_sys_nice), but acting on your own processes,
542 * then those actions should be allowed
543 * This is insufficient now since you can call code without suid, but
544 * yet with increased caps.
545 * So we check for increased caps on the target process.
547 static int cap_safe_nice(struct task_struct
*p
)
549 if (!cap_issubset(p
->cap_permitted
, current
->cap_permitted
) &&
550 !capable(CAP_SYS_NICE
))
555 int cap_task_setscheduler (struct task_struct
*p
, int policy
,
556 struct sched_param
*lp
)
558 return cap_safe_nice(p
);
561 int cap_task_setioprio (struct task_struct
*p
, int ioprio
)
563 return cap_safe_nice(p
);
566 int cap_task_setnice (struct task_struct
*p
, int nice
)
568 return cap_safe_nice(p
);
572 * called from kernel/sys.c for prctl(PR_CABSET_DROP)
573 * done without task_capability_lock() because it introduces
574 * no new races - i.e. only another task doing capget() on
575 * this task could get inconsistent info. There can be no
576 * racing writer bc a task can only change its own caps.
578 static long cap_prctl_drop(unsigned long cap
)
580 if (!capable(CAP_SETPCAP
))
584 cap_lower(current
->cap_bset
, cap
);
589 int cap_task_setscheduler (struct task_struct
*p
, int policy
,
590 struct sched_param
*lp
)
594 int cap_task_setioprio (struct task_struct
*p
, int ioprio
)
598 int cap_task_setnice (struct task_struct
*p
, int nice
)
604 int cap_task_prctl(int option
, unsigned long arg2
, unsigned long arg3
,
605 unsigned long arg4
, unsigned long arg5
, long *rc_p
)
610 case PR_CAPBSET_READ
:
611 if (!cap_valid(arg2
))
614 error
= !!cap_raised(current
->cap_bset
, arg2
);
616 #ifdef CONFIG_SECURITY_FILE_CAPABILITIES
617 case PR_CAPBSET_DROP
:
618 error
= cap_prctl_drop(arg2
);
622 * The next four prctl's remain to assist with transitioning a
623 * system from legacy UID=0 based privilege (when filesystem
624 * capabilities are not in use) to a system using filesystem
625 * capabilities only - as the POSIX.1e draft intended.
629 * PR_SET_SECUREBITS =
630 * issecure_mask(SECURE_KEEP_CAPS_LOCKED)
631 * | issecure_mask(SECURE_NOROOT)
632 * | issecure_mask(SECURE_NOROOT_LOCKED)
633 * | issecure_mask(SECURE_NO_SETUID_FIXUP)
634 * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
636 * will ensure that the current process and all of its
637 * children will be locked into a pure
638 * capability-based-privilege environment.
640 case PR_SET_SECUREBITS
:
641 if ((((current
->securebits
& SECURE_ALL_LOCKS
) >> 1)
642 & (current
->securebits
^ arg2
)) /*[1]*/
643 || ((current
->securebits
& SECURE_ALL_LOCKS
645 || (arg2
& ~(SECURE_ALL_LOCKS
| SECURE_ALL_BITS
)) /*[3]*/
646 || (cap_capable(current
, CAP_SETPCAP
) != 0)) { /*[4]*/
648 * [1] no changing of bits that are locked
649 * [2] no unlocking of locks
650 * [3] no setting of unsupported bits
651 * [4] doing anything requires privilege (go read about
652 * the "sendmail capabilities bug")
654 error
= -EPERM
; /* cannot change a locked bit */
656 current
->securebits
= arg2
;
659 case PR_GET_SECUREBITS
:
660 error
= current
->securebits
;
663 #endif /* def CONFIG_SECURITY_FILE_CAPABILITIES */
665 case PR_GET_KEEPCAPS
:
666 if (issecure(SECURE_KEEP_CAPS
))
669 case PR_SET_KEEPCAPS
:
670 if (arg2
> 1) /* Note, we rely on arg2 being unsigned here */
672 else if (issecure(SECURE_KEEP_CAPS_LOCKED
))
675 current
->securebits
|= issecure_mask(SECURE_KEEP_CAPS
);
677 current
->securebits
&=
678 ~issecure_mask(SECURE_KEEP_CAPS
);
682 /* No functionality available - continue with default */
686 /* Functionality provided */
691 void cap_task_reparent_to_init (struct task_struct
*p
)
693 cap_set_init_eff(p
->cap_effective
);
694 cap_clear(p
->cap_inheritable
);
695 cap_set_full(p
->cap_permitted
);
696 p
->securebits
= SECUREBITS_DEFAULT
;
700 int cap_syslog (int type
)
702 if ((type
!= 3 && type
!= 10) && !capable(CAP_SYS_ADMIN
))
707 int cap_vm_enough_memory(struct mm_struct
*mm
, long pages
)
709 int cap_sys_admin
= 0;
711 if (cap_capable(current
, CAP_SYS_ADMIN
) == 0)
713 return __vm_enough_memory(mm
, pages
, cap_sys_admin
);