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