1// SPDX-License-Identifier: GPL-2.0
2/*
3 * linux/kernel/capability.c
4 *
5 * Copyright (C) 1997 Andrew Main <zefram@fysh.org>
6 *
7 * Integrated into 2.1.97+, Andrew G. Morgan <morgan@kernel.org>
8 * 30 May 2002: Cleanup, Robert M. Love <rml@tech9.net>
9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/audit.h>
14#include <linux/capability.h>
15#include <linux/mm.h>
16#include <linux/export.h>
17#include <linux/security.h>
18#include <linux/syscalls.h>
19#include <linux/pid_namespace.h>
20#include <linux/user_namespace.h>
21#include <linux/uaccess.h>
22
23int file_caps_enabled = 1;
24
25static int __init file_caps_disable(char *str)
26{
27 file_caps_enabled = 0;
28 return 1;
29}
30__setup("no_file_caps", file_caps_disable);
31
32#ifdef CONFIG_MULTIUSER
33/*
34 * More recent versions of libcap are available from:
35 *
36 * http://www.kernel.org/pub/linux/libs/security/linux-privs/
37 */
38
39static void warn_legacy_capability_use(void)
40{
41 pr_info_once("warning: `%s' uses 32-bit capabilities (legacy support in use)\n",
42 current->comm);
43}
44
45/*
46 * Version 2 capabilities worked fine, but the linux/capability.h file
47 * that accompanied their introduction encouraged their use without
48 * the necessary user-space source code changes. As such, we have
49 * created a version 3 with equivalent functionality to version 2, but
50 * with a header change to protect legacy source code from using
51 * version 2 when it wanted to use version 1. If your system has code
52 * that trips the following warning, it is using version 2 specific
53 * capabilities and may be doing so insecurely.
54 *
55 * The remedy is to either upgrade your version of libcap (to 2.10+,
56 * if the application is linked against it), or recompile your
57 * application with modern kernel headers and this warning will go
58 * away.
59 */
60
61static void warn_deprecated_v2(void)
62{
63 pr_info_once("warning: `%s' uses deprecated v2 capabilities in a way that may be insecure\n",
64 current->comm);
65}
66
67/*
68 * Version check. Return the number of u32s in each capability flag
69 * array, or a negative value on error.
70 */
71static int cap_validate_magic(cap_user_header_t header, unsigned *tocopy)
72{
73 __u32 version;
74
75 if (get_user(version, &header->version))
76 return -EFAULT;
77
78 switch (version) {
79 case _LINUX_CAPABILITY_VERSION_1:
80 warn_legacy_capability_use();
81 *tocopy = _LINUX_CAPABILITY_U32S_1;
82 break;
83 case _LINUX_CAPABILITY_VERSION_2:
84 warn_deprecated_v2();
85 fallthrough; /* v3 is otherwise equivalent to v2 */
86 case _LINUX_CAPABILITY_VERSION_3:
87 *tocopy = _LINUX_CAPABILITY_U32S_3;
88 break;
89 default:
90 if (put_user((u32)_KERNEL_CAPABILITY_VERSION, &header->version))
91 return -EFAULT;
92 return -EINVAL;
93 }
94
95 return 0;
96}
97
98/*
99 * The only thing that can change the capabilities of the current
100 * process is the current process. As such, we can't be in this code
101 * at the same time as we are in the process of setting capabilities
102 * in this process. The net result is that we can limit our use of
103 * locks to when we are reading the caps of another process.
104 */
105static inline int cap_get_target_pid(pid_t pid, kernel_cap_t *pEp,
106 kernel_cap_t *pIp, kernel_cap_t *pPp)
107{
108 int ret;
109
110 if (pid && (pid != task_pid_vnr(current))) {
111 const struct task_struct *target;
112
113 rcu_read_lock();
114
115 target = find_task_by_vpid(nr: pid);
116 if (!target)
117 ret = -ESRCH;
118 else
119 ret = security_capget(target, effective: pEp, inheritable: pIp, permitted: pPp);
120
121 rcu_read_unlock();
122 } else
123 ret = security_capget(current, effective: pEp, inheritable: pIp, permitted: pPp);
124
125 return ret;
126}
127
128/**
129 * sys_capget - get the capabilities of a given process.
130 * @header: pointer to struct that contains capability version and
131 * target pid data
132 * @dataptr: pointer to struct that contains the effective, permitted,
133 * and inheritable capabilities that are returned
134 *
135 * Returns 0 on success and < 0 on error.
136 */
137SYSCALL_DEFINE2(capget, cap_user_header_t, header, cap_user_data_t, dataptr)
138{
139 int ret = 0;
140 pid_t pid;
141 unsigned tocopy;
142 kernel_cap_t pE, pI, pP;
143 struct __user_cap_data_struct kdata[2];
144
145 ret = cap_validate_magic(header, tocopy: &tocopy);
146 if ((dataptr == NULL) || (ret != 0))
147 return ((dataptr == NULL) && (ret == -EINVAL)) ? 0 : ret;
148
149 if (get_user(pid, &header->pid))
150 return -EFAULT;
151
152 if (pid < 0)
153 return -EINVAL;
154
155 ret = cap_get_target_pid(pid, pEp: &pE, pIp: &pI, pPp: &pP);
156 if (ret)
157 return ret;
158
159 /*
160 * Annoying legacy format with 64-bit capabilities exposed
161 * as two sets of 32-bit fields, so we need to split the
162 * capability values up.
163 */
164 kdata[0].effective = pE.val; kdata[1].effective = pE.val >> 32;
165 kdata[0].permitted = pP.val; kdata[1].permitted = pP.val >> 32;
166 kdata[0].inheritable = pI.val; kdata[1].inheritable = pI.val >> 32;
167
168 /*
169 * Note, in the case, tocopy < _KERNEL_CAPABILITY_U32S,
170 * we silently drop the upper capabilities here. This
171 * has the effect of making older libcap
172 * implementations implicitly drop upper capability
173 * bits when they perform a: capget/modify/capset
174 * sequence.
175 *
176 * This behavior is considered fail-safe
177 * behavior. Upgrading the application to a newer
178 * version of libcap will enable access to the newer
179 * capabilities.
180 *
181 * An alternative would be to return an error here
182 * (-ERANGE), but that causes legacy applications to
183 * unexpectedly fail; the capget/modify/capset aborts
184 * before modification is attempted and the application
185 * fails.
186 */
187 if (copy_to_user(to: dataptr, from: kdata, n: tocopy * sizeof(kdata[0])))
188 return -EFAULT;
189
190 return 0;
191}
192
193static kernel_cap_t mk_kernel_cap(u32 low, u32 high)
194{
195 return (kernel_cap_t) { (low | ((u64)high << 32)) & CAP_VALID_MASK };
196}
197
198/**
199 * sys_capset - set capabilities for a process or (*) a group of processes
200 * @header: pointer to struct that contains capability version and
201 * target pid data
202 * @data: pointer to struct that contains the effective, permitted,
203 * and inheritable capabilities
204 *
205 * Set capabilities for the current process only. The ability to any other
206 * process(es) has been deprecated and removed.
207 *
208 * The restrictions on setting capabilities are specified as:
209 *
210 * I: any raised capabilities must be a subset of the old permitted
211 * P: any raised capabilities must be a subset of the old permitted
212 * E: must be set to a subset of new permitted
213 *
214 * Returns 0 on success and < 0 on error.
215 */
216SYSCALL_DEFINE2(capset, cap_user_header_t, header, const cap_user_data_t, data)
217{
218 struct __user_cap_data_struct kdata[2] = { { 0, }, };
219 unsigned tocopy, copybytes;
220 kernel_cap_t inheritable, permitted, effective;
221 struct cred *new;
222 int ret;
223 pid_t pid;
224
225 ret = cap_validate_magic(header, tocopy: &tocopy);
226 if (ret != 0)
227 return ret;
228
229 if (get_user(pid, &header->pid))
230 return -EFAULT;
231
232 /* may only affect current now */
233 if (pid != 0 && pid != task_pid_vnr(current))
234 return -EPERM;
235
236 copybytes = tocopy * sizeof(struct __user_cap_data_struct);
237 if (copybytes > sizeof(kdata))
238 return -EFAULT;
239
240 if (copy_from_user(to: &kdata, from: data, n: copybytes))
241 return -EFAULT;
242
243 effective = mk_kernel_cap(low: kdata[0].effective, high: kdata[1].effective);
244 permitted = mk_kernel_cap(low: kdata[0].permitted, high: kdata[1].permitted);
245 inheritable = mk_kernel_cap(low: kdata[0].inheritable, high: kdata[1].inheritable);
246
247 new = prepare_creds();
248 if (!new)
249 return -ENOMEM;
250
251 ret = security_capset(new, current_cred(),
252 effective: &effective, inheritable: &inheritable, permitted: &permitted);
253 if (ret < 0)
254 goto error;
255
256 audit_log_capset(new, current_cred());
257
258 return commit_creds(new);
259
260error:
261 abort_creds(new);
262 return ret;
263}
264
265/**
266 * has_ns_capability - Does a task have a capability in a specific user ns
267 * @t: The task in question
268 * @ns: target user namespace
269 * @cap: The capability to be tested for
270 *
271 * Return true if the specified task has the given superior capability
272 * currently in effect to the specified user namespace, false if not.
273 *
274 * Note that this does not set PF_SUPERPRIV on the task.
275 */
276bool has_ns_capability(struct task_struct *t,
277 struct user_namespace *ns, int cap)
278{
279 int ret;
280
281 rcu_read_lock();
282 ret = security_capable(__task_cred(t), ns, cap, CAP_OPT_NONE);
283 rcu_read_unlock();
284
285 return (ret == 0);
286}
287
288/**
289 * has_ns_capability_noaudit - Does a task have a capability (unaudited)
290 * in a specific user ns.
291 * @t: The task in question
292 * @ns: target user namespace
293 * @cap: The capability to be tested for
294 *
295 * Return true if the specified task has the given superior capability
296 * currently in effect to the specified user namespace, false if not.
297 * Do not write an audit message for the check.
298 *
299 * Note that this does not set PF_SUPERPRIV on the task.
300 */
301bool has_ns_capability_noaudit(struct task_struct *t,
302 struct user_namespace *ns, int cap)
303{
304 int ret;
305
306 rcu_read_lock();
307 ret = security_capable(__task_cred(t), ns, cap, CAP_OPT_NOAUDIT);
308 rcu_read_unlock();
309
310 return (ret == 0);
311}
312
313/**
314 * has_capability_noaudit - Does a task have a capability (unaudited) in the
315 * initial user ns
316 * @t: The task in question
317 * @cap: The capability to be tested for
318 *
319 * Return true if the specified task has the given superior capability
320 * currently in effect to init_user_ns, false if not. Don't write an
321 * audit message for the check.
322 *
323 * Note that this does not set PF_SUPERPRIV on the task.
324 */
325bool has_capability_noaudit(struct task_struct *t, int cap)
326{
327 return has_ns_capability_noaudit(t, ns: &init_user_ns, cap);
328}
329EXPORT_SYMBOL(has_capability_noaudit);
330
331static bool ns_capable_common(struct user_namespace *ns,
332 int cap,
333 unsigned int opts)
334{
335 int capable;
336
337 if (unlikely(!cap_valid(cap))) {
338 pr_crit("capable() called with invalid cap=%u\n", cap);
339 BUG();
340 }
341
342 capable = security_capable(current_cred(), ns, cap, opts);
343 if (capable == 0) {
344 current->flags |= PF_SUPERPRIV;
345 return true;
346 }
347 return false;
348}
349
350/**
351 * ns_capable - Determine if the current task has a superior capability in effect
352 * @ns: The usernamespace we want the capability in
353 * @cap: The capability to be tested for
354 *
355 * Return true if the current task has the given superior capability currently
356 * available for use, false if not.
357 *
358 * This sets PF_SUPERPRIV on the task if the capability is available on the
359 * assumption that it's about to be used.
360 */
361bool ns_capable(struct user_namespace *ns, int cap)
362{
363 return ns_capable_common(ns, cap, CAP_OPT_NONE);
364}
365EXPORT_SYMBOL(ns_capable);
366
367/**
368 * ns_capable_noaudit - Determine if the current task has a superior capability
369 * (unaudited) in effect
370 * @ns: The usernamespace we want the capability in
371 * @cap: The capability to be tested for
372 *
373 * Return true if the current task has the given superior capability currently
374 * available for use, false if not.
375 *
376 * This sets PF_SUPERPRIV on the task if the capability is available on the
377 * assumption that it's about to be used.
378 */
379bool ns_capable_noaudit(struct user_namespace *ns, int cap)
380{
381 return ns_capable_common(ns, cap, CAP_OPT_NOAUDIT);
382}
383EXPORT_SYMBOL(ns_capable_noaudit);
384
385/**
386 * ns_capable_setid - Determine if the current task has a superior capability
387 * in effect, while signalling that this check is being done from within a
388 * setid or setgroups syscall.
389 * @ns: The usernamespace we want the capability in
390 * @cap: The capability to be tested for
391 *
392 * Return true if the current task has the given superior capability currently
393 * available for use, false if not.
394 *
395 * This sets PF_SUPERPRIV on the task if the capability is available on the
396 * assumption that it's about to be used.
397 */
398bool ns_capable_setid(struct user_namespace *ns, int cap)
399{
400 return ns_capable_common(ns, cap, CAP_OPT_INSETID);
401}
402EXPORT_SYMBOL(ns_capable_setid);
403
404/**
405 * capable - Determine if the current task has a superior capability in effect
406 * @cap: The capability to be tested for
407 *
408 * Return true if the current task has the given superior capability currently
409 * available for use, false if not.
410 *
411 * This sets PF_SUPERPRIV on the task if the capability is available on the
412 * assumption that it's about to be used.
413 */
414bool capable(int cap)
415{
416 return ns_capable(&init_user_ns, cap);
417}
418EXPORT_SYMBOL(capable);
419#endif /* CONFIG_MULTIUSER */
420
421/**
422 * file_ns_capable - Determine if the file's opener had a capability in effect
423 * @file: The file we want to check
424 * @ns: The usernamespace we want the capability in
425 * @cap: The capability to be tested for
426 *
427 * Return true if task that opened the file had a capability in effect
428 * when the file was opened.
429 *
430 * This does not set PF_SUPERPRIV because the caller may not
431 * actually be privileged.
432 */
433bool file_ns_capable(const struct file *file, struct user_namespace *ns,
434 int cap)
435{
436
437 if (WARN_ON_ONCE(!cap_valid(cap)))
438 return false;
439
440 if (security_capable(cred: file->f_cred, ns, cap, CAP_OPT_NONE) == 0)
441 return true;
442
443 return false;
444}
445EXPORT_SYMBOL(file_ns_capable);
446
447/**
448 * privileged_wrt_inode_uidgid - Do capabilities in the namespace work over the inode?
449 * @ns: The user namespace in question
450 * @idmap: idmap of the mount @inode was found from
451 * @inode: The inode in question
452 *
453 * Return true if the inode uid and gid are within the namespace.
454 */
455bool privileged_wrt_inode_uidgid(struct user_namespace *ns,
456 struct mnt_idmap *idmap,
457 const struct inode *inode)
458{
459 return vfsuid_has_mapping(userns: ns, vfsuid: i_uid_into_vfsuid(idmap, inode)) &&
460 vfsgid_has_mapping(userns: ns, vfsgid: i_gid_into_vfsgid(idmap, inode));
461}
462
463/**
464 * capable_wrt_inode_uidgid - Check nsown_capable and uid and gid mapped
465 * @idmap: idmap of the mount @inode was found from
466 * @inode: The inode in question
467 * @cap: The capability in question
468 *
469 * Return true if the current task has the given capability targeted at
470 * its own user namespace and that the given inode's uid and gid are
471 * mapped into the current user namespace.
472 */
473bool capable_wrt_inode_uidgid(struct mnt_idmap *idmap,
474 const struct inode *inode, int cap)
475{
476 struct user_namespace *ns = current_user_ns();
477
478 return ns_capable(ns, cap) &&
479 privileged_wrt_inode_uidgid(ns, idmap, inode);
480}
481EXPORT_SYMBOL(capable_wrt_inode_uidgid);
482
483/**
484 * ptracer_capable - Determine if the ptracer holds CAP_SYS_PTRACE in the namespace
485 * @tsk: The task that may be ptraced
486 * @ns: The user namespace to search for CAP_SYS_PTRACE in
487 *
488 * Return true if the task that is ptracing the current task had CAP_SYS_PTRACE
489 * in the specified user namespace.
490 */
491bool ptracer_capable(struct task_struct *tsk, struct user_namespace *ns)
492{
493 int ret = 0; /* An absent tracer adds no restrictions */
494 const struct cred *cred;
495
496 rcu_read_lock();
497 cred = rcu_dereference(tsk->ptracer_cred);
498 if (cred)
499 ret = security_capable(cred, ns, CAP_SYS_PTRACE,
500 CAP_OPT_NOAUDIT);
501 rcu_read_unlock();
502 return (ret == 0);
503}
504