Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
77b14db5 EB |
2 | /* |
3 | * /proc/sys support | |
4 | */ | |
1e0edd3f | 5 | #include <linux/init.h> |
77b14db5 | 6 | #include <linux/sysctl.h> |
f1ecf068 | 7 | #include <linux/poll.h> |
77b14db5 | 8 | #include <linux/proc_fs.h> |
87ebdc00 | 9 | #include <linux/printk.h> |
77b14db5 | 10 | #include <linux/security.h> |
40401530 | 11 | #include <linux/sched.h> |
5b825c3a | 12 | #include <linux/cred.h> |
34286d66 | 13 | #include <linux/namei.h> |
40401530 | 14 | #include <linux/mm.h> |
4bd6a735 | 15 | #include <linux/uio.h> |
1f87f0b5 | 16 | #include <linux/module.h> |
7b146ceb | 17 | #include <linux/bpf-cgroup.h> |
3db978d4 | 18 | #include <linux/mount.h> |
3ddd9a80 | 19 | #include <linux/kmemleak.h> |
77b14db5 EB |
20 | #include "internal.h" |
21 | ||
cb55f27a MT |
22 | #define list_for_each_table_entry(entry, table) \ |
23 | for ((entry) = (table); (entry)->procname; (entry)++) | |
24 | ||
d72f71eb | 25 | static const struct dentry_operations proc_sys_dentry_operations; |
77b14db5 | 26 | static const struct file_operations proc_sys_file_operations; |
03a44825 | 27 | static const struct inode_operations proc_sys_inode_operations; |
9043476f AV |
28 | static const struct file_operations proc_sys_dir_file_operations; |
29 | static const struct inode_operations proc_sys_dir_operations; | |
77b14db5 | 30 | |
f9bd6733 EB |
31 | /* Support for permanently empty directories */ |
32 | ||
33 | struct ctl_table sysctl_mount_point[] = { | |
34 | { } | |
35 | }; | |
36 | ||
ee9efac4 LC |
37 | /** |
38 | * register_sysctl_mount_point() - registers a sysctl mount point | |
39 | * @path: path for the mount point | |
40 | * | |
41 | * Used to create a permanently empty directory to serve as mount point. | |
42 | * There are some subtle but important permission checks this allows in the | |
43 | * case of unprivileged mounts. | |
44 | */ | |
45 | struct ctl_table_header *register_sysctl_mount_point(const char *path) | |
46 | { | |
47 | return register_sysctl(path, sysctl_mount_point); | |
48 | } | |
49 | EXPORT_SYMBOL(register_sysctl_mount_point); | |
50 | ||
f9bd6733 EB |
51 | static bool is_empty_dir(struct ctl_table_header *head) |
52 | { | |
53 | return head->ctl_table[0].child == sysctl_mount_point; | |
54 | } | |
55 | ||
56 | static void set_empty_dir(struct ctl_dir *dir) | |
57 | { | |
58 | dir->header.ctl_table[0].child = sysctl_mount_point; | |
59 | } | |
60 | ||
61 | static void clear_empty_dir(struct ctl_dir *dir) | |
62 | ||
63 | { | |
64 | dir->header.ctl_table[0].child = NULL; | |
65 | } | |
66 | ||
f1ecf068 LDM |
67 | void proc_sys_poll_notify(struct ctl_table_poll *poll) |
68 | { | |
69 | if (!poll) | |
70 | return; | |
71 | ||
72 | atomic_inc(&poll->event); | |
73 | wake_up_interruptible(&poll->wait); | |
74 | } | |
75 | ||
a194558e EB |
76 | static struct ctl_table root_table[] = { |
77 | { | |
78 | .procname = "", | |
7ec66d06 | 79 | .mode = S_IFDIR|S_IRUGO|S_IXUGO, |
a194558e EB |
80 | }, |
81 | { } | |
82 | }; | |
0e47c99d | 83 | static struct ctl_table_root sysctl_table_root = { |
0e47c99d | 84 | .default_set.dir.header = { |
7ec66d06 EB |
85 | {{.count = 1, |
86 | .nreg = 1, | |
ac13ac6f | 87 | .ctl_table = root_table }}, |
0e47c99d | 88 | .ctl_table_arg = root_table, |
7ec66d06 EB |
89 | .root = &sysctl_table_root, |
90 | .set = &sysctl_table_root.default_set, | |
91 | }, | |
1f87f0b5 | 92 | }; |
1f87f0b5 EB |
93 | |
94 | static DEFINE_SPINLOCK(sysctl_lock); | |
95 | ||
7ec66d06 | 96 | static void drop_sysctl_table(struct ctl_table_header *header); |
0e47c99d | 97 | static int sysctl_follow_link(struct ctl_table_header **phead, |
13bcc6a2 | 98 | struct ctl_table **pentry); |
0e47c99d EB |
99 | static int insert_links(struct ctl_table_header *head); |
100 | static void put_links(struct ctl_table_header *header); | |
7ec66d06 | 101 | |
6980128f EB |
102 | static void sysctl_print_dir(struct ctl_dir *dir) |
103 | { | |
104 | if (dir->header.parent) | |
105 | sysctl_print_dir(dir->header.parent); | |
87ebdc00 | 106 | pr_cont("%s/", dir->header.ctl_table[0].procname); |
6980128f EB |
107 | } |
108 | ||
076c3eed EB |
109 | static int namecmp(const char *name1, int len1, const char *name2, int len2) |
110 | { | |
076c3eed EB |
111 | int cmp; |
112 | ||
d8fc9b66 | 113 | cmp = memcmp(name1, name2, min(len1, len2)); |
076c3eed EB |
114 | if (cmp == 0) |
115 | cmp = len1 - len2; | |
116 | return cmp; | |
117 | } | |
118 | ||
60f126d9 | 119 | /* Called under sysctl_lock */ |
076c3eed | 120 | static struct ctl_table *find_entry(struct ctl_table_header **phead, |
0e47c99d | 121 | struct ctl_dir *dir, const char *name, int namelen) |
076c3eed EB |
122 | { |
123 | struct ctl_table_header *head; | |
124 | struct ctl_table *entry; | |
ac13ac6f | 125 | struct rb_node *node = dir->root.rb_node; |
076c3eed | 126 | |
ac13ac6f EB |
127 | while (node) |
128 | { | |
129 | struct ctl_node *ctl_node; | |
130 | const char *procname; | |
131 | int cmp; | |
132 | ||
133 | ctl_node = rb_entry(node, struct ctl_node, node); | |
134 | head = ctl_node->header; | |
135 | entry = &head->ctl_table[ctl_node - head->node]; | |
136 | procname = entry->procname; | |
137 | ||
138 | cmp = namecmp(name, namelen, procname, strlen(procname)); | |
139 | if (cmp < 0) | |
140 | node = node->rb_left; | |
141 | else if (cmp > 0) | |
142 | node = node->rb_right; | |
143 | else { | |
144 | *phead = head; | |
145 | return entry; | |
076c3eed EB |
146 | } |
147 | } | |
148 | return NULL; | |
149 | } | |
150 | ||
ac13ac6f EB |
151 | static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry) |
152 | { | |
153 | struct rb_node *node = &head->node[entry - head->ctl_table].node; | |
154 | struct rb_node **p = &head->parent->root.rb_node; | |
155 | struct rb_node *parent = NULL; | |
156 | const char *name = entry->procname; | |
157 | int namelen = strlen(name); | |
158 | ||
159 | while (*p) { | |
160 | struct ctl_table_header *parent_head; | |
161 | struct ctl_table *parent_entry; | |
162 | struct ctl_node *parent_node; | |
163 | const char *parent_name; | |
164 | int cmp; | |
165 | ||
166 | parent = *p; | |
167 | parent_node = rb_entry(parent, struct ctl_node, node); | |
168 | parent_head = parent_node->header; | |
169 | parent_entry = &parent_head->ctl_table[parent_node - parent_head->node]; | |
170 | parent_name = parent_entry->procname; | |
171 | ||
172 | cmp = namecmp(name, namelen, parent_name, strlen(parent_name)); | |
173 | if (cmp < 0) | |
174 | p = &(*p)->rb_left; | |
175 | else if (cmp > 0) | |
176 | p = &(*p)->rb_right; | |
177 | else { | |
87ebdc00 | 178 | pr_err("sysctl duplicate entry: "); |
ac13ac6f | 179 | sysctl_print_dir(head->parent); |
153ee1c4 | 180 | pr_cont("%s\n", entry->procname); |
ac13ac6f EB |
181 | return -EEXIST; |
182 | } | |
183 | } | |
184 | ||
185 | rb_link_node(node, parent, p); | |
ea5272f5 | 186 | rb_insert_color(node, &head->parent->root); |
ac13ac6f EB |
187 | return 0; |
188 | } | |
189 | ||
190 | static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry) | |
191 | { | |
192 | struct rb_node *node = &head->node[entry - head->ctl_table].node; | |
193 | ||
194 | rb_erase(node, &head->parent->root); | |
195 | } | |
196 | ||
e0d04529 EB |
197 | static void init_header(struct ctl_table_header *head, |
198 | struct ctl_table_root *root, struct ctl_table_set *set, | |
ac13ac6f | 199 | struct ctl_node *node, struct ctl_table *table) |
e0d04529 | 200 | { |
7ec66d06 | 201 | head->ctl_table = table; |
e0d04529 | 202 | head->ctl_table_arg = table; |
e0d04529 EB |
203 | head->used = 0; |
204 | head->count = 1; | |
205 | head->nreg = 1; | |
206 | head->unregistering = NULL; | |
207 | head->root = root; | |
208 | head->set = set; | |
209 | head->parent = NULL; | |
ac13ac6f | 210 | head->node = node; |
2fd1d2c4 | 211 | INIT_HLIST_HEAD(&head->inodes); |
ac13ac6f EB |
212 | if (node) { |
213 | struct ctl_table *entry; | |
cb55f27a MT |
214 | |
215 | list_for_each_table_entry(entry, table) { | |
ac13ac6f | 216 | node->header = head; |
cb55f27a MT |
217 | node++; |
218 | } | |
ac13ac6f | 219 | } |
e0d04529 EB |
220 | } |
221 | ||
8425d6aa EB |
222 | static void erase_header(struct ctl_table_header *head) |
223 | { | |
ac13ac6f | 224 | struct ctl_table *entry; |
cb55f27a MT |
225 | |
226 | list_for_each_table_entry(entry, head->ctl_table) | |
ac13ac6f | 227 | erase_entry(head, entry); |
8425d6aa EB |
228 | } |
229 | ||
0e47c99d | 230 | static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header) |
8425d6aa | 231 | { |
ac13ac6f | 232 | struct ctl_table *entry; |
0e47c99d EB |
233 | int err; |
234 | ||
f9bd6733 EB |
235 | /* Is this a permanently empty directory? */ |
236 | if (is_empty_dir(&dir->header)) | |
237 | return -EROFS; | |
238 | ||
239 | /* Am I creating a permanently empty directory? */ | |
240 | if (header->ctl_table == sysctl_mount_point) { | |
241 | if (!RB_EMPTY_ROOT(&dir->root)) | |
242 | return -EINVAL; | |
243 | set_empty_dir(dir); | |
244 | } | |
245 | ||
0e47c99d | 246 | dir->header.nreg++; |
7ec66d06 | 247 | header->parent = dir; |
0e47c99d EB |
248 | err = insert_links(header); |
249 | if (err) | |
250 | goto fail_links; | |
cb55f27a | 251 | list_for_each_table_entry(entry, header->ctl_table) { |
ac13ac6f EB |
252 | err = insert_entry(header, entry); |
253 | if (err) | |
254 | goto fail; | |
255 | } | |
0e47c99d | 256 | return 0; |
ac13ac6f EB |
257 | fail: |
258 | erase_header(header); | |
259 | put_links(header); | |
0e47c99d | 260 | fail_links: |
f9bd6733 EB |
261 | if (header->ctl_table == sysctl_mount_point) |
262 | clear_empty_dir(dir); | |
0e47c99d EB |
263 | header->parent = NULL; |
264 | drop_sysctl_table(&dir->header); | |
265 | return err; | |
8425d6aa EB |
266 | } |
267 | ||
1f87f0b5 EB |
268 | /* called under sysctl_lock */ |
269 | static int use_table(struct ctl_table_header *p) | |
270 | { | |
271 | if (unlikely(p->unregistering)) | |
272 | return 0; | |
273 | p->used++; | |
274 | return 1; | |
275 | } | |
276 | ||
277 | /* called under sysctl_lock */ | |
278 | static void unuse_table(struct ctl_table_header *p) | |
279 | { | |
280 | if (!--p->used) | |
281 | if (unlikely(p->unregistering)) | |
282 | complete(p->unregistering); | |
283 | } | |
284 | ||
f90f3caf | 285 | static void proc_sys_invalidate_dcache(struct ctl_table_header *head) |
d6cffbbe | 286 | { |
f90f3caf | 287 | proc_invalidate_siblings_dcache(&head->inodes, &sysctl_lock); |
d6cffbbe KK |
288 | } |
289 | ||
1f87f0b5 EB |
290 | /* called under sysctl_lock, will reacquire if has to wait */ |
291 | static void start_unregistering(struct ctl_table_header *p) | |
292 | { | |
293 | /* | |
294 | * if p->used is 0, nobody will ever touch that entry again; | |
295 | * we'll eliminate all paths to it before dropping sysctl_lock | |
296 | */ | |
297 | if (unlikely(p->used)) { | |
298 | struct completion wait; | |
299 | init_completion(&wait); | |
300 | p->unregistering = &wait; | |
301 | spin_unlock(&sysctl_lock); | |
302 | wait_for_completion(&wait); | |
1f87f0b5 EB |
303 | } else { |
304 | /* anything non-NULL; we'll never dereference it */ | |
305 | p->unregistering = ERR_PTR(-EINVAL); | |
ace0c791 | 306 | spin_unlock(&sysctl_lock); |
1f87f0b5 | 307 | } |
d6cffbbe | 308 | /* |
f90f3caf | 309 | * Invalidate dentries for unregistered sysctls: namespaced sysctls |
d6cffbbe KK |
310 | * can have duplicate names and contaminate dcache very badly. |
311 | */ | |
f90f3caf | 312 | proc_sys_invalidate_dcache(p); |
1f87f0b5 EB |
313 | /* |
314 | * do not remove from the list until nobody holds it; walking the | |
315 | * list in do_sysctl() relies on that. | |
316 | */ | |
ace0c791 | 317 | spin_lock(&sysctl_lock); |
8425d6aa | 318 | erase_header(p); |
1f87f0b5 EB |
319 | } |
320 | ||
1f87f0b5 EB |
321 | static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head) |
322 | { | |
ab4a1f24 | 323 | BUG_ON(!head); |
1f87f0b5 EB |
324 | spin_lock(&sysctl_lock); |
325 | if (!use_table(head)) | |
326 | head = ERR_PTR(-ENOENT); | |
327 | spin_unlock(&sysctl_lock); | |
328 | return head; | |
329 | } | |
330 | ||
331 | static void sysctl_head_finish(struct ctl_table_header *head) | |
332 | { | |
333 | if (!head) | |
334 | return; | |
335 | spin_lock(&sysctl_lock); | |
336 | unuse_table(head); | |
337 | spin_unlock(&sysctl_lock); | |
338 | } | |
339 | ||
340 | static struct ctl_table_set * | |
13bcc6a2 | 341 | lookup_header_set(struct ctl_table_root *root) |
1f87f0b5 EB |
342 | { |
343 | struct ctl_table_set *set = &root->default_set; | |
344 | if (root->lookup) | |
13bcc6a2 | 345 | set = root->lookup(root); |
1f87f0b5 EB |
346 | return set; |
347 | } | |
348 | ||
076c3eed | 349 | static struct ctl_table *lookup_entry(struct ctl_table_header **phead, |
7ec66d06 | 350 | struct ctl_dir *dir, |
076c3eed EB |
351 | const char *name, int namelen) |
352 | { | |
353 | struct ctl_table_header *head; | |
354 | struct ctl_table *entry; | |
076c3eed EB |
355 | |
356 | spin_lock(&sysctl_lock); | |
0e47c99d EB |
357 | entry = find_entry(&head, dir, name, namelen); |
358 | if (entry && use_table(head)) | |
359 | *phead = head; | |
360 | else | |
361 | entry = NULL; | |
076c3eed EB |
362 | spin_unlock(&sysctl_lock); |
363 | return entry; | |
364 | } | |
365 | ||
ac13ac6f | 366 | static struct ctl_node *first_usable_entry(struct rb_node *node) |
1f87f0b5 | 367 | { |
ac13ac6f | 368 | struct ctl_node *ctl_node; |
1f87f0b5 | 369 | |
ac13ac6f EB |
370 | for (;node; node = rb_next(node)) { |
371 | ctl_node = rb_entry(node, struct ctl_node, node); | |
372 | if (use_table(ctl_node->header)) | |
373 | return ctl_node; | |
1f87f0b5 | 374 | } |
1f87f0b5 EB |
375 | return NULL; |
376 | } | |
377 | ||
7ec66d06 | 378 | static void first_entry(struct ctl_dir *dir, |
6a75ce16 EB |
379 | struct ctl_table_header **phead, struct ctl_table **pentry) |
380 | { | |
ac13ac6f | 381 | struct ctl_table_header *head = NULL; |
7ec66d06 | 382 | struct ctl_table *entry = NULL; |
ac13ac6f | 383 | struct ctl_node *ctl_node; |
6a75ce16 EB |
384 | |
385 | spin_lock(&sysctl_lock); | |
ac13ac6f | 386 | ctl_node = first_usable_entry(rb_first(&dir->root)); |
6a75ce16 | 387 | spin_unlock(&sysctl_lock); |
ac13ac6f EB |
388 | if (ctl_node) { |
389 | head = ctl_node->header; | |
390 | entry = &head->ctl_table[ctl_node - head->node]; | |
391 | } | |
6a75ce16 EB |
392 | *phead = head; |
393 | *pentry = entry; | |
394 | } | |
395 | ||
7ec66d06 | 396 | static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry) |
1f87f0b5 | 397 | { |
6a75ce16 EB |
398 | struct ctl_table_header *head = *phead; |
399 | struct ctl_table *entry = *pentry; | |
ac13ac6f | 400 | struct ctl_node *ctl_node = &head->node[entry - head->ctl_table]; |
6a75ce16 | 401 | |
ac13ac6f EB |
402 | spin_lock(&sysctl_lock); |
403 | unuse_table(head); | |
404 | ||
405 | ctl_node = first_usable_entry(rb_next(&ctl_node->node)); | |
406 | spin_unlock(&sysctl_lock); | |
407 | head = NULL; | |
408 | if (ctl_node) { | |
409 | head = ctl_node->header; | |
410 | entry = &head->ctl_table[ctl_node - head->node]; | |
6a75ce16 EB |
411 | } |
412 | *phead = head; | |
413 | *pentry = entry; | |
1f87f0b5 EB |
414 | } |
415 | ||
1f87f0b5 EB |
416 | /* |
417 | * sysctl_perm does NOT grant the superuser all rights automatically, because | |
418 | * some sysctl variables are readonly even to root. | |
419 | */ | |
420 | ||
421 | static int test_perm(int mode, int op) | |
422 | { | |
091bd3ea | 423 | if (uid_eq(current_euid(), GLOBAL_ROOT_UID)) |
1f87f0b5 | 424 | mode >>= 6; |
091bd3ea | 425 | else if (in_egroup_p(GLOBAL_ROOT_GID)) |
1f87f0b5 EB |
426 | mode >>= 3; |
427 | if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0) | |
428 | return 0; | |
429 | return -EACCES; | |
430 | } | |
431 | ||
73f7ef43 | 432 | static int sysctl_perm(struct ctl_table_header *head, struct ctl_table *table, int op) |
1f87f0b5 | 433 | { |
73f7ef43 | 434 | struct ctl_table_root *root = head->root; |
1f87f0b5 EB |
435 | int mode; |
436 | ||
437 | if (root->permissions) | |
73f7ef43 | 438 | mode = root->permissions(head, table); |
1f87f0b5 EB |
439 | else |
440 | mode = table->mode; | |
441 | ||
442 | return test_perm(mode, op); | |
443 | } | |
444 | ||
9043476f AV |
445 | static struct inode *proc_sys_make_inode(struct super_block *sb, |
446 | struct ctl_table_header *head, struct ctl_table *table) | |
77b14db5 | 447 | { |
e79c6a4f | 448 | struct ctl_table_root *root = head->root; |
77b14db5 | 449 | struct inode *inode; |
9043476f | 450 | struct proc_inode *ei; |
77b14db5 | 451 | |
9043476f | 452 | inode = new_inode(sb); |
77b14db5 | 453 | if (!inode) |
ea5751cc | 454 | return ERR_PTR(-ENOMEM); |
77b14db5 | 455 | |
85fe4025 CH |
456 | inode->i_ino = get_next_ino(); |
457 | ||
77b14db5 | 458 | ei = PROC_I(inode); |
9043476f | 459 | |
d6cffbbe | 460 | spin_lock(&sysctl_lock); |
ace0c791 EB |
461 | if (unlikely(head->unregistering)) { |
462 | spin_unlock(&sysctl_lock); | |
463 | iput(inode); | |
ea5751cc | 464 | return ERR_PTR(-ENOENT); |
ace0c791 EB |
465 | } |
466 | ei->sysctl = head; | |
467 | ei->sysctl_entry = table; | |
0afa5ca8 | 468 | hlist_add_head_rcu(&ei->sibling_inodes, &head->inodes); |
d6cffbbe KK |
469 | head->count++; |
470 | spin_unlock(&sysctl_lock); | |
471 | ||
078cd827 | 472 | inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); |
9043476f | 473 | inode->i_mode = table->mode; |
7ec66d06 | 474 | if (!S_ISDIR(table->mode)) { |
9043476f AV |
475 | inode->i_mode |= S_IFREG; |
476 | inode->i_op = &proc_sys_inode_operations; | |
477 | inode->i_fop = &proc_sys_file_operations; | |
478 | } else { | |
479 | inode->i_mode |= S_IFDIR; | |
9043476f AV |
480 | inode->i_op = &proc_sys_dir_operations; |
481 | inode->i_fop = &proc_sys_dir_file_operations; | |
f9bd6733 EB |
482 | if (is_empty_dir(head)) |
483 | make_empty_dir_inode(inode); | |
9043476f | 484 | } |
e79c6a4f DT |
485 | |
486 | if (root->set_ownership) | |
487 | root->set_ownership(head, table, &inode->i_uid, &inode->i_gid); | |
5ec27ec7 RB |
488 | else { |
489 | inode->i_uid = GLOBAL_ROOT_UID; | |
490 | inode->i_gid = GLOBAL_ROOT_GID; | |
491 | } | |
e79c6a4f | 492 | |
77b14db5 EB |
493 | return inode; |
494 | } | |
495 | ||
d6cffbbe KK |
496 | void proc_sys_evict_inode(struct inode *inode, struct ctl_table_header *head) |
497 | { | |
498 | spin_lock(&sysctl_lock); | |
0afa5ca8 | 499 | hlist_del_init_rcu(&PROC_I(inode)->sibling_inodes); |
d6cffbbe KK |
500 | if (!--head->count) |
501 | kfree_rcu(head, rcu); | |
502 | spin_unlock(&sysctl_lock); | |
503 | } | |
504 | ||
81324364 | 505 | static struct ctl_table_header *grab_header(struct inode *inode) |
77b14db5 | 506 | { |
3cc3e046 EB |
507 | struct ctl_table_header *head = PROC_I(inode)->sysctl; |
508 | if (!head) | |
0e47c99d | 509 | head = &sysctl_table_root.default_set.dir.header; |
3cc3e046 | 510 | return sysctl_head_grab(head); |
9043476f | 511 | } |
77b14db5 | 512 | |
9043476f | 513 | static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry, |
00cd8dd3 | 514 | unsigned int flags) |
9043476f AV |
515 | { |
516 | struct ctl_table_header *head = grab_header(dir); | |
9043476f | 517 | struct ctl_table_header *h = NULL; |
dc12e909 | 518 | const struct qstr *name = &dentry->d_name; |
9043476f AV |
519 | struct ctl_table *p; |
520 | struct inode *inode; | |
521 | struct dentry *err = ERR_PTR(-ENOENT); | |
7ec66d06 | 522 | struct ctl_dir *ctl_dir; |
0e47c99d | 523 | int ret; |
77b14db5 | 524 | |
9043476f AV |
525 | if (IS_ERR(head)) |
526 | return ERR_CAST(head); | |
77b14db5 | 527 | |
7ec66d06 | 528 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 | 529 | |
7ec66d06 | 530 | p = lookup_entry(&h, ctl_dir, name->name, name->len); |
9043476f | 531 | if (!p) |
77b14db5 EB |
532 | goto out; |
533 | ||
4e757320 | 534 | if (S_ISLNK(p->mode)) { |
13bcc6a2 | 535 | ret = sysctl_follow_link(&h, &p); |
4e757320 EB |
536 | err = ERR_PTR(ret); |
537 | if (ret) | |
538 | goto out; | |
539 | } | |
0e47c99d | 540 | |
9043476f | 541 | inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p); |
ea5751cc ID |
542 | if (IS_ERR(inode)) { |
543 | err = ERR_CAST(inode); | |
77b14db5 | 544 | goto out; |
ea5751cc | 545 | } |
77b14db5 | 546 | |
fb045adb | 547 | d_set_d_op(dentry, &proc_sys_dentry_operations); |
888e2b03 | 548 | err = d_splice_alias(inode, dentry); |
77b14db5 EB |
549 | |
550 | out: | |
6bf61045 FR |
551 | if (h) |
552 | sysctl_head_finish(h); | |
77b14db5 EB |
553 | sysctl_head_finish(head); |
554 | return err; | |
555 | } | |
556 | ||
4bd6a735 MWO |
557 | static ssize_t proc_sys_call_handler(struct kiocb *iocb, struct iov_iter *iter, |
558 | int write) | |
77b14db5 | 559 | { |
4bd6a735 | 560 | struct inode *inode = file_inode(iocb->ki_filp); |
9043476f AV |
561 | struct ctl_table_header *head = grab_header(inode); |
562 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
4bd6a735 MWO |
563 | size_t count = iov_iter_count(iter); |
564 | char *kbuf; | |
2a2da53b | 565 | ssize_t error; |
77b14db5 | 566 | |
9043476f AV |
567 | if (IS_ERR(head)) |
568 | return PTR_ERR(head); | |
77b14db5 EB |
569 | |
570 | /* | |
571 | * At this point we know that the sysctl was not unregistered | |
572 | * and won't be until we finish. | |
573 | */ | |
574 | error = -EPERM; | |
73f7ef43 | 575 | if (sysctl_perm(head, table, write ? MAY_WRITE : MAY_READ)) |
77b14db5 EB |
576 | goto out; |
577 | ||
9043476f AV |
578 | /* if that can happen at all, it should be -EINVAL, not -EISDIR */ |
579 | error = -EINVAL; | |
580 | if (!table->proc_handler) | |
581 | goto out; | |
582 | ||
ef9d965b | 583 | /* don't even try if the size is too large */ |
d4d80e69 MWO |
584 | error = -ENOMEM; |
585 | if (count >= KMALLOC_MAX_SIZE) | |
586 | goto out; | |
45089437 | 587 | kbuf = kvzalloc(count + 1, GFP_KERNEL); |
4bd6a735 MWO |
588 | if (!kbuf) |
589 | goto out; | |
ef9d965b | 590 | |
32927393 | 591 | if (write) { |
4bd6a735 MWO |
592 | error = -EFAULT; |
593 | if (!copy_from_iter_full(kbuf, count, iter)) | |
594 | goto out_free_buf; | |
595 | kbuf[count] = '\0'; | |
32927393 CH |
596 | } |
597 | ||
598 | error = BPF_CGROUP_RUN_PROG_SYSCTL(head, table, write, &kbuf, &count, | |
4bd6a735 | 599 | &iocb->ki_pos); |
7b146ceb | 600 | if (error) |
32927393 | 601 | goto out_free_buf; |
7b146ceb | 602 | |
77b14db5 | 603 | /* careful: calling conventions are nasty here */ |
4bd6a735 | 604 | error = table->proc_handler(table, write, kbuf, &count, &iocb->ki_pos); |
32927393 CH |
605 | if (error) |
606 | goto out_free_buf; | |
607 | ||
608 | if (!write) { | |
609 | error = -EFAULT; | |
4bd6a735 | 610 | if (copy_to_iter(kbuf, count, iter) < count) |
32927393 | 611 | goto out_free_buf; |
4e63acdf AI |
612 | } |
613 | ||
32927393 CH |
614 | error = count; |
615 | out_free_buf: | |
45089437 | 616 | kvfree(kbuf); |
77b14db5 EB |
617 | out: |
618 | sysctl_head_finish(head); | |
619 | ||
620 | return error; | |
621 | } | |
622 | ||
4bd6a735 | 623 | static ssize_t proc_sys_read(struct kiocb *iocb, struct iov_iter *iter) |
77b14db5 | 624 | { |
4bd6a735 | 625 | return proc_sys_call_handler(iocb, iter, 0); |
7708bfb1 | 626 | } |
77b14db5 | 627 | |
4bd6a735 | 628 | static ssize_t proc_sys_write(struct kiocb *iocb, struct iov_iter *iter) |
7708bfb1 | 629 | { |
4bd6a735 | 630 | return proc_sys_call_handler(iocb, iter, 1); |
77b14db5 EB |
631 | } |
632 | ||
f1ecf068 LDM |
633 | static int proc_sys_open(struct inode *inode, struct file *filp) |
634 | { | |
4e474a00 | 635 | struct ctl_table_header *head = grab_header(inode); |
f1ecf068 LDM |
636 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; |
637 | ||
4e474a00 LDM |
638 | /* sysctl was unregistered */ |
639 | if (IS_ERR(head)) | |
640 | return PTR_ERR(head); | |
641 | ||
f1ecf068 LDM |
642 | if (table->poll) |
643 | filp->private_data = proc_sys_poll_event(table->poll); | |
644 | ||
4e474a00 LDM |
645 | sysctl_head_finish(head); |
646 | ||
f1ecf068 LDM |
647 | return 0; |
648 | } | |
649 | ||
076ccb76 | 650 | static __poll_t proc_sys_poll(struct file *filp, poll_table *wait) |
f1ecf068 | 651 | { |
496ad9aa | 652 | struct inode *inode = file_inode(filp); |
4e474a00 | 653 | struct ctl_table_header *head = grab_header(inode); |
f1ecf068 | 654 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; |
076ccb76 | 655 | __poll_t ret = DEFAULT_POLLMASK; |
4e474a00 LDM |
656 | unsigned long event; |
657 | ||
658 | /* sysctl was unregistered */ | |
659 | if (IS_ERR(head)) | |
a9a08845 | 660 | return EPOLLERR | EPOLLHUP; |
f1ecf068 LDM |
661 | |
662 | if (!table->proc_handler) | |
663 | goto out; | |
664 | ||
665 | if (!table->poll) | |
666 | goto out; | |
667 | ||
4e474a00 | 668 | event = (unsigned long)filp->private_data; |
f1ecf068 LDM |
669 | poll_wait(filp, &table->poll->wait, wait); |
670 | ||
671 | if (event != atomic_read(&table->poll->event)) { | |
672 | filp->private_data = proc_sys_poll_event(table->poll); | |
a9a08845 | 673 | ret = EPOLLIN | EPOLLRDNORM | EPOLLERR | EPOLLPRI; |
f1ecf068 LDM |
674 | } |
675 | ||
676 | out: | |
4e474a00 LDM |
677 | sysctl_head_finish(head); |
678 | ||
f1ecf068 LDM |
679 | return ret; |
680 | } | |
77b14db5 | 681 | |
f0c3b509 AV |
682 | static bool proc_sys_fill_cache(struct file *file, |
683 | struct dir_context *ctx, | |
9043476f AV |
684 | struct ctl_table_header *head, |
685 | struct ctl_table *table) | |
77b14db5 | 686 | { |
f0c3b509 | 687 | struct dentry *child, *dir = file->f_path.dentry; |
77b14db5 EB |
688 | struct inode *inode; |
689 | struct qstr qname; | |
690 | ino_t ino = 0; | |
691 | unsigned type = DT_UNKNOWN; | |
77b14db5 EB |
692 | |
693 | qname.name = table->procname; | |
694 | qname.len = strlen(table->procname); | |
8387ff25 | 695 | qname.hash = full_name_hash(dir, qname.name, qname.len); |
77b14db5 | 696 | |
77b14db5 EB |
697 | child = d_lookup(dir, &qname); |
698 | if (!child) { | |
76aab3ab AV |
699 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
700 | child = d_alloc_parallel(dir, &qname, &wq); | |
701 | if (IS_ERR(child)) | |
702 | return false; | |
703 | if (d_in_lookup(child)) { | |
888e2b03 | 704 | struct dentry *res; |
9043476f | 705 | inode = proc_sys_make_inode(dir->d_sb, head, table); |
ea5751cc | 706 | if (IS_ERR(inode)) { |
76aab3ab | 707 | d_lookup_done(child); |
9043476f | 708 | dput(child); |
f0c3b509 | 709 | return false; |
77b14db5 | 710 | } |
76aab3ab | 711 | d_set_d_op(child, &proc_sys_dentry_operations); |
888e2b03 AV |
712 | res = d_splice_alias(inode, child); |
713 | d_lookup_done(child); | |
714 | if (unlikely(res)) { | |
715 | if (IS_ERR(res)) { | |
716 | dput(child); | |
717 | return false; | |
718 | } | |
719 | dput(child); | |
720 | child = res; | |
721 | } | |
77b14db5 EB |
722 | } |
723 | } | |
2b0143b5 | 724 | inode = d_inode(child); |
9043476f AV |
725 | ino = inode->i_ino; |
726 | type = inode->i_mode >> 12; | |
77b14db5 | 727 | dput(child); |
f0c3b509 | 728 | return dir_emit(ctx, qname.name, qname.len, ino, type); |
9043476f AV |
729 | } |
730 | ||
f0c3b509 AV |
731 | static bool proc_sys_link_fill_cache(struct file *file, |
732 | struct dir_context *ctx, | |
0e47c99d EB |
733 | struct ctl_table_header *head, |
734 | struct ctl_table *table) | |
735 | { | |
f0c3b509 | 736 | bool ret = true; |
a0b0d1c3 | 737 | |
0e47c99d | 738 | head = sysctl_head_grab(head); |
a0b0d1c3 DK |
739 | if (IS_ERR(head)) |
740 | return false; | |
0e47c99d | 741 | |
835b94e0 DK |
742 | /* It is not an error if we can not follow the link ignore it */ |
743 | if (sysctl_follow_link(&head, &table)) | |
744 | goto out; | |
0e47c99d | 745 | |
f0c3b509 | 746 | ret = proc_sys_fill_cache(file, ctx, head, table); |
0e47c99d EB |
747 | out: |
748 | sysctl_head_finish(head); | |
749 | return ret; | |
750 | } | |
751 | ||
e5eea098 | 752 | static int scan(struct ctl_table_header *head, struct ctl_table *table, |
9043476f | 753 | unsigned long *pos, struct file *file, |
f0c3b509 | 754 | struct dir_context *ctx) |
9043476f | 755 | { |
f0c3b509 | 756 | bool res; |
9043476f | 757 | |
f0c3b509 AV |
758 | if ((*pos)++ < ctx->pos) |
759 | return true; | |
9043476f | 760 | |
0e47c99d | 761 | if (unlikely(S_ISLNK(table->mode))) |
f0c3b509 | 762 | res = proc_sys_link_fill_cache(file, ctx, head, table); |
0e47c99d | 763 | else |
f0c3b509 | 764 | res = proc_sys_fill_cache(file, ctx, head, table); |
9043476f | 765 | |
f0c3b509 AV |
766 | if (res) |
767 | ctx->pos = *pos; | |
9043476f | 768 | |
6a75ce16 | 769 | return res; |
77b14db5 EB |
770 | } |
771 | ||
f0c3b509 | 772 | static int proc_sys_readdir(struct file *file, struct dir_context *ctx) |
77b14db5 | 773 | { |
f0c3b509 | 774 | struct ctl_table_header *head = grab_header(file_inode(file)); |
9043476f | 775 | struct ctl_table_header *h = NULL; |
6a75ce16 | 776 | struct ctl_table *entry; |
7ec66d06 | 777 | struct ctl_dir *ctl_dir; |
77b14db5 | 778 | unsigned long pos; |
9043476f AV |
779 | |
780 | if (IS_ERR(head)) | |
781 | return PTR_ERR(head); | |
77b14db5 | 782 | |
7ec66d06 | 783 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 | 784 | |
f0c3b509 | 785 | if (!dir_emit_dots(file, ctx)) |
93362fa4 | 786 | goto out; |
f0c3b509 | 787 | |
77b14db5 EB |
788 | pos = 2; |
789 | ||
7ec66d06 | 790 | for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) { |
f0c3b509 | 791 | if (!scan(h, entry, &pos, file, ctx)) { |
9043476f AV |
792 | sysctl_head_finish(h); |
793 | break; | |
77b14db5 EB |
794 | } |
795 | } | |
93362fa4 | 796 | out: |
77b14db5 | 797 | sysctl_head_finish(head); |
f0c3b509 | 798 | return 0; |
77b14db5 EB |
799 | } |
800 | ||
4609e1f1 | 801 | static int proc_sys_permission(struct mnt_idmap *idmap, |
549c7297 | 802 | struct inode *inode, int mask) |
77b14db5 EB |
803 | { |
804 | /* | |
805 | * sysctl entries that are not writeable, | |
806 | * are _NOT_ writeable, capabilities or not. | |
807 | */ | |
f696a365 MS |
808 | struct ctl_table_header *head; |
809 | struct ctl_table *table; | |
77b14db5 EB |
810 | int error; |
811 | ||
f696a365 MS |
812 | /* Executable files are not allowed under /proc/sys/ */ |
813 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) | |
814 | return -EACCES; | |
815 | ||
816 | head = grab_header(inode); | |
9043476f AV |
817 | if (IS_ERR(head)) |
818 | return PTR_ERR(head); | |
77b14db5 | 819 | |
f696a365 | 820 | table = PROC_I(inode)->sysctl_entry; |
9043476f AV |
821 | if (!table) /* global root - r-xr-xr-x */ |
822 | error = mask & MAY_WRITE ? -EACCES : 0; | |
823 | else /* Use the permissions on the sysctl table entry */ | |
73f7ef43 | 824 | error = sysctl_perm(head, table, mask & ~MAY_NOT_BLOCK); |
77b14db5 | 825 | |
77b14db5 EB |
826 | sysctl_head_finish(head); |
827 | return error; | |
828 | } | |
829 | ||
c1632a0f | 830 | static int proc_sys_setattr(struct mnt_idmap *idmap, |
549c7297 | 831 | struct dentry *dentry, struct iattr *attr) |
77b14db5 | 832 | { |
2b0143b5 | 833 | struct inode *inode = d_inode(dentry); |
77b14db5 EB |
834 | int error; |
835 | ||
836 | if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) | |
837 | return -EPERM; | |
838 | ||
c1632a0f | 839 | error = setattr_prepare(&nop_mnt_idmap, dentry, attr); |
1025774c CH |
840 | if (error) |
841 | return error; | |
842 | ||
c1632a0f | 843 | setattr_copy(&nop_mnt_idmap, inode, attr); |
1025774c | 844 | return 0; |
77b14db5 EB |
845 | } |
846 | ||
b74d24f7 | 847 | static int proc_sys_getattr(struct mnt_idmap *idmap, |
549c7297 | 848 | const struct path *path, struct kstat *stat, |
a528d35e | 849 | u32 request_mask, unsigned int query_flags) |
9043476f | 850 | { |
a528d35e | 851 | struct inode *inode = d_inode(path->dentry); |
9043476f AV |
852 | struct ctl_table_header *head = grab_header(inode); |
853 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
854 | ||
855 | if (IS_ERR(head)) | |
856 | return PTR_ERR(head); | |
857 | ||
b74d24f7 | 858 | generic_fillattr(&nop_mnt_idmap, inode, stat); |
9043476f AV |
859 | if (table) |
860 | stat->mode = (stat->mode & S_IFMT) | table->mode; | |
861 | ||
862 | sysctl_head_finish(head); | |
863 | return 0; | |
864 | } | |
865 | ||
77b14db5 | 866 | static const struct file_operations proc_sys_file_operations = { |
f1ecf068 LDM |
867 | .open = proc_sys_open, |
868 | .poll = proc_sys_poll, | |
4bd6a735 MWO |
869 | .read_iter = proc_sys_read, |
870 | .write_iter = proc_sys_write, | |
b0072734 | 871 | .splice_read = copy_splice_read, |
4bd6a735 | 872 | .splice_write = iter_file_splice_write, |
6038f373 | 873 | .llseek = default_llseek, |
9043476f AV |
874 | }; |
875 | ||
876 | static const struct file_operations proc_sys_dir_file_operations = { | |
887df078 | 877 | .read = generic_read_dir, |
f50752ea | 878 | .iterate_shared = proc_sys_readdir, |
3222a3e5 | 879 | .llseek = generic_file_llseek, |
77b14db5 EB |
880 | }; |
881 | ||
03a44825 | 882 | static const struct inode_operations proc_sys_inode_operations = { |
9043476f AV |
883 | .permission = proc_sys_permission, |
884 | .setattr = proc_sys_setattr, | |
885 | .getattr = proc_sys_getattr, | |
886 | }; | |
887 | ||
888 | static const struct inode_operations proc_sys_dir_operations = { | |
77b14db5 EB |
889 | .lookup = proc_sys_lookup, |
890 | .permission = proc_sys_permission, | |
891 | .setattr = proc_sys_setattr, | |
9043476f | 892 | .getattr = proc_sys_getattr, |
77b14db5 EB |
893 | }; |
894 | ||
0b728e19 | 895 | static int proc_sys_revalidate(struct dentry *dentry, unsigned int flags) |
77b14db5 | 896 | { |
0b728e19 | 897 | if (flags & LOOKUP_RCU) |
34286d66 | 898 | return -ECHILD; |
2b0143b5 | 899 | return !PROC_I(d_inode(dentry))->sysctl->unregistering; |
9043476f AV |
900 | } |
901 | ||
fe15ce44 | 902 | static int proc_sys_delete(const struct dentry *dentry) |
9043476f | 903 | { |
2b0143b5 | 904 | return !!PROC_I(d_inode(dentry))->sysctl->unregistering; |
9043476f AV |
905 | } |
906 | ||
1f87f0b5 EB |
907 | static int sysctl_is_seen(struct ctl_table_header *p) |
908 | { | |
909 | struct ctl_table_set *set = p->set; | |
910 | int res; | |
911 | spin_lock(&sysctl_lock); | |
912 | if (p->unregistering) | |
913 | res = 0; | |
914 | else if (!set->is_seen) | |
915 | res = 1; | |
916 | else | |
917 | res = set->is_seen(set); | |
918 | spin_unlock(&sysctl_lock); | |
919 | return res; | |
920 | } | |
921 | ||
6fa67e70 | 922 | static int proc_sys_compare(const struct dentry *dentry, |
621e155a | 923 | unsigned int len, const char *str, const struct qstr *name) |
9043476f | 924 | { |
dfef6dcd | 925 | struct ctl_table_header *head; |
da53be12 LT |
926 | struct inode *inode; |
927 | ||
31e6b01f NP |
928 | /* Although proc doesn't have negative dentries, rcu-walk means |
929 | * that inode here can be NULL */ | |
dfef6dcd | 930 | /* AV: can it, indeed? */ |
2b0143b5 | 931 | inode = d_inode_rcu(dentry); |
31e6b01f | 932 | if (!inode) |
dfef6dcd | 933 | return 1; |
621e155a | 934 | if (name->len != len) |
9043476f | 935 | return 1; |
621e155a | 936 | if (memcmp(name->name, str, len)) |
9043476f | 937 | return 1; |
dfef6dcd AV |
938 | head = rcu_dereference(PROC_I(inode)->sysctl); |
939 | return !head || !sysctl_is_seen(head); | |
77b14db5 EB |
940 | } |
941 | ||
d72f71eb | 942 | static const struct dentry_operations proc_sys_dentry_operations = { |
77b14db5 | 943 | .d_revalidate = proc_sys_revalidate, |
9043476f AV |
944 | .d_delete = proc_sys_delete, |
945 | .d_compare = proc_sys_compare, | |
77b14db5 EB |
946 | }; |
947 | ||
0e47c99d EB |
948 | static struct ctl_dir *find_subdir(struct ctl_dir *dir, |
949 | const char *name, int namelen) | |
1f87f0b5 | 950 | { |
7ec66d06 EB |
951 | struct ctl_table_header *head; |
952 | struct ctl_table *entry; | |
1f87f0b5 | 953 | |
0e47c99d | 954 | entry = find_entry(&head, dir, name, namelen); |
7ec66d06 EB |
955 | if (!entry) |
956 | return ERR_PTR(-ENOENT); | |
51f72f4a EB |
957 | if (!S_ISDIR(entry->mode)) |
958 | return ERR_PTR(-ENOTDIR); | |
959 | return container_of(head, struct ctl_dir, header); | |
7ec66d06 EB |
960 | } |
961 | ||
962 | static struct ctl_dir *new_dir(struct ctl_table_set *set, | |
0e47c99d | 963 | const char *name, int namelen) |
7ec66d06 EB |
964 | { |
965 | struct ctl_table *table; | |
966 | struct ctl_dir *new; | |
ac13ac6f | 967 | struct ctl_node *node; |
7ec66d06 | 968 | char *new_name; |
1f87f0b5 | 969 | |
ac13ac6f EB |
970 | new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) + |
971 | sizeof(struct ctl_table)*2 + namelen + 1, | |
972 | GFP_KERNEL); | |
7ec66d06 | 973 | if (!new) |
1f87f0b5 EB |
974 | return NULL; |
975 | ||
ac13ac6f EB |
976 | node = (struct ctl_node *)(new + 1); |
977 | table = (struct ctl_table *)(node + 1); | |
7ec66d06 EB |
978 | new_name = (char *)(table + 2); |
979 | memcpy(new_name, name, namelen); | |
7ec66d06 EB |
980 | table[0].procname = new_name; |
981 | table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
ac13ac6f | 982 | init_header(&new->header, set->dir.header.root, set, node, table); |
7ec66d06 EB |
983 | |
984 | return new; | |
1f87f0b5 EB |
985 | } |
986 | ||
60f126d9 EB |
987 | /** |
988 | * get_subdir - find or create a subdir with the specified name. | |
989 | * @dir: Directory to create the subdirectory in | |
990 | * @name: The name of the subdirectory to find or create | |
991 | * @namelen: The length of name | |
992 | * | |
993 | * Takes a directory with an elevated reference count so we know that | |
994 | * if we drop the lock the directory will not go away. Upon success | |
995 | * the reference is moved from @dir to the returned subdirectory. | |
996 | * Upon error an error code is returned and the reference on @dir is | |
997 | * simply dropped. | |
998 | */ | |
0e47c99d EB |
999 | static struct ctl_dir *get_subdir(struct ctl_dir *dir, |
1000 | const char *name, int namelen) | |
1f87f0b5 | 1001 | { |
0e47c99d | 1002 | struct ctl_table_set *set = dir->header.set; |
7ec66d06 | 1003 | struct ctl_dir *subdir, *new = NULL; |
0eb97f38 | 1004 | int err; |
1f87f0b5 | 1005 | |
7ec66d06 | 1006 | spin_lock(&sysctl_lock); |
0e47c99d | 1007 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
1008 | if (!IS_ERR(subdir)) |
1009 | goto found; | |
1010 | if (PTR_ERR(subdir) != -ENOENT) | |
1011 | goto failed; | |
1012 | ||
1013 | spin_unlock(&sysctl_lock); | |
1014 | new = new_dir(set, name, namelen); | |
1015 | spin_lock(&sysctl_lock); | |
1016 | subdir = ERR_PTR(-ENOMEM); | |
1017 | if (!new) | |
1018 | goto failed; | |
1019 | ||
60f126d9 | 1020 | /* Was the subdir added while we dropped the lock? */ |
0e47c99d | 1021 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
1022 | if (!IS_ERR(subdir)) |
1023 | goto found; | |
1024 | if (PTR_ERR(subdir) != -ENOENT) | |
1025 | goto failed; | |
1026 | ||
60f126d9 | 1027 | /* Nope. Use the our freshly made directory entry. */ |
0eb97f38 EB |
1028 | err = insert_header(dir, &new->header); |
1029 | subdir = ERR_PTR(err); | |
1030 | if (err) | |
0e47c99d | 1031 | goto failed; |
7ec66d06 EB |
1032 | subdir = new; |
1033 | found: | |
1034 | subdir->header.nreg++; | |
1035 | failed: | |
a1c83681 | 1036 | if (IS_ERR(subdir)) { |
87ebdc00 | 1037 | pr_err("sysctl could not get directory: "); |
6980128f | 1038 | sysctl_print_dir(dir); |
153ee1c4 GU |
1039 | pr_cont("%*.*s %ld\n", namelen, namelen, name, |
1040 | PTR_ERR(subdir)); | |
1f87f0b5 | 1041 | } |
7ec66d06 EB |
1042 | drop_sysctl_table(&dir->header); |
1043 | if (new) | |
1044 | drop_sysctl_table(&new->header); | |
1045 | spin_unlock(&sysctl_lock); | |
1046 | return subdir; | |
1f87f0b5 EB |
1047 | } |
1048 | ||
0e47c99d EB |
1049 | static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir) |
1050 | { | |
1051 | struct ctl_dir *parent; | |
1052 | const char *procname; | |
1053 | if (!dir->header.parent) | |
1054 | return &set->dir; | |
1055 | parent = xlate_dir(set, dir->header.parent); | |
1056 | if (IS_ERR(parent)) | |
1057 | return parent; | |
1058 | procname = dir->header.ctl_table[0].procname; | |
1059 | return find_subdir(parent, procname, strlen(procname)); | |
1060 | } | |
1061 | ||
1062 | static int sysctl_follow_link(struct ctl_table_header **phead, | |
13bcc6a2 | 1063 | struct ctl_table **pentry) |
0e47c99d EB |
1064 | { |
1065 | struct ctl_table_header *head; | |
1066 | struct ctl_table_root *root; | |
1067 | struct ctl_table_set *set; | |
1068 | struct ctl_table *entry; | |
1069 | struct ctl_dir *dir; | |
1070 | int ret; | |
1071 | ||
0e47c99d EB |
1072 | spin_lock(&sysctl_lock); |
1073 | root = (*pentry)->data; | |
13bcc6a2 | 1074 | set = lookup_header_set(root); |
0e47c99d EB |
1075 | dir = xlate_dir(set, (*phead)->parent); |
1076 | if (IS_ERR(dir)) | |
1077 | ret = PTR_ERR(dir); | |
1078 | else { | |
1079 | const char *procname = (*pentry)->procname; | |
1080 | head = NULL; | |
1081 | entry = find_entry(&head, dir, procname, strlen(procname)); | |
1082 | ret = -ENOENT; | |
1083 | if (entry && use_table(head)) { | |
1084 | unuse_table(*phead); | |
1085 | *phead = head; | |
1086 | *pentry = entry; | |
1087 | ret = 0; | |
1088 | } | |
1089 | } | |
1090 | ||
1091 | spin_unlock(&sysctl_lock); | |
1092 | return ret; | |
1093 | } | |
1094 | ||
7c60c48f | 1095 | static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...) |
1f87f0b5 | 1096 | { |
7c60c48f EB |
1097 | struct va_format vaf; |
1098 | va_list args; | |
1f87f0b5 | 1099 | |
7c60c48f EB |
1100 | va_start(args, fmt); |
1101 | vaf.fmt = fmt; | |
1102 | vaf.va = &args; | |
1103 | ||
87ebdc00 AM |
1104 | pr_err("sysctl table check failed: %s/%s %pV\n", |
1105 | path, table->procname, &vaf); | |
1f87f0b5 | 1106 | |
7c60c48f EB |
1107 | va_end(args); |
1108 | return -EINVAL; | |
1f87f0b5 EB |
1109 | } |
1110 | ||
4f2fec00 LR |
1111 | static int sysctl_check_table_array(const char *path, struct ctl_table *table) |
1112 | { | |
1113 | int err = 0; | |
1114 | ||
61d9b56a LR |
1115 | if ((table->proc_handler == proc_douintvec) || |
1116 | (table->proc_handler == proc_douintvec_minmax)) { | |
4f2fec00 | 1117 | if (table->maxlen != sizeof(unsigned int)) |
64a11f3d | 1118 | err |= sysctl_err(path, table, "array not allowed"); |
4f2fec00 LR |
1119 | } |
1120 | ||
cb944413 ED |
1121 | if (table->proc_handler == proc_dou8vec_minmax) { |
1122 | if (table->maxlen != sizeof(u8)) | |
1123 | err |= sysctl_err(path, table, "array not allowed"); | |
1124 | } | |
1125 | ||
f1aa2eb5 OM |
1126 | if (table->proc_handler == proc_dobool) { |
1127 | if (table->maxlen != sizeof(bool)) | |
1128 | err |= sysctl_err(path, table, "array not allowed"); | |
1129 | } | |
1130 | ||
4f2fec00 LR |
1131 | return err; |
1132 | } | |
1133 | ||
7c60c48f | 1134 | static int sysctl_check_table(const char *path, struct ctl_table *table) |
1f87f0b5 | 1135 | { |
cb55f27a | 1136 | struct ctl_table *entry; |
7c60c48f | 1137 | int err = 0; |
cb55f27a MT |
1138 | list_for_each_table_entry(entry, table) { |
1139 | if (entry->child) | |
1140 | err |= sysctl_err(path, entry, "Not a file"); | |
1141 | ||
1142 | if ((entry->proc_handler == proc_dostring) || | |
f1aa2eb5 | 1143 | (entry->proc_handler == proc_dobool) || |
cb55f27a MT |
1144 | (entry->proc_handler == proc_dointvec) || |
1145 | (entry->proc_handler == proc_douintvec) || | |
1146 | (entry->proc_handler == proc_douintvec_minmax) || | |
1147 | (entry->proc_handler == proc_dointvec_minmax) || | |
1148 | (entry->proc_handler == proc_dou8vec_minmax) || | |
1149 | (entry->proc_handler == proc_dointvec_jiffies) || | |
1150 | (entry->proc_handler == proc_dointvec_userhz_jiffies) || | |
1151 | (entry->proc_handler == proc_dointvec_ms_jiffies) || | |
1152 | (entry->proc_handler == proc_doulongvec_minmax) || | |
1153 | (entry->proc_handler == proc_doulongvec_ms_jiffies_minmax)) { | |
1154 | if (!entry->data) | |
1155 | err |= sysctl_err(path, entry, "No data"); | |
1156 | if (!entry->maxlen) | |
1157 | err |= sysctl_err(path, entry, "No maxlen"); | |
4f2fec00 | 1158 | else |
cb55f27a | 1159 | err |= sysctl_check_table_array(path, entry); |
7c60c48f | 1160 | } |
cb55f27a MT |
1161 | if (!entry->proc_handler) |
1162 | err |= sysctl_err(path, entry, "No proc_handler"); | |
7c60c48f | 1163 | |
cb55f27a MT |
1164 | if ((entry->mode & (S_IRUGO|S_IWUGO)) != entry->mode) |
1165 | err |= sysctl_err(path, entry, "bogus .mode 0%o", | |
1166 | entry->mode); | |
1f87f0b5 | 1167 | } |
7c60c48f | 1168 | return err; |
1f87f0b5 | 1169 | } |
1f87f0b5 | 1170 | |
0e47c99d EB |
1171 | static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table *table, |
1172 | struct ctl_table_root *link_root) | |
1173 | { | |
1174 | struct ctl_table *link_table, *entry, *link; | |
1175 | struct ctl_table_header *links; | |
ac13ac6f | 1176 | struct ctl_node *node; |
0e47c99d EB |
1177 | char *link_name; |
1178 | int nr_entries, name_bytes; | |
1179 | ||
1180 | name_bytes = 0; | |
1181 | nr_entries = 0; | |
cb55f27a | 1182 | list_for_each_table_entry(entry, table) { |
0e47c99d EB |
1183 | nr_entries++; |
1184 | name_bytes += strlen(entry->procname) + 1; | |
1185 | } | |
1186 | ||
1187 | links = kzalloc(sizeof(struct ctl_table_header) + | |
ac13ac6f | 1188 | sizeof(struct ctl_node)*nr_entries + |
0e47c99d EB |
1189 | sizeof(struct ctl_table)*(nr_entries + 1) + |
1190 | name_bytes, | |
1191 | GFP_KERNEL); | |
1192 | ||
1193 | if (!links) | |
1194 | return NULL; | |
1195 | ||
ac13ac6f EB |
1196 | node = (struct ctl_node *)(links + 1); |
1197 | link_table = (struct ctl_table *)(node + nr_entries); | |
0e47c99d | 1198 | link_name = (char *)&link_table[nr_entries + 1]; |
cb55f27a | 1199 | link = link_table; |
0e47c99d | 1200 | |
cb55f27a | 1201 | list_for_each_table_entry(entry, table) { |
0e47c99d EB |
1202 | int len = strlen(entry->procname) + 1; |
1203 | memcpy(link_name, entry->procname, len); | |
1204 | link->procname = link_name; | |
1205 | link->mode = S_IFLNK|S_IRWXUGO; | |
1206 | link->data = link_root; | |
1207 | link_name += len; | |
cb55f27a | 1208 | link++; |
0e47c99d | 1209 | } |
ac13ac6f | 1210 | init_header(links, dir->header.root, dir->header.set, node, link_table); |
0e47c99d EB |
1211 | links->nreg = nr_entries; |
1212 | ||
1213 | return links; | |
1214 | } | |
1215 | ||
1216 | static bool get_links(struct ctl_dir *dir, | |
1217 | struct ctl_table *table, struct ctl_table_root *link_root) | |
1218 | { | |
1219 | struct ctl_table_header *head; | |
1220 | struct ctl_table *entry, *link; | |
1221 | ||
1222 | /* Are there links available for every entry in table? */ | |
cb55f27a | 1223 | list_for_each_table_entry(entry, table) { |
0e47c99d EB |
1224 | const char *procname = entry->procname; |
1225 | link = find_entry(&head, dir, procname, strlen(procname)); | |
1226 | if (!link) | |
1227 | return false; | |
1228 | if (S_ISDIR(link->mode) && S_ISDIR(entry->mode)) | |
1229 | continue; | |
1230 | if (S_ISLNK(link->mode) && (link->data == link_root)) | |
1231 | continue; | |
1232 | return false; | |
1233 | } | |
1234 | ||
1235 | /* The checks passed. Increase the registration count on the links */ | |
cb55f27a | 1236 | list_for_each_table_entry(entry, table) { |
0e47c99d EB |
1237 | const char *procname = entry->procname; |
1238 | link = find_entry(&head, dir, procname, strlen(procname)); | |
1239 | head->nreg++; | |
1240 | } | |
1241 | return true; | |
1242 | } | |
1243 | ||
1244 | static int insert_links(struct ctl_table_header *head) | |
1245 | { | |
1246 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
9a521359 | 1247 | struct ctl_dir *core_parent; |
0e47c99d EB |
1248 | struct ctl_table_header *links; |
1249 | int err; | |
1250 | ||
1251 | if (head->set == root_set) | |
1252 | return 0; | |
1253 | ||
1254 | core_parent = xlate_dir(root_set, head->parent); | |
1255 | if (IS_ERR(core_parent)) | |
1256 | return 0; | |
1257 | ||
1258 | if (get_links(core_parent, head->ctl_table, head->root)) | |
1259 | return 0; | |
1260 | ||
1261 | core_parent->header.nreg++; | |
1262 | spin_unlock(&sysctl_lock); | |
1263 | ||
1264 | links = new_links(core_parent, head->ctl_table, head->root); | |
1265 | ||
1266 | spin_lock(&sysctl_lock); | |
1267 | err = -ENOMEM; | |
1268 | if (!links) | |
1269 | goto out; | |
1270 | ||
1271 | err = 0; | |
1272 | if (get_links(core_parent, head->ctl_table, head->root)) { | |
1273 | kfree(links); | |
1274 | goto out; | |
1275 | } | |
1276 | ||
1277 | err = insert_header(core_parent, links); | |
1278 | if (err) | |
1279 | kfree(links); | |
1280 | out: | |
1281 | drop_sysctl_table(&core_parent->header); | |
1282 | return err; | |
1283 | } | |
1284 | ||
b2f56e55 LC |
1285 | /* Find the directory for the ctl_table. If one is not found create it. */ |
1286 | static struct ctl_dir *sysctl_mkdir_p(struct ctl_dir *dir, const char *path) | |
1287 | { | |
1288 | const char *name, *nextname; | |
1289 | ||
1290 | for (name = path; name; name = nextname) { | |
1291 | int namelen; | |
1292 | nextname = strchr(name, '/'); | |
1293 | if (nextname) { | |
1294 | namelen = nextname - name; | |
1295 | nextname++; | |
1296 | } else { | |
1297 | namelen = strlen(name); | |
1298 | } | |
1299 | if (namelen == 0) | |
1300 | continue; | |
1301 | ||
1302 | /* | |
1303 | * namelen ensures if name is "foo/bar/yay" only foo is | |
1304 | * registered first. We traverse as if using mkdir -p and | |
1305 | * return a ctl_dir for the last directory entry. | |
1306 | */ | |
1307 | dir = get_subdir(dir, name, namelen); | |
1308 | if (IS_ERR(dir)) | |
1309 | break; | |
1310 | } | |
1311 | return dir; | |
1312 | } | |
1313 | ||
1f87f0b5 | 1314 | /** |
f728019b | 1315 | * __register_sysctl_table - register a leaf sysctl table |
60a47a2e | 1316 | * @set: Sysctl tree to register on |
1f87f0b5 | 1317 | * @path: The path to the directory the sysctl table is in. |
1dc8689e LC |
1318 | * @table: the top-level table structure without any child. This table |
1319 | * should not be free'd after registration. So it should not be | |
1320 | * used on stack. It can either be a global or dynamically allocated | |
1321 | * by the caller and free'd later after sysctl unregistration. | |
1f87f0b5 EB |
1322 | * |
1323 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1324 | * array. A completely 0 filled entry terminates the table. | |
1325 | * | |
1326 | * The members of the &struct ctl_table structure are used as follows: | |
1327 | * | |
1328 | * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not | |
1329 | * enter a sysctl file | |
1330 | * | |
1331 | * data - a pointer to data for use by proc_handler | |
1332 | * | |
1333 | * maxlen - the maximum size in bytes of the data | |
1334 | * | |
f728019b | 1335 | * mode - the file permissions for the /proc/sys file |
1f87f0b5 | 1336 | * |
f728019b | 1337 | * child - must be %NULL. |
1f87f0b5 EB |
1338 | * |
1339 | * proc_handler - the text handler routine (described below) | |
1340 | * | |
1f87f0b5 | 1341 | * extra1, extra2 - extra pointers usable by the proc handler routines |
67ff3228 LC |
1342 | * XXX: we should eventually modify these to use long min / max [0] |
1343 | * [0] https://lkml.kernel.org/87zgpte9o4.fsf@email.froward.int.ebiederm.org | |
1f87f0b5 EB |
1344 | * |
1345 | * Leaf nodes in the sysctl tree will be represented by a single file | |
67ff3228 LC |
1346 | * under /proc; non-leaf nodes (where child is not NULL) are not allowed, |
1347 | * sysctl_check_table() verifies this. | |
1f87f0b5 | 1348 | * |
f728019b EB |
1349 | * There must be a proc_handler routine for any terminal nodes. |
1350 | * Several default handlers are available to cover common cases - | |
1f87f0b5 EB |
1351 | * |
1352 | * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), | |
1353 | * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), | |
1354 | * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() | |
1355 | * | |
1356 | * It is the handler's job to read the input buffer from user memory | |
1357 | * and process it. The handler should return 0 on success. | |
1358 | * | |
1359 | * This routine returns %NULL on a failure to register, and a pointer | |
1360 | * to the table header on success. | |
1361 | */ | |
6e9d5164 | 1362 | struct ctl_table_header *__register_sysctl_table( |
60a47a2e | 1363 | struct ctl_table_set *set, |
6e9d5164 | 1364 | const char *path, struct ctl_table *table) |
1f87f0b5 | 1365 | { |
60a47a2e | 1366 | struct ctl_table_root *root = set->dir.header.root; |
1f87f0b5 | 1367 | struct ctl_table_header *header; |
7ec66d06 | 1368 | struct ctl_dir *dir; |
ac13ac6f EB |
1369 | struct ctl_table *entry; |
1370 | struct ctl_node *node; | |
1371 | int nr_entries = 0; | |
1372 | ||
cb55f27a | 1373 | list_for_each_table_entry(entry, table) |
ac13ac6f | 1374 | nr_entries++; |
1f87f0b5 | 1375 | |
ac13ac6f | 1376 | header = kzalloc(sizeof(struct ctl_table_header) + |
425b9c7f | 1377 | sizeof(struct ctl_node)*nr_entries, GFP_KERNEL_ACCOUNT); |
1f87f0b5 EB |
1378 | if (!header) |
1379 | return NULL; | |
1380 | ||
ac13ac6f EB |
1381 | node = (struct ctl_node *)(header + 1); |
1382 | init_header(header, root, set, node, table); | |
7ec66d06 EB |
1383 | if (sysctl_check_table(path, table)) |
1384 | goto fail; | |
1385 | ||
1386 | spin_lock(&sysctl_lock); | |
0e47c99d | 1387 | dir = &set->dir; |
67ff3228 | 1388 | /* Reference moved down the directory tree get_subdir */ |
7ec66d06 EB |
1389 | dir->header.nreg++; |
1390 | spin_unlock(&sysctl_lock); | |
1f87f0b5 | 1391 | |
b2f56e55 LC |
1392 | dir = sysctl_mkdir_p(dir, path); |
1393 | if (IS_ERR(dir)) | |
1394 | goto fail; | |
1f87f0b5 | 1395 | spin_lock(&sysctl_lock); |
0e47c99d | 1396 | if (insert_header(dir, header)) |
7ec66d06 | 1397 | goto fail_put_dir_locked; |
0e47c99d | 1398 | |
7ec66d06 | 1399 | drop_sysctl_table(&dir->header); |
1f87f0b5 EB |
1400 | spin_unlock(&sysctl_lock); |
1401 | ||
1402 | return header; | |
0e47c99d | 1403 | |
7ec66d06 EB |
1404 | fail_put_dir_locked: |
1405 | drop_sysctl_table(&dir->header); | |
7c60c48f EB |
1406 | spin_unlock(&sysctl_lock); |
1407 | fail: | |
1408 | kfree(header); | |
1409 | dump_stack(); | |
1410 | return NULL; | |
1f87f0b5 EB |
1411 | } |
1412 | ||
fea478d4 EB |
1413 | /** |
1414 | * register_sysctl - register a sysctl table | |
1dc8689e LC |
1415 | * @path: The path to the directory the sysctl table is in. If the path |
1416 | * doesn't exist we will create it for you. | |
1417 | * @table: the table structure. The calller must ensure the life of the @table | |
1418 | * will be kept during the lifetime use of the syctl. It must not be freed | |
1419 | * until unregister_sysctl_table() is called with the given returned table | |
1420 | * with this registration. If your code is non modular then you don't need | |
1421 | * to call unregister_sysctl_table() and can instead use something like | |
1422 | * register_sysctl_init() which does not care for the result of the syctl | |
1423 | * registration. | |
fea478d4 EB |
1424 | * |
1425 | * Register a sysctl table. @table should be a filled in ctl_table | |
1426 | * array. A completely 0 filled entry terminates the table. | |
1427 | * | |
1428 | * See __register_sysctl_table for more details. | |
1429 | */ | |
1430 | struct ctl_table_header *register_sysctl(const char *path, struct ctl_table *table) | |
1431 | { | |
1432 | return __register_sysctl_table(&sysctl_table_root.default_set, | |
1433 | path, table); | |
1434 | } | |
1435 | EXPORT_SYMBOL(register_sysctl); | |
1436 | ||
3ddd9a80 XN |
1437 | /** |
1438 | * __register_sysctl_init() - register sysctl table to path | |
1dc8689e LC |
1439 | * @path: path name for sysctl base. If that path doesn't exist we will create |
1440 | * it for you. | |
1441 | * @table: This is the sysctl table that needs to be registered to the path. | |
1442 | * The caller must ensure the life of the @table will be kept during the | |
1443 | * lifetime use of the sysctl. | |
3ddd9a80 XN |
1444 | * @table_name: The name of sysctl table, only used for log printing when |
1445 | * registration fails | |
1446 | * | |
1447 | * The sysctl interface is used by userspace to query or modify at runtime | |
1448 | * a predefined value set on a variable. These variables however have default | |
1449 | * values pre-set. Code which depends on these variables will always work even | |
1450 | * if register_sysctl() fails. If register_sysctl() fails you'd just loose the | |
1451 | * ability to query or modify the sysctls dynamically at run time. Chances of | |
1452 | * register_sysctl() failing on init are extremely low, and so for both reasons | |
1453 | * this function does not return any error as it is used by initialization code. | |
1454 | * | |
228b09de | 1455 | * Context: if your base directory does not exist it will be created for you. |
3ddd9a80 XN |
1456 | */ |
1457 | void __init __register_sysctl_init(const char *path, struct ctl_table *table, | |
1458 | const char *table_name) | |
1459 | { | |
1460 | struct ctl_table_header *hdr = register_sysctl(path, table); | |
1461 | ||
1462 | if (unlikely(!hdr)) { | |
1463 | pr_err("failed when register_sysctl %s to %s\n", table_name, path); | |
1464 | return; | |
1465 | } | |
1466 | kmemleak_not_leak(hdr); | |
1467 | } | |
1468 | ||
6e9d5164 EB |
1469 | static char *append_path(const char *path, char *pos, const char *name) |
1470 | { | |
1471 | int namelen; | |
1472 | namelen = strlen(name); | |
1473 | if (((pos - path) + namelen + 2) >= PATH_MAX) | |
1474 | return NULL; | |
1475 | memcpy(pos, name, namelen); | |
1476 | pos[namelen] = '/'; | |
1477 | pos[namelen + 1] = '\0'; | |
1478 | pos += namelen + 1; | |
1479 | return pos; | |
1480 | } | |
1481 | ||
f728019b EB |
1482 | static int count_subheaders(struct ctl_table *table) |
1483 | { | |
1484 | int has_files = 0; | |
1485 | int nr_subheaders = 0; | |
1486 | struct ctl_table *entry; | |
1487 | ||
1488 | /* special case: no directory and empty directory */ | |
1489 | if (!table || !table->procname) | |
1490 | return 1; | |
1491 | ||
cb55f27a | 1492 | list_for_each_table_entry(entry, table) { |
f728019b EB |
1493 | if (entry->child) |
1494 | nr_subheaders += count_subheaders(entry->child); | |
1495 | else | |
1496 | has_files = 1; | |
1497 | } | |
1498 | return nr_subheaders + has_files; | |
1499 | } | |
1500 | ||
1501 | static int register_leaf_sysctl_tables(const char *path, char *pos, | |
60a47a2e | 1502 | struct ctl_table_header ***subheader, struct ctl_table_set *set, |
f728019b EB |
1503 | struct ctl_table *table) |
1504 | { | |
1505 | struct ctl_table *ctl_table_arg = NULL; | |
1506 | struct ctl_table *entry, *files; | |
1507 | int nr_files = 0; | |
1508 | int nr_dirs = 0; | |
1509 | int err = -ENOMEM; | |
1510 | ||
cb55f27a | 1511 | list_for_each_table_entry(entry, table) { |
f728019b EB |
1512 | if (entry->child) |
1513 | nr_dirs++; | |
1514 | else | |
1515 | nr_files++; | |
1516 | } | |
1517 | ||
1518 | files = table; | |
1519 | /* If there are mixed files and directories we need a new table */ | |
1520 | if (nr_dirs && nr_files) { | |
1521 | struct ctl_table *new; | |
6396bb22 | 1522 | files = kcalloc(nr_files + 1, sizeof(struct ctl_table), |
f728019b EB |
1523 | GFP_KERNEL); |
1524 | if (!files) | |
1525 | goto out; | |
1526 | ||
1527 | ctl_table_arg = files; | |
cb55f27a MT |
1528 | new = files; |
1529 | ||
1530 | list_for_each_table_entry(entry, table) { | |
f728019b EB |
1531 | if (entry->child) |
1532 | continue; | |
1533 | *new = *entry; | |
1534 | new++; | |
1535 | } | |
1536 | } | |
1537 | ||
1538 | /* Register everything except a directory full of subdirectories */ | |
1539 | if (nr_files || !nr_dirs) { | |
1540 | struct ctl_table_header *header; | |
60a47a2e | 1541 | header = __register_sysctl_table(set, path, files); |
f728019b EB |
1542 | if (!header) { |
1543 | kfree(ctl_table_arg); | |
1544 | goto out; | |
1545 | } | |
1546 | ||
1547 | /* Remember if we need to free the file table */ | |
1548 | header->ctl_table_arg = ctl_table_arg; | |
1549 | **subheader = header; | |
1550 | (*subheader)++; | |
1551 | } | |
1552 | ||
1553 | /* Recurse into the subdirectories. */ | |
cb55f27a | 1554 | list_for_each_table_entry(entry, table) { |
f728019b EB |
1555 | char *child_pos; |
1556 | ||
1557 | if (!entry->child) | |
1558 | continue; | |
1559 | ||
1560 | err = -ENAMETOOLONG; | |
1561 | child_pos = append_path(path, pos, entry->procname); | |
1562 | if (!child_pos) | |
1563 | goto out; | |
1564 | ||
1565 | err = register_leaf_sysctl_tables(path, child_pos, subheader, | |
60a47a2e | 1566 | set, entry->child); |
f728019b EB |
1567 | pos[0] = '\0'; |
1568 | if (err) | |
1569 | goto out; | |
1570 | } | |
1571 | err = 0; | |
1572 | out: | |
1573 | /* On failure our caller will unregister all registered subheaders */ | |
1574 | return err; | |
1575 | } | |
1576 | ||
6e9d5164 | 1577 | /** |
0199849a | 1578 | * register_sysctl_table - register a sysctl table hierarchy |
6e9d5164 EB |
1579 | * @table: the top-level table structure |
1580 | * | |
1581 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1582 | * array. A completely 0 filled entry terminates the table. | |
1dc8689e | 1583 | * We are slowly deprecating this call so avoid its use. |
6e9d5164 | 1584 | */ |
0199849a | 1585 | struct ctl_table_header *register_sysctl_table(struct ctl_table *table) |
6e9d5164 | 1586 | { |
ec6a5266 | 1587 | struct ctl_table *ctl_table_arg = table; |
f728019b EB |
1588 | int nr_subheaders = count_subheaders(table); |
1589 | struct ctl_table_header *header = NULL, **subheaders, **subheader; | |
6e9d5164 EB |
1590 | char *new_path, *pos; |
1591 | ||
1592 | pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL); | |
1593 | if (!new_path) | |
1594 | return NULL; | |
1595 | ||
1596 | pos[0] = '\0'; | |
ec6a5266 EB |
1597 | while (table->procname && table->child && !table[1].procname) { |
1598 | pos = append_path(new_path, pos, table->procname); | |
1599 | if (!pos) | |
1600 | goto out; | |
1601 | table = table->child; | |
1602 | } | |
f728019b | 1603 | if (nr_subheaders == 1) { |
0199849a | 1604 | header = __register_sysctl_table(&sysctl_table_root.default_set, new_path, table); |
f728019b EB |
1605 | if (header) |
1606 | header->ctl_table_arg = ctl_table_arg; | |
1607 | } else { | |
1608 | header = kzalloc(sizeof(*header) + | |
1609 | sizeof(*subheaders)*nr_subheaders, GFP_KERNEL); | |
1610 | if (!header) | |
1611 | goto out; | |
1612 | ||
1613 | subheaders = (struct ctl_table_header **) (header + 1); | |
1614 | subheader = subheaders; | |
ec6a5266 | 1615 | header->ctl_table_arg = ctl_table_arg; |
f728019b EB |
1616 | |
1617 | if (register_leaf_sysctl_tables(new_path, pos, &subheader, | |
0199849a | 1618 | &sysctl_table_root.default_set, table)) |
f728019b EB |
1619 | goto err_register_leaves; |
1620 | } | |
1621 | ||
6e9d5164 EB |
1622 | out: |
1623 | kfree(new_path); | |
1624 | return header; | |
f728019b EB |
1625 | |
1626 | err_register_leaves: | |
1627 | while (subheader > subheaders) { | |
1628 | struct ctl_table_header *subh = *(--subheader); | |
1629 | struct ctl_table *table = subh->ctl_table_arg; | |
1630 | unregister_sysctl_table(subh); | |
1631 | kfree(table); | |
1632 | } | |
1633 | kfree(header); | |
1634 | header = NULL; | |
1635 | goto out; | |
6e9d5164 | 1636 | } |
1f87f0b5 EB |
1637 | EXPORT_SYMBOL(register_sysctl_table); |
1638 | ||
51cb8dfc LC |
1639 | int __register_sysctl_base(struct ctl_table *base_table) |
1640 | { | |
1641 | struct ctl_table_header *hdr; | |
1642 | ||
1643 | hdr = register_sysctl_table(base_table); | |
1644 | kmemleak_not_leak(hdr); | |
1645 | return 0; | |
1646 | } | |
1647 | ||
0e47c99d EB |
1648 | static void put_links(struct ctl_table_header *header) |
1649 | { | |
1650 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
1651 | struct ctl_table_root *root = header->root; | |
1652 | struct ctl_dir *parent = header->parent; | |
1653 | struct ctl_dir *core_parent; | |
1654 | struct ctl_table *entry; | |
1655 | ||
1656 | if (header->set == root_set) | |
1657 | return; | |
1658 | ||
1659 | core_parent = xlate_dir(root_set, parent); | |
1660 | if (IS_ERR(core_parent)) | |
1661 | return; | |
1662 | ||
cb55f27a | 1663 | list_for_each_table_entry(entry, header->ctl_table) { |
0e47c99d EB |
1664 | struct ctl_table_header *link_head; |
1665 | struct ctl_table *link; | |
1666 | const char *name = entry->procname; | |
1667 | ||
1668 | link = find_entry(&link_head, core_parent, name, strlen(name)); | |
1669 | if (link && | |
1670 | ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) || | |
1671 | (S_ISLNK(link->mode) && (link->data == root)))) { | |
1672 | drop_sysctl_table(link_head); | |
1673 | } | |
1674 | else { | |
87ebdc00 | 1675 | pr_err("sysctl link missing during unregister: "); |
0e47c99d | 1676 | sysctl_print_dir(parent); |
153ee1c4 | 1677 | pr_cont("%s\n", name); |
0e47c99d EB |
1678 | } |
1679 | } | |
1680 | } | |
1681 | ||
938aaa4f EB |
1682 | static void drop_sysctl_table(struct ctl_table_header *header) |
1683 | { | |
7ec66d06 EB |
1684 | struct ctl_dir *parent = header->parent; |
1685 | ||
938aaa4f EB |
1686 | if (--header->nreg) |
1687 | return; | |
1688 | ||
89189557 | 1689 | if (parent) { |
23da9588 | 1690 | put_links(header); |
89189557 Y |
1691 | start_unregistering(header); |
1692 | } | |
1693 | ||
938aaa4f EB |
1694 | if (!--header->count) |
1695 | kfree_rcu(header, rcu); | |
7ec66d06 EB |
1696 | |
1697 | if (parent) | |
1698 | drop_sysctl_table(&parent->header); | |
938aaa4f EB |
1699 | } |
1700 | ||
1f87f0b5 EB |
1701 | /** |
1702 | * unregister_sysctl_table - unregister a sysctl table hierarchy | |
1703 | * @header: the header returned from register_sysctl_table | |
1704 | * | |
1705 | * Unregisters the sysctl table and all children. proc entries may not | |
1706 | * actually be removed until they are no longer used by anyone. | |
1707 | */ | |
1708 | void unregister_sysctl_table(struct ctl_table_header * header) | |
1709 | { | |
f728019b | 1710 | int nr_subheaders; |
1f87f0b5 EB |
1711 | might_sleep(); |
1712 | ||
1713 | if (header == NULL) | |
1714 | return; | |
1715 | ||
f728019b EB |
1716 | nr_subheaders = count_subheaders(header->ctl_table_arg); |
1717 | if (unlikely(nr_subheaders > 1)) { | |
1718 | struct ctl_table_header **subheaders; | |
1719 | int i; | |
1720 | ||
1721 | subheaders = (struct ctl_table_header **)(header + 1); | |
1722 | for (i = nr_subheaders -1; i >= 0; i--) { | |
1723 | struct ctl_table_header *subh = subheaders[i]; | |
1724 | struct ctl_table *table = subh->ctl_table_arg; | |
1725 | unregister_sysctl_table(subh); | |
1726 | kfree(table); | |
1727 | } | |
1728 | kfree(header); | |
1729 | return; | |
1730 | } | |
1731 | ||
1f87f0b5 | 1732 | spin_lock(&sysctl_lock); |
938aaa4f | 1733 | drop_sysctl_table(header); |
1f87f0b5 EB |
1734 | spin_unlock(&sysctl_lock); |
1735 | } | |
1736 | EXPORT_SYMBOL(unregister_sysctl_table); | |
1737 | ||
0e47c99d | 1738 | void setup_sysctl_set(struct ctl_table_set *set, |
9eb47c26 | 1739 | struct ctl_table_root *root, |
1f87f0b5 EB |
1740 | int (*is_seen)(struct ctl_table_set *)) |
1741 | { | |
1347440d | 1742 | memset(set, 0, sizeof(*set)); |
0e47c99d | 1743 | set->is_seen = is_seen; |
ac13ac6f | 1744 | init_header(&set->dir.header, root, set, NULL, root_table); |
1f87f0b5 EB |
1745 | } |
1746 | ||
97324cd8 EB |
1747 | void retire_sysctl_set(struct ctl_table_set *set) |
1748 | { | |
ac13ac6f | 1749 | WARN_ON(!RB_EMPTY_ROOT(&set->dir.root)); |
97324cd8 | 1750 | } |
1f87f0b5 | 1751 | |
1e0edd3f | 1752 | int __init proc_sys_init(void) |
77b14db5 | 1753 | { |
e1675231 AD |
1754 | struct proc_dir_entry *proc_sys_root; |
1755 | ||
77b14db5 | 1756 | proc_sys_root = proc_mkdir("sys", NULL); |
9043476f | 1757 | proc_sys_root->proc_iops = &proc_sys_dir_operations; |
d56c0d45 | 1758 | proc_sys_root->proc_dir_ops = &proc_sys_dir_file_operations; |
77b14db5 | 1759 | proc_sys_root->nlink = 0; |
de4e83bd | 1760 | |
d8c0418a | 1761 | return sysctl_init_bases(); |
77b14db5 | 1762 | } |
3db978d4 | 1763 | |
0a477e1a VB |
1764 | struct sysctl_alias { |
1765 | const char *kernel_param; | |
1766 | const char *sysctl_param; | |
1767 | }; | |
1768 | ||
1769 | /* | |
1770 | * Historically some settings had both sysctl and a command line parameter. | |
1771 | * With the generic sysctl. parameter support, we can handle them at a single | |
1772 | * place and only keep the historical name for compatibility. This is not meant | |
1773 | * to add brand new aliases. When adding existing aliases, consider whether | |
1774 | * the possibly different moment of changing the value (e.g. from early_param | |
1775 | * to the moment do_sysctl_args() is called) is an issue for the specific | |
1776 | * parameter. | |
1777 | */ | |
1778 | static const struct sysctl_alias sysctl_aliases[] = { | |
f117955a GP |
1779 | {"hardlockup_all_cpu_backtrace", "kernel.hardlockup_all_cpu_backtrace" }, |
1780 | {"hung_task_panic", "kernel.hung_task_panic" }, | |
1781 | {"numa_zonelist_order", "vm.numa_zonelist_order" }, | |
1782 | {"softlockup_all_cpu_backtrace", "kernel.softlockup_all_cpu_backtrace" }, | |
1783 | {"softlockup_panic", "kernel.softlockup_panic" }, | |
0a477e1a VB |
1784 | { } |
1785 | }; | |
1786 | ||
1787 | static const char *sysctl_find_alias(char *param) | |
1788 | { | |
1789 | const struct sysctl_alias *alias; | |
1790 | ||
1791 | for (alias = &sysctl_aliases[0]; alias->kernel_param != NULL; alias++) { | |
1792 | if (strcmp(alias->kernel_param, param) == 0) | |
1793 | return alias->sysctl_param; | |
1794 | } | |
1795 | ||
1796 | return NULL; | |
1797 | } | |
1798 | ||
3db978d4 VB |
1799 | /* Set sysctl value passed on kernel command line. */ |
1800 | static int process_sysctl_arg(char *param, char *val, | |
1801 | const char *unused, void *arg) | |
1802 | { | |
1803 | char *path; | |
1804 | struct vfsmount **proc_mnt = arg; | |
1805 | struct file_system_type *proc_fs_type; | |
1806 | struct file *file; | |
1807 | int len; | |
1808 | int err; | |
1809 | loff_t pos = 0; | |
1810 | ssize_t wret; | |
1811 | ||
0a477e1a VB |
1812 | if (strncmp(param, "sysctl", sizeof("sysctl") - 1) == 0) { |
1813 | param += sizeof("sysctl") - 1; | |
3db978d4 | 1814 | |
0a477e1a VB |
1815 | if (param[0] != '/' && param[0] != '.') |
1816 | return 0; | |
3db978d4 | 1817 | |
0a477e1a VB |
1818 | param++; |
1819 | } else { | |
1820 | param = (char *) sysctl_find_alias(param); | |
1821 | if (!param) | |
1822 | return 0; | |
1823 | } | |
3db978d4 | 1824 | |
697edcb0 XN |
1825 | if (!val) |
1826 | return -EINVAL; | |
1827 | len = strlen(val); | |
1828 | if (len == 0) | |
1829 | return -EINVAL; | |
1830 | ||
3db978d4 VB |
1831 | /* |
1832 | * To set sysctl options, we use a temporary mount of proc, look up the | |
1833 | * respective sys/ file and write to it. To avoid mounting it when no | |
1834 | * options were given, we mount it only when the first sysctl option is | |
1835 | * found. Why not a persistent mount? There are problems with a | |
1836 | * persistent mount of proc in that it forces userspace not to use any | |
1837 | * proc mount options. | |
1838 | */ | |
1839 | if (!*proc_mnt) { | |
1840 | proc_fs_type = get_fs_type("proc"); | |
1841 | if (!proc_fs_type) { | |
1842 | pr_err("Failed to find procfs to set sysctl from command line\n"); | |
1843 | return 0; | |
1844 | } | |
1845 | *proc_mnt = kern_mount(proc_fs_type); | |
1846 | put_filesystem(proc_fs_type); | |
1847 | if (IS_ERR(*proc_mnt)) { | |
1848 | pr_err("Failed to mount procfs to set sysctl from command line\n"); | |
1849 | return 0; | |
1850 | } | |
1851 | } | |
1852 | ||
1853 | path = kasprintf(GFP_KERNEL, "sys/%s", param); | |
1854 | if (!path) | |
1855 | panic("%s: Failed to allocate path for %s\n", __func__, param); | |
1856 | strreplace(path, '.', '/'); | |
1857 | ||
ffb37ca3 | 1858 | file = file_open_root_mnt(*proc_mnt, path, O_WRONLY, 0); |
3db978d4 VB |
1859 | if (IS_ERR(file)) { |
1860 | err = PTR_ERR(file); | |
1861 | if (err == -ENOENT) | |
1862 | pr_err("Failed to set sysctl parameter '%s=%s': parameter not found\n", | |
1863 | param, val); | |
1864 | else if (err == -EACCES) | |
1865 | pr_err("Failed to set sysctl parameter '%s=%s': permission denied (read-only?)\n", | |
1866 | param, val); | |
1867 | else | |
1868 | pr_err("Error %pe opening proc file to set sysctl parameter '%s=%s'\n", | |
1869 | file, param, val); | |
1870 | goto out; | |
1871 | } | |
3db978d4 VB |
1872 | wret = kernel_write(file, val, len, &pos); |
1873 | if (wret < 0) { | |
1874 | err = wret; | |
1875 | if (err == -EINVAL) | |
1876 | pr_err("Failed to set sysctl parameter '%s=%s': invalid value\n", | |
1877 | param, val); | |
1878 | else | |
1879 | pr_err("Error %pe writing to proc file to set sysctl parameter '%s=%s'\n", | |
1880 | ERR_PTR(err), param, val); | |
1881 | } else if (wret != len) { | |
1882 | pr_err("Wrote only %zd bytes of %d writing to proc file %s to set sysctl parameter '%s=%s\n", | |
1883 | wret, len, path, param, val); | |
1884 | } | |
1885 | ||
1886 | err = filp_close(file, NULL); | |
1887 | if (err) | |
1888 | pr_err("Error %pe closing proc file to set sysctl parameter '%s=%s\n", | |
1889 | ERR_PTR(err), param, val); | |
1890 | out: | |
1891 | kfree(path); | |
1892 | return 0; | |
1893 | } | |
1894 | ||
1895 | void do_sysctl_args(void) | |
1896 | { | |
1897 | char *command_line; | |
1898 | struct vfsmount *proc_mnt = NULL; | |
1899 | ||
1900 | command_line = kstrdup(saved_command_line, GFP_KERNEL); | |
1901 | if (!command_line) | |
1902 | panic("%s: Failed to allocate copy of command line\n", __func__); | |
1903 | ||
1904 | parse_args("Setting sysctl args", command_line, | |
1905 | NULL, 0, -1, -1, &proc_mnt, process_sysctl_arg); | |
1906 | ||
1907 | if (proc_mnt) | |
1908 | kern_unmount(proc_mnt); | |
1909 | ||
1910 | kfree(command_line); | |
1911 | } |