]>
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> | |
b1d9e6b0 | 15 | #include <linux/lsm_hooks.h> |
1da177e4 LT |
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> |
40401530 | 31 | #include <linux/binfmts.h> |
51b79bee | 32 | #include <linux/personality.h> |
72c2d582 | 33 | |
b5f22a59 SH |
34 | /* |
35 | * If a non-root user executes a setuid-root binary in | |
36 | * !secure(SECURE_NOROOT) mode, then we raise capabilities. | |
37 | * However if fE is also set, then the intent is for only | |
38 | * the file capabilities to be applied, and the setuid-root | |
39 | * bit is left on either to change the uid (plausible) or | |
40 | * to get full privilege on a kernel without file capabilities | |
41 | * support. So in that case we do not raise capabilities. | |
42 | * | |
43 | * Warn if that happens, once per boot. | |
44 | */ | |
d7627467 | 45 | static void warn_setuid_and_fcaps_mixed(const char *fname) |
b5f22a59 SH |
46 | { |
47 | static int warned; | |
48 | if (!warned) { | |
49 | printk(KERN_INFO "warning: `%s' has both setuid-root and" | |
50 | " effective capabilities. Therefore not raising all" | |
51 | " capabilities.\n", fname); | |
52 | warned = 1; | |
53 | } | |
54 | } | |
55 | ||
1d045980 DH |
56 | /** |
57 | * cap_capable - Determine whether a task has a particular effective capability | |
3699c53c | 58 | * @cred: The credentials to use |
3486740a | 59 | * @ns: The user namespace in which we need the capability |
1d045980 DH |
60 | * @cap: The capability to check for |
61 | * @audit: Whether to write an audit message or not | |
62 | * | |
63 | * Determine whether the nominated task has the specified capability amongst | |
64 | * its effective set, returning 0 if it does, -ve if it does not. | |
65 | * | |
3699c53c DH |
66 | * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable() |
67 | * and has_capability() functions. That is, it has the reverse semantics: | |
68 | * cap_has_capability() returns 0 when a task has a capability, but the | |
69 | * kernel's capable() and has_capability() returns 1 for this case. | |
a6dbb1ef | 70 | */ |
6a9de491 EP |
71 | int cap_capable(const struct cred *cred, struct user_namespace *targ_ns, |
72 | int cap, int audit) | |
1da177e4 | 73 | { |
520d9eab | 74 | struct user_namespace *ns = targ_ns; |
3486740a | 75 | |
520d9eab EB |
76 | /* See if cred has the capability in the target user namespace |
77 | * by examining the target user namespace and all of the target | |
78 | * user namespace's parents. | |
79 | */ | |
80 | for (;;) { | |
3486740a | 81 | /* Do we have the necessary capabilities? */ |
520d9eab | 82 | if (ns == cred->user_ns) |
3486740a SH |
83 | return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM; |
84 | ||
64db4c7f KT |
85 | /* |
86 | * If we're already at a lower level than we're looking for, | |
87 | * we're done searching. | |
88 | */ | |
89 | if (ns->level <= cred->user_ns->level) | |
3486740a SH |
90 | return -EPERM; |
91 | ||
520d9eab EB |
92 | /* |
93 | * The owner of the user namespace in the parent of the | |
94 | * user namespace has all caps. | |
95 | */ | |
96 | if ((ns->parent == cred->user_ns) && uid_eq(ns->owner, cred->euid)) | |
97 | return 0; | |
98 | ||
3486740a | 99 | /* |
520d9eab | 100 | * If you have a capability in a parent user ns, then you have |
3486740a SH |
101 | * it over all children user namespaces as well. |
102 | */ | |
520d9eab | 103 | ns = ns->parent; |
3486740a SH |
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 | */ | |
457db29b | 117 | int cap_settime(const struct timespec64 *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; |
caaee623 | 143 | const kernel_cap_t *caller_caps; |
c69e8d9c DH |
144 | |
145 | rcu_read_lock(); | |
8409cca7 SH |
146 | cred = current_cred(); |
147 | child_cred = __task_cred(child); | |
caaee623 JH |
148 | if (mode & PTRACE_MODE_FSCREDS) |
149 | caller_caps = &cred->cap_effective; | |
150 | else | |
151 | caller_caps = &cred->cap_permitted; | |
c4a4d603 | 152 | if (cred->user_ns == child_cred->user_ns && |
caaee623 | 153 | cap_issubset(child_cred->cap_permitted, *caller_caps)) |
8409cca7 | 154 | goto out; |
c4a4d603 | 155 | if (ns_capable(child_cred->user_ns, CAP_SYS_PTRACE)) |
8409cca7 SH |
156 | goto out; |
157 | ret = -EPERM; | |
158 | out: | |
c69e8d9c DH |
159 | rcu_read_unlock(); |
160 | return ret; | |
5cd9c58f DH |
161 | } |
162 | ||
1d045980 DH |
163 | /** |
164 | * cap_ptrace_traceme - Determine whether another process may trace the current | |
165 | * @parent: The task proposed to be the tracer | |
166 | * | |
8409cca7 SH |
167 | * If parent is in the same or an ancestor user_ns and has all current's |
168 | * capabilities, then ptrace access is allowed. | |
169 | * If parent has the ptrace capability to current's user_ns, then ptrace | |
170 | * access is allowed. | |
171 | * Else denied. | |
172 | * | |
1d045980 DH |
173 | * Determine whether the nominated task is permitted to trace the current |
174 | * process, returning 0 if permission is granted, -ve if denied. | |
175 | */ | |
5cd9c58f DH |
176 | int cap_ptrace_traceme(struct task_struct *parent) |
177 | { | |
c69e8d9c | 178 | int ret = 0; |
8409cca7 | 179 | const struct cred *cred, *child_cred; |
c69e8d9c DH |
180 | |
181 | rcu_read_lock(); | |
8409cca7 SH |
182 | cred = __task_cred(parent); |
183 | child_cred = current_cred(); | |
c4a4d603 | 184 | if (cred->user_ns == child_cred->user_ns && |
8409cca7 SH |
185 | cap_issubset(child_cred->cap_permitted, cred->cap_permitted)) |
186 | goto out; | |
c4a4d603 | 187 | if (has_ns_capability(parent, child_cred->user_ns, CAP_SYS_PTRACE)) |
8409cca7 SH |
188 | goto out; |
189 | ret = -EPERM; | |
190 | out: | |
c69e8d9c DH |
191 | rcu_read_unlock(); |
192 | return ret; | |
1da177e4 LT |
193 | } |
194 | ||
1d045980 DH |
195 | /** |
196 | * cap_capget - Retrieve a task's capability sets | |
197 | * @target: The task from which to retrieve the capability sets | |
198 | * @effective: The place to record the effective set | |
199 | * @inheritable: The place to record the inheritable set | |
200 | * @permitted: The place to record the permitted set | |
201 | * | |
202 | * This function retrieves the capabilities of the nominated task and returns | |
203 | * them to the caller. | |
204 | */ | |
205 | int cap_capget(struct task_struct *target, kernel_cap_t *effective, | |
206 | kernel_cap_t *inheritable, kernel_cap_t *permitted) | |
1da177e4 | 207 | { |
c69e8d9c | 208 | const struct cred *cred; |
b6dff3ec | 209 | |
1da177e4 | 210 | /* Derived from kernel/capability.c:sys_capget. */ |
c69e8d9c DH |
211 | rcu_read_lock(); |
212 | cred = __task_cred(target); | |
b6dff3ec DH |
213 | *effective = cred->cap_effective; |
214 | *inheritable = cred->cap_inheritable; | |
215 | *permitted = cred->cap_permitted; | |
c69e8d9c | 216 | rcu_read_unlock(); |
1da177e4 LT |
217 | return 0; |
218 | } | |
219 | ||
1d045980 DH |
220 | /* |
221 | * Determine whether the inheritable capabilities are limited to the old | |
222 | * permitted set. Returns 1 if they are limited, 0 if they are not. | |
223 | */ | |
72c2d582 AM |
224 | static inline int cap_inh_is_capped(void) |
225 | { | |
72c2d582 | 226 | |
1d045980 DH |
227 | /* they are so limited unless the current task has the CAP_SETPCAP |
228 | * capability | |
229 | */ | |
c4a4d603 | 230 | if (cap_capable(current_cred(), current_cred()->user_ns, |
6a9de491 | 231 | CAP_SETPCAP, SECURITY_CAP_AUDIT) == 0) |
1d045980 | 232 | return 0; |
1d045980 | 233 | return 1; |
1209726c | 234 | } |
72c2d582 | 235 | |
1d045980 DH |
236 | /** |
237 | * cap_capset - Validate and apply proposed changes to current's capabilities | |
238 | * @new: The proposed new credentials; alterations should be made here | |
239 | * @old: The current task's current credentials | |
240 | * @effective: A pointer to the proposed new effective capabilities set | |
241 | * @inheritable: A pointer to the proposed new inheritable capabilities set | |
242 | * @permitted: A pointer to the proposed new permitted capabilities set | |
243 | * | |
244 | * This function validates and applies a proposed mass change to the current | |
245 | * process's capability sets. The changes are made to the proposed new | |
246 | * credentials, and assuming no error, will be committed by the caller of LSM. | |
247 | */ | |
d84f4f99 DH |
248 | int cap_capset(struct cred *new, |
249 | const struct cred *old, | |
250 | const kernel_cap_t *effective, | |
251 | const kernel_cap_t *inheritable, | |
252 | const kernel_cap_t *permitted) | |
1da177e4 | 253 | { |
d84f4f99 DH |
254 | if (cap_inh_is_capped() && |
255 | !cap_issubset(*inheritable, | |
256 | cap_combine(old->cap_inheritable, | |
257 | old->cap_permitted))) | |
72c2d582 | 258 | /* incapable of using this inheritable set */ |
1da177e4 | 259 | return -EPERM; |
d84f4f99 | 260 | |
3b7391de | 261 | if (!cap_issubset(*inheritable, |
d84f4f99 DH |
262 | cap_combine(old->cap_inheritable, |
263 | old->cap_bset))) | |
3b7391de SH |
264 | /* no new pI capabilities outside bounding set */ |
265 | return -EPERM; | |
1da177e4 LT |
266 | |
267 | /* verify restrictions on target's new Permitted set */ | |
d84f4f99 | 268 | if (!cap_issubset(*permitted, old->cap_permitted)) |
1da177e4 | 269 | return -EPERM; |
1da177e4 LT |
270 | |
271 | /* verify the _new_Effective_ is a subset of the _new_Permitted_ */ | |
d84f4f99 | 272 | if (!cap_issubset(*effective, *permitted)) |
1da177e4 | 273 | return -EPERM; |
1da177e4 | 274 | |
d84f4f99 DH |
275 | new->cap_effective = *effective; |
276 | new->cap_inheritable = *inheritable; | |
277 | new->cap_permitted = *permitted; | |
58319057 AL |
278 | |
279 | /* | |
280 | * Mask off ambient bits that are no longer both permitted and | |
281 | * inheritable. | |
282 | */ | |
283 | new->cap_ambient = cap_intersect(new->cap_ambient, | |
284 | cap_intersect(*permitted, | |
285 | *inheritable)); | |
286 | if (WARN_ON(!cap_ambient_invariant_ok(new))) | |
287 | return -EINVAL; | |
1da177e4 LT |
288 | return 0; |
289 | } | |
290 | ||
1d045980 DH |
291 | /** |
292 | * cap_inode_need_killpriv - Determine if inode change affects privileges | |
293 | * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV | |
294 | * | |
295 | * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV | |
296 | * affects the security markings on that inode, and if it is, should | |
ab5348c9 | 297 | * inode_killpriv() be invoked or the change rejected. |
1d045980 | 298 | * |
ab5348c9 SB |
299 | * Returns 1 if security.capability has a value, meaning inode_killpriv() |
300 | * is required, 0 otherwise, meaning inode_killpriv() is not required. | |
1d045980 | 301 | */ |
b5376771 SH |
302 | int cap_inode_need_killpriv(struct dentry *dentry) |
303 | { | |
c6f493d6 | 304 | struct inode *inode = d_backing_inode(dentry); |
b5376771 SH |
305 | int error; |
306 | ||
5d6c3191 AG |
307 | error = __vfs_getxattr(dentry, inode, XATTR_NAME_CAPS, NULL, 0); |
308 | return error > 0; | |
b5376771 SH |
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 | { | |
5d6c3191 | 321 | int error; |
b5376771 | 322 | |
5d6c3191 AG |
323 | error = __vfs_removexattr(dentry, XATTR_NAME_CAPS); |
324 | if (error == -EOPNOTSUPP) | |
325 | error = 0; | |
326 | return error; | |
b5376771 SH |
327 | } |
328 | ||
8db6c34f SH |
329 | static bool rootid_owns_currentns(kuid_t kroot) |
330 | { | |
331 | struct user_namespace *ns; | |
332 | ||
333 | if (!uid_valid(kroot)) | |
334 | return false; | |
335 | ||
336 | for (ns = current_user_ns(); ; ns = ns->parent) { | |
337 | if (from_kuid(ns, kroot) == 0) | |
338 | return true; | |
339 | if (ns == &init_user_ns) | |
340 | break; | |
341 | } | |
342 | ||
343 | return false; | |
344 | } | |
345 | ||
346 | static __u32 sansflags(__u32 m) | |
347 | { | |
348 | return m & ~VFS_CAP_FLAGS_EFFECTIVE; | |
349 | } | |
350 | ||
351 | static bool is_v2header(size_t size, __le32 magic) | |
352 | { | |
353 | __u32 m = le32_to_cpu(magic); | |
354 | if (size != XATTR_CAPS_SZ_2) | |
355 | return false; | |
356 | return sansflags(m) == VFS_CAP_REVISION_2; | |
357 | } | |
358 | ||
359 | static bool is_v3header(size_t size, __le32 magic) | |
360 | { | |
361 | __u32 m = le32_to_cpu(magic); | |
362 | ||
363 | if (size != XATTR_CAPS_SZ_3) | |
364 | return false; | |
365 | return sansflags(m) == VFS_CAP_REVISION_3; | |
366 | } | |
367 | ||
368 | /* | |
369 | * getsecurity: We are called for security.* before any attempt to read the | |
370 | * xattr from the inode itself. | |
371 | * | |
372 | * This gives us a chance to read the on-disk value and convert it. If we | |
373 | * return -EOPNOTSUPP, then vfs_getxattr() will call the i_op handler. | |
374 | * | |
375 | * Note we are not called by vfs_getxattr_alloc(), but that is only called | |
376 | * by the integrity subsystem, which really wants the unconverted values - | |
377 | * so that's good. | |
378 | */ | |
379 | int cap_inode_getsecurity(struct inode *inode, const char *name, void **buffer, | |
380 | bool alloc) | |
381 | { | |
382 | int size, ret; | |
383 | kuid_t kroot; | |
384 | uid_t root, mappedroot; | |
385 | char *tmpbuf = NULL; | |
386 | struct vfs_cap_data *cap; | |
387 | struct vfs_ns_cap_data *nscap; | |
388 | struct dentry *dentry; | |
389 | struct user_namespace *fs_ns; | |
390 | ||
391 | if (strcmp(name, "capability") != 0) | |
392 | return -EOPNOTSUPP; | |
393 | ||
394 | dentry = d_find_alias(inode); | |
395 | if (!dentry) | |
396 | return -EINVAL; | |
397 | ||
398 | size = sizeof(struct vfs_ns_cap_data); | |
399 | ret = (int) vfs_getxattr_alloc(dentry, XATTR_NAME_CAPS, | |
400 | &tmpbuf, size, GFP_NOFS); | |
401 | dput(dentry); | |
402 | ||
403 | if (ret < 0) | |
404 | return ret; | |
405 | ||
406 | fs_ns = inode->i_sb->s_user_ns; | |
407 | cap = (struct vfs_cap_data *) tmpbuf; | |
408 | if (is_v2header((size_t) ret, cap->magic_etc)) { | |
409 | /* If this is sizeof(vfs_cap_data) then we're ok with the | |
410 | * on-disk value, so return that. */ | |
411 | if (alloc) | |
412 | *buffer = tmpbuf; | |
413 | else | |
414 | kfree(tmpbuf); | |
415 | return ret; | |
416 | } else if (!is_v3header((size_t) ret, cap->magic_etc)) { | |
417 | kfree(tmpbuf); | |
418 | return -EINVAL; | |
419 | } | |
420 | ||
421 | nscap = (struct vfs_ns_cap_data *) tmpbuf; | |
422 | root = le32_to_cpu(nscap->rootid); | |
423 | kroot = make_kuid(fs_ns, root); | |
424 | ||
425 | /* If the root kuid maps to a valid uid in current ns, then return | |
426 | * this as a nscap. */ | |
427 | mappedroot = from_kuid(current_user_ns(), kroot); | |
428 | if (mappedroot != (uid_t)-1 && mappedroot != (uid_t)0) { | |
429 | if (alloc) { | |
430 | *buffer = tmpbuf; | |
431 | nscap->rootid = cpu_to_le32(mappedroot); | |
432 | } else | |
433 | kfree(tmpbuf); | |
434 | return size; | |
435 | } | |
436 | ||
437 | if (!rootid_owns_currentns(kroot)) { | |
438 | kfree(tmpbuf); | |
439 | return -EOPNOTSUPP; | |
440 | } | |
441 | ||
442 | /* This comes from a parent namespace. Return as a v2 capability */ | |
443 | size = sizeof(struct vfs_cap_data); | |
444 | if (alloc) { | |
445 | *buffer = kmalloc(size, GFP_ATOMIC); | |
446 | if (*buffer) { | |
447 | struct vfs_cap_data *cap = *buffer; | |
448 | __le32 nsmagic, magic; | |
449 | magic = VFS_CAP_REVISION_2; | |
450 | nsmagic = le32_to_cpu(nscap->magic_etc); | |
451 | if (nsmagic & VFS_CAP_FLAGS_EFFECTIVE) | |
452 | magic |= VFS_CAP_FLAGS_EFFECTIVE; | |
453 | memcpy(&cap->data, &nscap->data, sizeof(__le32) * 2 * VFS_CAP_U32); | |
454 | cap->magic_etc = cpu_to_le32(magic); | |
455 | } | |
456 | } | |
457 | kfree(tmpbuf); | |
458 | return size; | |
459 | } | |
460 | ||
461 | static kuid_t rootid_from_xattr(const void *value, size_t size, | |
462 | struct user_namespace *task_ns) | |
463 | { | |
464 | const struct vfs_ns_cap_data *nscap = value; | |
465 | uid_t rootid = 0; | |
466 | ||
467 | if (size == XATTR_CAPS_SZ_3) | |
468 | rootid = le32_to_cpu(nscap->rootid); | |
469 | ||
470 | return make_kuid(task_ns, rootid); | |
471 | } | |
472 | ||
473 | static bool validheader(size_t size, __le32 magic) | |
474 | { | |
475 | return is_v2header(size, magic) || is_v3header(size, magic); | |
476 | } | |
477 | ||
478 | /* | |
479 | * User requested a write of security.capability. If needed, update the | |
480 | * xattr to change from v2 to v3, or to fixup the v3 rootid. | |
481 | * | |
482 | * If all is ok, we return the new size, on error return < 0. | |
483 | */ | |
484 | int cap_convert_nscap(struct dentry *dentry, void **ivalue, size_t size) | |
485 | { | |
486 | struct vfs_ns_cap_data *nscap; | |
487 | uid_t nsrootid; | |
488 | const struct vfs_cap_data *cap = *ivalue; | |
489 | __u32 magic, nsmagic; | |
490 | struct inode *inode = d_backing_inode(dentry); | |
491 | struct user_namespace *task_ns = current_user_ns(), | |
492 | *fs_ns = inode->i_sb->s_user_ns; | |
493 | kuid_t rootid; | |
494 | size_t newsize; | |
495 | ||
496 | if (!*ivalue) | |
497 | return -EINVAL; | |
498 | if (!validheader(size, cap->magic_etc)) | |
499 | return -EINVAL; | |
500 | if (!capable_wrt_inode_uidgid(inode, CAP_SETFCAP)) | |
501 | return -EPERM; | |
502 | if (size == XATTR_CAPS_SZ_2) | |
503 | if (ns_capable(inode->i_sb->s_user_ns, CAP_SETFCAP)) | |
504 | /* user is privileged, just write the v2 */ | |
505 | return size; | |
506 | ||
507 | rootid = rootid_from_xattr(*ivalue, size, task_ns); | |
508 | if (!uid_valid(rootid)) | |
509 | return -EINVAL; | |
510 | ||
511 | nsrootid = from_kuid(fs_ns, rootid); | |
512 | if (nsrootid == -1) | |
513 | return -EINVAL; | |
514 | ||
515 | newsize = sizeof(struct vfs_ns_cap_data); | |
516 | nscap = kmalloc(newsize, GFP_ATOMIC); | |
517 | if (!nscap) | |
518 | return -ENOMEM; | |
519 | nscap->rootid = cpu_to_le32(nsrootid); | |
520 | nsmagic = VFS_CAP_REVISION_3; | |
521 | magic = le32_to_cpu(cap->magic_etc); | |
522 | if (magic & VFS_CAP_FLAGS_EFFECTIVE) | |
523 | nsmagic |= VFS_CAP_FLAGS_EFFECTIVE; | |
524 | nscap->magic_etc = cpu_to_le32(nsmagic); | |
525 | memcpy(&nscap->data, &cap->data, sizeof(__le32) * 2 * VFS_CAP_U32); | |
526 | ||
527 | kvfree(*ivalue); | |
528 | *ivalue = nscap; | |
529 | return newsize; | |
530 | } | |
531 | ||
1d045980 DH |
532 | /* |
533 | * Calculate the new process capability sets from the capability sets attached | |
534 | * to a file. | |
535 | */ | |
c0b00441 | 536 | static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps, |
a6f76f23 | 537 | struct linux_binprm *bprm, |
4d49f671 ZL |
538 | bool *effective, |
539 | bool *has_cap) | |
b5376771 | 540 | { |
a6f76f23 | 541 | struct cred *new = bprm->cred; |
c0b00441 EP |
542 | unsigned i; |
543 | int ret = 0; | |
544 | ||
545 | if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE) | |
a6f76f23 | 546 | *effective = true; |
c0b00441 | 547 | |
4d49f671 ZL |
548 | if (caps->magic_etc & VFS_CAP_REVISION_MASK) |
549 | *has_cap = true; | |
550 | ||
c0b00441 EP |
551 | CAP_FOR_EACH_U32(i) { |
552 | __u32 permitted = caps->permitted.cap[i]; | |
553 | __u32 inheritable = caps->inheritable.cap[i]; | |
554 | ||
555 | /* | |
556 | * pP' = (X & fP) | (pI & fI) | |
58319057 | 557 | * The addition of pA' is handled later. |
c0b00441 | 558 | */ |
a6f76f23 DH |
559 | new->cap_permitted.cap[i] = |
560 | (new->cap_bset.cap[i] & permitted) | | |
561 | (new->cap_inheritable.cap[i] & inheritable); | |
c0b00441 | 562 | |
a6f76f23 DH |
563 | if (permitted & ~new->cap_permitted.cap[i]) |
564 | /* insufficient to execute correctly */ | |
c0b00441 | 565 | ret = -EPERM; |
c0b00441 EP |
566 | } |
567 | ||
568 | /* | |
569 | * For legacy apps, with no internal support for recognizing they | |
570 | * do not have enough capabilities, we return an error if they are | |
571 | * missing some "forced" (aka file-permitted) capabilities. | |
572 | */ | |
a6f76f23 | 573 | return *effective ? ret : 0; |
c0b00441 EP |
574 | } |
575 | ||
1d045980 DH |
576 | /* |
577 | * Extract the on-exec-apply capability sets for an executable file. | |
578 | */ | |
c0b00441 EP |
579 | int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps) |
580 | { | |
c6f493d6 | 581 | struct inode *inode = d_backing_inode(dentry); |
b5376771 | 582 | __u32 magic_etc; |
e338d263 | 583 | unsigned tocopy, i; |
c0b00441 | 584 | int size; |
8db6c34f SH |
585 | struct vfs_ns_cap_data data, *nscaps = &data; |
586 | struct vfs_cap_data *caps = (struct vfs_cap_data *) &data; | |
587 | kuid_t rootkuid; | |
588 | struct user_namespace *fs_ns = inode->i_sb->s_user_ns; | |
c0b00441 EP |
589 | |
590 | memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data)); | |
591 | ||
5d6c3191 | 592 | if (!inode) |
c0b00441 EP |
593 | return -ENODATA; |
594 | ||
5d6c3191 | 595 | size = __vfs_getxattr((struct dentry *)dentry, inode, |
8db6c34f | 596 | XATTR_NAME_CAPS, &data, XATTR_CAPS_SZ); |
a6f76f23 | 597 | if (size == -ENODATA || size == -EOPNOTSUPP) |
c0b00441 EP |
598 | /* no data, that's ok */ |
599 | return -ENODATA; | |
8db6c34f | 600 | |
c0b00441 EP |
601 | if (size < 0) |
602 | return size; | |
b5376771 | 603 | |
e338d263 | 604 | if (size < sizeof(magic_etc)) |
b5376771 SH |
605 | return -EINVAL; |
606 | ||
8db6c34f | 607 | cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps->magic_etc); |
b5376771 | 608 | |
8db6c34f | 609 | rootkuid = make_kuid(fs_ns, 0); |
a6f76f23 | 610 | switch (magic_etc & VFS_CAP_REVISION_MASK) { |
e338d263 AM |
611 | case VFS_CAP_REVISION_1: |
612 | if (size != XATTR_CAPS_SZ_1) | |
613 | return -EINVAL; | |
614 | tocopy = VFS_CAP_U32_1; | |
615 | break; | |
616 | case VFS_CAP_REVISION_2: | |
617 | if (size != XATTR_CAPS_SZ_2) | |
618 | return -EINVAL; | |
619 | tocopy = VFS_CAP_U32_2; | |
620 | break; | |
8db6c34f SH |
621 | case VFS_CAP_REVISION_3: |
622 | if (size != XATTR_CAPS_SZ_3) | |
623 | return -EINVAL; | |
624 | tocopy = VFS_CAP_U32_3; | |
625 | rootkuid = make_kuid(fs_ns, le32_to_cpu(nscaps->rootid)); | |
626 | break; | |
627 | ||
b5376771 SH |
628 | default: |
629 | return -EINVAL; | |
630 | } | |
8db6c34f SH |
631 | /* Limit the caps to the mounter of the filesystem |
632 | * or the more limited uid specified in the xattr. | |
633 | */ | |
634 | if (!rootid_owns_currentns(rootkuid)) | |
635 | return -ENODATA; | |
e338d263 | 636 | |
5459c164 | 637 | CAP_FOR_EACH_U32(i) { |
c0b00441 EP |
638 | if (i >= tocopy) |
639 | break; | |
8db6c34f SH |
640 | cpu_caps->permitted.cap[i] = le32_to_cpu(caps->data[i].permitted); |
641 | cpu_caps->inheritable.cap[i] = le32_to_cpu(caps->data[i].inheritable); | |
e338d263 | 642 | } |
a6f76f23 | 643 | |
7d8b6c63 EP |
644 | cpu_caps->permitted.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK; |
645 | cpu_caps->inheritable.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK; | |
646 | ||
c0b00441 | 647 | return 0; |
b5376771 SH |
648 | } |
649 | ||
1d045980 DH |
650 | /* |
651 | * Attempt to get the on-exec apply capability sets for an executable file from | |
652 | * its xattrs and, if present, apply them to the proposed credentials being | |
653 | * constructed by execve(). | |
654 | */ | |
4d49f671 | 655 | static int get_file_caps(struct linux_binprm *bprm, bool *effective, bool *has_cap) |
b5376771 | 656 | { |
b5376771 | 657 | int rc = 0; |
c0b00441 | 658 | struct cpu_vfs_cap_data vcaps; |
b5376771 | 659 | |
ee67ae7e | 660 | cap_clear(bprm->cred->cap_permitted); |
3318a386 | 661 | |
1f29fae2 SH |
662 | if (!file_caps_enabled) |
663 | return 0; | |
664 | ||
380cf5ba | 665 | if (!mnt_may_suid(bprm->file->f_path.mnt)) |
b5376771 | 666 | return 0; |
380cf5ba AL |
667 | |
668 | /* | |
669 | * This check is redundant with mnt_may_suid() but is kept to make | |
670 | * explicit that capability bits are limited to s_user_ns and its | |
671 | * descendants. | |
672 | */ | |
d07b846f SF |
673 | if (!current_in_userns(bprm->file->f_path.mnt->mnt_sb->s_user_ns)) |
674 | return 0; | |
b5376771 | 675 | |
f4a4a8b1 | 676 | rc = get_vfs_caps_from_disk(bprm->file->f_path.dentry, &vcaps); |
c0b00441 EP |
677 | if (rc < 0) { |
678 | if (rc == -EINVAL) | |
8db6c34f SH |
679 | printk(KERN_NOTICE "Invalid argument reading file caps for %s\n", |
680 | bprm->filename); | |
c0b00441 EP |
681 | else if (rc == -ENODATA) |
682 | rc = 0; | |
b5376771 SH |
683 | goto out; |
684 | } | |
b5376771 | 685 | |
4d49f671 | 686 | rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective, has_cap); |
a6f76f23 DH |
687 | if (rc == -EINVAL) |
688 | printk(KERN_NOTICE "%s: cap_from_disk returned %d for %s\n", | |
689 | __func__, rc, bprm->filename); | |
b5376771 SH |
690 | |
691 | out: | |
b5376771 | 692 | if (rc) |
ee67ae7e | 693 | cap_clear(bprm->cred->cap_permitted); |
b5376771 SH |
694 | |
695 | return rc; | |
696 | } | |
697 | ||
db1a8922 RGB |
698 | /* |
699 | * handle_privileged_root - Handle case of privileged root | |
700 | * @bprm: The execution parameters, including the proposed creds | |
701 | * @has_fcap: Are any file capabilities set? | |
702 | * @effective: Do we have effective root privilege? | |
703 | * @root_uid: This namespace' root UID WRT initial USER namespace | |
704 | * | |
705 | * Handle the case where root is privileged and hasn't been neutered by | |
706 | * SECURE_NOROOT. If file capabilities are set, they won't be combined with | |
707 | * set UID root and nothing is changed. If we are root, cap_permitted is | |
708 | * updated. If we have become set UID root, the effective bit is set. | |
709 | */ | |
710 | static void handle_privileged_root(struct linux_binprm *bprm, bool has_cap, | |
711 | bool *effective, kuid_t root_uid) | |
712 | { | |
713 | const struct cred *old = current_cred(); | |
714 | struct cred *new = bprm->cred; | |
715 | ||
716 | if (issecure(SECURE_NOROOT)) | |
717 | return; | |
718 | /* | |
719 | * If the legacy file capability is set, then don't set privs | |
720 | * for a setuid root binary run by a non-root user. Do set it | |
721 | * for a root user just to cause least surprise to an admin. | |
722 | */ | |
723 | if (has_cap && !uid_eq(new->uid, root_uid) && uid_eq(new->euid, root_uid)) { | |
724 | warn_setuid_and_fcaps_mixed(bprm->filename); | |
725 | return; | |
726 | } | |
727 | /* | |
728 | * To support inheritance of root-permissions and suid-root | |
729 | * executables under compatibility mode, we override the | |
730 | * capability sets for the file. | |
731 | */ | |
732 | if (uid_eq(new->euid, root_uid) || uid_eq(new->uid, root_uid)) { | |
733 | /* pP' = (cap_bset & ~0) | (pI & ~0) */ | |
734 | new->cap_permitted = cap_combine(old->cap_bset, | |
735 | old->cap_inheritable); | |
736 | } | |
737 | /* | |
738 | * If only the real uid is 0, we do not set the effective bit. | |
739 | */ | |
740 | if (uid_eq(new->euid, root_uid)) | |
741 | *effective = true; | |
742 | } | |
743 | ||
1d045980 DH |
744 | /** |
745 | * cap_bprm_set_creds - Set up the proposed credentials for execve(). | |
746 | * @bprm: The execution parameters, including the proposed creds | |
747 | * | |
748 | * Set up the proposed credentials for a new execution context being | |
749 | * constructed by execve(). The proposed creds in @bprm->cred is altered, | |
750 | * which won't take effect immediately. Returns 0 if successful, -ve on error. | |
a6f76f23 DH |
751 | */ |
752 | int cap_bprm_set_creds(struct linux_binprm *bprm) | |
1da177e4 | 753 | { |
a6f76f23 DH |
754 | const struct cred *old = current_cred(); |
755 | struct cred *new = bprm->cred; | |
db1a8922 | 756 | bool effective = false, has_cap = false, is_setid; |
b5376771 | 757 | int ret; |
18815a18 | 758 | kuid_t root_uid; |
1da177e4 | 759 | |
58319057 AL |
760 | if (WARN_ON(!cap_ambient_invariant_ok(old))) |
761 | return -EPERM; | |
762 | ||
4d49f671 | 763 | ret = get_file_caps(bprm, &effective, &has_cap); |
a6f76f23 DH |
764 | if (ret < 0) |
765 | return ret; | |
1da177e4 | 766 | |
18815a18 EB |
767 | root_uid = make_kuid(new->user_ns, 0); |
768 | ||
db1a8922 | 769 | handle_privileged_root(bprm, has_cap, &effective, root_uid); |
b5376771 | 770 | |
d52fc5dd EP |
771 | /* if we have fs caps, clear dangerous personality flags */ |
772 | if (!cap_issubset(new->cap_permitted, old->cap_permitted)) | |
773 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
774 | ||
775 | ||
a6f76f23 | 776 | /* Don't let someone trace a set[ug]id/setpcap binary with the revised |
259e5e6c AL |
777 | * credentials unless they have the appropriate permit. |
778 | * | |
779 | * In addition, if NO_NEW_PRIVS, then ensure we get no new privs. | |
a6f76f23 | 780 | */ |
58319057 AL |
781 | is_setid = !uid_eq(new->euid, old->uid) || !gid_eq(new->egid, old->gid); |
782 | ||
783 | if ((is_setid || | |
a6f76f23 | 784 | !cap_issubset(new->cap_permitted, old->cap_permitted)) && |
9227dd2a | 785 | ((bprm->unsafe & ~LSM_UNSAFE_PTRACE) || |
20523132 | 786 | !ptracer_capable(current, new->user_ns))) { |
a6f76f23 | 787 | /* downgrade; they get no more than they had, and maybe less */ |
70169420 | 788 | if (!ns_capable(new->user_ns, CAP_SETUID) || |
259e5e6c | 789 | (bprm->unsafe & LSM_UNSAFE_NO_NEW_PRIVS)) { |
a6f76f23 DH |
790 | new->euid = new->uid; |
791 | new->egid = new->gid; | |
1da177e4 | 792 | } |
b3a222e5 SH |
793 | new->cap_permitted = cap_intersect(new->cap_permitted, |
794 | old->cap_permitted); | |
1da177e4 LT |
795 | } |
796 | ||
a6f76f23 DH |
797 | new->suid = new->fsuid = new->euid; |
798 | new->sgid = new->fsgid = new->egid; | |
1da177e4 | 799 | |
58319057 AL |
800 | /* File caps or setid cancels ambient. */ |
801 | if (has_cap || is_setid) | |
802 | cap_clear(new->cap_ambient); | |
803 | ||
804 | /* | |
805 | * Now that we've computed pA', update pP' to give: | |
806 | * pP' = (X & fP) | (pI & fI) | pA' | |
807 | */ | |
808 | new->cap_permitted = cap_combine(new->cap_permitted, new->cap_ambient); | |
809 | ||
810 | /* | |
811 | * Set pE' = (fE ? pP' : pA'). Because pA' is zero if fE is set, | |
812 | * this is the same as pE' = (fE ? pP' : 0) | pA'. | |
813 | */ | |
4bf2ea77 EP |
814 | if (effective) |
815 | new->cap_effective = new->cap_permitted; | |
816 | else | |
58319057 AL |
817 | new->cap_effective = new->cap_ambient; |
818 | ||
819 | if (WARN_ON(!cap_ambient_invariant_ok(new))) | |
820 | return -EPERM; | |
821 | ||
3fc689e9 EP |
822 | /* |
823 | * Audit candidate if current->cap_effective is set | |
824 | * | |
825 | * We do not bother to audit if 3 things are true: | |
826 | * 1) cap_effective has all caps | |
827 | * 2) we are root | |
828 | * 3) root is supposed to have all caps (SECURE_NOROOT) | |
829 | * Since this is just a normal root execing a process. | |
830 | * | |
831 | * Number 1 above might fail if you don't have a full bset, but I think | |
832 | * that is interesting information to audit. | |
833 | */ | |
58319057 | 834 | if (!cap_issubset(new->cap_effective, new->cap_ambient)) { |
d84f4f99 | 835 | if (!cap_issubset(CAP_FULL_SET, new->cap_effective) || |
18815a18 | 836 | !uid_eq(new->euid, root_uid) || !uid_eq(new->uid, root_uid) || |
a6f76f23 DH |
837 | issecure(SECURE_NOROOT)) { |
838 | ret = audit_log_bprm_fcaps(bprm, new, old); | |
839 | if (ret < 0) | |
840 | return ret; | |
841 | } | |
3fc689e9 | 842 | } |
1da177e4 | 843 | |
d84f4f99 | 844 | new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS); |
58319057 AL |
845 | |
846 | if (WARN_ON(!cap_ambient_invariant_ok(new))) | |
847 | return -EPERM; | |
848 | ||
46d98eb4 | 849 | /* Check for privilege-elevated exec. */ |
ee67ae7e KC |
850 | bprm->cap_elevated = 0; |
851 | if (is_setid) { | |
852 | bprm->cap_elevated = 1; | |
853 | } else if (!uid_eq(new->uid, root_uid)) { | |
854 | if (effective || | |
855 | !cap_issubset(new->cap_permitted, new->cap_ambient)) | |
856 | bprm->cap_elevated = 1; | |
b5376771 SH |
857 | } |
858 | ||
ee67ae7e | 859 | return 0; |
1da177e4 LT |
860 | } |
861 | ||
1d045980 DH |
862 | /** |
863 | * cap_inode_setxattr - Determine whether an xattr may be altered | |
864 | * @dentry: The inode/dentry being altered | |
865 | * @name: The name of the xattr to be changed | |
866 | * @value: The value that the xattr will be changed to | |
867 | * @size: The size of value | |
868 | * @flags: The replacement flag | |
869 | * | |
870 | * Determine whether an xattr may be altered or set on an inode, returning 0 if | |
871 | * permission is granted, -ve if denied. | |
872 | * | |
873 | * This is used to make sure security xattrs don't get updated or set by those | |
874 | * who aren't privileged to do so. | |
875 | */ | |
8f0cfa52 DH |
876 | int cap_inode_setxattr(struct dentry *dentry, const char *name, |
877 | const void *value, size_t size, int flags) | |
1da177e4 | 878 | { |
8db6c34f SH |
879 | /* Ignore non-security xattrs */ |
880 | if (strncmp(name, XATTR_SECURITY_PREFIX, | |
881 | sizeof(XATTR_SECURITY_PREFIX) - 1) != 0) | |
882 | return 0; | |
883 | ||
884 | /* | |
885 | * For XATTR_NAME_CAPS the check will be done in | |
886 | * cap_convert_nscap(), called by setxattr() | |
887 | */ | |
888 | if (strcmp(name, XATTR_NAME_CAPS) == 0) | |
b5376771 | 889 | return 0; |
1d045980 | 890 | |
8db6c34f | 891 | if (!capable(CAP_SYS_ADMIN)) |
1da177e4 LT |
892 | return -EPERM; |
893 | return 0; | |
894 | } | |
895 | ||
1d045980 DH |
896 | /** |
897 | * cap_inode_removexattr - Determine whether an xattr may be removed | |
898 | * @dentry: The inode/dentry being altered | |
899 | * @name: The name of the xattr to be changed | |
900 | * | |
901 | * Determine whether an xattr may be removed from an inode, returning 0 if | |
902 | * permission is granted, -ve if denied. | |
903 | * | |
904 | * This is used to make sure security xattrs don't get removed by those who | |
905 | * aren't privileged to remove them. | |
906 | */ | |
8f0cfa52 | 907 | int cap_inode_removexattr(struct dentry *dentry, const char *name) |
1da177e4 | 908 | { |
8db6c34f SH |
909 | /* Ignore non-security xattrs */ |
910 | if (strncmp(name, XATTR_SECURITY_PREFIX, | |
911 | sizeof(XATTR_SECURITY_PREFIX) - 1) != 0) | |
912 | return 0; | |
913 | ||
914 | if (strcmp(name, XATTR_NAME_CAPS) == 0) { | |
915 | /* security.capability gets namespaced */ | |
916 | struct inode *inode = d_backing_inode(dentry); | |
917 | if (!inode) | |
918 | return -EINVAL; | |
919 | if (!capable_wrt_inode_uidgid(inode, CAP_SETFCAP)) | |
b5376771 SH |
920 | return -EPERM; |
921 | return 0; | |
1d045980 DH |
922 | } |
923 | ||
8db6c34f | 924 | if (!capable(CAP_SYS_ADMIN)) |
1da177e4 LT |
925 | return -EPERM; |
926 | return 0; | |
927 | } | |
928 | ||
a6f76f23 | 929 | /* |
1da177e4 LT |
930 | * cap_emulate_setxuid() fixes the effective / permitted capabilities of |
931 | * a process after a call to setuid, setreuid, or setresuid. | |
932 | * | |
933 | * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of | |
934 | * {r,e,s}uid != 0, the permitted and effective capabilities are | |
935 | * cleared. | |
936 | * | |
937 | * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective | |
938 | * capabilities of the process are cleared. | |
939 | * | |
940 | * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective | |
941 | * capabilities are set to the permitted capabilities. | |
942 | * | |
a6f76f23 | 943 | * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should |
1da177e4 LT |
944 | * never happen. |
945 | * | |
a6f76f23 | 946 | * -astor |
1da177e4 LT |
947 | * |
948 | * cevans - New behaviour, Oct '99 | |
949 | * A process may, via prctl(), elect to keep its capabilities when it | |
950 | * calls setuid() and switches away from uid==0. Both permitted and | |
951 | * effective sets will be retained. | |
952 | * Without this change, it was impossible for a daemon to drop only some | |
953 | * of its privilege. The call to setuid(!=0) would drop all privileges! | |
954 | * Keeping uid 0 is not an option because uid 0 owns too many vital | |
955 | * files.. | |
956 | * Thanks to Olaf Kirch and Peter Benie for spotting this. | |
957 | */ | |
d84f4f99 | 958 | static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old) |
1da177e4 | 959 | { |
18815a18 EB |
960 | kuid_t root_uid = make_kuid(old->user_ns, 0); |
961 | ||
962 | if ((uid_eq(old->uid, root_uid) || | |
963 | uid_eq(old->euid, root_uid) || | |
964 | uid_eq(old->suid, root_uid)) && | |
965 | (!uid_eq(new->uid, root_uid) && | |
966 | !uid_eq(new->euid, root_uid) && | |
58319057 AL |
967 | !uid_eq(new->suid, root_uid))) { |
968 | if (!issecure(SECURE_KEEP_CAPS)) { | |
969 | cap_clear(new->cap_permitted); | |
970 | cap_clear(new->cap_effective); | |
971 | } | |
972 | ||
973 | /* | |
974 | * Pre-ambient programs expect setresuid to nonroot followed | |
975 | * by exec to drop capabilities. We should make sure that | |
976 | * this remains the case. | |
977 | */ | |
978 | cap_clear(new->cap_ambient); | |
1da177e4 | 979 | } |
18815a18 | 980 | if (uid_eq(old->euid, root_uid) && !uid_eq(new->euid, root_uid)) |
d84f4f99 | 981 | cap_clear(new->cap_effective); |
18815a18 | 982 | if (!uid_eq(old->euid, root_uid) && uid_eq(new->euid, root_uid)) |
d84f4f99 | 983 | new->cap_effective = new->cap_permitted; |
1da177e4 LT |
984 | } |
985 | ||
1d045980 DH |
986 | /** |
987 | * cap_task_fix_setuid - Fix up the results of setuid() call | |
988 | * @new: The proposed credentials | |
989 | * @old: The current task's current credentials | |
990 | * @flags: Indications of what has changed | |
991 | * | |
992 | * Fix up the results of setuid() call before the credential changes are | |
993 | * actually applied, returning 0 to grant the changes, -ve to deny them. | |
994 | */ | |
d84f4f99 | 995 | int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags) |
1da177e4 LT |
996 | { |
997 | switch (flags) { | |
998 | case LSM_SETID_RE: | |
999 | case LSM_SETID_ID: | |
1000 | case LSM_SETID_RES: | |
1d045980 DH |
1001 | /* juggle the capabilities to follow [RES]UID changes unless |
1002 | * otherwise suppressed */ | |
d84f4f99 DH |
1003 | if (!issecure(SECURE_NO_SETUID_FIXUP)) |
1004 | cap_emulate_setxuid(new, old); | |
1da177e4 | 1005 | break; |
1da177e4 | 1006 | |
1d045980 DH |
1007 | case LSM_SETID_FS: |
1008 | /* juggle the capabilties to follow FSUID changes, unless | |
1009 | * otherwise suppressed | |
1010 | * | |
d84f4f99 DH |
1011 | * FIXME - is fsuser used for all CAP_FS_MASK capabilities? |
1012 | * if not, we might be a bit too harsh here. | |
1013 | */ | |
1014 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { | |
18815a18 EB |
1015 | kuid_t root_uid = make_kuid(old->user_ns, 0); |
1016 | if (uid_eq(old->fsuid, root_uid) && !uid_eq(new->fsuid, root_uid)) | |
d84f4f99 DH |
1017 | new->cap_effective = |
1018 | cap_drop_fs_set(new->cap_effective); | |
1d045980 | 1019 | |
18815a18 | 1020 | if (!uid_eq(old->fsuid, root_uid) && uid_eq(new->fsuid, root_uid)) |
d84f4f99 DH |
1021 | new->cap_effective = |
1022 | cap_raise_fs_set(new->cap_effective, | |
1023 | new->cap_permitted); | |
1da177e4 | 1024 | } |
d84f4f99 | 1025 | break; |
1d045980 | 1026 | |
1da177e4 LT |
1027 | default: |
1028 | return -EINVAL; | |
1029 | } | |
1030 | ||
1031 | return 0; | |
1032 | } | |
1033 | ||
b5376771 SH |
1034 | /* |
1035 | * Rationale: code calling task_setscheduler, task_setioprio, and | |
1036 | * task_setnice, assumes that | |
1037 | * . if capable(cap_sys_nice), then those actions should be allowed | |
1038 | * . if not capable(cap_sys_nice), but acting on your own processes, | |
1039 | * then those actions should be allowed | |
1040 | * This is insufficient now since you can call code without suid, but | |
1041 | * yet with increased caps. | |
1042 | * So we check for increased caps on the target process. | |
1043 | */ | |
de45e806 | 1044 | static int cap_safe_nice(struct task_struct *p) |
b5376771 | 1045 | { |
f54fb863 | 1046 | int is_subset, ret = 0; |
c69e8d9c DH |
1047 | |
1048 | rcu_read_lock(); | |
1049 | is_subset = cap_issubset(__task_cred(p)->cap_permitted, | |
1050 | current_cred()->cap_permitted); | |
f54fb863 SH |
1051 | if (!is_subset && !ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) |
1052 | ret = -EPERM; | |
c69e8d9c DH |
1053 | rcu_read_unlock(); |
1054 | ||
f54fb863 | 1055 | return ret; |
b5376771 SH |
1056 | } |
1057 | ||
1d045980 DH |
1058 | /** |
1059 | * cap_task_setscheduler - Detemine if scheduler policy change is permitted | |
1060 | * @p: The task to affect | |
1d045980 DH |
1061 | * |
1062 | * Detemine if the requested scheduler policy change is permitted for the | |
1063 | * specified task, returning 0 if permission is granted, -ve if denied. | |
1064 | */ | |
b0ae1981 | 1065 | int cap_task_setscheduler(struct task_struct *p) |
b5376771 SH |
1066 | { |
1067 | return cap_safe_nice(p); | |
1068 | } | |
1069 | ||
1d045980 DH |
1070 | /** |
1071 | * cap_task_ioprio - Detemine if I/O priority change is permitted | |
1072 | * @p: The task to affect | |
1073 | * @ioprio: The I/O priority to set | |
1074 | * | |
1075 | * Detemine if the requested I/O priority change is permitted for the specified | |
1076 | * task, returning 0 if permission is granted, -ve if denied. | |
1077 | */ | |
1078 | int cap_task_setioprio(struct task_struct *p, int ioprio) | |
b5376771 SH |
1079 | { |
1080 | return cap_safe_nice(p); | |
1081 | } | |
1082 | ||
1d045980 DH |
1083 | /** |
1084 | * cap_task_ioprio - Detemine if task priority change is permitted | |
1085 | * @p: The task to affect | |
1086 | * @nice: The nice value to set | |
1087 | * | |
1088 | * Detemine if the requested task priority change is permitted for the | |
1089 | * specified task, returning 0 if permission is granted, -ve if denied. | |
1090 | */ | |
1091 | int cap_task_setnice(struct task_struct *p, int nice) | |
b5376771 SH |
1092 | { |
1093 | return cap_safe_nice(p); | |
1094 | } | |
1095 | ||
3b7391de | 1096 | /* |
1d045980 DH |
1097 | * Implement PR_CAPBSET_DROP. Attempt to remove the specified capability from |
1098 | * the current task's bounding set. Returns 0 on success, -ve on error. | |
3b7391de | 1099 | */ |
6d6f3328 | 1100 | static int cap_prctl_drop(unsigned long cap) |
3b7391de | 1101 | { |
6d6f3328 TH |
1102 | struct cred *new; |
1103 | ||
160da84d | 1104 | if (!ns_capable(current_user_ns(), CAP_SETPCAP)) |
3b7391de SH |
1105 | return -EPERM; |
1106 | if (!cap_valid(cap)) | |
1107 | return -EINVAL; | |
d84f4f99 | 1108 | |
6d6f3328 TH |
1109 | new = prepare_creds(); |
1110 | if (!new) | |
1111 | return -ENOMEM; | |
d84f4f99 | 1112 | cap_lower(new->cap_bset, cap); |
6d6f3328 | 1113 | return commit_creds(new); |
3b7391de | 1114 | } |
3898b1b4 | 1115 | |
1d045980 DH |
1116 | /** |
1117 | * cap_task_prctl - Implement process control functions for this security module | |
1118 | * @option: The process control function requested | |
1119 | * @arg2, @arg3, @arg4, @arg5: The argument data for this function | |
1120 | * | |
1121 | * Allow process control functions (sys_prctl()) to alter capabilities; may | |
1122 | * also deny access to other functions not otherwise implemented here. | |
1123 | * | |
1124 | * Returns 0 or +ve on success, -ENOSYS if this function is not implemented | |
1125 | * here, other -ve on error. If -ENOSYS is returned, sys_prctl() and other LSM | |
1126 | * modules will consider performing the function. | |
1127 | */ | |
3898b1b4 | 1128 | int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3, |
d84f4f99 | 1129 | unsigned long arg4, unsigned long arg5) |
3898b1b4 | 1130 | { |
6d6f3328 | 1131 | const struct cred *old = current_cred(); |
d84f4f99 | 1132 | struct cred *new; |
d84f4f99 | 1133 | |
3898b1b4 AM |
1134 | switch (option) { |
1135 | case PR_CAPBSET_READ: | |
1136 | if (!cap_valid(arg2)) | |
6d6f3328 TH |
1137 | return -EINVAL; |
1138 | return !!cap_raised(old->cap_bset, arg2); | |
d84f4f99 | 1139 | |
3898b1b4 | 1140 | case PR_CAPBSET_DROP: |
6d6f3328 | 1141 | return cap_prctl_drop(arg2); |
3898b1b4 AM |
1142 | |
1143 | /* | |
1144 | * The next four prctl's remain to assist with transitioning a | |
1145 | * system from legacy UID=0 based privilege (when filesystem | |
1146 | * capabilities are not in use) to a system using filesystem | |
1147 | * capabilities only - as the POSIX.1e draft intended. | |
1148 | * | |
1149 | * Note: | |
1150 | * | |
1151 | * PR_SET_SECUREBITS = | |
1152 | * issecure_mask(SECURE_KEEP_CAPS_LOCKED) | |
1153 | * | issecure_mask(SECURE_NOROOT) | |
1154 | * | issecure_mask(SECURE_NOROOT_LOCKED) | |
1155 | * | issecure_mask(SECURE_NO_SETUID_FIXUP) | |
1156 | * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED) | |
1157 | * | |
1158 | * will ensure that the current process and all of its | |
1159 | * children will be locked into a pure | |
1160 | * capability-based-privilege environment. | |
1161 | */ | |
1162 | case PR_SET_SECUREBITS: | |
6d6f3328 TH |
1163 | if ((((old->securebits & SECURE_ALL_LOCKS) >> 1) |
1164 | & (old->securebits ^ arg2)) /*[1]*/ | |
1165 | || ((old->securebits & SECURE_ALL_LOCKS & ~arg2)) /*[2]*/ | |
d84f4f99 | 1166 | || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS)) /*[3]*/ |
6a9de491 | 1167 | || (cap_capable(current_cred(), |
c4a4d603 | 1168 | current_cred()->user_ns, CAP_SETPCAP, |
3699c53c | 1169 | SECURITY_CAP_AUDIT) != 0) /*[4]*/ |
3898b1b4 AM |
1170 | /* |
1171 | * [1] no changing of bits that are locked | |
1172 | * [2] no unlocking of locks | |
1173 | * [3] no setting of unsupported bits | |
1174 | * [4] doing anything requires privilege (go read about | |
1175 | * the "sendmail capabilities bug") | |
1176 | */ | |
d84f4f99 DH |
1177 | ) |
1178 | /* cannot change a locked bit */ | |
6d6f3328 TH |
1179 | return -EPERM; |
1180 | ||
1181 | new = prepare_creds(); | |
1182 | if (!new) | |
1183 | return -ENOMEM; | |
d84f4f99 | 1184 | new->securebits = arg2; |
6d6f3328 | 1185 | return commit_creds(new); |
d84f4f99 | 1186 | |
3898b1b4 | 1187 | case PR_GET_SECUREBITS: |
6d6f3328 | 1188 | return old->securebits; |
3898b1b4 | 1189 | |
3898b1b4 | 1190 | case PR_GET_KEEPCAPS: |
6d6f3328 | 1191 | return !!issecure(SECURE_KEEP_CAPS); |
d84f4f99 | 1192 | |
3898b1b4 AM |
1193 | case PR_SET_KEEPCAPS: |
1194 | if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */ | |
6d6f3328 | 1195 | return -EINVAL; |
d84f4f99 | 1196 | if (issecure(SECURE_KEEP_CAPS_LOCKED)) |
6d6f3328 TH |
1197 | return -EPERM; |
1198 | ||
1199 | new = prepare_creds(); | |
1200 | if (!new) | |
1201 | return -ENOMEM; | |
d84f4f99 DH |
1202 | if (arg2) |
1203 | new->securebits |= issecure_mask(SECURE_KEEP_CAPS); | |
3898b1b4 | 1204 | else |
d84f4f99 | 1205 | new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS); |
6d6f3328 | 1206 | return commit_creds(new); |
3898b1b4 | 1207 | |
58319057 AL |
1208 | case PR_CAP_AMBIENT: |
1209 | if (arg2 == PR_CAP_AMBIENT_CLEAR_ALL) { | |
1210 | if (arg3 | arg4 | arg5) | |
1211 | return -EINVAL; | |
1212 | ||
1213 | new = prepare_creds(); | |
1214 | if (!new) | |
1215 | return -ENOMEM; | |
1216 | cap_clear(new->cap_ambient); | |
1217 | return commit_creds(new); | |
1218 | } | |
1219 | ||
1220 | if (((!cap_valid(arg3)) | arg4 | arg5)) | |
1221 | return -EINVAL; | |
1222 | ||
1223 | if (arg2 == PR_CAP_AMBIENT_IS_SET) { | |
1224 | return !!cap_raised(current_cred()->cap_ambient, arg3); | |
1225 | } else if (arg2 != PR_CAP_AMBIENT_RAISE && | |
1226 | arg2 != PR_CAP_AMBIENT_LOWER) { | |
1227 | return -EINVAL; | |
1228 | } else { | |
1229 | if (arg2 == PR_CAP_AMBIENT_RAISE && | |
1230 | (!cap_raised(current_cred()->cap_permitted, arg3) || | |
1231 | !cap_raised(current_cred()->cap_inheritable, | |
746bf6d6 AL |
1232 | arg3) || |
1233 | issecure(SECURE_NO_CAP_AMBIENT_RAISE))) | |
58319057 AL |
1234 | return -EPERM; |
1235 | ||
1236 | new = prepare_creds(); | |
1237 | if (!new) | |
1238 | return -ENOMEM; | |
1239 | if (arg2 == PR_CAP_AMBIENT_RAISE) | |
1240 | cap_raise(new->cap_ambient, arg3); | |
1241 | else | |
1242 | cap_lower(new->cap_ambient, arg3); | |
1243 | return commit_creds(new); | |
1244 | } | |
1245 | ||
3898b1b4 AM |
1246 | default: |
1247 | /* No functionality available - continue with default */ | |
6d6f3328 | 1248 | return -ENOSYS; |
3898b1b4 | 1249 | } |
1da177e4 LT |
1250 | } |
1251 | ||
1d045980 DH |
1252 | /** |
1253 | * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted | |
1254 | * @mm: The VM space in which the new mapping is to be made | |
1255 | * @pages: The size of the mapping | |
1256 | * | |
1257 | * Determine whether the allocation of a new virtual mapping by the current | |
b1d9e6b0 | 1258 | * task is permitted, returning 1 if permission is granted, 0 if not. |
1d045980 | 1259 | */ |
34b4e4aa | 1260 | int cap_vm_enough_memory(struct mm_struct *mm, long pages) |
1da177e4 LT |
1261 | { |
1262 | int cap_sys_admin = 0; | |
1263 | ||
6a9de491 | 1264 | if (cap_capable(current_cred(), &init_user_ns, CAP_SYS_ADMIN, |
3699c53c | 1265 | SECURITY_CAP_NOAUDIT) == 0) |
1da177e4 | 1266 | cap_sys_admin = 1; |
b1d9e6b0 | 1267 | return cap_sys_admin; |
1da177e4 | 1268 | } |
7c73875e EP |
1269 | |
1270 | /* | |
d007794a | 1271 | * cap_mmap_addr - check if able to map given addr |
7c73875e | 1272 | * @addr: address attempting to be mapped |
7c73875e | 1273 | * |
6f262d8e | 1274 | * If the process is attempting to map memory below dac_mmap_min_addr they need |
7c73875e EP |
1275 | * CAP_SYS_RAWIO. The other parameters to this function are unused by the |
1276 | * capability security module. Returns 0 if this mapping should be allowed | |
1277 | * -EPERM if not. | |
1278 | */ | |
d007794a | 1279 | int cap_mmap_addr(unsigned long addr) |
7c73875e EP |
1280 | { |
1281 | int ret = 0; | |
1282 | ||
a2551df7 | 1283 | if (addr < dac_mmap_min_addr) { |
6a9de491 | 1284 | ret = cap_capable(current_cred(), &init_user_ns, CAP_SYS_RAWIO, |
7c73875e EP |
1285 | SECURITY_CAP_AUDIT); |
1286 | /* set PF_SUPERPRIV if it turns out we allow the low mmap */ | |
1287 | if (ret == 0) | |
1288 | current->flags |= PF_SUPERPRIV; | |
1289 | } | |
1290 | return ret; | |
1291 | } | |
d007794a | 1292 | |
e5467859 AV |
1293 | int cap_mmap_file(struct file *file, unsigned long reqprot, |
1294 | unsigned long prot, unsigned long flags) | |
d007794a | 1295 | { |
e5467859 | 1296 | return 0; |
d007794a | 1297 | } |
b1d9e6b0 CS |
1298 | |
1299 | #ifdef CONFIG_SECURITY | |
1300 | ||
ca97d939 | 1301 | struct security_hook_list capability_hooks[] __lsm_ro_after_init = { |
b1d9e6b0 CS |
1302 | LSM_HOOK_INIT(capable, cap_capable), |
1303 | LSM_HOOK_INIT(settime, cap_settime), | |
1304 | LSM_HOOK_INIT(ptrace_access_check, cap_ptrace_access_check), | |
1305 | LSM_HOOK_INIT(ptrace_traceme, cap_ptrace_traceme), | |
1306 | LSM_HOOK_INIT(capget, cap_capget), | |
1307 | LSM_HOOK_INIT(capset, cap_capset), | |
1308 | LSM_HOOK_INIT(bprm_set_creds, cap_bprm_set_creds), | |
b1d9e6b0 CS |
1309 | LSM_HOOK_INIT(inode_need_killpriv, cap_inode_need_killpriv), |
1310 | LSM_HOOK_INIT(inode_killpriv, cap_inode_killpriv), | |
8db6c34f | 1311 | LSM_HOOK_INIT(inode_getsecurity, cap_inode_getsecurity), |
b1d9e6b0 CS |
1312 | LSM_HOOK_INIT(mmap_addr, cap_mmap_addr), |
1313 | LSM_HOOK_INIT(mmap_file, cap_mmap_file), | |
1314 | LSM_HOOK_INIT(task_fix_setuid, cap_task_fix_setuid), | |
1315 | LSM_HOOK_INIT(task_prctl, cap_task_prctl), | |
1316 | LSM_HOOK_INIT(task_setscheduler, cap_task_setscheduler), | |
1317 | LSM_HOOK_INIT(task_setioprio, cap_task_setioprio), | |
1318 | LSM_HOOK_INIT(task_setnice, cap_task_setnice), | |
1319 | LSM_HOOK_INIT(vm_enough_memory, cap_vm_enough_memory), | |
1320 | }; | |
1321 | ||
1322 | void __init capability_add_hooks(void) | |
1323 | { | |
d69dece5 CS |
1324 | security_add_hooks(capability_hooks, ARRAY_SIZE(capability_hooks), |
1325 | "capability"); | |
b1d9e6b0 CS |
1326 | } |
1327 | ||
1328 | #endif /* CONFIG_SECURITY */ |