move mount_capable() calls to vfs_get_tree()
[linux-2.6-block.git] / fs / fsopen.c
CommitLineData
24dcb3d9
DH
1/* Filesystem access-by-fd.
2 *
3 * Copyright (C) 2017 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/fs_context.h>
ecdab150 13#include <linux/fs_parser.h>
24dcb3d9
DH
14#include <linux/slab.h>
15#include <linux/uaccess.h>
16#include <linux/syscalls.h>
17#include <linux/security.h>
18#include <linux/anon_inodes.h>
19#include <linux/namei.h>
20#include <linux/file.h>
21#include <uapi/linux/mount.h>
ecdab150 22#include "internal.h"
24dcb3d9
DH
23#include "mount.h"
24
007ec26c
DH
25/*
26 * Allow the user to read back any error, warning or informational messages.
27 */
28static ssize_t fscontext_read(struct file *file,
29 char __user *_buf, size_t len, loff_t *pos)
30{
31 struct fs_context *fc = file->private_data;
32 struct fc_log *log = fc->log;
33 unsigned int logsize = ARRAY_SIZE(log->buffer);
34 ssize_t ret;
35 char *p;
36 bool need_free;
37 int index, n;
38
39 ret = mutex_lock_interruptible(&fc->uapi_mutex);
40 if (ret < 0)
41 return ret;
42
43 if (log->head == log->tail) {
44 mutex_unlock(&fc->uapi_mutex);
45 return -ENODATA;
46 }
47
48 index = log->tail & (logsize - 1);
49 p = log->buffer[index];
50 need_free = log->need_free & (1 << index);
51 log->buffer[index] = NULL;
52 log->need_free &= ~(1 << index);
53 log->tail++;
54 mutex_unlock(&fc->uapi_mutex);
55
56 ret = -EMSGSIZE;
57 n = strlen(p);
58 if (n > len)
59 goto err_free;
60 ret = -EFAULT;
61 if (copy_to_user(_buf, p, n) != 0)
62 goto err_free;
63 ret = n;
64
65err_free:
66 if (need_free)
67 kfree(p);
68 return ret;
69}
70
24dcb3d9
DH
71static int fscontext_release(struct inode *inode, struct file *file)
72{
73 struct fs_context *fc = file->private_data;
74
75 if (fc) {
76 file->private_data = NULL;
77 put_fs_context(fc);
78 }
79 return 0;
80}
81
82const struct file_operations fscontext_fops = {
007ec26c 83 .read = fscontext_read,
24dcb3d9
DH
84 .release = fscontext_release,
85 .llseek = no_llseek,
86};
87
88/*
89 * Attach a filesystem context to a file and an fd.
90 */
91static int fscontext_create_fd(struct fs_context *fc, unsigned int o_flags)
92{
93 int fd;
94
1cdc415f 95 fd = anon_inode_getfd("[fscontext]", &fscontext_fops, fc,
24dcb3d9
DH
96 O_RDWR | o_flags);
97 if (fd < 0)
98 put_fs_context(fc);
99 return fd;
100}
101
007ec26c
DH
102static int fscontext_alloc_log(struct fs_context *fc)
103{
104 fc->log = kzalloc(sizeof(*fc->log), GFP_KERNEL);
105 if (!fc->log)
106 return -ENOMEM;
107 refcount_set(&fc->log->usage, 1);
108 fc->log->owner = fc->fs_type->owner;
109 return 0;
110}
111
24dcb3d9
DH
112/*
113 * Open a filesystem by name so that it can be configured for mounting.
114 *
115 * We are allowed to specify a container in which the filesystem will be
116 * opened, thereby indicating which namespaces will be used (notably, which
117 * network namespace will be used for network filesystems).
118 */
119SYSCALL_DEFINE2(fsopen, const char __user *, _fs_name, unsigned int, flags)
120{
121 struct file_system_type *fs_type;
122 struct fs_context *fc;
123 const char *fs_name;
007ec26c 124 int ret;
24dcb3d9
DH
125
126 if (!ns_capable(current->nsproxy->mnt_ns->user_ns, CAP_SYS_ADMIN))
127 return -EPERM;
128
129 if (flags & ~FSOPEN_CLOEXEC)
130 return -EINVAL;
131
132 fs_name = strndup_user(_fs_name, PAGE_SIZE);
133 if (IS_ERR(fs_name))
134 return PTR_ERR(fs_name);
135
136 fs_type = get_fs_type(fs_name);
137 kfree(fs_name);
138 if (!fs_type)
139 return -ENODEV;
140
141 fc = fs_context_for_mount(fs_type, 0);
142 put_filesystem(fs_type);
143 if (IS_ERR(fc))
144 return PTR_ERR(fc);
145
146 fc->phase = FS_CONTEXT_CREATE_PARAMS;
007ec26c
DH
147
148 ret = fscontext_alloc_log(fc);
149 if (ret < 0)
150 goto err_fc;
151
24dcb3d9 152 return fscontext_create_fd(fc, flags & FSOPEN_CLOEXEC ? O_CLOEXEC : 0);
007ec26c
DH
153
154err_fc:
155 put_fs_context(fc);
156 return ret;
24dcb3d9 157}
ecdab150 158
cf3cba4a
DH
159/*
160 * Pick a superblock into a context for reconfiguration.
161 */
162SYSCALL_DEFINE3(fspick, int, dfd, const char __user *, path, unsigned int, flags)
163{
164 struct fs_context *fc;
165 struct path target;
166 unsigned int lookup_flags;
167 int ret;
168
169 if (!ns_capable(current->nsproxy->mnt_ns->user_ns, CAP_SYS_ADMIN))
170 return -EPERM;
171
172 if ((flags & ~(FSPICK_CLOEXEC |
173 FSPICK_SYMLINK_NOFOLLOW |
174 FSPICK_NO_AUTOMOUNT |
175 FSPICK_EMPTY_PATH)) != 0)
176 return -EINVAL;
177
178 lookup_flags = LOOKUP_FOLLOW | LOOKUP_AUTOMOUNT;
179 if (flags & FSPICK_SYMLINK_NOFOLLOW)
180 lookup_flags &= ~LOOKUP_FOLLOW;
181 if (flags & FSPICK_NO_AUTOMOUNT)
182 lookup_flags &= ~LOOKUP_AUTOMOUNT;
183 if (flags & FSPICK_EMPTY_PATH)
184 lookup_flags |= LOOKUP_EMPTY;
185 ret = user_path_at(dfd, path, lookup_flags, &target);
186 if (ret < 0)
187 goto err;
188
189 ret = -EINVAL;
190 if (target.mnt->mnt_root != target.dentry)
191 goto err_path;
192
193 fc = fs_context_for_reconfigure(target.dentry, 0, 0);
194 if (IS_ERR(fc)) {
195 ret = PTR_ERR(fc);
196 goto err_path;
197 }
198
199 fc->phase = FS_CONTEXT_RECONF_PARAMS;
200
201 ret = fscontext_alloc_log(fc);
202 if (ret < 0)
203 goto err_fc;
204
205 path_put(&target);
206 return fscontext_create_fd(fc, flags & FSPICK_CLOEXEC ? O_CLOEXEC : 0);
207
208err_fc:
209 put_fs_context(fc);
210err_path:
211 path_put(&target);
212err:
213 return ret;
214}
215
ecdab150
DH
216/*
217 * Check the state and apply the configuration. Note that this function is
218 * allowed to 'steal' the value by setting param->xxx to NULL before returning.
219 */
220static int vfs_fsconfig_locked(struct fs_context *fc, int cmd,
221 struct fs_parameter *param)
222{
223 struct super_block *sb;
224 int ret;
225
226 ret = finish_clean_context(fc);
227 if (ret)
228 return ret;
229 switch (cmd) {
230 case FSCONFIG_CMD_CREATE:
231 if (fc->phase != FS_CONTEXT_CREATE_PARAMS)
232 return -EBUSY;
233 fc->phase = FS_CONTEXT_CREATING;
234 ret = vfs_get_tree(fc);
235 if (ret)
236 break;
237 sb = fc->root->d_sb;
238 ret = security_sb_kern_mount(sb);
239 if (unlikely(ret)) {
240 fc_drop_locked(fc);
241 break;
242 }
243 up_write(&sb->s_umount);
244 fc->phase = FS_CONTEXT_AWAITING_MOUNT;
245 return 0;
246 case FSCONFIG_CMD_RECONFIGURE:
247 if (fc->phase != FS_CONTEXT_RECONF_PARAMS)
248 return -EBUSY;
249 fc->phase = FS_CONTEXT_RECONFIGURING;
250 sb = fc->root->d_sb;
251 if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN)) {
252 ret = -EPERM;
253 break;
254 }
255 down_write(&sb->s_umount);
256 ret = reconfigure_super(fc);
257 up_write(&sb->s_umount);
258 if (ret)
259 break;
260 vfs_clean_context(fc);
261 return 0;
262 default:
263 if (fc->phase != FS_CONTEXT_CREATE_PARAMS &&
264 fc->phase != FS_CONTEXT_RECONF_PARAMS)
265 return -EBUSY;
266
267 return vfs_parse_fs_param(fc, param);
268 }
269 fc->phase = FS_CONTEXT_FAILED;
270 return ret;
271}
272
273/**
274 * sys_fsconfig - Set parameters and trigger actions on a context
275 * @fd: The filesystem context to act upon
276 * @cmd: The action to take
277 * @_key: Where appropriate, the parameter key to set
278 * @_value: Where appropriate, the parameter value to set
279 * @aux: Additional information for the value
280 *
281 * This system call is used to set parameters on a context, including
282 * superblock settings, data source and security labelling.
283 *
284 * Actions include triggering the creation of a superblock and the
285 * reconfiguration of the superblock attached to the specified context.
286 *
287 * When setting a parameter, @cmd indicates the type of value being proposed
288 * and @_key indicates the parameter to be altered.
289 *
290 * @_value and @aux are used to specify the value, should a value be required:
291 *
292 * (*) fsconfig_set_flag: No value is specified. The parameter must be boolean
293 * in nature. The key may be prefixed with "no" to invert the
294 * setting. @_value must be NULL and @aux must be 0.
295 *
296 * (*) fsconfig_set_string: A string value is specified. The parameter can be
297 * expecting boolean, integer, string or take a path. A conversion to an
298 * appropriate type will be attempted (which may include looking up as a
299 * path). @_value points to a NUL-terminated string and @aux must be 0.
300 *
301 * (*) fsconfig_set_binary: A binary blob is specified. @_value points to the
302 * blob and @aux indicates its size. The parameter must be expecting a
303 * blob.
304 *
305 * (*) fsconfig_set_path: A non-empty path is specified. The parameter must be
306 * expecting a path object. @_value points to a NUL-terminated string that
307 * is the path and @aux is a file descriptor at which to start a relative
308 * lookup or AT_FDCWD.
309 *
310 * (*) fsconfig_set_path_empty: As fsconfig_set_path, but with AT_EMPTY_PATH
311 * implied.
312 *
313 * (*) fsconfig_set_fd: An open file descriptor is specified. @_value must be
314 * NULL and @aux indicates the file descriptor.
315 */
316SYSCALL_DEFINE5(fsconfig,
317 int, fd,
318 unsigned int, cmd,
319 const char __user *, _key,
320 const void __user *, _value,
321 int, aux)
322{
323 struct fs_context *fc;
324 struct fd f;
325 int ret;
326
327 struct fs_parameter param = {
328 .type = fs_value_is_undefined,
329 };
330
331 if (fd < 0)
332 return -EINVAL;
333
334 switch (cmd) {
335 case FSCONFIG_SET_FLAG:
336 if (!_key || _value || aux)
337 return -EINVAL;
338 break;
339 case FSCONFIG_SET_STRING:
340 if (!_key || !_value || aux)
341 return -EINVAL;
342 break;
343 case FSCONFIG_SET_BINARY:
344 if (!_key || !_value || aux <= 0 || aux > 1024 * 1024)
345 return -EINVAL;
346 break;
347 case FSCONFIG_SET_PATH:
348 case FSCONFIG_SET_PATH_EMPTY:
349 if (!_key || !_value || (aux != AT_FDCWD && aux < 0))
350 return -EINVAL;
351 break;
352 case FSCONFIG_SET_FD:
353 if (!_key || _value || aux < 0)
354 return -EINVAL;
355 break;
356 case FSCONFIG_CMD_CREATE:
357 case FSCONFIG_CMD_RECONFIGURE:
358 if (_key || _value || aux)
359 return -EINVAL;
360 break;
361 default:
362 return -EOPNOTSUPP;
363 }
364
365 f = fdget(fd);
366 if (!f.file)
367 return -EBADF;
368 ret = -EINVAL;
369 if (f.file->f_op != &fscontext_fops)
370 goto out_f;
371
372 fc = f.file->private_data;
373 if (fc->ops == &legacy_fs_context_ops) {
374 switch (cmd) {
375 case FSCONFIG_SET_BINARY:
376 case FSCONFIG_SET_PATH:
377 case FSCONFIG_SET_PATH_EMPTY:
378 case FSCONFIG_SET_FD:
379 ret = -EOPNOTSUPP;
380 goto out_f;
381 }
382 }
383
384 if (_key) {
385 param.key = strndup_user(_key, 256);
386 if (IS_ERR(param.key)) {
387 ret = PTR_ERR(param.key);
388 goto out_f;
389 }
390 }
391
392 switch (cmd) {
393 case FSCONFIG_SET_FLAG:
394 param.type = fs_value_is_flag;
395 break;
396 case FSCONFIG_SET_STRING:
397 param.type = fs_value_is_string;
398 param.string = strndup_user(_value, 256);
399 if (IS_ERR(param.string)) {
400 ret = PTR_ERR(param.string);
401 goto out_key;
402 }
403 param.size = strlen(param.string);
404 break;
405 case FSCONFIG_SET_BINARY:
406 param.type = fs_value_is_blob;
407 param.size = aux;
408 param.blob = memdup_user_nul(_value, aux);
409 if (IS_ERR(param.blob)) {
410 ret = PTR_ERR(param.blob);
411 goto out_key;
412 }
413 break;
414 case FSCONFIG_SET_PATH:
415 param.type = fs_value_is_filename;
416 param.name = getname_flags(_value, 0, NULL);
417 if (IS_ERR(param.name)) {
418 ret = PTR_ERR(param.name);
419 goto out_key;
420 }
421 param.dirfd = aux;
422 param.size = strlen(param.name->name);
423 break;
424 case FSCONFIG_SET_PATH_EMPTY:
425 param.type = fs_value_is_filename_empty;
426 param.name = getname_flags(_value, LOOKUP_EMPTY, NULL);
427 if (IS_ERR(param.name)) {
428 ret = PTR_ERR(param.name);
429 goto out_key;
430 }
431 param.dirfd = aux;
432 param.size = strlen(param.name->name);
433 break;
434 case FSCONFIG_SET_FD:
435 param.type = fs_value_is_file;
436 ret = -EBADF;
437 param.file = fget(aux);
438 if (!param.file)
439 goto out_key;
440 break;
441 default:
442 break;
443 }
444
445 ret = mutex_lock_interruptible(&fc->uapi_mutex);
446 if (ret == 0) {
447 ret = vfs_fsconfig_locked(fc, cmd, &param);
448 mutex_unlock(&fc->uapi_mutex);
449 }
450
451 /* Clean up the our record of any value that we obtained from
452 * userspace. Note that the value may have been stolen by the LSM or
453 * filesystem, in which case the value pointer will have been cleared.
454 */
455 switch (cmd) {
456 case FSCONFIG_SET_STRING:
457 case FSCONFIG_SET_BINARY:
458 kfree(param.string);
459 break;
460 case FSCONFIG_SET_PATH:
461 case FSCONFIG_SET_PATH_EMPTY:
462 if (param.name)
463 putname(param.name);
464 break;
465 case FSCONFIG_SET_FD:
466 if (param.file)
467 fput(param.file);
468 break;
469 default:
470 break;
471 }
472out_key:
473 kfree(param.key);
474out_f:
475 fdput(f);
476 return ret;
477}