Commit | Line | Data |
---|---|---|
3e1c2515 | 1 | /* Common capabilities, needed by capability.o. |
1da177e4 LT |
2 | * |
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. | |
7 | * | |
8 | */ | |
9 | ||
c59ede7b | 10 | #include <linux/capability.h> |
3fc689e9 | 11 | #include <linux/audit.h> |
1da177e4 LT |
12 | #include <linux/module.h> |
13 | #include <linux/init.h> | |
14 | #include <linux/kernel.h> | |
15 | #include <linux/security.h> | |
16 | #include <linux/file.h> | |
17 | #include <linux/mm.h> | |
18 | #include <linux/mman.h> | |
19 | #include <linux/pagemap.h> | |
20 | #include <linux/swap.h> | |
1da177e4 LT |
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> | |
b5376771 | 26 | #include <linux/mount.h> |
b460cbc5 | 27 | #include <linux/sched.h> |
3898b1b4 AM |
28 | #include <linux/prctl.h> |
29 | #include <linux/securebits.h> | |
3486740a | 30 | #include <linux/user_namespace.h> |
72c2d582 | 31 | |
b5f22a59 SH |
32 | /* |
33 | * If a non-root user executes a setuid-root binary in | |
34 | * !secure(SECURE_NOROOT) mode, then we raise capabilities. | |
35 | * However if fE is also set, then the intent is for only | |
36 | * the file capabilities to be applied, and the setuid-root | |
37 | * bit is left on either to change the uid (plausible) or | |
38 | * to get full privilege on a kernel without file capabilities | |
39 | * support. So in that case we do not raise capabilities. | |
40 | * | |
41 | * Warn if that happens, once per boot. | |
42 | */ | |
d7627467 | 43 | static void warn_setuid_and_fcaps_mixed(const char *fname) |
b5f22a59 SH |
44 | { |
45 | static int warned; | |
46 | if (!warned) { | |
47 | printk(KERN_INFO "warning: `%s' has both setuid-root and" | |
48 | " effective capabilities. Therefore not raising all" | |
49 | " capabilities.\n", fname); | |
50 | warned = 1; | |
51 | } | |
52 | } | |
53 | ||
1da177e4 LT |
54 | int cap_netlink_send(struct sock *sk, struct sk_buff *skb) |
55 | { | |
1da177e4 LT |
56 | return 0; |
57 | } | |
58 | ||
c7bdb545 | 59 | int cap_netlink_recv(struct sk_buff *skb, int cap) |
1da177e4 | 60 | { |
01a16b21 | 61 | if (!cap_raised(current_cap(), cap)) |
1da177e4 LT |
62 | return -EPERM; |
63 | return 0; | |
64 | } | |
1da177e4 LT |
65 | EXPORT_SYMBOL(cap_netlink_recv); |
66 | ||
1d045980 DH |
67 | /** |
68 | * cap_capable - Determine whether a task has a particular effective capability | |
69 | * @tsk: The task to query | |
3699c53c | 70 | * @cred: The credentials to use |
3486740a | 71 | * @ns: The user namespace in which we need the capability |
1d045980 DH |
72 | * @cap: The capability to check for |
73 | * @audit: Whether to write an audit message or not | |
74 | * | |
75 | * Determine whether the nominated task has the specified capability amongst | |
76 | * its effective set, returning 0 if it does, -ve if it does not. | |
77 | * | |
3699c53c DH |
78 | * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable() |
79 | * and has_capability() functions. That is, it has the reverse semantics: | |
80 | * cap_has_capability() returns 0 when a task has a capability, but the | |
81 | * kernel's capable() and has_capability() returns 1 for this case. | |
a6dbb1ef | 82 | */ |
3486740a SH |
83 | int cap_capable(struct task_struct *tsk, const struct cred *cred, |
84 | struct user_namespace *targ_ns, int cap, int audit) | |
1da177e4 | 85 | { |
3486740a SH |
86 | for (;;) { |
87 | /* The creator of the user namespace has all caps. */ | |
88 | if (targ_ns != &init_user_ns && targ_ns->creator == cred->user) | |
89 | return 0; | |
90 | ||
91 | /* Do we have the necessary capabilities? */ | |
92 | if (targ_ns == cred->user->user_ns) | |
93 | return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM; | |
94 | ||
95 | /* Have we tried all of the parent namespaces? */ | |
96 | if (targ_ns == &init_user_ns) | |
97 | return -EPERM; | |
98 | ||
99 | /* | |
100 | *If you have a capability in a parent user ns, then you have | |
101 | * it over all children user namespaces as well. | |
102 | */ | |
103 | targ_ns = targ_ns->creator->user_ns; | |
104 | } | |
105 | ||
106 | /* We never get here */ | |
1da177e4 LT |
107 | } |
108 | ||
1d045980 DH |
109 | /** |
110 | * cap_settime - Determine whether the current process may set the system clock | |
111 | * @ts: The time to set | |
112 | * @tz: The timezone to set | |
113 | * | |
114 | * Determine whether the current process may set the system clock and timezone | |
115 | * information, returning 0 if permission granted, -ve if denied. | |
116 | */ | |
1e6d7679 | 117 | int cap_settime(const struct timespec *ts, const struct timezone *tz) |
1da177e4 LT |
118 | { |
119 | if (!capable(CAP_SYS_TIME)) | |
120 | return -EPERM; | |
121 | return 0; | |
122 | } | |
123 | ||
1d045980 | 124 | /** |
9e48858f | 125 | * cap_ptrace_access_check - Determine whether the current process may access |
1d045980 DH |
126 | * another |
127 | * @child: The process to be accessed | |
128 | * @mode: The mode of attachment. | |
129 | * | |
130 | * Determine whether a process may access another, returning 0 if permission | |
131 | * granted, -ve if denied. | |
132 | */ | |
9e48858f | 133 | int cap_ptrace_access_check(struct task_struct *child, unsigned int mode) |
1da177e4 | 134 | { |
c69e8d9c DH |
135 | int ret = 0; |
136 | ||
137 | rcu_read_lock(); | |
d84f4f99 DH |
138 | if (!cap_issubset(__task_cred(child)->cap_permitted, |
139 | current_cred()->cap_permitted) && | |
c69e8d9c DH |
140 | !capable(CAP_SYS_PTRACE)) |
141 | ret = -EPERM; | |
142 | rcu_read_unlock(); | |
143 | return ret; | |
5cd9c58f DH |
144 | } |
145 | ||
1d045980 DH |
146 | /** |
147 | * cap_ptrace_traceme - Determine whether another process may trace the current | |
148 | * @parent: The task proposed to be the tracer | |
149 | * | |
150 | * Determine whether the nominated task is permitted to trace the current | |
151 | * process, returning 0 if permission is granted, -ve if denied. | |
152 | */ | |
5cd9c58f DH |
153 | int cap_ptrace_traceme(struct task_struct *parent) |
154 | { | |
c69e8d9c DH |
155 | int ret = 0; |
156 | ||
157 | rcu_read_lock(); | |
d84f4f99 DH |
158 | if (!cap_issubset(current_cred()->cap_permitted, |
159 | __task_cred(parent)->cap_permitted) && | |
c69e8d9c DH |
160 | !has_capability(parent, CAP_SYS_PTRACE)) |
161 | ret = -EPERM; | |
162 | rcu_read_unlock(); | |
163 | return ret; | |
1da177e4 LT |
164 | } |
165 | ||
1d045980 DH |
166 | /** |
167 | * cap_capget - Retrieve a task's capability sets | |
168 | * @target: The task from which to retrieve the capability sets | |
169 | * @effective: The place to record the effective set | |
170 | * @inheritable: The place to record the inheritable set | |
171 | * @permitted: The place to record the permitted set | |
172 | * | |
173 | * This function retrieves the capabilities of the nominated task and returns | |
174 | * them to the caller. | |
175 | */ | |
176 | int cap_capget(struct task_struct *target, kernel_cap_t *effective, | |
177 | kernel_cap_t *inheritable, kernel_cap_t *permitted) | |
1da177e4 | 178 | { |
c69e8d9c | 179 | const struct cred *cred; |
b6dff3ec | 180 | |
1da177e4 | 181 | /* Derived from kernel/capability.c:sys_capget. */ |
c69e8d9c DH |
182 | rcu_read_lock(); |
183 | cred = __task_cred(target); | |
b6dff3ec DH |
184 | *effective = cred->cap_effective; |
185 | *inheritable = cred->cap_inheritable; | |
186 | *permitted = cred->cap_permitted; | |
c69e8d9c | 187 | rcu_read_unlock(); |
1da177e4 LT |
188 | return 0; |
189 | } | |
190 | ||
1d045980 DH |
191 | /* |
192 | * Determine whether the inheritable capabilities are limited to the old | |
193 | * permitted set. Returns 1 if they are limited, 0 if they are not. | |
194 | */ | |
72c2d582 AM |
195 | static inline int cap_inh_is_capped(void) |
196 | { | |
72c2d582 | 197 | |
1d045980 DH |
198 | /* they are so limited unless the current task has the CAP_SETPCAP |
199 | * capability | |
200 | */ | |
3486740a SH |
201 | if (cap_capable(current, current_cred(), |
202 | current_cred()->user->user_ns, CAP_SETPCAP, | |
3699c53c | 203 | SECURITY_CAP_AUDIT) == 0) |
1d045980 | 204 | return 0; |
1d045980 | 205 | return 1; |
1209726c | 206 | } |
72c2d582 | 207 | |
1d045980 DH |
208 | /** |
209 | * cap_capset - Validate and apply proposed changes to current's capabilities | |
210 | * @new: The proposed new credentials; alterations should be made here | |
211 | * @old: The current task's current credentials | |
212 | * @effective: A pointer to the proposed new effective capabilities set | |
213 | * @inheritable: A pointer to the proposed new inheritable capabilities set | |
214 | * @permitted: A pointer to the proposed new permitted capabilities set | |
215 | * | |
216 | * This function validates and applies a proposed mass change to the current | |
217 | * process's capability sets. The changes are made to the proposed new | |
218 | * credentials, and assuming no error, will be committed by the caller of LSM. | |
219 | */ | |
d84f4f99 DH |
220 | int cap_capset(struct cred *new, |
221 | const struct cred *old, | |
222 | const kernel_cap_t *effective, | |
223 | const kernel_cap_t *inheritable, | |
224 | const kernel_cap_t *permitted) | |
1da177e4 | 225 | { |
d84f4f99 DH |
226 | if (cap_inh_is_capped() && |
227 | !cap_issubset(*inheritable, | |
228 | cap_combine(old->cap_inheritable, | |
229 | old->cap_permitted))) | |
72c2d582 | 230 | /* incapable of using this inheritable set */ |
1da177e4 | 231 | return -EPERM; |
d84f4f99 | 232 | |
3b7391de | 233 | if (!cap_issubset(*inheritable, |
d84f4f99 DH |
234 | cap_combine(old->cap_inheritable, |
235 | old->cap_bset))) | |
3b7391de SH |
236 | /* no new pI capabilities outside bounding set */ |
237 | return -EPERM; | |
1da177e4 LT |
238 | |
239 | /* verify restrictions on target's new Permitted set */ | |
d84f4f99 | 240 | if (!cap_issubset(*permitted, old->cap_permitted)) |
1da177e4 | 241 | return -EPERM; |
1da177e4 LT |
242 | |
243 | /* verify the _new_Effective_ is a subset of the _new_Permitted_ */ | |
d84f4f99 | 244 | if (!cap_issubset(*effective, *permitted)) |
1da177e4 | 245 | return -EPERM; |
1da177e4 | 246 | |
d84f4f99 DH |
247 | new->cap_effective = *effective; |
248 | new->cap_inheritable = *inheritable; | |
249 | new->cap_permitted = *permitted; | |
1da177e4 LT |
250 | return 0; |
251 | } | |
252 | ||
1d045980 DH |
253 | /* |
254 | * Clear proposed capability sets for execve(). | |
255 | */ | |
b5376771 SH |
256 | static inline void bprm_clear_caps(struct linux_binprm *bprm) |
257 | { | |
a6f76f23 | 258 | cap_clear(bprm->cred->cap_permitted); |
b5376771 SH |
259 | bprm->cap_effective = false; |
260 | } | |
261 | ||
1d045980 DH |
262 | /** |
263 | * cap_inode_need_killpriv - Determine if inode change affects privileges | |
264 | * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV | |
265 | * | |
266 | * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV | |
267 | * affects the security markings on that inode, and if it is, should | |
268 | * inode_killpriv() be invoked or the change rejected? | |
269 | * | |
270 | * Returns 0 if granted; +ve if granted, but inode_killpriv() is required; and | |
271 | * -ve to deny the change. | |
272 | */ | |
b5376771 SH |
273 | int cap_inode_need_killpriv(struct dentry *dentry) |
274 | { | |
275 | struct inode *inode = dentry->d_inode; | |
276 | int error; | |
277 | ||
acfa4380 | 278 | if (!inode->i_op->getxattr) |
b5376771 SH |
279 | return 0; |
280 | ||
281 | error = inode->i_op->getxattr(dentry, XATTR_NAME_CAPS, NULL, 0); | |
282 | if (error <= 0) | |
283 | return 0; | |
284 | return 1; | |
285 | } | |
286 | ||
1d045980 DH |
287 | /** |
288 | * cap_inode_killpriv - Erase the security markings on an inode | |
289 | * @dentry: The inode/dentry to alter | |
290 | * | |
291 | * Erase the privilege-enhancing security markings on an inode. | |
292 | * | |
293 | * Returns 0 if successful, -ve on error. | |
294 | */ | |
b5376771 SH |
295 | int cap_inode_killpriv(struct dentry *dentry) |
296 | { | |
297 | struct inode *inode = dentry->d_inode; | |
298 | ||
acfa4380 | 299 | if (!inode->i_op->removexattr) |
b5376771 SH |
300 | return 0; |
301 | ||
302 | return inode->i_op->removexattr(dentry, XATTR_NAME_CAPS); | |
303 | } | |
304 | ||
1d045980 DH |
305 | /* |
306 | * Calculate the new process capability sets from the capability sets attached | |
307 | * to a file. | |
308 | */ | |
c0b00441 | 309 | static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps, |
a6f76f23 DH |
310 | struct linux_binprm *bprm, |
311 | bool *effective) | |
b5376771 | 312 | { |
a6f76f23 | 313 | struct cred *new = bprm->cred; |
c0b00441 EP |
314 | unsigned i; |
315 | int ret = 0; | |
316 | ||
317 | if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE) | |
a6f76f23 | 318 | *effective = true; |
c0b00441 EP |
319 | |
320 | CAP_FOR_EACH_U32(i) { | |
321 | __u32 permitted = caps->permitted.cap[i]; | |
322 | __u32 inheritable = caps->inheritable.cap[i]; | |
323 | ||
324 | /* | |
325 | * pP' = (X & fP) | (pI & fI) | |
326 | */ | |
a6f76f23 DH |
327 | new->cap_permitted.cap[i] = |
328 | (new->cap_bset.cap[i] & permitted) | | |
329 | (new->cap_inheritable.cap[i] & inheritable); | |
c0b00441 | 330 | |
a6f76f23 DH |
331 | if (permitted & ~new->cap_permitted.cap[i]) |
332 | /* insufficient to execute correctly */ | |
c0b00441 | 333 | ret = -EPERM; |
c0b00441 EP |
334 | } |
335 | ||
336 | /* | |
337 | * For legacy apps, with no internal support for recognizing they | |
338 | * do not have enough capabilities, we return an error if they are | |
339 | * missing some "forced" (aka file-permitted) capabilities. | |
340 | */ | |
a6f76f23 | 341 | return *effective ? ret : 0; |
c0b00441 EP |
342 | } |
343 | ||
1d045980 DH |
344 | /* |
345 | * Extract the on-exec-apply capability sets for an executable file. | |
346 | */ | |
c0b00441 EP |
347 | int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps) |
348 | { | |
349 | struct inode *inode = dentry->d_inode; | |
b5376771 | 350 | __u32 magic_etc; |
e338d263 | 351 | unsigned tocopy, i; |
c0b00441 EP |
352 | int size; |
353 | struct vfs_cap_data caps; | |
354 | ||
355 | memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data)); | |
356 | ||
acfa4380 | 357 | if (!inode || !inode->i_op->getxattr) |
c0b00441 EP |
358 | return -ENODATA; |
359 | ||
360 | size = inode->i_op->getxattr((struct dentry *)dentry, XATTR_NAME_CAPS, &caps, | |
361 | XATTR_CAPS_SZ); | |
a6f76f23 | 362 | if (size == -ENODATA || size == -EOPNOTSUPP) |
c0b00441 EP |
363 | /* no data, that's ok */ |
364 | return -ENODATA; | |
c0b00441 EP |
365 | if (size < 0) |
366 | return size; | |
b5376771 | 367 | |
e338d263 | 368 | if (size < sizeof(magic_etc)) |
b5376771 SH |
369 | return -EINVAL; |
370 | ||
c0b00441 | 371 | cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps.magic_etc); |
b5376771 | 372 | |
a6f76f23 | 373 | switch (magic_etc & VFS_CAP_REVISION_MASK) { |
e338d263 AM |
374 | case VFS_CAP_REVISION_1: |
375 | if (size != XATTR_CAPS_SZ_1) | |
376 | return -EINVAL; | |
377 | tocopy = VFS_CAP_U32_1; | |
378 | break; | |
379 | case VFS_CAP_REVISION_2: | |
380 | if (size != XATTR_CAPS_SZ_2) | |
381 | return -EINVAL; | |
382 | tocopy = VFS_CAP_U32_2; | |
383 | break; | |
b5376771 SH |
384 | default: |
385 | return -EINVAL; | |
386 | } | |
e338d263 | 387 | |
5459c164 | 388 | CAP_FOR_EACH_U32(i) { |
c0b00441 EP |
389 | if (i >= tocopy) |
390 | break; | |
391 | cpu_caps->permitted.cap[i] = le32_to_cpu(caps.data[i].permitted); | |
392 | cpu_caps->inheritable.cap[i] = le32_to_cpu(caps.data[i].inheritable); | |
e338d263 | 393 | } |
a6f76f23 | 394 | |
c0b00441 | 395 | return 0; |
b5376771 SH |
396 | } |
397 | ||
1d045980 DH |
398 | /* |
399 | * Attempt to get the on-exec apply capability sets for an executable file from | |
400 | * its xattrs and, if present, apply them to the proposed credentials being | |
401 | * constructed by execve(). | |
402 | */ | |
a6f76f23 | 403 | static int get_file_caps(struct linux_binprm *bprm, bool *effective) |
b5376771 SH |
404 | { |
405 | struct dentry *dentry; | |
406 | int rc = 0; | |
c0b00441 | 407 | struct cpu_vfs_cap_data vcaps; |
b5376771 | 408 | |
3318a386 SH |
409 | bprm_clear_caps(bprm); |
410 | ||
1f29fae2 SH |
411 | if (!file_caps_enabled) |
412 | return 0; | |
413 | ||
3318a386 | 414 | if (bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID) |
b5376771 | 415 | return 0; |
b5376771 SH |
416 | |
417 | dentry = dget(bprm->file->f_dentry); | |
b5376771 | 418 | |
c0b00441 EP |
419 | rc = get_vfs_caps_from_disk(dentry, &vcaps); |
420 | if (rc < 0) { | |
421 | if (rc == -EINVAL) | |
422 | printk(KERN_NOTICE "%s: get_vfs_caps_from_disk returned %d for %s\n", | |
423 | __func__, rc, bprm->filename); | |
424 | else if (rc == -ENODATA) | |
425 | rc = 0; | |
b5376771 SH |
426 | goto out; |
427 | } | |
b5376771 | 428 | |
a6f76f23 DH |
429 | rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective); |
430 | if (rc == -EINVAL) | |
431 | printk(KERN_NOTICE "%s: cap_from_disk returned %d for %s\n", | |
432 | __func__, rc, bprm->filename); | |
b5376771 SH |
433 | |
434 | out: | |
435 | dput(dentry); | |
436 | if (rc) | |
437 | bprm_clear_caps(bprm); | |
438 | ||
439 | return rc; | |
440 | } | |
441 | ||
1d045980 DH |
442 | /** |
443 | * cap_bprm_set_creds - Set up the proposed credentials for execve(). | |
444 | * @bprm: The execution parameters, including the proposed creds | |
445 | * | |
446 | * Set up the proposed credentials for a new execution context being | |
447 | * constructed by execve(). The proposed creds in @bprm->cred is altered, | |
448 | * which won't take effect immediately. Returns 0 if successful, -ve on error. | |
a6f76f23 DH |
449 | */ |
450 | int cap_bprm_set_creds(struct linux_binprm *bprm) | |
1da177e4 | 451 | { |
a6f76f23 DH |
452 | const struct cred *old = current_cred(); |
453 | struct cred *new = bprm->cred; | |
454 | bool effective; | |
b5376771 | 455 | int ret; |
1da177e4 | 456 | |
a6f76f23 DH |
457 | effective = false; |
458 | ret = get_file_caps(bprm, &effective); | |
459 | if (ret < 0) | |
460 | return ret; | |
1da177e4 | 461 | |
5459c164 | 462 | if (!issecure(SECURE_NOROOT)) { |
b5f22a59 SH |
463 | /* |
464 | * If the legacy file capability is set, then don't set privs | |
465 | * for a setuid root binary run by a non-root user. Do set it | |
466 | * for a root user just to cause least surprise to an admin. | |
467 | */ | |
468 | if (effective && new->uid != 0 && new->euid == 0) { | |
469 | warn_setuid_and_fcaps_mixed(bprm->filename); | |
470 | goto skip; | |
471 | } | |
5459c164 AM |
472 | /* |
473 | * To support inheritance of root-permissions and suid-root | |
474 | * executables under compatibility mode, we override the | |
475 | * capability sets for the file. | |
476 | * | |
a6f76f23 | 477 | * If only the real uid is 0, we do not set the effective bit. |
5459c164 | 478 | */ |
a6f76f23 | 479 | if (new->euid == 0 || new->uid == 0) { |
5459c164 | 480 | /* pP' = (cap_bset & ~0) | (pI & ~0) */ |
a6f76f23 DH |
481 | new->cap_permitted = cap_combine(old->cap_bset, |
482 | old->cap_inheritable); | |
1da177e4 | 483 | } |
a6f76f23 DH |
484 | if (new->euid == 0) |
485 | effective = true; | |
1da177e4 | 486 | } |
b5f22a59 | 487 | skip: |
b5376771 | 488 | |
a6f76f23 DH |
489 | /* Don't let someone trace a set[ug]id/setpcap binary with the revised |
490 | * credentials unless they have the appropriate permit | |
491 | */ | |
492 | if ((new->euid != old->uid || | |
493 | new->egid != old->gid || | |
494 | !cap_issubset(new->cap_permitted, old->cap_permitted)) && | |
495 | bprm->unsafe & ~LSM_UNSAFE_PTRACE_CAP) { | |
496 | /* downgrade; they get no more than they had, and maybe less */ | |
497 | if (!capable(CAP_SETUID)) { | |
498 | new->euid = new->uid; | |
499 | new->egid = new->gid; | |
1da177e4 | 500 | } |
b3a222e5 SH |
501 | new->cap_permitted = cap_intersect(new->cap_permitted, |
502 | old->cap_permitted); | |
1da177e4 LT |
503 | } |
504 | ||
a6f76f23 DH |
505 | new->suid = new->fsuid = new->euid; |
506 | new->sgid = new->fsgid = new->egid; | |
1da177e4 | 507 | |
a6f76f23 DH |
508 | /* For init, we want to retain the capabilities set in the initial |
509 | * task. Thus we skip the usual capability rules | |
510 | */ | |
b460cbc5 | 511 | if (!is_global_init(current)) { |
a6f76f23 DH |
512 | if (effective) |
513 | new->cap_effective = new->cap_permitted; | |
e338d263 | 514 | else |
d84f4f99 | 515 | cap_clear(new->cap_effective); |
1da177e4 | 516 | } |
a6f76f23 | 517 | bprm->cap_effective = effective; |
1da177e4 | 518 | |
3fc689e9 EP |
519 | /* |
520 | * Audit candidate if current->cap_effective is set | |
521 | * | |
522 | * We do not bother to audit if 3 things are true: | |
523 | * 1) cap_effective has all caps | |
524 | * 2) we are root | |
525 | * 3) root is supposed to have all caps (SECURE_NOROOT) | |
526 | * Since this is just a normal root execing a process. | |
527 | * | |
528 | * Number 1 above might fail if you don't have a full bset, but I think | |
529 | * that is interesting information to audit. | |
530 | */ | |
d84f4f99 DH |
531 | if (!cap_isclear(new->cap_effective)) { |
532 | if (!cap_issubset(CAP_FULL_SET, new->cap_effective) || | |
a6f76f23 DH |
533 | new->euid != 0 || new->uid != 0 || |
534 | issecure(SECURE_NOROOT)) { | |
535 | ret = audit_log_bprm_fcaps(bprm, new, old); | |
536 | if (ret < 0) | |
537 | return ret; | |
538 | } | |
3fc689e9 | 539 | } |
1da177e4 | 540 | |
d84f4f99 | 541 | new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS); |
a6f76f23 | 542 | return 0; |
1da177e4 LT |
543 | } |
544 | ||
1d045980 DH |
545 | /** |
546 | * cap_bprm_secureexec - Determine whether a secure execution is required | |
547 | * @bprm: The execution parameters | |
548 | * | |
549 | * Determine whether a secure execution is required, return 1 if it is, and 0 | |
550 | * if it is not. | |
551 | * | |
552 | * The credentials have been committed by this point, and so are no longer | |
553 | * available through @bprm->cred. | |
a6f76f23 DH |
554 | */ |
555 | int cap_bprm_secureexec(struct linux_binprm *bprm) | |
1da177e4 | 556 | { |
c69e8d9c | 557 | const struct cred *cred = current_cred(); |
b6dff3ec DH |
558 | |
559 | if (cred->uid != 0) { | |
b5376771 SH |
560 | if (bprm->cap_effective) |
561 | return 1; | |
a6f76f23 | 562 | if (!cap_isclear(cred->cap_permitted)) |
b5376771 SH |
563 | return 1; |
564 | } | |
565 | ||
b6dff3ec DH |
566 | return (cred->euid != cred->uid || |
567 | cred->egid != cred->gid); | |
1da177e4 LT |
568 | } |
569 | ||
1d045980 DH |
570 | /** |
571 | * cap_inode_setxattr - Determine whether an xattr may be altered | |
572 | * @dentry: The inode/dentry being altered | |
573 | * @name: The name of the xattr to be changed | |
574 | * @value: The value that the xattr will be changed to | |
575 | * @size: The size of value | |
576 | * @flags: The replacement flag | |
577 | * | |
578 | * Determine whether an xattr may be altered or set on an inode, returning 0 if | |
579 | * permission is granted, -ve if denied. | |
580 | * | |
581 | * This is used to make sure security xattrs don't get updated or set by those | |
582 | * who aren't privileged to do so. | |
583 | */ | |
8f0cfa52 DH |
584 | int cap_inode_setxattr(struct dentry *dentry, const char *name, |
585 | const void *value, size_t size, int flags) | |
1da177e4 | 586 | { |
b5376771 SH |
587 | if (!strcmp(name, XATTR_NAME_CAPS)) { |
588 | if (!capable(CAP_SETFCAP)) | |
589 | return -EPERM; | |
590 | return 0; | |
1d045980 DH |
591 | } |
592 | ||
593 | if (!strncmp(name, XATTR_SECURITY_PREFIX, | |
c5b60b5e | 594 | sizeof(XATTR_SECURITY_PREFIX) - 1) && |
1da177e4 LT |
595 | !capable(CAP_SYS_ADMIN)) |
596 | return -EPERM; | |
597 | return 0; | |
598 | } | |
599 | ||
1d045980 DH |
600 | /** |
601 | * cap_inode_removexattr - Determine whether an xattr may be removed | |
602 | * @dentry: The inode/dentry being altered | |
603 | * @name: The name of the xattr to be changed | |
604 | * | |
605 | * Determine whether an xattr may be removed from an inode, returning 0 if | |
606 | * permission is granted, -ve if denied. | |
607 | * | |
608 | * This is used to make sure security xattrs don't get removed by those who | |
609 | * aren't privileged to remove them. | |
610 | */ | |
8f0cfa52 | 611 | int cap_inode_removexattr(struct dentry *dentry, const char *name) |
1da177e4 | 612 | { |
b5376771 SH |
613 | if (!strcmp(name, XATTR_NAME_CAPS)) { |
614 | if (!capable(CAP_SETFCAP)) | |
615 | return -EPERM; | |
616 | return 0; | |
1d045980 DH |
617 | } |
618 | ||
619 | if (!strncmp(name, XATTR_SECURITY_PREFIX, | |
c5b60b5e | 620 | sizeof(XATTR_SECURITY_PREFIX) - 1) && |
1da177e4 LT |
621 | !capable(CAP_SYS_ADMIN)) |
622 | return -EPERM; | |
623 | return 0; | |
624 | } | |
625 | ||
a6f76f23 | 626 | /* |
1da177e4 LT |
627 | * cap_emulate_setxuid() fixes the effective / permitted capabilities of |
628 | * a process after a call to setuid, setreuid, or setresuid. | |
629 | * | |
630 | * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of | |
631 | * {r,e,s}uid != 0, the permitted and effective capabilities are | |
632 | * cleared. | |
633 | * | |
634 | * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective | |
635 | * capabilities of the process are cleared. | |
636 | * | |
637 | * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective | |
638 | * capabilities are set to the permitted capabilities. | |
639 | * | |
a6f76f23 | 640 | * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should |
1da177e4 LT |
641 | * never happen. |
642 | * | |
a6f76f23 | 643 | * -astor |
1da177e4 LT |
644 | * |
645 | * cevans - New behaviour, Oct '99 | |
646 | * A process may, via prctl(), elect to keep its capabilities when it | |
647 | * calls setuid() and switches away from uid==0. Both permitted and | |
648 | * effective sets will be retained. | |
649 | * Without this change, it was impossible for a daemon to drop only some | |
650 | * of its privilege. The call to setuid(!=0) would drop all privileges! | |
651 | * Keeping uid 0 is not an option because uid 0 owns too many vital | |
652 | * files.. | |
653 | * Thanks to Olaf Kirch and Peter Benie for spotting this. | |
654 | */ | |
d84f4f99 | 655 | static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old) |
1da177e4 | 656 | { |
d84f4f99 DH |
657 | if ((old->uid == 0 || old->euid == 0 || old->suid == 0) && |
658 | (new->uid != 0 && new->euid != 0 && new->suid != 0) && | |
3898b1b4 | 659 | !issecure(SECURE_KEEP_CAPS)) { |
d84f4f99 DH |
660 | cap_clear(new->cap_permitted); |
661 | cap_clear(new->cap_effective); | |
1da177e4 | 662 | } |
d84f4f99 DH |
663 | if (old->euid == 0 && new->euid != 0) |
664 | cap_clear(new->cap_effective); | |
665 | if (old->euid != 0 && new->euid == 0) | |
666 | new->cap_effective = new->cap_permitted; | |
1da177e4 LT |
667 | } |
668 | ||
1d045980 DH |
669 | /** |
670 | * cap_task_fix_setuid - Fix up the results of setuid() call | |
671 | * @new: The proposed credentials | |
672 | * @old: The current task's current credentials | |
673 | * @flags: Indications of what has changed | |
674 | * | |
675 | * Fix up the results of setuid() call before the credential changes are | |
676 | * actually applied, returning 0 to grant the changes, -ve to deny them. | |
677 | */ | |
d84f4f99 | 678 | int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags) |
1da177e4 LT |
679 | { |
680 | switch (flags) { | |
681 | case LSM_SETID_RE: | |
682 | case LSM_SETID_ID: | |
683 | case LSM_SETID_RES: | |
1d045980 DH |
684 | /* juggle the capabilities to follow [RES]UID changes unless |
685 | * otherwise suppressed */ | |
d84f4f99 DH |
686 | if (!issecure(SECURE_NO_SETUID_FIXUP)) |
687 | cap_emulate_setxuid(new, old); | |
1da177e4 | 688 | break; |
1da177e4 | 689 | |
1d045980 DH |
690 | case LSM_SETID_FS: |
691 | /* juggle the capabilties to follow FSUID changes, unless | |
692 | * otherwise suppressed | |
693 | * | |
d84f4f99 DH |
694 | * FIXME - is fsuser used for all CAP_FS_MASK capabilities? |
695 | * if not, we might be a bit too harsh here. | |
696 | */ | |
697 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { | |
1d045980 | 698 | if (old->fsuid == 0 && new->fsuid != 0) |
d84f4f99 DH |
699 | new->cap_effective = |
700 | cap_drop_fs_set(new->cap_effective); | |
1d045980 DH |
701 | |
702 | if (old->fsuid != 0 && new->fsuid == 0) | |
d84f4f99 DH |
703 | new->cap_effective = |
704 | cap_raise_fs_set(new->cap_effective, | |
705 | new->cap_permitted); | |
1da177e4 | 706 | } |
d84f4f99 | 707 | break; |
1d045980 | 708 | |
1da177e4 LT |
709 | default: |
710 | return -EINVAL; | |
711 | } | |
712 | ||
713 | return 0; | |
714 | } | |
715 | ||
b5376771 SH |
716 | /* |
717 | * Rationale: code calling task_setscheduler, task_setioprio, and | |
718 | * task_setnice, assumes that | |
719 | * . if capable(cap_sys_nice), then those actions should be allowed | |
720 | * . if not capable(cap_sys_nice), but acting on your own processes, | |
721 | * then those actions should be allowed | |
722 | * This is insufficient now since you can call code without suid, but | |
723 | * yet with increased caps. | |
724 | * So we check for increased caps on the target process. | |
725 | */ | |
de45e806 | 726 | static int cap_safe_nice(struct task_struct *p) |
b5376771 | 727 | { |
c69e8d9c DH |
728 | int is_subset; |
729 | ||
730 | rcu_read_lock(); | |
731 | is_subset = cap_issubset(__task_cred(p)->cap_permitted, | |
732 | current_cred()->cap_permitted); | |
733 | rcu_read_unlock(); | |
734 | ||
735 | if (!is_subset && !capable(CAP_SYS_NICE)) | |
b5376771 SH |
736 | return -EPERM; |
737 | return 0; | |
738 | } | |
739 | ||
1d045980 DH |
740 | /** |
741 | * cap_task_setscheduler - Detemine if scheduler policy change is permitted | |
742 | * @p: The task to affect | |
1d045980 DH |
743 | * |
744 | * Detemine if the requested scheduler policy change is permitted for the | |
745 | * specified task, returning 0 if permission is granted, -ve if denied. | |
746 | */ | |
b0ae1981 | 747 | int cap_task_setscheduler(struct task_struct *p) |
b5376771 SH |
748 | { |
749 | return cap_safe_nice(p); | |
750 | } | |
751 | ||
1d045980 DH |
752 | /** |
753 | * cap_task_ioprio - Detemine if I/O priority change is permitted | |
754 | * @p: The task to affect | |
755 | * @ioprio: The I/O priority to set | |
756 | * | |
757 | * Detemine if the requested I/O priority change is permitted for the specified | |
758 | * task, returning 0 if permission is granted, -ve if denied. | |
759 | */ | |
760 | int cap_task_setioprio(struct task_struct *p, int ioprio) | |
b5376771 SH |
761 | { |
762 | return cap_safe_nice(p); | |
763 | } | |
764 | ||
1d045980 DH |
765 | /** |
766 | * cap_task_ioprio - Detemine if task priority change is permitted | |
767 | * @p: The task to affect | |
768 | * @nice: The nice value to set | |
769 | * | |
770 | * Detemine if the requested task priority change is permitted for the | |
771 | * specified task, returning 0 if permission is granted, -ve if denied. | |
772 | */ | |
773 | int cap_task_setnice(struct task_struct *p, int nice) | |
b5376771 SH |
774 | { |
775 | return cap_safe_nice(p); | |
776 | } | |
777 | ||
3b7391de | 778 | /* |
1d045980 DH |
779 | * Implement PR_CAPBSET_DROP. Attempt to remove the specified capability from |
780 | * the current task's bounding set. Returns 0 on success, -ve on error. | |
3b7391de | 781 | */ |
d84f4f99 | 782 | static long cap_prctl_drop(struct cred *new, unsigned long cap) |
3b7391de SH |
783 | { |
784 | if (!capable(CAP_SETPCAP)) | |
785 | return -EPERM; | |
786 | if (!cap_valid(cap)) | |
787 | return -EINVAL; | |
d84f4f99 DH |
788 | |
789 | cap_lower(new->cap_bset, cap); | |
3b7391de SH |
790 | return 0; |
791 | } | |
3898b1b4 | 792 | |
1d045980 DH |
793 | /** |
794 | * cap_task_prctl - Implement process control functions for this security module | |
795 | * @option: The process control function requested | |
796 | * @arg2, @arg3, @arg4, @arg5: The argument data for this function | |
797 | * | |
798 | * Allow process control functions (sys_prctl()) to alter capabilities; may | |
799 | * also deny access to other functions not otherwise implemented here. | |
800 | * | |
801 | * Returns 0 or +ve on success, -ENOSYS if this function is not implemented | |
802 | * here, other -ve on error. If -ENOSYS is returned, sys_prctl() and other LSM | |
803 | * modules will consider performing the function. | |
804 | */ | |
3898b1b4 | 805 | int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3, |
d84f4f99 | 806 | unsigned long arg4, unsigned long arg5) |
3898b1b4 | 807 | { |
d84f4f99 | 808 | struct cred *new; |
3898b1b4 AM |
809 | long error = 0; |
810 | ||
d84f4f99 DH |
811 | new = prepare_creds(); |
812 | if (!new) | |
813 | return -ENOMEM; | |
814 | ||
3898b1b4 AM |
815 | switch (option) { |
816 | case PR_CAPBSET_READ: | |
d84f4f99 | 817 | error = -EINVAL; |
3898b1b4 | 818 | if (!cap_valid(arg2)) |
d84f4f99 DH |
819 | goto error; |
820 | error = !!cap_raised(new->cap_bset, arg2); | |
821 | goto no_change; | |
822 | ||
3898b1b4 | 823 | case PR_CAPBSET_DROP: |
d84f4f99 DH |
824 | error = cap_prctl_drop(new, arg2); |
825 | if (error < 0) | |
826 | goto error; | |
827 | goto changed; | |
3898b1b4 AM |
828 | |
829 | /* | |
830 | * The next four prctl's remain to assist with transitioning a | |
831 | * system from legacy UID=0 based privilege (when filesystem | |
832 | * capabilities are not in use) to a system using filesystem | |
833 | * capabilities only - as the POSIX.1e draft intended. | |
834 | * | |
835 | * Note: | |
836 | * | |
837 | * PR_SET_SECUREBITS = | |
838 | * issecure_mask(SECURE_KEEP_CAPS_LOCKED) | |
839 | * | issecure_mask(SECURE_NOROOT) | |
840 | * | issecure_mask(SECURE_NOROOT_LOCKED) | |
841 | * | issecure_mask(SECURE_NO_SETUID_FIXUP) | |
842 | * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED) | |
843 | * | |
844 | * will ensure that the current process and all of its | |
845 | * children will be locked into a pure | |
846 | * capability-based-privilege environment. | |
847 | */ | |
848 | case PR_SET_SECUREBITS: | |
d84f4f99 DH |
849 | error = -EPERM; |
850 | if ((((new->securebits & SECURE_ALL_LOCKS) >> 1) | |
851 | & (new->securebits ^ arg2)) /*[1]*/ | |
852 | || ((new->securebits & SECURE_ALL_LOCKS & ~arg2)) /*[2]*/ | |
853 | || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS)) /*[3]*/ | |
3486740a SH |
854 | || (cap_capable(current, current_cred(), |
855 | current_cred()->user->user_ns, CAP_SETPCAP, | |
3699c53c | 856 | SECURITY_CAP_AUDIT) != 0) /*[4]*/ |
3898b1b4 AM |
857 | /* |
858 | * [1] no changing of bits that are locked | |
859 | * [2] no unlocking of locks | |
860 | * [3] no setting of unsupported bits | |
861 | * [4] doing anything requires privilege (go read about | |
862 | * the "sendmail capabilities bug") | |
863 | */ | |
d84f4f99 DH |
864 | ) |
865 | /* cannot change a locked bit */ | |
866 | goto error; | |
867 | new->securebits = arg2; | |
868 | goto changed; | |
869 | ||
3898b1b4 | 870 | case PR_GET_SECUREBITS: |
d84f4f99 DH |
871 | error = new->securebits; |
872 | goto no_change; | |
3898b1b4 | 873 | |
3898b1b4 AM |
874 | case PR_GET_KEEPCAPS: |
875 | if (issecure(SECURE_KEEP_CAPS)) | |
876 | error = 1; | |
d84f4f99 DH |
877 | goto no_change; |
878 | ||
3898b1b4 | 879 | case PR_SET_KEEPCAPS: |
d84f4f99 | 880 | error = -EINVAL; |
3898b1b4 | 881 | if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */ |
d84f4f99 DH |
882 | goto error; |
883 | error = -EPERM; | |
884 | if (issecure(SECURE_KEEP_CAPS_LOCKED)) | |
885 | goto error; | |
886 | if (arg2) | |
887 | new->securebits |= issecure_mask(SECURE_KEEP_CAPS); | |
3898b1b4 | 888 | else |
d84f4f99 DH |
889 | new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS); |
890 | goto changed; | |
3898b1b4 AM |
891 | |
892 | default: | |
893 | /* No functionality available - continue with default */ | |
d84f4f99 DH |
894 | error = -ENOSYS; |
895 | goto error; | |
3898b1b4 AM |
896 | } |
897 | ||
898 | /* Functionality provided */ | |
d84f4f99 DH |
899 | changed: |
900 | return commit_creds(new); | |
901 | ||
902 | no_change: | |
d84f4f99 DH |
903 | error: |
904 | abort_creds(new); | |
905 | return error; | |
1da177e4 LT |
906 | } |
907 | ||
1d045980 DH |
908 | /** |
909 | * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted | |
910 | * @mm: The VM space in which the new mapping is to be made | |
911 | * @pages: The size of the mapping | |
912 | * | |
913 | * Determine whether the allocation of a new virtual mapping by the current | |
914 | * task is permitted, returning 0 if permission is granted, -ve if not. | |
915 | */ | |
34b4e4aa | 916 | int cap_vm_enough_memory(struct mm_struct *mm, long pages) |
1da177e4 LT |
917 | { |
918 | int cap_sys_admin = 0; | |
919 | ||
3486740a | 920 | if (cap_capable(current, current_cred(), &init_user_ns, CAP_SYS_ADMIN, |
3699c53c | 921 | SECURITY_CAP_NOAUDIT) == 0) |
1da177e4 | 922 | cap_sys_admin = 1; |
34b4e4aa | 923 | return __vm_enough_memory(mm, pages, cap_sys_admin); |
1da177e4 | 924 | } |
7c73875e EP |
925 | |
926 | /* | |
927 | * cap_file_mmap - check if able to map given addr | |
928 | * @file: unused | |
929 | * @reqprot: unused | |
930 | * @prot: unused | |
931 | * @flags: unused | |
932 | * @addr: address attempting to be mapped | |
933 | * @addr_only: unused | |
934 | * | |
6f262d8e | 935 | * If the process is attempting to map memory below dac_mmap_min_addr they need |
7c73875e EP |
936 | * CAP_SYS_RAWIO. The other parameters to this function are unused by the |
937 | * capability security module. Returns 0 if this mapping should be allowed | |
938 | * -EPERM if not. | |
939 | */ | |
940 | int cap_file_mmap(struct file *file, unsigned long reqprot, | |
941 | unsigned long prot, unsigned long flags, | |
942 | unsigned long addr, unsigned long addr_only) | |
943 | { | |
944 | int ret = 0; | |
945 | ||
a2551df7 | 946 | if (addr < dac_mmap_min_addr) { |
3486740a | 947 | ret = cap_capable(current, current_cred(), &init_user_ns, CAP_SYS_RAWIO, |
7c73875e EP |
948 | SECURITY_CAP_AUDIT); |
949 | /* set PF_SUPERPRIV if it turns out we allow the low mmap */ | |
950 | if (ret == 0) | |
951 | current->flags |= PF_SUPERPRIV; | |
952 | } | |
953 | return ret; | |
954 | } |