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 | * | |
8409cca7 SH |
130 | * If we are in the same or an ancestor user_ns and have all the target |
131 | * task's capabilities, then ptrace access is allowed. | |
132 | * If we have the ptrace capability to the target user_ns, then ptrace | |
133 | * access is allowed. | |
134 | * Else denied. | |
135 | * | |
1d045980 DH |
136 | * Determine whether a process may access another, returning 0 if permission |
137 | * granted, -ve if denied. | |
138 | */ | |
9e48858f | 139 | int cap_ptrace_access_check(struct task_struct *child, unsigned int mode) |
1da177e4 | 140 | { |
c69e8d9c | 141 | int ret = 0; |
8409cca7 | 142 | const struct cred *cred, *child_cred; |
c69e8d9c DH |
143 | |
144 | rcu_read_lock(); | |
8409cca7 SH |
145 | cred = current_cred(); |
146 | child_cred = __task_cred(child); | |
147 | if (cred->user->user_ns == child_cred->user->user_ns && | |
148 | cap_issubset(child_cred->cap_permitted, cred->cap_permitted)) | |
149 | goto out; | |
150 | if (ns_capable(child_cred->user->user_ns, CAP_SYS_PTRACE)) | |
151 | goto out; | |
152 | ret = -EPERM; | |
153 | out: | |
c69e8d9c DH |
154 | rcu_read_unlock(); |
155 | return ret; | |
5cd9c58f DH |
156 | } |
157 | ||
1d045980 DH |
158 | /** |
159 | * cap_ptrace_traceme - Determine whether another process may trace the current | |
160 | * @parent: The task proposed to be the tracer | |
161 | * | |
8409cca7 SH |
162 | * If parent is in the same or an ancestor user_ns and has all current's |
163 | * capabilities, then ptrace access is allowed. | |
164 | * If parent has the ptrace capability to current's user_ns, then ptrace | |
165 | * access is allowed. | |
166 | * Else denied. | |
167 | * | |
1d045980 DH |
168 | * Determine whether the nominated task is permitted to trace the current |
169 | * process, returning 0 if permission is granted, -ve if denied. | |
170 | */ | |
5cd9c58f DH |
171 | int cap_ptrace_traceme(struct task_struct *parent) |
172 | { | |
c69e8d9c | 173 | int ret = 0; |
8409cca7 | 174 | const struct cred *cred, *child_cred; |
c69e8d9c DH |
175 | |
176 | rcu_read_lock(); | |
8409cca7 SH |
177 | cred = __task_cred(parent); |
178 | child_cred = current_cred(); | |
179 | if (cred->user->user_ns == child_cred->user->user_ns && | |
180 | cap_issubset(child_cred->cap_permitted, cred->cap_permitted)) | |
181 | goto out; | |
182 | if (has_ns_capability(parent, child_cred->user->user_ns, CAP_SYS_PTRACE)) | |
183 | goto out; | |
184 | ret = -EPERM; | |
185 | out: | |
c69e8d9c DH |
186 | rcu_read_unlock(); |
187 | return ret; | |
1da177e4 LT |
188 | } |
189 | ||
1d045980 DH |
190 | /** |
191 | * cap_capget - Retrieve a task's capability sets | |
192 | * @target: The task from which to retrieve the capability sets | |
193 | * @effective: The place to record the effective set | |
194 | * @inheritable: The place to record the inheritable set | |
195 | * @permitted: The place to record the permitted set | |
196 | * | |
197 | * This function retrieves the capabilities of the nominated task and returns | |
198 | * them to the caller. | |
199 | */ | |
200 | int cap_capget(struct task_struct *target, kernel_cap_t *effective, | |
201 | kernel_cap_t *inheritable, kernel_cap_t *permitted) | |
1da177e4 | 202 | { |
c69e8d9c | 203 | const struct cred *cred; |
b6dff3ec | 204 | |
1da177e4 | 205 | /* Derived from kernel/capability.c:sys_capget. */ |
c69e8d9c DH |
206 | rcu_read_lock(); |
207 | cred = __task_cred(target); | |
b6dff3ec DH |
208 | *effective = cred->cap_effective; |
209 | *inheritable = cred->cap_inheritable; | |
210 | *permitted = cred->cap_permitted; | |
c69e8d9c | 211 | rcu_read_unlock(); |
1da177e4 LT |
212 | return 0; |
213 | } | |
214 | ||
1d045980 DH |
215 | /* |
216 | * Determine whether the inheritable capabilities are limited to the old | |
217 | * permitted set. Returns 1 if they are limited, 0 if they are not. | |
218 | */ | |
72c2d582 AM |
219 | static inline int cap_inh_is_capped(void) |
220 | { | |
72c2d582 | 221 | |
1d045980 DH |
222 | /* they are so limited unless the current task has the CAP_SETPCAP |
223 | * capability | |
224 | */ | |
3486740a SH |
225 | if (cap_capable(current, current_cred(), |
226 | current_cred()->user->user_ns, CAP_SETPCAP, | |
3699c53c | 227 | SECURITY_CAP_AUDIT) == 0) |
1d045980 | 228 | return 0; |
1d045980 | 229 | return 1; |
1209726c | 230 | } |
72c2d582 | 231 | |
1d045980 DH |
232 | /** |
233 | * cap_capset - Validate and apply proposed changes to current's capabilities | |
234 | * @new: The proposed new credentials; alterations should be made here | |
235 | * @old: The current task's current credentials | |
236 | * @effective: A pointer to the proposed new effective capabilities set | |
237 | * @inheritable: A pointer to the proposed new inheritable capabilities set | |
238 | * @permitted: A pointer to the proposed new permitted capabilities set | |
239 | * | |
240 | * This function validates and applies a proposed mass change to the current | |
241 | * process's capability sets. The changes are made to the proposed new | |
242 | * credentials, and assuming no error, will be committed by the caller of LSM. | |
243 | */ | |
d84f4f99 DH |
244 | int cap_capset(struct cred *new, |
245 | const struct cred *old, | |
246 | const kernel_cap_t *effective, | |
247 | const kernel_cap_t *inheritable, | |
248 | const kernel_cap_t *permitted) | |
1da177e4 | 249 | { |
d84f4f99 DH |
250 | if (cap_inh_is_capped() && |
251 | !cap_issubset(*inheritable, | |
252 | cap_combine(old->cap_inheritable, | |
253 | old->cap_permitted))) | |
72c2d582 | 254 | /* incapable of using this inheritable set */ |
1da177e4 | 255 | return -EPERM; |
d84f4f99 | 256 | |
3b7391de | 257 | if (!cap_issubset(*inheritable, |
d84f4f99 DH |
258 | cap_combine(old->cap_inheritable, |
259 | old->cap_bset))) | |
3b7391de SH |
260 | /* no new pI capabilities outside bounding set */ |
261 | return -EPERM; | |
1da177e4 LT |
262 | |
263 | /* verify restrictions on target's new Permitted set */ | |
d84f4f99 | 264 | if (!cap_issubset(*permitted, old->cap_permitted)) |
1da177e4 | 265 | return -EPERM; |
1da177e4 LT |
266 | |
267 | /* verify the _new_Effective_ is a subset of the _new_Permitted_ */ | |
d84f4f99 | 268 | if (!cap_issubset(*effective, *permitted)) |
1da177e4 | 269 | return -EPERM; |
1da177e4 | 270 | |
d84f4f99 DH |
271 | new->cap_effective = *effective; |
272 | new->cap_inheritable = *inheritable; | |
273 | new->cap_permitted = *permitted; | |
1da177e4 LT |
274 | return 0; |
275 | } | |
276 | ||
1d045980 DH |
277 | /* |
278 | * Clear proposed capability sets for execve(). | |
279 | */ | |
b5376771 SH |
280 | static inline void bprm_clear_caps(struct linux_binprm *bprm) |
281 | { | |
a6f76f23 | 282 | cap_clear(bprm->cred->cap_permitted); |
b5376771 SH |
283 | bprm->cap_effective = false; |
284 | } | |
285 | ||
1d045980 DH |
286 | /** |
287 | * cap_inode_need_killpriv - Determine if inode change affects privileges | |
288 | * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV | |
289 | * | |
290 | * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV | |
291 | * affects the security markings on that inode, and if it is, should | |
292 | * inode_killpriv() be invoked or the change rejected? | |
293 | * | |
294 | * Returns 0 if granted; +ve if granted, but inode_killpriv() is required; and | |
295 | * -ve to deny the change. | |
296 | */ | |
b5376771 SH |
297 | int cap_inode_need_killpriv(struct dentry *dentry) |
298 | { | |
299 | struct inode *inode = dentry->d_inode; | |
300 | int error; | |
301 | ||
acfa4380 | 302 | if (!inode->i_op->getxattr) |
b5376771 SH |
303 | return 0; |
304 | ||
305 | error = inode->i_op->getxattr(dentry, XATTR_NAME_CAPS, NULL, 0); | |
306 | if (error <= 0) | |
307 | return 0; | |
308 | return 1; | |
309 | } | |
310 | ||
1d045980 DH |
311 | /** |
312 | * cap_inode_killpriv - Erase the security markings on an inode | |
313 | * @dentry: The inode/dentry to alter | |
314 | * | |
315 | * Erase the privilege-enhancing security markings on an inode. | |
316 | * | |
317 | * Returns 0 if successful, -ve on error. | |
318 | */ | |
b5376771 SH |
319 | int cap_inode_killpriv(struct dentry *dentry) |
320 | { | |
321 | struct inode *inode = dentry->d_inode; | |
322 | ||
acfa4380 | 323 | if (!inode->i_op->removexattr) |
b5376771 SH |
324 | return 0; |
325 | ||
326 | return inode->i_op->removexattr(dentry, XATTR_NAME_CAPS); | |
327 | } | |
328 | ||
1d045980 DH |
329 | /* |
330 | * Calculate the new process capability sets from the capability sets attached | |
331 | * to a file. | |
332 | */ | |
c0b00441 | 333 | static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps, |
a6f76f23 | 334 | struct linux_binprm *bprm, |
4d49f671 ZL |
335 | bool *effective, |
336 | bool *has_cap) | |
b5376771 | 337 | { |
a6f76f23 | 338 | struct cred *new = bprm->cred; |
c0b00441 EP |
339 | unsigned i; |
340 | int ret = 0; | |
341 | ||
342 | if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE) | |
a6f76f23 | 343 | *effective = true; |
c0b00441 | 344 | |
4d49f671 ZL |
345 | if (caps->magic_etc & VFS_CAP_REVISION_MASK) |
346 | *has_cap = true; | |
347 | ||
c0b00441 EP |
348 | CAP_FOR_EACH_U32(i) { |
349 | __u32 permitted = caps->permitted.cap[i]; | |
350 | __u32 inheritable = caps->inheritable.cap[i]; | |
351 | ||
352 | /* | |
353 | * pP' = (X & fP) | (pI & fI) | |
354 | */ | |
a6f76f23 DH |
355 | new->cap_permitted.cap[i] = |
356 | (new->cap_bset.cap[i] & permitted) | | |
357 | (new->cap_inheritable.cap[i] & inheritable); | |
c0b00441 | 358 | |
a6f76f23 DH |
359 | if (permitted & ~new->cap_permitted.cap[i]) |
360 | /* insufficient to execute correctly */ | |
c0b00441 | 361 | ret = -EPERM; |
c0b00441 EP |
362 | } |
363 | ||
364 | /* | |
365 | * For legacy apps, with no internal support for recognizing they | |
366 | * do not have enough capabilities, we return an error if they are | |
367 | * missing some "forced" (aka file-permitted) capabilities. | |
368 | */ | |
a6f76f23 | 369 | return *effective ? ret : 0; |
c0b00441 EP |
370 | } |
371 | ||
1d045980 DH |
372 | /* |
373 | * Extract the on-exec-apply capability sets for an executable file. | |
374 | */ | |
c0b00441 EP |
375 | int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps) |
376 | { | |
377 | struct inode *inode = dentry->d_inode; | |
b5376771 | 378 | __u32 magic_etc; |
e338d263 | 379 | unsigned tocopy, i; |
c0b00441 EP |
380 | int size; |
381 | struct vfs_cap_data caps; | |
382 | ||
383 | memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data)); | |
384 | ||
acfa4380 | 385 | if (!inode || !inode->i_op->getxattr) |
c0b00441 EP |
386 | return -ENODATA; |
387 | ||
388 | size = inode->i_op->getxattr((struct dentry *)dentry, XATTR_NAME_CAPS, &caps, | |
389 | XATTR_CAPS_SZ); | |
a6f76f23 | 390 | if (size == -ENODATA || size == -EOPNOTSUPP) |
c0b00441 EP |
391 | /* no data, that's ok */ |
392 | return -ENODATA; | |
c0b00441 EP |
393 | if (size < 0) |
394 | return size; | |
b5376771 | 395 | |
e338d263 | 396 | if (size < sizeof(magic_etc)) |
b5376771 SH |
397 | return -EINVAL; |
398 | ||
c0b00441 | 399 | cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps.magic_etc); |
b5376771 | 400 | |
a6f76f23 | 401 | switch (magic_etc & VFS_CAP_REVISION_MASK) { |
e338d263 AM |
402 | case VFS_CAP_REVISION_1: |
403 | if (size != XATTR_CAPS_SZ_1) | |
404 | return -EINVAL; | |
405 | tocopy = VFS_CAP_U32_1; | |
406 | break; | |
407 | case VFS_CAP_REVISION_2: | |
408 | if (size != XATTR_CAPS_SZ_2) | |
409 | return -EINVAL; | |
410 | tocopy = VFS_CAP_U32_2; | |
411 | break; | |
b5376771 SH |
412 | default: |
413 | return -EINVAL; | |
414 | } | |
e338d263 | 415 | |
5459c164 | 416 | CAP_FOR_EACH_U32(i) { |
c0b00441 EP |
417 | if (i >= tocopy) |
418 | break; | |
419 | cpu_caps->permitted.cap[i] = le32_to_cpu(caps.data[i].permitted); | |
420 | cpu_caps->inheritable.cap[i] = le32_to_cpu(caps.data[i].inheritable); | |
e338d263 | 421 | } |
a6f76f23 | 422 | |
c0b00441 | 423 | return 0; |
b5376771 SH |
424 | } |
425 | ||
1d045980 DH |
426 | /* |
427 | * Attempt to get the on-exec apply capability sets for an executable file from | |
428 | * its xattrs and, if present, apply them to the proposed credentials being | |
429 | * constructed by execve(). | |
430 | */ | |
4d49f671 | 431 | static int get_file_caps(struct linux_binprm *bprm, bool *effective, bool *has_cap) |
b5376771 SH |
432 | { |
433 | struct dentry *dentry; | |
434 | int rc = 0; | |
c0b00441 | 435 | struct cpu_vfs_cap_data vcaps; |
b5376771 | 436 | |
3318a386 SH |
437 | bprm_clear_caps(bprm); |
438 | ||
1f29fae2 SH |
439 | if (!file_caps_enabled) |
440 | return 0; | |
441 | ||
3318a386 | 442 | if (bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID) |
b5376771 | 443 | return 0; |
b5376771 SH |
444 | |
445 | dentry = dget(bprm->file->f_dentry); | |
b5376771 | 446 | |
c0b00441 EP |
447 | rc = get_vfs_caps_from_disk(dentry, &vcaps); |
448 | if (rc < 0) { | |
449 | if (rc == -EINVAL) | |
450 | printk(KERN_NOTICE "%s: get_vfs_caps_from_disk returned %d for %s\n", | |
451 | __func__, rc, bprm->filename); | |
452 | else if (rc == -ENODATA) | |
453 | rc = 0; | |
b5376771 SH |
454 | goto out; |
455 | } | |
b5376771 | 456 | |
4d49f671 | 457 | rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective, has_cap); |
a6f76f23 DH |
458 | if (rc == -EINVAL) |
459 | printk(KERN_NOTICE "%s: cap_from_disk returned %d for %s\n", | |
460 | __func__, rc, bprm->filename); | |
b5376771 SH |
461 | |
462 | out: | |
463 | dput(dentry); | |
464 | if (rc) | |
465 | bprm_clear_caps(bprm); | |
466 | ||
467 | return rc; | |
468 | } | |
469 | ||
1d045980 DH |
470 | /** |
471 | * cap_bprm_set_creds - Set up the proposed credentials for execve(). | |
472 | * @bprm: The execution parameters, including the proposed creds | |
473 | * | |
474 | * Set up the proposed credentials for a new execution context being | |
475 | * constructed by execve(). The proposed creds in @bprm->cred is altered, | |
476 | * which won't take effect immediately. Returns 0 if successful, -ve on error. | |
a6f76f23 DH |
477 | */ |
478 | int cap_bprm_set_creds(struct linux_binprm *bprm) | |
1da177e4 | 479 | { |
a6f76f23 DH |
480 | const struct cred *old = current_cred(); |
481 | struct cred *new = bprm->cred; | |
7d8db180 | 482 | bool effective, has_cap = false; |
b5376771 | 483 | int ret; |
1da177e4 | 484 | |
a6f76f23 | 485 | effective = false; |
4d49f671 | 486 | ret = get_file_caps(bprm, &effective, &has_cap); |
a6f76f23 DH |
487 | if (ret < 0) |
488 | return ret; | |
1da177e4 | 489 | |
5459c164 | 490 | if (!issecure(SECURE_NOROOT)) { |
b5f22a59 SH |
491 | /* |
492 | * If the legacy file capability is set, then don't set privs | |
493 | * for a setuid root binary run by a non-root user. Do set it | |
494 | * for a root user just to cause least surprise to an admin. | |
495 | */ | |
4d49f671 | 496 | if (has_cap && new->uid != 0 && new->euid == 0) { |
b5f22a59 SH |
497 | warn_setuid_and_fcaps_mixed(bprm->filename); |
498 | goto skip; | |
499 | } | |
5459c164 AM |
500 | /* |
501 | * To support inheritance of root-permissions and suid-root | |
502 | * executables under compatibility mode, we override the | |
503 | * capability sets for the file. | |
504 | * | |
a6f76f23 | 505 | * If only the real uid is 0, we do not set the effective bit. |
5459c164 | 506 | */ |
a6f76f23 | 507 | if (new->euid == 0 || new->uid == 0) { |
5459c164 | 508 | /* pP' = (cap_bset & ~0) | (pI & ~0) */ |
a6f76f23 DH |
509 | new->cap_permitted = cap_combine(old->cap_bset, |
510 | old->cap_inheritable); | |
1da177e4 | 511 | } |
a6f76f23 DH |
512 | if (new->euid == 0) |
513 | effective = true; | |
1da177e4 | 514 | } |
b5f22a59 | 515 | skip: |
b5376771 | 516 | |
a6f76f23 DH |
517 | /* Don't let someone trace a set[ug]id/setpcap binary with the revised |
518 | * credentials unless they have the appropriate permit | |
519 | */ | |
520 | if ((new->euid != old->uid || | |
521 | new->egid != old->gid || | |
522 | !cap_issubset(new->cap_permitted, old->cap_permitted)) && | |
523 | bprm->unsafe & ~LSM_UNSAFE_PTRACE_CAP) { | |
524 | /* downgrade; they get no more than they had, and maybe less */ | |
525 | if (!capable(CAP_SETUID)) { | |
526 | new->euid = new->uid; | |
527 | new->egid = new->gid; | |
1da177e4 | 528 | } |
b3a222e5 SH |
529 | new->cap_permitted = cap_intersect(new->cap_permitted, |
530 | old->cap_permitted); | |
1da177e4 LT |
531 | } |
532 | ||
a6f76f23 DH |
533 | new->suid = new->fsuid = new->euid; |
534 | new->sgid = new->fsgid = new->egid; | |
1da177e4 | 535 | |
4bf2ea77 EP |
536 | if (effective) |
537 | new->cap_effective = new->cap_permitted; | |
538 | else | |
539 | cap_clear(new->cap_effective); | |
a6f76f23 | 540 | bprm->cap_effective = effective; |
1da177e4 | 541 | |
3fc689e9 EP |
542 | /* |
543 | * Audit candidate if current->cap_effective is set | |
544 | * | |
545 | * We do not bother to audit if 3 things are true: | |
546 | * 1) cap_effective has all caps | |
547 | * 2) we are root | |
548 | * 3) root is supposed to have all caps (SECURE_NOROOT) | |
549 | * Since this is just a normal root execing a process. | |
550 | * | |
551 | * Number 1 above might fail if you don't have a full bset, but I think | |
552 | * that is interesting information to audit. | |
553 | */ | |
d84f4f99 DH |
554 | if (!cap_isclear(new->cap_effective)) { |
555 | if (!cap_issubset(CAP_FULL_SET, new->cap_effective) || | |
a6f76f23 DH |
556 | new->euid != 0 || new->uid != 0 || |
557 | issecure(SECURE_NOROOT)) { | |
558 | ret = audit_log_bprm_fcaps(bprm, new, old); | |
559 | if (ret < 0) | |
560 | return ret; | |
561 | } | |
3fc689e9 | 562 | } |
1da177e4 | 563 | |
d84f4f99 | 564 | new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS); |
a6f76f23 | 565 | return 0; |
1da177e4 LT |
566 | } |
567 | ||
1d045980 DH |
568 | /** |
569 | * cap_bprm_secureexec - Determine whether a secure execution is required | |
570 | * @bprm: The execution parameters | |
571 | * | |
572 | * Determine whether a secure execution is required, return 1 if it is, and 0 | |
573 | * if it is not. | |
574 | * | |
575 | * The credentials have been committed by this point, and so are no longer | |
576 | * available through @bprm->cred. | |
a6f76f23 DH |
577 | */ |
578 | int cap_bprm_secureexec(struct linux_binprm *bprm) | |
1da177e4 | 579 | { |
c69e8d9c | 580 | const struct cred *cred = current_cred(); |
b6dff3ec DH |
581 | |
582 | if (cred->uid != 0) { | |
b5376771 SH |
583 | if (bprm->cap_effective) |
584 | return 1; | |
a6f76f23 | 585 | if (!cap_isclear(cred->cap_permitted)) |
b5376771 SH |
586 | return 1; |
587 | } | |
588 | ||
b6dff3ec DH |
589 | return (cred->euid != cred->uid || |
590 | cred->egid != cred->gid); | |
1da177e4 LT |
591 | } |
592 | ||
1d045980 DH |
593 | /** |
594 | * cap_inode_setxattr - Determine whether an xattr may be altered | |
595 | * @dentry: The inode/dentry being altered | |
596 | * @name: The name of the xattr to be changed | |
597 | * @value: The value that the xattr will be changed to | |
598 | * @size: The size of value | |
599 | * @flags: The replacement flag | |
600 | * | |
601 | * Determine whether an xattr may be altered or set on an inode, returning 0 if | |
602 | * permission is granted, -ve if denied. | |
603 | * | |
604 | * This is used to make sure security xattrs don't get updated or set by those | |
605 | * who aren't privileged to do so. | |
606 | */ | |
8f0cfa52 DH |
607 | int cap_inode_setxattr(struct dentry *dentry, const char *name, |
608 | const void *value, size_t size, int flags) | |
1da177e4 | 609 | { |
b5376771 SH |
610 | if (!strcmp(name, XATTR_NAME_CAPS)) { |
611 | if (!capable(CAP_SETFCAP)) | |
612 | return -EPERM; | |
613 | return 0; | |
1d045980 DH |
614 | } |
615 | ||
616 | if (!strncmp(name, XATTR_SECURITY_PREFIX, | |
c5b60b5e | 617 | sizeof(XATTR_SECURITY_PREFIX) - 1) && |
1da177e4 LT |
618 | !capable(CAP_SYS_ADMIN)) |
619 | return -EPERM; | |
620 | return 0; | |
621 | } | |
622 | ||
1d045980 DH |
623 | /** |
624 | * cap_inode_removexattr - Determine whether an xattr may be removed | |
625 | * @dentry: The inode/dentry being altered | |
626 | * @name: The name of the xattr to be changed | |
627 | * | |
628 | * Determine whether an xattr may be removed from an inode, returning 0 if | |
629 | * permission is granted, -ve if denied. | |
630 | * | |
631 | * This is used to make sure security xattrs don't get removed by those who | |
632 | * aren't privileged to remove them. | |
633 | */ | |
8f0cfa52 | 634 | int cap_inode_removexattr(struct dentry *dentry, const char *name) |
1da177e4 | 635 | { |
b5376771 SH |
636 | if (!strcmp(name, XATTR_NAME_CAPS)) { |
637 | if (!capable(CAP_SETFCAP)) | |
638 | return -EPERM; | |
639 | return 0; | |
1d045980 DH |
640 | } |
641 | ||
642 | if (!strncmp(name, XATTR_SECURITY_PREFIX, | |
c5b60b5e | 643 | sizeof(XATTR_SECURITY_PREFIX) - 1) && |
1da177e4 LT |
644 | !capable(CAP_SYS_ADMIN)) |
645 | return -EPERM; | |
646 | return 0; | |
647 | } | |
648 | ||
a6f76f23 | 649 | /* |
1da177e4 LT |
650 | * cap_emulate_setxuid() fixes the effective / permitted capabilities of |
651 | * a process after a call to setuid, setreuid, or setresuid. | |
652 | * | |
653 | * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of | |
654 | * {r,e,s}uid != 0, the permitted and effective capabilities are | |
655 | * cleared. | |
656 | * | |
657 | * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective | |
658 | * capabilities of the process are cleared. | |
659 | * | |
660 | * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective | |
661 | * capabilities are set to the permitted capabilities. | |
662 | * | |
a6f76f23 | 663 | * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should |
1da177e4 LT |
664 | * never happen. |
665 | * | |
a6f76f23 | 666 | * -astor |
1da177e4 LT |
667 | * |
668 | * cevans - New behaviour, Oct '99 | |
669 | * A process may, via prctl(), elect to keep its capabilities when it | |
670 | * calls setuid() and switches away from uid==0. Both permitted and | |
671 | * effective sets will be retained. | |
672 | * Without this change, it was impossible for a daemon to drop only some | |
673 | * of its privilege. The call to setuid(!=0) would drop all privileges! | |
674 | * Keeping uid 0 is not an option because uid 0 owns too many vital | |
675 | * files.. | |
676 | * Thanks to Olaf Kirch and Peter Benie for spotting this. | |
677 | */ | |
d84f4f99 | 678 | static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old) |
1da177e4 | 679 | { |
d84f4f99 DH |
680 | if ((old->uid == 0 || old->euid == 0 || old->suid == 0) && |
681 | (new->uid != 0 && new->euid != 0 && new->suid != 0) && | |
3898b1b4 | 682 | !issecure(SECURE_KEEP_CAPS)) { |
d84f4f99 DH |
683 | cap_clear(new->cap_permitted); |
684 | cap_clear(new->cap_effective); | |
1da177e4 | 685 | } |
d84f4f99 DH |
686 | if (old->euid == 0 && new->euid != 0) |
687 | cap_clear(new->cap_effective); | |
688 | if (old->euid != 0 && new->euid == 0) | |
689 | new->cap_effective = new->cap_permitted; | |
1da177e4 LT |
690 | } |
691 | ||
1d045980 DH |
692 | /** |
693 | * cap_task_fix_setuid - Fix up the results of setuid() call | |
694 | * @new: The proposed credentials | |
695 | * @old: The current task's current credentials | |
696 | * @flags: Indications of what has changed | |
697 | * | |
698 | * Fix up the results of setuid() call before the credential changes are | |
699 | * actually applied, returning 0 to grant the changes, -ve to deny them. | |
700 | */ | |
d84f4f99 | 701 | int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags) |
1da177e4 LT |
702 | { |
703 | switch (flags) { | |
704 | case LSM_SETID_RE: | |
705 | case LSM_SETID_ID: | |
706 | case LSM_SETID_RES: | |
1d045980 DH |
707 | /* juggle the capabilities to follow [RES]UID changes unless |
708 | * otherwise suppressed */ | |
d84f4f99 DH |
709 | if (!issecure(SECURE_NO_SETUID_FIXUP)) |
710 | cap_emulate_setxuid(new, old); | |
1da177e4 | 711 | break; |
1da177e4 | 712 | |
1d045980 DH |
713 | case LSM_SETID_FS: |
714 | /* juggle the capabilties to follow FSUID changes, unless | |
715 | * otherwise suppressed | |
716 | * | |
d84f4f99 DH |
717 | * FIXME - is fsuser used for all CAP_FS_MASK capabilities? |
718 | * if not, we might be a bit too harsh here. | |
719 | */ | |
720 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { | |
1d045980 | 721 | if (old->fsuid == 0 && new->fsuid != 0) |
d84f4f99 DH |
722 | new->cap_effective = |
723 | cap_drop_fs_set(new->cap_effective); | |
1d045980 DH |
724 | |
725 | if (old->fsuid != 0 && new->fsuid == 0) | |
d84f4f99 DH |
726 | new->cap_effective = |
727 | cap_raise_fs_set(new->cap_effective, | |
728 | new->cap_permitted); | |
1da177e4 | 729 | } |
d84f4f99 | 730 | break; |
1d045980 | 731 | |
1da177e4 LT |
732 | default: |
733 | return -EINVAL; | |
734 | } | |
735 | ||
736 | return 0; | |
737 | } | |
738 | ||
b5376771 SH |
739 | /* |
740 | * Rationale: code calling task_setscheduler, task_setioprio, and | |
741 | * task_setnice, assumes that | |
742 | * . if capable(cap_sys_nice), then those actions should be allowed | |
743 | * . if not capable(cap_sys_nice), but acting on your own processes, | |
744 | * then those actions should be allowed | |
745 | * This is insufficient now since you can call code without suid, but | |
746 | * yet with increased caps. | |
747 | * So we check for increased caps on the target process. | |
748 | */ | |
de45e806 | 749 | static int cap_safe_nice(struct task_struct *p) |
b5376771 | 750 | { |
c69e8d9c DH |
751 | int is_subset; |
752 | ||
753 | rcu_read_lock(); | |
754 | is_subset = cap_issubset(__task_cred(p)->cap_permitted, | |
755 | current_cred()->cap_permitted); | |
756 | rcu_read_unlock(); | |
757 | ||
758 | if (!is_subset && !capable(CAP_SYS_NICE)) | |
b5376771 SH |
759 | return -EPERM; |
760 | return 0; | |
761 | } | |
762 | ||
1d045980 DH |
763 | /** |
764 | * cap_task_setscheduler - Detemine if scheduler policy change is permitted | |
765 | * @p: The task to affect | |
1d045980 DH |
766 | * |
767 | * Detemine if the requested scheduler policy change is permitted for the | |
768 | * specified task, returning 0 if permission is granted, -ve if denied. | |
769 | */ | |
b0ae1981 | 770 | int cap_task_setscheduler(struct task_struct *p) |
b5376771 SH |
771 | { |
772 | return cap_safe_nice(p); | |
773 | } | |
774 | ||
1d045980 DH |
775 | /** |
776 | * cap_task_ioprio - Detemine if I/O priority change is permitted | |
777 | * @p: The task to affect | |
778 | * @ioprio: The I/O priority to set | |
779 | * | |
780 | * Detemine if the requested I/O priority change is permitted for the specified | |
781 | * task, returning 0 if permission is granted, -ve if denied. | |
782 | */ | |
783 | int cap_task_setioprio(struct task_struct *p, int ioprio) | |
b5376771 SH |
784 | { |
785 | return cap_safe_nice(p); | |
786 | } | |
787 | ||
1d045980 DH |
788 | /** |
789 | * cap_task_ioprio - Detemine if task priority change is permitted | |
790 | * @p: The task to affect | |
791 | * @nice: The nice value to set | |
792 | * | |
793 | * Detemine if the requested task priority change is permitted for the | |
794 | * specified task, returning 0 if permission is granted, -ve if denied. | |
795 | */ | |
796 | int cap_task_setnice(struct task_struct *p, int nice) | |
b5376771 SH |
797 | { |
798 | return cap_safe_nice(p); | |
799 | } | |
800 | ||
3b7391de | 801 | /* |
1d045980 DH |
802 | * Implement PR_CAPBSET_DROP. Attempt to remove the specified capability from |
803 | * the current task's bounding set. Returns 0 on success, -ve on error. | |
3b7391de | 804 | */ |
d84f4f99 | 805 | static long cap_prctl_drop(struct cred *new, unsigned long cap) |
3b7391de SH |
806 | { |
807 | if (!capable(CAP_SETPCAP)) | |
808 | return -EPERM; | |
809 | if (!cap_valid(cap)) | |
810 | return -EINVAL; | |
d84f4f99 DH |
811 | |
812 | cap_lower(new->cap_bset, cap); | |
3b7391de SH |
813 | return 0; |
814 | } | |
3898b1b4 | 815 | |
1d045980 DH |
816 | /** |
817 | * cap_task_prctl - Implement process control functions for this security module | |
818 | * @option: The process control function requested | |
819 | * @arg2, @arg3, @arg4, @arg5: The argument data for this function | |
820 | * | |
821 | * Allow process control functions (sys_prctl()) to alter capabilities; may | |
822 | * also deny access to other functions not otherwise implemented here. | |
823 | * | |
824 | * Returns 0 or +ve on success, -ENOSYS if this function is not implemented | |
825 | * here, other -ve on error. If -ENOSYS is returned, sys_prctl() and other LSM | |
826 | * modules will consider performing the function. | |
827 | */ | |
3898b1b4 | 828 | int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3, |
d84f4f99 | 829 | unsigned long arg4, unsigned long arg5) |
3898b1b4 | 830 | { |
d84f4f99 | 831 | struct cred *new; |
3898b1b4 AM |
832 | long error = 0; |
833 | ||
d84f4f99 DH |
834 | new = prepare_creds(); |
835 | if (!new) | |
836 | return -ENOMEM; | |
837 | ||
3898b1b4 AM |
838 | switch (option) { |
839 | case PR_CAPBSET_READ: | |
d84f4f99 | 840 | error = -EINVAL; |
3898b1b4 | 841 | if (!cap_valid(arg2)) |
d84f4f99 DH |
842 | goto error; |
843 | error = !!cap_raised(new->cap_bset, arg2); | |
844 | goto no_change; | |
845 | ||
3898b1b4 | 846 | case PR_CAPBSET_DROP: |
d84f4f99 DH |
847 | error = cap_prctl_drop(new, arg2); |
848 | if (error < 0) | |
849 | goto error; | |
850 | goto changed; | |
3898b1b4 AM |
851 | |
852 | /* | |
853 | * The next four prctl's remain to assist with transitioning a | |
854 | * system from legacy UID=0 based privilege (when filesystem | |
855 | * capabilities are not in use) to a system using filesystem | |
856 | * capabilities only - as the POSIX.1e draft intended. | |
857 | * | |
858 | * Note: | |
859 | * | |
860 | * PR_SET_SECUREBITS = | |
861 | * issecure_mask(SECURE_KEEP_CAPS_LOCKED) | |
862 | * | issecure_mask(SECURE_NOROOT) | |
863 | * | issecure_mask(SECURE_NOROOT_LOCKED) | |
864 | * | issecure_mask(SECURE_NO_SETUID_FIXUP) | |
865 | * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED) | |
866 | * | |
867 | * will ensure that the current process and all of its | |
868 | * children will be locked into a pure | |
869 | * capability-based-privilege environment. | |
870 | */ | |
871 | case PR_SET_SECUREBITS: | |
d84f4f99 DH |
872 | error = -EPERM; |
873 | if ((((new->securebits & SECURE_ALL_LOCKS) >> 1) | |
874 | & (new->securebits ^ arg2)) /*[1]*/ | |
875 | || ((new->securebits & SECURE_ALL_LOCKS & ~arg2)) /*[2]*/ | |
876 | || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS)) /*[3]*/ | |
3486740a SH |
877 | || (cap_capable(current, current_cred(), |
878 | current_cred()->user->user_ns, CAP_SETPCAP, | |
3699c53c | 879 | SECURITY_CAP_AUDIT) != 0) /*[4]*/ |
3898b1b4 AM |
880 | /* |
881 | * [1] no changing of bits that are locked | |
882 | * [2] no unlocking of locks | |
883 | * [3] no setting of unsupported bits | |
884 | * [4] doing anything requires privilege (go read about | |
885 | * the "sendmail capabilities bug") | |
886 | */ | |
d84f4f99 DH |
887 | ) |
888 | /* cannot change a locked bit */ | |
889 | goto error; | |
890 | new->securebits = arg2; | |
891 | goto changed; | |
892 | ||
3898b1b4 | 893 | case PR_GET_SECUREBITS: |
d84f4f99 DH |
894 | error = new->securebits; |
895 | goto no_change; | |
3898b1b4 | 896 | |
3898b1b4 AM |
897 | case PR_GET_KEEPCAPS: |
898 | if (issecure(SECURE_KEEP_CAPS)) | |
899 | error = 1; | |
d84f4f99 DH |
900 | goto no_change; |
901 | ||
3898b1b4 | 902 | case PR_SET_KEEPCAPS: |
d84f4f99 | 903 | error = -EINVAL; |
3898b1b4 | 904 | if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */ |
d84f4f99 DH |
905 | goto error; |
906 | error = -EPERM; | |
907 | if (issecure(SECURE_KEEP_CAPS_LOCKED)) | |
908 | goto error; | |
909 | if (arg2) | |
910 | new->securebits |= issecure_mask(SECURE_KEEP_CAPS); | |
3898b1b4 | 911 | else |
d84f4f99 DH |
912 | new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS); |
913 | goto changed; | |
3898b1b4 AM |
914 | |
915 | default: | |
916 | /* No functionality available - continue with default */ | |
d84f4f99 DH |
917 | error = -ENOSYS; |
918 | goto error; | |
3898b1b4 AM |
919 | } |
920 | ||
921 | /* Functionality provided */ | |
d84f4f99 DH |
922 | changed: |
923 | return commit_creds(new); | |
924 | ||
925 | no_change: | |
d84f4f99 DH |
926 | error: |
927 | abort_creds(new); | |
928 | return error; | |
1da177e4 LT |
929 | } |
930 | ||
1d045980 DH |
931 | /** |
932 | * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted | |
933 | * @mm: The VM space in which the new mapping is to be made | |
934 | * @pages: The size of the mapping | |
935 | * | |
936 | * Determine whether the allocation of a new virtual mapping by the current | |
937 | * task is permitted, returning 0 if permission is granted, -ve if not. | |
938 | */ | |
34b4e4aa | 939 | int cap_vm_enough_memory(struct mm_struct *mm, long pages) |
1da177e4 LT |
940 | { |
941 | int cap_sys_admin = 0; | |
942 | ||
3486740a | 943 | if (cap_capable(current, current_cred(), &init_user_ns, CAP_SYS_ADMIN, |
3699c53c | 944 | SECURITY_CAP_NOAUDIT) == 0) |
1da177e4 | 945 | cap_sys_admin = 1; |
34b4e4aa | 946 | return __vm_enough_memory(mm, pages, cap_sys_admin); |
1da177e4 | 947 | } |
7c73875e EP |
948 | |
949 | /* | |
950 | * cap_file_mmap - check if able to map given addr | |
951 | * @file: unused | |
952 | * @reqprot: unused | |
953 | * @prot: unused | |
954 | * @flags: unused | |
955 | * @addr: address attempting to be mapped | |
956 | * @addr_only: unused | |
957 | * | |
6f262d8e | 958 | * If the process is attempting to map memory below dac_mmap_min_addr they need |
7c73875e EP |
959 | * CAP_SYS_RAWIO. The other parameters to this function are unused by the |
960 | * capability security module. Returns 0 if this mapping should be allowed | |
961 | * -EPERM if not. | |
962 | */ | |
963 | int cap_file_mmap(struct file *file, unsigned long reqprot, | |
964 | unsigned long prot, unsigned long flags, | |
965 | unsigned long addr, unsigned long addr_only) | |
966 | { | |
967 | int ret = 0; | |
968 | ||
a2551df7 | 969 | if (addr < dac_mmap_min_addr) { |
3486740a | 970 | ret = cap_capable(current, current_cred(), &init_user_ns, CAP_SYS_RAWIO, |
7c73875e EP |
971 | SECURITY_CAP_AUDIT); |
972 | /* set PF_SUPERPRIV if it turns out we allow the low mmap */ | |
973 | if (ret == 0) | |
974 | current->flags |= PF_SUPERPRIV; | |
975 | } | |
976 | return ret; | |
977 | } |