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 | ||
53f3811d JG |
22 | #define list_for_each_table_entry(entry, header) \ |
23 | entry = header->ctl_table; \ | |
24 | for (size_t i = 0 ; i < header->ctl_table_size && entry->procname; ++i, entry++) | |
cb55f27a | 25 | |
d72f71eb | 26 | static const struct dentry_operations proc_sys_dentry_operations; |
77b14db5 | 27 | static const struct file_operations proc_sys_file_operations; |
03a44825 | 28 | static const struct inode_operations proc_sys_inode_operations; |
9043476f AV |
29 | static const struct file_operations proc_sys_dir_file_operations; |
30 | static const struct inode_operations proc_sys_dir_operations; | |
77b14db5 | 31 | |
f9bd6733 | 32 | /* Support for permanently empty directories */ |
7fffbc71 | 33 | static struct ctl_table sysctl_mount_point[] = { |
2f2665c1 | 34 | {.type = SYSCTL_TABLE_TYPE_PERMANENTLY_EMPTY } |
f9bd6733 EB |
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 | { | |
31555231 | 47 | return register_sysctl(path, sysctl_mount_point); |
ee9efac4 LC |
48 | } |
49 | EXPORT_SYMBOL(register_sysctl_mount_point); | |
50 | ||
2f2665c1 JG |
51 | #define sysctl_is_perm_empty_ctl_table(tptr) \ |
52 | (tptr[0].type == SYSCTL_TABLE_TYPE_PERMANENTLY_EMPTY) | |
53 | #define sysctl_is_perm_empty_ctl_header(hptr) \ | |
54 | (sysctl_is_perm_empty_ctl_table(hptr->ctl_table)) | |
55 | #define sysctl_set_perm_empty_ctl_header(hptr) \ | |
56 | (hptr->ctl_table[0].type = SYSCTL_TABLE_TYPE_PERMANENTLY_EMPTY) | |
57 | #define sysctl_clear_perm_empty_ctl_header(hptr) \ | |
58 | (hptr->ctl_table[0].type = SYSCTL_TABLE_TYPE_DEFAULT) | |
f9bd6733 | 59 | |
f1ecf068 LDM |
60 | void proc_sys_poll_notify(struct ctl_table_poll *poll) |
61 | { | |
62 | if (!poll) | |
63 | return; | |
64 | ||
65 | atomic_inc(&poll->event); | |
66 | wake_up_interruptible(&poll->wait); | |
67 | } | |
68 | ||
a194558e EB |
69 | static struct ctl_table root_table[] = { |
70 | { | |
71 | .procname = "", | |
7ec66d06 | 72 | .mode = S_IFDIR|S_IRUGO|S_IXUGO, |
a194558e | 73 | }, |
a194558e | 74 | }; |
0e47c99d | 75 | static struct ctl_table_root sysctl_table_root = { |
0e47c99d | 76 | .default_set.dir.header = { |
7ec66d06 EB |
77 | {{.count = 1, |
78 | .nreg = 1, | |
ac13ac6f | 79 | .ctl_table = root_table }}, |
0e47c99d | 80 | .ctl_table_arg = root_table, |
7ec66d06 EB |
81 | .root = &sysctl_table_root, |
82 | .set = &sysctl_table_root.default_set, | |
83 | }, | |
1f87f0b5 | 84 | }; |
1f87f0b5 EB |
85 | |
86 | static DEFINE_SPINLOCK(sysctl_lock); | |
87 | ||
7ec66d06 | 88 | static void drop_sysctl_table(struct ctl_table_header *header); |
0e47c99d | 89 | static int sysctl_follow_link(struct ctl_table_header **phead, |
13bcc6a2 | 90 | struct ctl_table **pentry); |
0e47c99d EB |
91 | static int insert_links(struct ctl_table_header *head); |
92 | static void put_links(struct ctl_table_header *header); | |
7ec66d06 | 93 | |
6980128f EB |
94 | static void sysctl_print_dir(struct ctl_dir *dir) |
95 | { | |
96 | if (dir->header.parent) | |
97 | sysctl_print_dir(dir->header.parent); | |
87ebdc00 | 98 | pr_cont("%s/", dir->header.ctl_table[0].procname); |
6980128f EB |
99 | } |
100 | ||
076c3eed EB |
101 | static int namecmp(const char *name1, int len1, const char *name2, int len2) |
102 | { | |
076c3eed EB |
103 | int cmp; |
104 | ||
d8fc9b66 | 105 | cmp = memcmp(name1, name2, min(len1, len2)); |
076c3eed EB |
106 | if (cmp == 0) |
107 | cmp = len1 - len2; | |
108 | return cmp; | |
109 | } | |
110 | ||
60f126d9 | 111 | /* Called under sysctl_lock */ |
076c3eed | 112 | static struct ctl_table *find_entry(struct ctl_table_header **phead, |
0e47c99d | 113 | struct ctl_dir *dir, const char *name, int namelen) |
076c3eed EB |
114 | { |
115 | struct ctl_table_header *head; | |
116 | struct ctl_table *entry; | |
ac13ac6f | 117 | struct rb_node *node = dir->root.rb_node; |
076c3eed | 118 | |
ac13ac6f EB |
119 | while (node) |
120 | { | |
121 | struct ctl_node *ctl_node; | |
122 | const char *procname; | |
123 | int cmp; | |
124 | ||
125 | ctl_node = rb_entry(node, struct ctl_node, node); | |
126 | head = ctl_node->header; | |
127 | entry = &head->ctl_table[ctl_node - head->node]; | |
128 | procname = entry->procname; | |
129 | ||
130 | cmp = namecmp(name, namelen, procname, strlen(procname)); | |
131 | if (cmp < 0) | |
132 | node = node->rb_left; | |
133 | else if (cmp > 0) | |
134 | node = node->rb_right; | |
135 | else { | |
136 | *phead = head; | |
137 | return entry; | |
076c3eed EB |
138 | } |
139 | } | |
140 | return NULL; | |
141 | } | |
142 | ||
ac13ac6f EB |
143 | static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry) |
144 | { | |
145 | struct rb_node *node = &head->node[entry - head->ctl_table].node; | |
146 | struct rb_node **p = &head->parent->root.rb_node; | |
147 | struct rb_node *parent = NULL; | |
148 | const char *name = entry->procname; | |
149 | int namelen = strlen(name); | |
150 | ||
151 | while (*p) { | |
152 | struct ctl_table_header *parent_head; | |
153 | struct ctl_table *parent_entry; | |
154 | struct ctl_node *parent_node; | |
155 | const char *parent_name; | |
156 | int cmp; | |
157 | ||
158 | parent = *p; | |
159 | parent_node = rb_entry(parent, struct ctl_node, node); | |
160 | parent_head = parent_node->header; | |
161 | parent_entry = &parent_head->ctl_table[parent_node - parent_head->node]; | |
162 | parent_name = parent_entry->procname; | |
163 | ||
164 | cmp = namecmp(name, namelen, parent_name, strlen(parent_name)); | |
165 | if (cmp < 0) | |
166 | p = &(*p)->rb_left; | |
167 | else if (cmp > 0) | |
168 | p = &(*p)->rb_right; | |
169 | else { | |
87ebdc00 | 170 | pr_err("sysctl duplicate entry: "); |
ac13ac6f | 171 | sysctl_print_dir(head->parent); |
153ee1c4 | 172 | pr_cont("%s\n", entry->procname); |
ac13ac6f EB |
173 | return -EEXIST; |
174 | } | |
175 | } | |
176 | ||
177 | rb_link_node(node, parent, p); | |
ea5272f5 | 178 | rb_insert_color(node, &head->parent->root); |
ac13ac6f EB |
179 | return 0; |
180 | } | |
181 | ||
182 | static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry) | |
183 | { | |
184 | struct rb_node *node = &head->node[entry - head->ctl_table].node; | |
185 | ||
186 | rb_erase(node, &head->parent->root); | |
187 | } | |
188 | ||
e0d04529 EB |
189 | static void init_header(struct ctl_table_header *head, |
190 | struct ctl_table_root *root, struct ctl_table_set *set, | |
b1f01e2b | 191 | struct ctl_node *node, struct ctl_table *table, size_t table_size) |
e0d04529 | 192 | { |
7ec66d06 | 193 | head->ctl_table = table; |
b1f01e2b | 194 | head->ctl_table_size = table_size; |
e0d04529 | 195 | head->ctl_table_arg = table; |
e0d04529 EB |
196 | head->used = 0; |
197 | head->count = 1; | |
198 | head->nreg = 1; | |
199 | head->unregistering = NULL; | |
200 | head->root = root; | |
201 | head->set = set; | |
202 | head->parent = NULL; | |
ac13ac6f | 203 | head->node = node; |
2fd1d2c4 | 204 | INIT_HLIST_HEAD(&head->inodes); |
ac13ac6f EB |
205 | if (node) { |
206 | struct ctl_table *entry; | |
cb55f27a | 207 | |
18d4b42e | 208 | list_for_each_table_entry(entry, head) { |
ac13ac6f | 209 | node->header = head; |
cb55f27a MT |
210 | node++; |
211 | } | |
ac13ac6f | 212 | } |
e0d04529 EB |
213 | } |
214 | ||
8425d6aa EB |
215 | static void erase_header(struct ctl_table_header *head) |
216 | { | |
ac13ac6f | 217 | struct ctl_table *entry; |
cb55f27a | 218 | |
18d4b42e | 219 | list_for_each_table_entry(entry, head) |
ac13ac6f | 220 | erase_entry(head, entry); |
8425d6aa EB |
221 | } |
222 | ||
0e47c99d | 223 | static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header) |
8425d6aa | 224 | { |
ac13ac6f | 225 | struct ctl_table *entry; |
2f2665c1 | 226 | struct ctl_table_header *dir_h = &dir->header; |
0e47c99d EB |
227 | int err; |
228 | ||
2f2665c1 | 229 | |
f9bd6733 | 230 | /* Is this a permanently empty directory? */ |
2f2665c1 | 231 | if (sysctl_is_perm_empty_ctl_header(dir_h)) |
f9bd6733 EB |
232 | return -EROFS; |
233 | ||
234 | /* Am I creating a permanently empty directory? */ | |
31555231 JG |
235 | if (header->ctl_table_size > 0 && |
236 | sysctl_is_perm_empty_ctl_table(header->ctl_table)) { | |
f9bd6733 EB |
237 | if (!RB_EMPTY_ROOT(&dir->root)) |
238 | return -EINVAL; | |
2f2665c1 | 239 | sysctl_set_perm_empty_ctl_header(dir_h); |
f9bd6733 EB |
240 | } |
241 | ||
2f2665c1 | 242 | dir_h->nreg++; |
7ec66d06 | 243 | header->parent = dir; |
0e47c99d EB |
244 | err = insert_links(header); |
245 | if (err) | |
246 | goto fail_links; | |
18d4b42e | 247 | list_for_each_table_entry(entry, header) { |
ac13ac6f EB |
248 | err = insert_entry(header, entry); |
249 | if (err) | |
250 | goto fail; | |
251 | } | |
0e47c99d | 252 | return 0; |
ac13ac6f EB |
253 | fail: |
254 | erase_header(header); | |
255 | put_links(header); | |
0e47c99d | 256 | fail_links: |
f9bd6733 | 257 | if (header->ctl_table == sysctl_mount_point) |
2f2665c1 | 258 | sysctl_clear_perm_empty_ctl_header(dir_h); |
0e47c99d | 259 | header->parent = NULL; |
2f2665c1 | 260 | drop_sysctl_table(dir_h); |
0e47c99d | 261 | return err; |
8425d6aa EB |
262 | } |
263 | ||
1f87f0b5 EB |
264 | /* called under sysctl_lock */ |
265 | static int use_table(struct ctl_table_header *p) | |
266 | { | |
267 | if (unlikely(p->unregistering)) | |
268 | return 0; | |
269 | p->used++; | |
270 | return 1; | |
271 | } | |
272 | ||
273 | /* called under sysctl_lock */ | |
274 | static void unuse_table(struct ctl_table_header *p) | |
275 | { | |
276 | if (!--p->used) | |
277 | if (unlikely(p->unregistering)) | |
278 | complete(p->unregistering); | |
279 | } | |
280 | ||
f90f3caf | 281 | static void proc_sys_invalidate_dcache(struct ctl_table_header *head) |
d6cffbbe | 282 | { |
f90f3caf | 283 | proc_invalidate_siblings_dcache(&head->inodes, &sysctl_lock); |
d6cffbbe KK |
284 | } |
285 | ||
1f87f0b5 EB |
286 | /* called under sysctl_lock, will reacquire if has to wait */ |
287 | static void start_unregistering(struct ctl_table_header *p) | |
288 | { | |
289 | /* | |
290 | * if p->used is 0, nobody will ever touch that entry again; | |
291 | * we'll eliminate all paths to it before dropping sysctl_lock | |
292 | */ | |
293 | if (unlikely(p->used)) { | |
294 | struct completion wait; | |
295 | init_completion(&wait); | |
296 | p->unregistering = &wait; | |
297 | spin_unlock(&sysctl_lock); | |
298 | wait_for_completion(&wait); | |
1f87f0b5 EB |
299 | } else { |
300 | /* anything non-NULL; we'll never dereference it */ | |
301 | p->unregistering = ERR_PTR(-EINVAL); | |
ace0c791 | 302 | spin_unlock(&sysctl_lock); |
1f87f0b5 | 303 | } |
d6cffbbe | 304 | /* |
f90f3caf | 305 | * Invalidate dentries for unregistered sysctls: namespaced sysctls |
d6cffbbe KK |
306 | * can have duplicate names and contaminate dcache very badly. |
307 | */ | |
f90f3caf | 308 | proc_sys_invalidate_dcache(p); |
1f87f0b5 EB |
309 | /* |
310 | * do not remove from the list until nobody holds it; walking the | |
311 | * list in do_sysctl() relies on that. | |
312 | */ | |
ace0c791 | 313 | spin_lock(&sysctl_lock); |
8425d6aa | 314 | erase_header(p); |
1f87f0b5 EB |
315 | } |
316 | ||
1f87f0b5 EB |
317 | static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head) |
318 | { | |
ab4a1f24 | 319 | BUG_ON(!head); |
1f87f0b5 EB |
320 | spin_lock(&sysctl_lock); |
321 | if (!use_table(head)) | |
322 | head = ERR_PTR(-ENOENT); | |
323 | spin_unlock(&sysctl_lock); | |
324 | return head; | |
325 | } | |
326 | ||
327 | static void sysctl_head_finish(struct ctl_table_header *head) | |
328 | { | |
329 | if (!head) | |
330 | return; | |
331 | spin_lock(&sysctl_lock); | |
332 | unuse_table(head); | |
333 | spin_unlock(&sysctl_lock); | |
334 | } | |
335 | ||
336 | static struct ctl_table_set * | |
13bcc6a2 | 337 | lookup_header_set(struct ctl_table_root *root) |
1f87f0b5 EB |
338 | { |
339 | struct ctl_table_set *set = &root->default_set; | |
340 | if (root->lookup) | |
13bcc6a2 | 341 | set = root->lookup(root); |
1f87f0b5 EB |
342 | return set; |
343 | } | |
344 | ||
076c3eed | 345 | static struct ctl_table *lookup_entry(struct ctl_table_header **phead, |
7ec66d06 | 346 | struct ctl_dir *dir, |
076c3eed EB |
347 | const char *name, int namelen) |
348 | { | |
349 | struct ctl_table_header *head; | |
350 | struct ctl_table *entry; | |
076c3eed EB |
351 | |
352 | spin_lock(&sysctl_lock); | |
0e47c99d EB |
353 | entry = find_entry(&head, dir, name, namelen); |
354 | if (entry && use_table(head)) | |
355 | *phead = head; | |
356 | else | |
357 | entry = NULL; | |
076c3eed EB |
358 | spin_unlock(&sysctl_lock); |
359 | return entry; | |
360 | } | |
361 | ||
ac13ac6f | 362 | static struct ctl_node *first_usable_entry(struct rb_node *node) |
1f87f0b5 | 363 | { |
ac13ac6f | 364 | struct ctl_node *ctl_node; |
1f87f0b5 | 365 | |
ac13ac6f EB |
366 | for (;node; node = rb_next(node)) { |
367 | ctl_node = rb_entry(node, struct ctl_node, node); | |
368 | if (use_table(ctl_node->header)) | |
369 | return ctl_node; | |
1f87f0b5 | 370 | } |
1f87f0b5 EB |
371 | return NULL; |
372 | } | |
373 | ||
7ec66d06 | 374 | static void first_entry(struct ctl_dir *dir, |
6a75ce16 EB |
375 | struct ctl_table_header **phead, struct ctl_table **pentry) |
376 | { | |
ac13ac6f | 377 | struct ctl_table_header *head = NULL; |
7ec66d06 | 378 | struct ctl_table *entry = NULL; |
ac13ac6f | 379 | struct ctl_node *ctl_node; |
6a75ce16 EB |
380 | |
381 | spin_lock(&sysctl_lock); | |
ac13ac6f | 382 | ctl_node = first_usable_entry(rb_first(&dir->root)); |
6a75ce16 | 383 | spin_unlock(&sysctl_lock); |
ac13ac6f EB |
384 | if (ctl_node) { |
385 | head = ctl_node->header; | |
386 | entry = &head->ctl_table[ctl_node - head->node]; | |
387 | } | |
6a75ce16 EB |
388 | *phead = head; |
389 | *pentry = entry; | |
390 | } | |
391 | ||
7ec66d06 | 392 | static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry) |
1f87f0b5 | 393 | { |
6a75ce16 EB |
394 | struct ctl_table_header *head = *phead; |
395 | struct ctl_table *entry = *pentry; | |
ac13ac6f | 396 | struct ctl_node *ctl_node = &head->node[entry - head->ctl_table]; |
6a75ce16 | 397 | |
ac13ac6f EB |
398 | spin_lock(&sysctl_lock); |
399 | unuse_table(head); | |
400 | ||
401 | ctl_node = first_usable_entry(rb_next(&ctl_node->node)); | |
402 | spin_unlock(&sysctl_lock); | |
403 | head = NULL; | |
404 | if (ctl_node) { | |
405 | head = ctl_node->header; | |
406 | entry = &head->ctl_table[ctl_node - head->node]; | |
6a75ce16 EB |
407 | } |
408 | *phead = head; | |
409 | *pentry = entry; | |
1f87f0b5 EB |
410 | } |
411 | ||
1f87f0b5 EB |
412 | /* |
413 | * sysctl_perm does NOT grant the superuser all rights automatically, because | |
414 | * some sysctl variables are readonly even to root. | |
415 | */ | |
416 | ||
417 | static int test_perm(int mode, int op) | |
418 | { | |
091bd3ea | 419 | if (uid_eq(current_euid(), GLOBAL_ROOT_UID)) |
1f87f0b5 | 420 | mode >>= 6; |
091bd3ea | 421 | else if (in_egroup_p(GLOBAL_ROOT_GID)) |
1f87f0b5 EB |
422 | mode >>= 3; |
423 | if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0) | |
424 | return 0; | |
425 | return -EACCES; | |
426 | } | |
427 | ||
73f7ef43 | 428 | static int sysctl_perm(struct ctl_table_header *head, struct ctl_table *table, int op) |
1f87f0b5 | 429 | { |
73f7ef43 | 430 | struct ctl_table_root *root = head->root; |
1f87f0b5 EB |
431 | int mode; |
432 | ||
433 | if (root->permissions) | |
73f7ef43 | 434 | mode = root->permissions(head, table); |
1f87f0b5 EB |
435 | else |
436 | mode = table->mode; | |
437 | ||
438 | return test_perm(mode, op); | |
439 | } | |
440 | ||
9043476f AV |
441 | static struct inode *proc_sys_make_inode(struct super_block *sb, |
442 | struct ctl_table_header *head, struct ctl_table *table) | |
77b14db5 | 443 | { |
e79c6a4f | 444 | struct ctl_table_root *root = head->root; |
77b14db5 | 445 | struct inode *inode; |
9043476f | 446 | struct proc_inode *ei; |
77b14db5 | 447 | |
9043476f | 448 | inode = new_inode(sb); |
77b14db5 | 449 | if (!inode) |
ea5751cc | 450 | return ERR_PTR(-ENOMEM); |
77b14db5 | 451 | |
85fe4025 CH |
452 | inode->i_ino = get_next_ino(); |
453 | ||
77b14db5 | 454 | ei = PROC_I(inode); |
9043476f | 455 | |
d6cffbbe | 456 | spin_lock(&sysctl_lock); |
ace0c791 EB |
457 | if (unlikely(head->unregistering)) { |
458 | spin_unlock(&sysctl_lock); | |
459 | iput(inode); | |
ea5751cc | 460 | return ERR_PTR(-ENOENT); |
ace0c791 EB |
461 | } |
462 | ei->sysctl = head; | |
463 | ei->sysctl_entry = table; | |
0afa5ca8 | 464 | hlist_add_head_rcu(&ei->sibling_inodes, &head->inodes); |
d6cffbbe KK |
465 | head->count++; |
466 | spin_unlock(&sysctl_lock); | |
467 | ||
200d9421 | 468 | simple_inode_init_ts(inode); |
9043476f | 469 | inode->i_mode = table->mode; |
7ec66d06 | 470 | if (!S_ISDIR(table->mode)) { |
9043476f AV |
471 | inode->i_mode |= S_IFREG; |
472 | inode->i_op = &proc_sys_inode_operations; | |
473 | inode->i_fop = &proc_sys_file_operations; | |
474 | } else { | |
475 | inode->i_mode |= S_IFDIR; | |
9043476f AV |
476 | inode->i_op = &proc_sys_dir_operations; |
477 | inode->i_fop = &proc_sys_dir_file_operations; | |
2f2665c1 | 478 | if (sysctl_is_perm_empty_ctl_header(head)) |
f9bd6733 | 479 | make_empty_dir_inode(inode); |
9043476f | 480 | } |
e79c6a4f DT |
481 | |
482 | if (root->set_ownership) | |
483 | root->set_ownership(head, table, &inode->i_uid, &inode->i_gid); | |
5ec27ec7 RB |
484 | else { |
485 | inode->i_uid = GLOBAL_ROOT_UID; | |
486 | inode->i_gid = GLOBAL_ROOT_GID; | |
487 | } | |
e79c6a4f | 488 | |
77b14db5 EB |
489 | return inode; |
490 | } | |
491 | ||
d6cffbbe KK |
492 | void proc_sys_evict_inode(struct inode *inode, struct ctl_table_header *head) |
493 | { | |
494 | spin_lock(&sysctl_lock); | |
0afa5ca8 | 495 | hlist_del_init_rcu(&PROC_I(inode)->sibling_inodes); |
d6cffbbe KK |
496 | if (!--head->count) |
497 | kfree_rcu(head, rcu); | |
498 | spin_unlock(&sysctl_lock); | |
499 | } | |
500 | ||
81324364 | 501 | static struct ctl_table_header *grab_header(struct inode *inode) |
77b14db5 | 502 | { |
3cc3e046 EB |
503 | struct ctl_table_header *head = PROC_I(inode)->sysctl; |
504 | if (!head) | |
0e47c99d | 505 | head = &sysctl_table_root.default_set.dir.header; |
3cc3e046 | 506 | return sysctl_head_grab(head); |
9043476f | 507 | } |
77b14db5 | 508 | |
9043476f | 509 | static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry, |
00cd8dd3 | 510 | unsigned int flags) |
9043476f AV |
511 | { |
512 | struct ctl_table_header *head = grab_header(dir); | |
9043476f | 513 | struct ctl_table_header *h = NULL; |
dc12e909 | 514 | const struct qstr *name = &dentry->d_name; |
9043476f AV |
515 | struct ctl_table *p; |
516 | struct inode *inode; | |
517 | struct dentry *err = ERR_PTR(-ENOENT); | |
7ec66d06 | 518 | struct ctl_dir *ctl_dir; |
0e47c99d | 519 | int ret; |
77b14db5 | 520 | |
9043476f AV |
521 | if (IS_ERR(head)) |
522 | return ERR_CAST(head); | |
77b14db5 | 523 | |
7ec66d06 | 524 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 | 525 | |
7ec66d06 | 526 | p = lookup_entry(&h, ctl_dir, name->name, name->len); |
9043476f | 527 | if (!p) |
77b14db5 EB |
528 | goto out; |
529 | ||
4e757320 | 530 | if (S_ISLNK(p->mode)) { |
13bcc6a2 | 531 | ret = sysctl_follow_link(&h, &p); |
4e757320 EB |
532 | err = ERR_PTR(ret); |
533 | if (ret) | |
534 | goto out; | |
535 | } | |
0e47c99d | 536 | |
fb045adb | 537 | d_set_d_op(dentry, &proc_sys_dentry_operations); |
1eae9a47 | 538 | inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p); |
888e2b03 | 539 | err = d_splice_alias(inode, dentry); |
77b14db5 EB |
540 | |
541 | out: | |
6bf61045 FR |
542 | if (h) |
543 | sysctl_head_finish(h); | |
77b14db5 EB |
544 | sysctl_head_finish(head); |
545 | return err; | |
546 | } | |
547 | ||
4bd6a735 MWO |
548 | static ssize_t proc_sys_call_handler(struct kiocb *iocb, struct iov_iter *iter, |
549 | int write) | |
77b14db5 | 550 | { |
4bd6a735 | 551 | struct inode *inode = file_inode(iocb->ki_filp); |
9043476f AV |
552 | struct ctl_table_header *head = grab_header(inode); |
553 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
4bd6a735 MWO |
554 | size_t count = iov_iter_count(iter); |
555 | char *kbuf; | |
2a2da53b | 556 | ssize_t error; |
77b14db5 | 557 | |
9043476f AV |
558 | if (IS_ERR(head)) |
559 | return PTR_ERR(head); | |
77b14db5 EB |
560 | |
561 | /* | |
562 | * At this point we know that the sysctl was not unregistered | |
563 | * and won't be until we finish. | |
564 | */ | |
565 | error = -EPERM; | |
73f7ef43 | 566 | if (sysctl_perm(head, table, write ? MAY_WRITE : MAY_READ)) |
77b14db5 EB |
567 | goto out; |
568 | ||
9043476f AV |
569 | /* if that can happen at all, it should be -EINVAL, not -EISDIR */ |
570 | error = -EINVAL; | |
571 | if (!table->proc_handler) | |
572 | goto out; | |
573 | ||
ef9d965b | 574 | /* don't even try if the size is too large */ |
d4d80e69 MWO |
575 | error = -ENOMEM; |
576 | if (count >= KMALLOC_MAX_SIZE) | |
577 | goto out; | |
45089437 | 578 | kbuf = kvzalloc(count + 1, GFP_KERNEL); |
4bd6a735 MWO |
579 | if (!kbuf) |
580 | goto out; | |
ef9d965b | 581 | |
32927393 | 582 | if (write) { |
4bd6a735 MWO |
583 | error = -EFAULT; |
584 | if (!copy_from_iter_full(kbuf, count, iter)) | |
585 | goto out_free_buf; | |
586 | kbuf[count] = '\0'; | |
32927393 CH |
587 | } |
588 | ||
589 | error = BPF_CGROUP_RUN_PROG_SYSCTL(head, table, write, &kbuf, &count, | |
4bd6a735 | 590 | &iocb->ki_pos); |
7b146ceb | 591 | if (error) |
32927393 | 592 | goto out_free_buf; |
7b146ceb | 593 | |
77b14db5 | 594 | /* careful: calling conventions are nasty here */ |
4bd6a735 | 595 | error = table->proc_handler(table, write, kbuf, &count, &iocb->ki_pos); |
32927393 CH |
596 | if (error) |
597 | goto out_free_buf; | |
598 | ||
599 | if (!write) { | |
600 | error = -EFAULT; | |
4bd6a735 | 601 | if (copy_to_iter(kbuf, count, iter) < count) |
32927393 | 602 | goto out_free_buf; |
4e63acdf AI |
603 | } |
604 | ||
32927393 CH |
605 | error = count; |
606 | out_free_buf: | |
45089437 | 607 | kvfree(kbuf); |
77b14db5 EB |
608 | out: |
609 | sysctl_head_finish(head); | |
610 | ||
611 | return error; | |
612 | } | |
613 | ||
4bd6a735 | 614 | static ssize_t proc_sys_read(struct kiocb *iocb, struct iov_iter *iter) |
77b14db5 | 615 | { |
4bd6a735 | 616 | return proc_sys_call_handler(iocb, iter, 0); |
7708bfb1 | 617 | } |
77b14db5 | 618 | |
4bd6a735 | 619 | static ssize_t proc_sys_write(struct kiocb *iocb, struct iov_iter *iter) |
7708bfb1 | 620 | { |
4bd6a735 | 621 | return proc_sys_call_handler(iocb, iter, 1); |
77b14db5 EB |
622 | } |
623 | ||
f1ecf068 LDM |
624 | static int proc_sys_open(struct inode *inode, struct file *filp) |
625 | { | |
4e474a00 | 626 | struct ctl_table_header *head = grab_header(inode); |
f1ecf068 LDM |
627 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; |
628 | ||
4e474a00 LDM |
629 | /* sysctl was unregistered */ |
630 | if (IS_ERR(head)) | |
631 | return PTR_ERR(head); | |
632 | ||
f1ecf068 LDM |
633 | if (table->poll) |
634 | filp->private_data = proc_sys_poll_event(table->poll); | |
635 | ||
4e474a00 LDM |
636 | sysctl_head_finish(head); |
637 | ||
f1ecf068 LDM |
638 | return 0; |
639 | } | |
640 | ||
076ccb76 | 641 | static __poll_t proc_sys_poll(struct file *filp, poll_table *wait) |
f1ecf068 | 642 | { |
496ad9aa | 643 | struct inode *inode = file_inode(filp); |
4e474a00 | 644 | struct ctl_table_header *head = grab_header(inode); |
f1ecf068 | 645 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; |
076ccb76 | 646 | __poll_t ret = DEFAULT_POLLMASK; |
4e474a00 LDM |
647 | unsigned long event; |
648 | ||
649 | /* sysctl was unregistered */ | |
650 | if (IS_ERR(head)) | |
a9a08845 | 651 | return EPOLLERR | EPOLLHUP; |
f1ecf068 LDM |
652 | |
653 | if (!table->proc_handler) | |
654 | goto out; | |
655 | ||
656 | if (!table->poll) | |
657 | goto out; | |
658 | ||
4e474a00 | 659 | event = (unsigned long)filp->private_data; |
f1ecf068 LDM |
660 | poll_wait(filp, &table->poll->wait, wait); |
661 | ||
662 | if (event != atomic_read(&table->poll->event)) { | |
663 | filp->private_data = proc_sys_poll_event(table->poll); | |
a9a08845 | 664 | ret = EPOLLIN | EPOLLRDNORM | EPOLLERR | EPOLLPRI; |
f1ecf068 LDM |
665 | } |
666 | ||
667 | out: | |
4e474a00 LDM |
668 | sysctl_head_finish(head); |
669 | ||
f1ecf068 LDM |
670 | return ret; |
671 | } | |
77b14db5 | 672 | |
f0c3b509 AV |
673 | static bool proc_sys_fill_cache(struct file *file, |
674 | struct dir_context *ctx, | |
9043476f AV |
675 | struct ctl_table_header *head, |
676 | struct ctl_table *table) | |
77b14db5 | 677 | { |
f0c3b509 | 678 | struct dentry *child, *dir = file->f_path.dentry; |
77b14db5 EB |
679 | struct inode *inode; |
680 | struct qstr qname; | |
681 | ino_t ino = 0; | |
682 | unsigned type = DT_UNKNOWN; | |
77b14db5 EB |
683 | |
684 | qname.name = table->procname; | |
685 | qname.len = strlen(table->procname); | |
8387ff25 | 686 | qname.hash = full_name_hash(dir, qname.name, qname.len); |
77b14db5 | 687 | |
77b14db5 EB |
688 | child = d_lookup(dir, &qname); |
689 | if (!child) { | |
76aab3ab AV |
690 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
691 | child = d_alloc_parallel(dir, &qname, &wq); | |
692 | if (IS_ERR(child)) | |
693 | return false; | |
694 | if (d_in_lookup(child)) { | |
888e2b03 | 695 | struct dentry *res; |
76aab3ab | 696 | d_set_d_op(child, &proc_sys_dentry_operations); |
1eae9a47 | 697 | inode = proc_sys_make_inode(dir->d_sb, head, table); |
888e2b03 AV |
698 | res = d_splice_alias(inode, child); |
699 | d_lookup_done(child); | |
700 | if (unlikely(res)) { | |
701 | if (IS_ERR(res)) { | |
702 | dput(child); | |
703 | return false; | |
704 | } | |
705 | dput(child); | |
706 | child = res; | |
707 | } | |
77b14db5 EB |
708 | } |
709 | } | |
2b0143b5 | 710 | inode = d_inode(child); |
9043476f AV |
711 | ino = inode->i_ino; |
712 | type = inode->i_mode >> 12; | |
77b14db5 | 713 | dput(child); |
f0c3b509 | 714 | return dir_emit(ctx, qname.name, qname.len, ino, type); |
9043476f AV |
715 | } |
716 | ||
f0c3b509 AV |
717 | static bool proc_sys_link_fill_cache(struct file *file, |
718 | struct dir_context *ctx, | |
0e47c99d EB |
719 | struct ctl_table_header *head, |
720 | struct ctl_table *table) | |
721 | { | |
f0c3b509 | 722 | bool ret = true; |
a0b0d1c3 | 723 | |
0e47c99d | 724 | head = sysctl_head_grab(head); |
a0b0d1c3 DK |
725 | if (IS_ERR(head)) |
726 | return false; | |
0e47c99d | 727 | |
835b94e0 DK |
728 | /* It is not an error if we can not follow the link ignore it */ |
729 | if (sysctl_follow_link(&head, &table)) | |
730 | goto out; | |
0e47c99d | 731 | |
f0c3b509 | 732 | ret = proc_sys_fill_cache(file, ctx, head, table); |
0e47c99d EB |
733 | out: |
734 | sysctl_head_finish(head); | |
735 | return ret; | |
736 | } | |
737 | ||
e5eea098 | 738 | static int scan(struct ctl_table_header *head, struct ctl_table *table, |
9043476f | 739 | unsigned long *pos, struct file *file, |
f0c3b509 | 740 | struct dir_context *ctx) |
9043476f | 741 | { |
f0c3b509 | 742 | bool res; |
9043476f | 743 | |
f0c3b509 AV |
744 | if ((*pos)++ < ctx->pos) |
745 | return true; | |
9043476f | 746 | |
0e47c99d | 747 | if (unlikely(S_ISLNK(table->mode))) |
f0c3b509 | 748 | res = proc_sys_link_fill_cache(file, ctx, head, table); |
0e47c99d | 749 | else |
f0c3b509 | 750 | res = proc_sys_fill_cache(file, ctx, head, table); |
9043476f | 751 | |
f0c3b509 AV |
752 | if (res) |
753 | ctx->pos = *pos; | |
9043476f | 754 | |
6a75ce16 | 755 | return res; |
77b14db5 EB |
756 | } |
757 | ||
f0c3b509 | 758 | static int proc_sys_readdir(struct file *file, struct dir_context *ctx) |
77b14db5 | 759 | { |
f0c3b509 | 760 | struct ctl_table_header *head = grab_header(file_inode(file)); |
9043476f | 761 | struct ctl_table_header *h = NULL; |
6a75ce16 | 762 | struct ctl_table *entry; |
7ec66d06 | 763 | struct ctl_dir *ctl_dir; |
77b14db5 | 764 | unsigned long pos; |
9043476f AV |
765 | |
766 | if (IS_ERR(head)) | |
767 | return PTR_ERR(head); | |
77b14db5 | 768 | |
7ec66d06 | 769 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 | 770 | |
f0c3b509 | 771 | if (!dir_emit_dots(file, ctx)) |
93362fa4 | 772 | goto out; |
f0c3b509 | 773 | |
77b14db5 EB |
774 | pos = 2; |
775 | ||
7ec66d06 | 776 | for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) { |
f0c3b509 | 777 | if (!scan(h, entry, &pos, file, ctx)) { |
9043476f AV |
778 | sysctl_head_finish(h); |
779 | break; | |
77b14db5 EB |
780 | } |
781 | } | |
93362fa4 | 782 | out: |
77b14db5 | 783 | sysctl_head_finish(head); |
f0c3b509 | 784 | return 0; |
77b14db5 EB |
785 | } |
786 | ||
4609e1f1 | 787 | static int proc_sys_permission(struct mnt_idmap *idmap, |
549c7297 | 788 | struct inode *inode, int mask) |
77b14db5 EB |
789 | { |
790 | /* | |
791 | * sysctl entries that are not writeable, | |
792 | * are _NOT_ writeable, capabilities or not. | |
793 | */ | |
f696a365 MS |
794 | struct ctl_table_header *head; |
795 | struct ctl_table *table; | |
77b14db5 EB |
796 | int error; |
797 | ||
f696a365 MS |
798 | /* Executable files are not allowed under /proc/sys/ */ |
799 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) | |
800 | return -EACCES; | |
801 | ||
802 | head = grab_header(inode); | |
9043476f AV |
803 | if (IS_ERR(head)) |
804 | return PTR_ERR(head); | |
77b14db5 | 805 | |
f696a365 | 806 | table = PROC_I(inode)->sysctl_entry; |
9043476f AV |
807 | if (!table) /* global root - r-xr-xr-x */ |
808 | error = mask & MAY_WRITE ? -EACCES : 0; | |
809 | else /* Use the permissions on the sysctl table entry */ | |
73f7ef43 | 810 | error = sysctl_perm(head, table, mask & ~MAY_NOT_BLOCK); |
77b14db5 | 811 | |
77b14db5 EB |
812 | sysctl_head_finish(head); |
813 | return error; | |
814 | } | |
815 | ||
c1632a0f | 816 | static int proc_sys_setattr(struct mnt_idmap *idmap, |
549c7297 | 817 | struct dentry *dentry, struct iattr *attr) |
77b14db5 | 818 | { |
2b0143b5 | 819 | struct inode *inode = d_inode(dentry); |
77b14db5 EB |
820 | int error; |
821 | ||
822 | if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) | |
823 | return -EPERM; | |
824 | ||
c1632a0f | 825 | error = setattr_prepare(&nop_mnt_idmap, dentry, attr); |
1025774c CH |
826 | if (error) |
827 | return error; | |
828 | ||
c1632a0f | 829 | setattr_copy(&nop_mnt_idmap, inode, attr); |
1025774c | 830 | return 0; |
77b14db5 EB |
831 | } |
832 | ||
b74d24f7 | 833 | static int proc_sys_getattr(struct mnt_idmap *idmap, |
549c7297 | 834 | const struct path *path, struct kstat *stat, |
a528d35e | 835 | u32 request_mask, unsigned int query_flags) |
9043476f | 836 | { |
a528d35e | 837 | struct inode *inode = d_inode(path->dentry); |
9043476f AV |
838 | struct ctl_table_header *head = grab_header(inode); |
839 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
840 | ||
841 | if (IS_ERR(head)) | |
842 | return PTR_ERR(head); | |
843 | ||
0d72b928 | 844 | generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat); |
9043476f AV |
845 | if (table) |
846 | stat->mode = (stat->mode & S_IFMT) | table->mode; | |
847 | ||
848 | sysctl_head_finish(head); | |
849 | return 0; | |
850 | } | |
851 | ||
77b14db5 | 852 | static const struct file_operations proc_sys_file_operations = { |
f1ecf068 LDM |
853 | .open = proc_sys_open, |
854 | .poll = proc_sys_poll, | |
4bd6a735 MWO |
855 | .read_iter = proc_sys_read, |
856 | .write_iter = proc_sys_write, | |
b0072734 | 857 | .splice_read = copy_splice_read, |
4bd6a735 | 858 | .splice_write = iter_file_splice_write, |
6038f373 | 859 | .llseek = default_llseek, |
9043476f AV |
860 | }; |
861 | ||
862 | static const struct file_operations proc_sys_dir_file_operations = { | |
887df078 | 863 | .read = generic_read_dir, |
f50752ea | 864 | .iterate_shared = proc_sys_readdir, |
3222a3e5 | 865 | .llseek = generic_file_llseek, |
77b14db5 EB |
866 | }; |
867 | ||
03a44825 | 868 | static const struct inode_operations proc_sys_inode_operations = { |
9043476f AV |
869 | .permission = proc_sys_permission, |
870 | .setattr = proc_sys_setattr, | |
871 | .getattr = proc_sys_getattr, | |
872 | }; | |
873 | ||
874 | static const struct inode_operations proc_sys_dir_operations = { | |
77b14db5 EB |
875 | .lookup = proc_sys_lookup, |
876 | .permission = proc_sys_permission, | |
877 | .setattr = proc_sys_setattr, | |
9043476f | 878 | .getattr = proc_sys_getattr, |
77b14db5 EB |
879 | }; |
880 | ||
0b728e19 | 881 | static int proc_sys_revalidate(struct dentry *dentry, unsigned int flags) |
77b14db5 | 882 | { |
0b728e19 | 883 | if (flags & LOOKUP_RCU) |
34286d66 | 884 | return -ECHILD; |
2b0143b5 | 885 | return !PROC_I(d_inode(dentry))->sysctl->unregistering; |
9043476f AV |
886 | } |
887 | ||
fe15ce44 | 888 | static int proc_sys_delete(const struct dentry *dentry) |
9043476f | 889 | { |
2b0143b5 | 890 | return !!PROC_I(d_inode(dentry))->sysctl->unregistering; |
9043476f AV |
891 | } |
892 | ||
1f87f0b5 EB |
893 | static int sysctl_is_seen(struct ctl_table_header *p) |
894 | { | |
895 | struct ctl_table_set *set = p->set; | |
896 | int res; | |
897 | spin_lock(&sysctl_lock); | |
898 | if (p->unregistering) | |
899 | res = 0; | |
900 | else if (!set->is_seen) | |
901 | res = 1; | |
902 | else | |
903 | res = set->is_seen(set); | |
904 | spin_unlock(&sysctl_lock); | |
905 | return res; | |
906 | } | |
907 | ||
6fa67e70 | 908 | static int proc_sys_compare(const struct dentry *dentry, |
621e155a | 909 | unsigned int len, const char *str, const struct qstr *name) |
9043476f | 910 | { |
dfef6dcd | 911 | struct ctl_table_header *head; |
da53be12 LT |
912 | struct inode *inode; |
913 | ||
31e6b01f NP |
914 | /* Although proc doesn't have negative dentries, rcu-walk means |
915 | * that inode here can be NULL */ | |
dfef6dcd | 916 | /* AV: can it, indeed? */ |
2b0143b5 | 917 | inode = d_inode_rcu(dentry); |
31e6b01f | 918 | if (!inode) |
dfef6dcd | 919 | return 1; |
621e155a | 920 | if (name->len != len) |
9043476f | 921 | return 1; |
621e155a | 922 | if (memcmp(name->name, str, len)) |
9043476f | 923 | return 1; |
dfef6dcd AV |
924 | head = rcu_dereference(PROC_I(inode)->sysctl); |
925 | return !head || !sysctl_is_seen(head); | |
77b14db5 EB |
926 | } |
927 | ||
d72f71eb | 928 | static const struct dentry_operations proc_sys_dentry_operations = { |
77b14db5 | 929 | .d_revalidate = proc_sys_revalidate, |
9043476f AV |
930 | .d_delete = proc_sys_delete, |
931 | .d_compare = proc_sys_compare, | |
77b14db5 EB |
932 | }; |
933 | ||
0e47c99d EB |
934 | static struct ctl_dir *find_subdir(struct ctl_dir *dir, |
935 | const char *name, int namelen) | |
1f87f0b5 | 936 | { |
7ec66d06 EB |
937 | struct ctl_table_header *head; |
938 | struct ctl_table *entry; | |
1f87f0b5 | 939 | |
0e47c99d | 940 | entry = find_entry(&head, dir, name, namelen); |
7ec66d06 EB |
941 | if (!entry) |
942 | return ERR_PTR(-ENOENT); | |
51f72f4a EB |
943 | if (!S_ISDIR(entry->mode)) |
944 | return ERR_PTR(-ENOTDIR); | |
945 | return container_of(head, struct ctl_dir, header); | |
7ec66d06 EB |
946 | } |
947 | ||
948 | static struct ctl_dir *new_dir(struct ctl_table_set *set, | |
0e47c99d | 949 | const char *name, int namelen) |
7ec66d06 EB |
950 | { |
951 | struct ctl_table *table; | |
952 | struct ctl_dir *new; | |
ac13ac6f | 953 | struct ctl_node *node; |
7ec66d06 | 954 | char *new_name; |
1f87f0b5 | 955 | |
ac13ac6f EB |
956 | new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) + |
957 | sizeof(struct ctl_table)*2 + namelen + 1, | |
958 | GFP_KERNEL); | |
7ec66d06 | 959 | if (!new) |
1f87f0b5 EB |
960 | return NULL; |
961 | ||
ac13ac6f EB |
962 | node = (struct ctl_node *)(new + 1); |
963 | table = (struct ctl_table *)(node + 1); | |
7ec66d06 EB |
964 | new_name = (char *)(table + 2); |
965 | memcpy(new_name, name, namelen); | |
7ec66d06 EB |
966 | table[0].procname = new_name; |
967 | table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
b1f01e2b | 968 | init_header(&new->header, set->dir.header.root, set, node, table, 1); |
7ec66d06 EB |
969 | |
970 | return new; | |
1f87f0b5 EB |
971 | } |
972 | ||
60f126d9 EB |
973 | /** |
974 | * get_subdir - find or create a subdir with the specified name. | |
975 | * @dir: Directory to create the subdirectory in | |
976 | * @name: The name of the subdirectory to find or create | |
977 | * @namelen: The length of name | |
978 | * | |
979 | * Takes a directory with an elevated reference count so we know that | |
980 | * if we drop the lock the directory will not go away. Upon success | |
981 | * the reference is moved from @dir to the returned subdirectory. | |
982 | * Upon error an error code is returned and the reference on @dir is | |
983 | * simply dropped. | |
984 | */ | |
0e47c99d EB |
985 | static struct ctl_dir *get_subdir(struct ctl_dir *dir, |
986 | const char *name, int namelen) | |
1f87f0b5 | 987 | { |
0e47c99d | 988 | struct ctl_table_set *set = dir->header.set; |
7ec66d06 | 989 | struct ctl_dir *subdir, *new = NULL; |
0eb97f38 | 990 | int err; |
1f87f0b5 | 991 | |
7ec66d06 | 992 | spin_lock(&sysctl_lock); |
0e47c99d | 993 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
994 | if (!IS_ERR(subdir)) |
995 | goto found; | |
996 | if (PTR_ERR(subdir) != -ENOENT) | |
997 | goto failed; | |
998 | ||
999 | spin_unlock(&sysctl_lock); | |
1000 | new = new_dir(set, name, namelen); | |
1001 | spin_lock(&sysctl_lock); | |
1002 | subdir = ERR_PTR(-ENOMEM); | |
1003 | if (!new) | |
1004 | goto failed; | |
1005 | ||
60f126d9 | 1006 | /* Was the subdir added while we dropped the 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 | ||
60f126d9 | 1013 | /* Nope. Use the our freshly made directory entry. */ |
0eb97f38 EB |
1014 | err = insert_header(dir, &new->header); |
1015 | subdir = ERR_PTR(err); | |
1016 | if (err) | |
0e47c99d | 1017 | goto failed; |
7ec66d06 EB |
1018 | subdir = new; |
1019 | found: | |
1020 | subdir->header.nreg++; | |
1021 | failed: | |
a1c83681 | 1022 | if (IS_ERR(subdir)) { |
87ebdc00 | 1023 | pr_err("sysctl could not get directory: "); |
6980128f | 1024 | sysctl_print_dir(dir); |
153ee1c4 GU |
1025 | pr_cont("%*.*s %ld\n", namelen, namelen, name, |
1026 | PTR_ERR(subdir)); | |
1f87f0b5 | 1027 | } |
7ec66d06 EB |
1028 | drop_sysctl_table(&dir->header); |
1029 | if (new) | |
1030 | drop_sysctl_table(&new->header); | |
1031 | spin_unlock(&sysctl_lock); | |
1032 | return subdir; | |
1f87f0b5 EB |
1033 | } |
1034 | ||
0e47c99d EB |
1035 | static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir) |
1036 | { | |
1037 | struct ctl_dir *parent; | |
1038 | const char *procname; | |
1039 | if (!dir->header.parent) | |
1040 | return &set->dir; | |
1041 | parent = xlate_dir(set, dir->header.parent); | |
1042 | if (IS_ERR(parent)) | |
1043 | return parent; | |
1044 | procname = dir->header.ctl_table[0].procname; | |
1045 | return find_subdir(parent, procname, strlen(procname)); | |
1046 | } | |
1047 | ||
1048 | static int sysctl_follow_link(struct ctl_table_header **phead, | |
13bcc6a2 | 1049 | struct ctl_table **pentry) |
0e47c99d EB |
1050 | { |
1051 | struct ctl_table_header *head; | |
1052 | struct ctl_table_root *root; | |
1053 | struct ctl_table_set *set; | |
1054 | struct ctl_table *entry; | |
1055 | struct ctl_dir *dir; | |
1056 | int ret; | |
1057 | ||
0e47c99d EB |
1058 | spin_lock(&sysctl_lock); |
1059 | root = (*pentry)->data; | |
13bcc6a2 | 1060 | set = lookup_header_set(root); |
0e47c99d EB |
1061 | dir = xlate_dir(set, (*phead)->parent); |
1062 | if (IS_ERR(dir)) | |
1063 | ret = PTR_ERR(dir); | |
1064 | else { | |
1065 | const char *procname = (*pentry)->procname; | |
1066 | head = NULL; | |
1067 | entry = find_entry(&head, dir, procname, strlen(procname)); | |
1068 | ret = -ENOENT; | |
1069 | if (entry && use_table(head)) { | |
1070 | unuse_table(*phead); | |
1071 | *phead = head; | |
1072 | *pentry = entry; | |
1073 | ret = 0; | |
1074 | } | |
1075 | } | |
1076 | ||
1077 | spin_unlock(&sysctl_lock); | |
1078 | return ret; | |
1079 | } | |
1080 | ||
7c60c48f | 1081 | static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...) |
1f87f0b5 | 1082 | { |
7c60c48f EB |
1083 | struct va_format vaf; |
1084 | va_list args; | |
1f87f0b5 | 1085 | |
7c60c48f EB |
1086 | va_start(args, fmt); |
1087 | vaf.fmt = fmt; | |
1088 | vaf.va = &args; | |
1089 | ||
87ebdc00 AM |
1090 | pr_err("sysctl table check failed: %s/%s %pV\n", |
1091 | path, table->procname, &vaf); | |
1f87f0b5 | 1092 | |
7c60c48f EB |
1093 | va_end(args); |
1094 | return -EINVAL; | |
1f87f0b5 EB |
1095 | } |
1096 | ||
4f2fec00 LR |
1097 | static int sysctl_check_table_array(const char *path, struct ctl_table *table) |
1098 | { | |
1099 | int err = 0; | |
1100 | ||
61d9b56a LR |
1101 | if ((table->proc_handler == proc_douintvec) || |
1102 | (table->proc_handler == proc_douintvec_minmax)) { | |
4f2fec00 | 1103 | if (table->maxlen != sizeof(unsigned int)) |
64a11f3d | 1104 | err |= sysctl_err(path, table, "array not allowed"); |
4f2fec00 LR |
1105 | } |
1106 | ||
cb944413 ED |
1107 | if (table->proc_handler == proc_dou8vec_minmax) { |
1108 | if (table->maxlen != sizeof(u8)) | |
1109 | err |= sysctl_err(path, table, "array not allowed"); | |
1110 | } | |
1111 | ||
f1aa2eb5 OM |
1112 | if (table->proc_handler == proc_dobool) { |
1113 | if (table->maxlen != sizeof(bool)) | |
1114 | err |= sysctl_err(path, table, "array not allowed"); | |
1115 | } | |
1116 | ||
4f2fec00 LR |
1117 | return err; |
1118 | } | |
1119 | ||
cc9f7ee0 | 1120 | static int sysctl_check_table(const char *path, struct ctl_table_header *header) |
1f87f0b5 | 1121 | { |
cb55f27a | 1122 | struct ctl_table *entry; |
7c60c48f | 1123 | int err = 0; |
18d4b42e | 1124 | list_for_each_table_entry(entry, header) { |
cb55f27a | 1125 | if ((entry->proc_handler == proc_dostring) || |
f1aa2eb5 | 1126 | (entry->proc_handler == proc_dobool) || |
cb55f27a MT |
1127 | (entry->proc_handler == proc_dointvec) || |
1128 | (entry->proc_handler == proc_douintvec) || | |
1129 | (entry->proc_handler == proc_douintvec_minmax) || | |
1130 | (entry->proc_handler == proc_dointvec_minmax) || | |
1131 | (entry->proc_handler == proc_dou8vec_minmax) || | |
1132 | (entry->proc_handler == proc_dointvec_jiffies) || | |
1133 | (entry->proc_handler == proc_dointvec_userhz_jiffies) || | |
1134 | (entry->proc_handler == proc_dointvec_ms_jiffies) || | |
1135 | (entry->proc_handler == proc_doulongvec_minmax) || | |
1136 | (entry->proc_handler == proc_doulongvec_ms_jiffies_minmax)) { | |
1137 | if (!entry->data) | |
1138 | err |= sysctl_err(path, entry, "No data"); | |
1139 | if (!entry->maxlen) | |
1140 | err |= sysctl_err(path, entry, "No maxlen"); | |
4f2fec00 | 1141 | else |
cb55f27a | 1142 | err |= sysctl_check_table_array(path, entry); |
7c60c48f | 1143 | } |
cb55f27a MT |
1144 | if (!entry->proc_handler) |
1145 | err |= sysctl_err(path, entry, "No proc_handler"); | |
7c60c48f | 1146 | |
cb55f27a MT |
1147 | if ((entry->mode & (S_IRUGO|S_IWUGO)) != entry->mode) |
1148 | err |= sysctl_err(path, entry, "bogus .mode 0%o", | |
1149 | entry->mode); | |
1f87f0b5 | 1150 | } |
7c60c48f | 1151 | return err; |
1f87f0b5 | 1152 | } |
1f87f0b5 | 1153 | |
cc9f7ee0 | 1154 | static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table_header *head) |
0e47c99d EB |
1155 | { |
1156 | struct ctl_table *link_table, *entry, *link; | |
1157 | struct ctl_table_header *links; | |
ac13ac6f | 1158 | struct ctl_node *node; |
0e47c99d EB |
1159 | char *link_name; |
1160 | int nr_entries, name_bytes; | |
1161 | ||
1162 | name_bytes = 0; | |
1163 | nr_entries = 0; | |
18d4b42e | 1164 | list_for_each_table_entry(entry, head) { |
0e47c99d EB |
1165 | nr_entries++; |
1166 | name_bytes += strlen(entry->procname) + 1; | |
1167 | } | |
1168 | ||
1169 | links = kzalloc(sizeof(struct ctl_table_header) + | |
ac13ac6f | 1170 | sizeof(struct ctl_node)*nr_entries + |
0e47c99d EB |
1171 | sizeof(struct ctl_table)*(nr_entries + 1) + |
1172 | name_bytes, | |
1173 | GFP_KERNEL); | |
1174 | ||
1175 | if (!links) | |
1176 | return NULL; | |
1177 | ||
ac13ac6f EB |
1178 | node = (struct ctl_node *)(links + 1); |
1179 | link_table = (struct ctl_table *)(node + nr_entries); | |
0e47c99d | 1180 | link_name = (char *)&link_table[nr_entries + 1]; |
cb55f27a | 1181 | link = link_table; |
0e47c99d | 1182 | |
18d4b42e | 1183 | list_for_each_table_entry(entry, head) { |
0e47c99d EB |
1184 | int len = strlen(entry->procname) + 1; |
1185 | memcpy(link_name, entry->procname, len); | |
1186 | link->procname = link_name; | |
1187 | link->mode = S_IFLNK|S_IRWXUGO; | |
cc9f7ee0 | 1188 | link->data = head->root; |
0e47c99d | 1189 | link_name += len; |
cb55f27a | 1190 | link++; |
0e47c99d | 1191 | } |
b1f01e2b JG |
1192 | init_header(links, dir->header.root, dir->header.set, node, link_table, |
1193 | head->ctl_table_size); | |
0e47c99d EB |
1194 | links->nreg = nr_entries; |
1195 | ||
1196 | return links; | |
1197 | } | |
1198 | ||
1199 | static bool get_links(struct ctl_dir *dir, | |
cc9f7ee0 JG |
1200 | struct ctl_table_header *header, |
1201 | struct ctl_table_root *link_root) | |
0e47c99d | 1202 | { |
cc9f7ee0 | 1203 | struct ctl_table_header *tmp_head; |
0e47c99d EB |
1204 | struct ctl_table *entry, *link; |
1205 | ||
31555231 JG |
1206 | if (header->ctl_table_size == 0 || |
1207 | sysctl_is_perm_empty_ctl_table(header->ctl_table)) | |
1208 | return true; | |
1209 | ||
0e47c99d | 1210 | /* Are there links available for every entry in table? */ |
18d4b42e | 1211 | list_for_each_table_entry(entry, header) { |
0e47c99d | 1212 | const char *procname = entry->procname; |
cc9f7ee0 | 1213 | link = find_entry(&tmp_head, dir, procname, strlen(procname)); |
0e47c99d EB |
1214 | if (!link) |
1215 | return false; | |
1216 | if (S_ISDIR(link->mode) && S_ISDIR(entry->mode)) | |
1217 | continue; | |
1218 | if (S_ISLNK(link->mode) && (link->data == link_root)) | |
1219 | continue; | |
1220 | return false; | |
1221 | } | |
1222 | ||
1223 | /* The checks passed. Increase the registration count on the links */ | |
18d4b42e | 1224 | list_for_each_table_entry(entry, header) { |
0e47c99d | 1225 | const char *procname = entry->procname; |
cc9f7ee0 JG |
1226 | link = find_entry(&tmp_head, dir, procname, strlen(procname)); |
1227 | tmp_head->nreg++; | |
0e47c99d EB |
1228 | } |
1229 | return true; | |
1230 | } | |
1231 | ||
1232 | static int insert_links(struct ctl_table_header *head) | |
1233 | { | |
1234 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
9a521359 | 1235 | struct ctl_dir *core_parent; |
0e47c99d EB |
1236 | struct ctl_table_header *links; |
1237 | int err; | |
1238 | ||
1239 | if (head->set == root_set) | |
1240 | return 0; | |
1241 | ||
1242 | core_parent = xlate_dir(root_set, head->parent); | |
1243 | if (IS_ERR(core_parent)) | |
1244 | return 0; | |
1245 | ||
cc9f7ee0 | 1246 | if (get_links(core_parent, head, head->root)) |
0e47c99d EB |
1247 | return 0; |
1248 | ||
1249 | core_parent->header.nreg++; | |
1250 | spin_unlock(&sysctl_lock); | |
1251 | ||
cc9f7ee0 | 1252 | links = new_links(core_parent, head); |
0e47c99d EB |
1253 | |
1254 | spin_lock(&sysctl_lock); | |
1255 | err = -ENOMEM; | |
1256 | if (!links) | |
1257 | goto out; | |
1258 | ||
1259 | err = 0; | |
cc9f7ee0 | 1260 | if (get_links(core_parent, head, head->root)) { |
0e47c99d EB |
1261 | kfree(links); |
1262 | goto out; | |
1263 | } | |
1264 | ||
1265 | err = insert_header(core_parent, links); | |
1266 | if (err) | |
1267 | kfree(links); | |
1268 | out: | |
1269 | drop_sysctl_table(&core_parent->header); | |
1270 | return err; | |
1271 | } | |
1272 | ||
b2f56e55 LC |
1273 | /* Find the directory for the ctl_table. If one is not found create it. */ |
1274 | static struct ctl_dir *sysctl_mkdir_p(struct ctl_dir *dir, const char *path) | |
1275 | { | |
1276 | const char *name, *nextname; | |
1277 | ||
1278 | for (name = path; name; name = nextname) { | |
1279 | int namelen; | |
1280 | nextname = strchr(name, '/'); | |
1281 | if (nextname) { | |
1282 | namelen = nextname - name; | |
1283 | nextname++; | |
1284 | } else { | |
1285 | namelen = strlen(name); | |
1286 | } | |
1287 | if (namelen == 0) | |
1288 | continue; | |
1289 | ||
1290 | /* | |
1291 | * namelen ensures if name is "foo/bar/yay" only foo is | |
1292 | * registered first. We traverse as if using mkdir -p and | |
1293 | * return a ctl_dir for the last directory entry. | |
1294 | */ | |
1295 | dir = get_subdir(dir, name, namelen); | |
1296 | if (IS_ERR(dir)) | |
1297 | break; | |
1298 | } | |
1299 | return dir; | |
1300 | } | |
1301 | ||
1f87f0b5 | 1302 | /** |
f728019b | 1303 | * __register_sysctl_table - register a leaf sysctl table |
60a47a2e | 1304 | * @set: Sysctl tree to register on |
1f87f0b5 | 1305 | * @path: The path to the directory the sysctl table is in. |
1dc8689e LC |
1306 | * @table: the top-level table structure without any child. This table |
1307 | * should not be free'd after registration. So it should not be | |
1308 | * used on stack. It can either be a global or dynamically allocated | |
1309 | * by the caller and free'd later after sysctl unregistration. | |
bff97cf1 | 1310 | * @table_size : The number of elements in table |
1f87f0b5 EB |
1311 | * |
1312 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1313 | * array. A completely 0 filled entry terminates the table. | |
1314 | * | |
1315 | * The members of the &struct ctl_table structure are used as follows: | |
1316 | * | |
1317 | * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not | |
1318 | * enter a sysctl file | |
1319 | * | |
1320 | * data - a pointer to data for use by proc_handler | |
1321 | * | |
1322 | * maxlen - the maximum size in bytes of the data | |
1323 | * | |
f728019b | 1324 | * mode - the file permissions for the /proc/sys file |
1f87f0b5 | 1325 | * |
f728019b | 1326 | * child - must be %NULL. |
1f87f0b5 EB |
1327 | * |
1328 | * proc_handler - the text handler routine (described below) | |
1329 | * | |
1f87f0b5 | 1330 | * extra1, extra2 - extra pointers usable by the proc handler routines |
67ff3228 LC |
1331 | * XXX: we should eventually modify these to use long min / max [0] |
1332 | * [0] https://lkml.kernel.org/87zgpte9o4.fsf@email.froward.int.ebiederm.org | |
1f87f0b5 EB |
1333 | * |
1334 | * Leaf nodes in the sysctl tree will be represented by a single file | |
67ff3228 LC |
1335 | * under /proc; non-leaf nodes (where child is not NULL) are not allowed, |
1336 | * sysctl_check_table() verifies this. | |
1f87f0b5 | 1337 | * |
f728019b EB |
1338 | * There must be a proc_handler routine for any terminal nodes. |
1339 | * Several default handlers are available to cover common cases - | |
1f87f0b5 EB |
1340 | * |
1341 | * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), | |
1342 | * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), | |
1343 | * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() | |
1344 | * | |
1345 | * It is the handler's job to read the input buffer from user memory | |
1346 | * and process it. The handler should return 0 on success. | |
1347 | * | |
1348 | * This routine returns %NULL on a failure to register, and a pointer | |
1349 | * to the table header on success. | |
1350 | */ | |
6e9d5164 | 1351 | struct ctl_table_header *__register_sysctl_table( |
60a47a2e | 1352 | struct ctl_table_set *set, |
bff97cf1 | 1353 | const char *path, struct ctl_table *table, size_t table_size) |
1f87f0b5 | 1354 | { |
60a47a2e | 1355 | struct ctl_table_root *root = set->dir.header.root; |
1f87f0b5 | 1356 | struct ctl_table_header *header; |
7ec66d06 | 1357 | struct ctl_dir *dir; |
ac13ac6f | 1358 | struct ctl_node *node; |
1f87f0b5 | 1359 | |
ac13ac6f | 1360 | header = kzalloc(sizeof(struct ctl_table_header) + |
bff97cf1 | 1361 | sizeof(struct ctl_node)*table_size, GFP_KERNEL_ACCOUNT); |
1f87f0b5 EB |
1362 | if (!header) |
1363 | return NULL; | |
1364 | ||
ac13ac6f | 1365 | node = (struct ctl_node *)(header + 1); |
bff97cf1 | 1366 | init_header(header, root, set, node, table, table_size); |
cc9f7ee0 | 1367 | if (sysctl_check_table(path, header)) |
7ec66d06 EB |
1368 | goto fail; |
1369 | ||
1370 | spin_lock(&sysctl_lock); | |
0e47c99d | 1371 | dir = &set->dir; |
67ff3228 | 1372 | /* Reference moved down the directory tree get_subdir */ |
7ec66d06 EB |
1373 | dir->header.nreg++; |
1374 | spin_unlock(&sysctl_lock); | |
1f87f0b5 | 1375 | |
b2f56e55 LC |
1376 | dir = sysctl_mkdir_p(dir, path); |
1377 | if (IS_ERR(dir)) | |
1378 | goto fail; | |
1f87f0b5 | 1379 | spin_lock(&sysctl_lock); |
0e47c99d | 1380 | if (insert_header(dir, header)) |
7ec66d06 | 1381 | goto fail_put_dir_locked; |
0e47c99d | 1382 | |
7ec66d06 | 1383 | drop_sysctl_table(&dir->header); |
1f87f0b5 EB |
1384 | spin_unlock(&sysctl_lock); |
1385 | ||
1386 | return header; | |
0e47c99d | 1387 | |
7ec66d06 EB |
1388 | fail_put_dir_locked: |
1389 | drop_sysctl_table(&dir->header); | |
7c60c48f EB |
1390 | spin_unlock(&sysctl_lock); |
1391 | fail: | |
1392 | kfree(header); | |
7c60c48f | 1393 | return NULL; |
1f87f0b5 EB |
1394 | } |
1395 | ||
fea478d4 | 1396 | /** |
9edbfe92 | 1397 | * register_sysctl_sz - register a sysctl table |
1dc8689e LC |
1398 | * @path: The path to the directory the sysctl table is in. If the path |
1399 | * doesn't exist we will create it for you. | |
1400 | * @table: the table structure. The calller must ensure the life of the @table | |
1401 | * will be kept during the lifetime use of the syctl. It must not be freed | |
1402 | * until unregister_sysctl_table() is called with the given returned table | |
1403 | * with this registration. If your code is non modular then you don't need | |
1404 | * to call unregister_sysctl_table() and can instead use something like | |
1405 | * register_sysctl_init() which does not care for the result of the syctl | |
1406 | * registration. | |
9edbfe92 | 1407 | * @table_size: The number of elements in table. |
fea478d4 EB |
1408 | * |
1409 | * Register a sysctl table. @table should be a filled in ctl_table | |
1410 | * array. A completely 0 filled entry terminates the table. | |
1411 | * | |
1412 | * See __register_sysctl_table for more details. | |
1413 | */ | |
9edbfe92 JG |
1414 | struct ctl_table_header *register_sysctl_sz(const char *path, struct ctl_table *table, |
1415 | size_t table_size) | |
fea478d4 EB |
1416 | { |
1417 | return __register_sysctl_table(&sysctl_table_root.default_set, | |
9edbfe92 | 1418 | path, table, table_size); |
fea478d4 | 1419 | } |
9edbfe92 | 1420 | EXPORT_SYMBOL(register_sysctl_sz); |
fea478d4 | 1421 | |
3ddd9a80 XN |
1422 | /** |
1423 | * __register_sysctl_init() - register sysctl table to path | |
1dc8689e LC |
1424 | * @path: path name for sysctl base. If that path doesn't exist we will create |
1425 | * it for you. | |
1426 | * @table: This is the sysctl table that needs to be registered to the path. | |
1427 | * The caller must ensure the life of the @table will be kept during the | |
1428 | * lifetime use of the sysctl. | |
3ddd9a80 XN |
1429 | * @table_name: The name of sysctl table, only used for log printing when |
1430 | * registration fails | |
3bc269cf | 1431 | * @table_size: The number of elements in table |
3ddd9a80 XN |
1432 | * |
1433 | * The sysctl interface is used by userspace to query or modify at runtime | |
1434 | * a predefined value set on a variable. These variables however have default | |
1435 | * values pre-set. Code which depends on these variables will always work even | |
1436 | * if register_sysctl() fails. If register_sysctl() fails you'd just loose the | |
1437 | * ability to query or modify the sysctls dynamically at run time. Chances of | |
1438 | * register_sysctl() failing on init are extremely low, and so for both reasons | |
1439 | * this function does not return any error as it is used by initialization code. | |
1440 | * | |
228b09de | 1441 | * Context: if your base directory does not exist it will be created for you. |
3ddd9a80 XN |
1442 | */ |
1443 | void __init __register_sysctl_init(const char *path, struct ctl_table *table, | |
3bc269cf | 1444 | const char *table_name, size_t table_size) |
3ddd9a80 | 1445 | { |
3bc269cf | 1446 | struct ctl_table_header *hdr = register_sysctl_sz(path, table, table_size); |
3ddd9a80 XN |
1447 | |
1448 | if (unlikely(!hdr)) { | |
9edbfe92 | 1449 | pr_err("failed when register_sysctl_sz %s to %s\n", table_name, path); |
3ddd9a80 XN |
1450 | return; |
1451 | } | |
1452 | kmemleak_not_leak(hdr); | |
1453 | } | |
1454 | ||
0e47c99d EB |
1455 | static void put_links(struct ctl_table_header *header) |
1456 | { | |
1457 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
1458 | struct ctl_table_root *root = header->root; | |
1459 | struct ctl_dir *parent = header->parent; | |
1460 | struct ctl_dir *core_parent; | |
1461 | struct ctl_table *entry; | |
1462 | ||
1463 | if (header->set == root_set) | |
1464 | return; | |
1465 | ||
1466 | core_parent = xlate_dir(root_set, parent); | |
1467 | if (IS_ERR(core_parent)) | |
1468 | return; | |
1469 | ||
18d4b42e | 1470 | list_for_each_table_entry(entry, header) { |
0e47c99d EB |
1471 | struct ctl_table_header *link_head; |
1472 | struct ctl_table *link; | |
1473 | const char *name = entry->procname; | |
1474 | ||
1475 | link = find_entry(&link_head, core_parent, name, strlen(name)); | |
1476 | if (link && | |
1477 | ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) || | |
1478 | (S_ISLNK(link->mode) && (link->data == root)))) { | |
1479 | drop_sysctl_table(link_head); | |
1480 | } | |
1481 | else { | |
87ebdc00 | 1482 | pr_err("sysctl link missing during unregister: "); |
0e47c99d | 1483 | sysctl_print_dir(parent); |
153ee1c4 | 1484 | pr_cont("%s\n", name); |
0e47c99d EB |
1485 | } |
1486 | } | |
1487 | } | |
1488 | ||
938aaa4f EB |
1489 | static void drop_sysctl_table(struct ctl_table_header *header) |
1490 | { | |
7ec66d06 EB |
1491 | struct ctl_dir *parent = header->parent; |
1492 | ||
938aaa4f EB |
1493 | if (--header->nreg) |
1494 | return; | |
1495 | ||
89189557 | 1496 | if (parent) { |
23da9588 | 1497 | put_links(header); |
89189557 Y |
1498 | start_unregistering(header); |
1499 | } | |
1500 | ||
938aaa4f EB |
1501 | if (!--header->count) |
1502 | kfree_rcu(header, rcu); | |
7ec66d06 EB |
1503 | |
1504 | if (parent) | |
1505 | drop_sysctl_table(&parent->header); | |
938aaa4f EB |
1506 | } |
1507 | ||
1f87f0b5 EB |
1508 | /** |
1509 | * unregister_sysctl_table - unregister a sysctl table hierarchy | |
19c4e618 | 1510 | * @header: the header returned from register_sysctl or __register_sysctl_table |
1f87f0b5 EB |
1511 | * |
1512 | * Unregisters the sysctl table and all children. proc entries may not | |
1513 | * actually be removed until they are no longer used by anyone. | |
1514 | */ | |
1515 | void unregister_sysctl_table(struct ctl_table_header * header) | |
1516 | { | |
1517 | might_sleep(); | |
1518 | ||
1519 | if (header == NULL) | |
1520 | return; | |
1521 | ||
1522 | spin_lock(&sysctl_lock); | |
938aaa4f | 1523 | drop_sysctl_table(header); |
1f87f0b5 EB |
1524 | spin_unlock(&sysctl_lock); |
1525 | } | |
1526 | EXPORT_SYMBOL(unregister_sysctl_table); | |
1527 | ||
0e47c99d | 1528 | void setup_sysctl_set(struct ctl_table_set *set, |
9eb47c26 | 1529 | struct ctl_table_root *root, |
1f87f0b5 EB |
1530 | int (*is_seen)(struct ctl_table_set *)) |
1531 | { | |
1347440d | 1532 | memset(set, 0, sizeof(*set)); |
0e47c99d | 1533 | set->is_seen = is_seen; |
b1f01e2b | 1534 | init_header(&set->dir.header, root, set, NULL, root_table, 1); |
1f87f0b5 EB |
1535 | } |
1536 | ||
97324cd8 EB |
1537 | void retire_sysctl_set(struct ctl_table_set *set) |
1538 | { | |
ac13ac6f | 1539 | WARN_ON(!RB_EMPTY_ROOT(&set->dir.root)); |
97324cd8 | 1540 | } |
1f87f0b5 | 1541 | |
1e0edd3f | 1542 | int __init proc_sys_init(void) |
77b14db5 | 1543 | { |
e1675231 AD |
1544 | struct proc_dir_entry *proc_sys_root; |
1545 | ||
77b14db5 | 1546 | proc_sys_root = proc_mkdir("sys", NULL); |
9043476f | 1547 | proc_sys_root->proc_iops = &proc_sys_dir_operations; |
d56c0d45 | 1548 | proc_sys_root->proc_dir_ops = &proc_sys_dir_file_operations; |
77b14db5 | 1549 | proc_sys_root->nlink = 0; |
de4e83bd | 1550 | |
d8c0418a | 1551 | return sysctl_init_bases(); |
77b14db5 | 1552 | } |
3db978d4 | 1553 | |
0a477e1a VB |
1554 | struct sysctl_alias { |
1555 | const char *kernel_param; | |
1556 | const char *sysctl_param; | |
1557 | }; | |
1558 | ||
1559 | /* | |
1560 | * Historically some settings had both sysctl and a command line parameter. | |
1561 | * With the generic sysctl. parameter support, we can handle them at a single | |
1562 | * place and only keep the historical name for compatibility. This is not meant | |
1563 | * to add brand new aliases. When adding existing aliases, consider whether | |
1564 | * the possibly different moment of changing the value (e.g. from early_param | |
1565 | * to the moment do_sysctl_args() is called) is an issue for the specific | |
1566 | * parameter. | |
1567 | */ | |
1568 | static const struct sysctl_alias sysctl_aliases[] = { | |
f117955a GP |
1569 | {"hardlockup_all_cpu_backtrace", "kernel.hardlockup_all_cpu_backtrace" }, |
1570 | {"hung_task_panic", "kernel.hung_task_panic" }, | |
1571 | {"numa_zonelist_order", "vm.numa_zonelist_order" }, | |
1572 | {"softlockup_all_cpu_backtrace", "kernel.softlockup_all_cpu_backtrace" }, | |
0a477e1a VB |
1573 | { } |
1574 | }; | |
1575 | ||
1576 | static const char *sysctl_find_alias(char *param) | |
1577 | { | |
1578 | const struct sysctl_alias *alias; | |
1579 | ||
1580 | for (alias = &sysctl_aliases[0]; alias->kernel_param != NULL; alias++) { | |
1581 | if (strcmp(alias->kernel_param, param) == 0) | |
1582 | return alias->sysctl_param; | |
1583 | } | |
1584 | ||
1585 | return NULL; | |
1586 | } | |
1587 | ||
8001f493 KJ |
1588 | bool sysctl_is_alias(char *param) |
1589 | { | |
1590 | const char *alias = sysctl_find_alias(param); | |
1591 | ||
1592 | return alias != NULL; | |
1593 | } | |
1594 | ||
3db978d4 VB |
1595 | /* Set sysctl value passed on kernel command line. */ |
1596 | static int process_sysctl_arg(char *param, char *val, | |
1597 | const char *unused, void *arg) | |
1598 | { | |
1599 | char *path; | |
1600 | struct vfsmount **proc_mnt = arg; | |
1601 | struct file_system_type *proc_fs_type; | |
1602 | struct file *file; | |
1603 | int len; | |
1604 | int err; | |
1605 | loff_t pos = 0; | |
1606 | ssize_t wret; | |
1607 | ||
0a477e1a VB |
1608 | if (strncmp(param, "sysctl", sizeof("sysctl") - 1) == 0) { |
1609 | param += sizeof("sysctl") - 1; | |
3db978d4 | 1610 | |
0a477e1a VB |
1611 | if (param[0] != '/' && param[0] != '.') |
1612 | return 0; | |
3db978d4 | 1613 | |
0a477e1a VB |
1614 | param++; |
1615 | } else { | |
1616 | param = (char *) sysctl_find_alias(param); | |
1617 | if (!param) | |
1618 | return 0; | |
1619 | } | |
3db978d4 | 1620 | |
697edcb0 XN |
1621 | if (!val) |
1622 | return -EINVAL; | |
1623 | len = strlen(val); | |
1624 | if (len == 0) | |
1625 | return -EINVAL; | |
1626 | ||
3db978d4 VB |
1627 | /* |
1628 | * To set sysctl options, we use a temporary mount of proc, look up the | |
1629 | * respective sys/ file and write to it. To avoid mounting it when no | |
1630 | * options were given, we mount it only when the first sysctl option is | |
1631 | * found. Why not a persistent mount? There are problems with a | |
1632 | * persistent mount of proc in that it forces userspace not to use any | |
1633 | * proc mount options. | |
1634 | */ | |
1635 | if (!*proc_mnt) { | |
1636 | proc_fs_type = get_fs_type("proc"); | |
1637 | if (!proc_fs_type) { | |
1638 | pr_err("Failed to find procfs to set sysctl from command line\n"); | |
1639 | return 0; | |
1640 | } | |
1641 | *proc_mnt = kern_mount(proc_fs_type); | |
1642 | put_filesystem(proc_fs_type); | |
1643 | if (IS_ERR(*proc_mnt)) { | |
1644 | pr_err("Failed to mount procfs to set sysctl from command line\n"); | |
1645 | return 0; | |
1646 | } | |
1647 | } | |
1648 | ||
1649 | path = kasprintf(GFP_KERNEL, "sys/%s", param); | |
1650 | if (!path) | |
1651 | panic("%s: Failed to allocate path for %s\n", __func__, param); | |
1652 | strreplace(path, '.', '/'); | |
1653 | ||
ffb37ca3 | 1654 | file = file_open_root_mnt(*proc_mnt, path, O_WRONLY, 0); |
3db978d4 VB |
1655 | if (IS_ERR(file)) { |
1656 | err = PTR_ERR(file); | |
1657 | if (err == -ENOENT) | |
1658 | pr_err("Failed to set sysctl parameter '%s=%s': parameter not found\n", | |
1659 | param, val); | |
1660 | else if (err == -EACCES) | |
1661 | pr_err("Failed to set sysctl parameter '%s=%s': permission denied (read-only?)\n", | |
1662 | param, val); | |
1663 | else | |
1664 | pr_err("Error %pe opening proc file to set sysctl parameter '%s=%s'\n", | |
1665 | file, param, val); | |
1666 | goto out; | |
1667 | } | |
3db978d4 VB |
1668 | wret = kernel_write(file, val, len, &pos); |
1669 | if (wret < 0) { | |
1670 | err = wret; | |
1671 | if (err == -EINVAL) | |
1672 | pr_err("Failed to set sysctl parameter '%s=%s': invalid value\n", | |
1673 | param, val); | |
1674 | else | |
1675 | pr_err("Error %pe writing to proc file to set sysctl parameter '%s=%s'\n", | |
1676 | ERR_PTR(err), param, val); | |
1677 | } else if (wret != len) { | |
1678 | pr_err("Wrote only %zd bytes of %d writing to proc file %s to set sysctl parameter '%s=%s\n", | |
1679 | wret, len, path, param, val); | |
1680 | } | |
1681 | ||
1682 | err = filp_close(file, NULL); | |
1683 | if (err) | |
1684 | pr_err("Error %pe closing proc file to set sysctl parameter '%s=%s\n", | |
1685 | ERR_PTR(err), param, val); | |
1686 | out: | |
1687 | kfree(path); | |
1688 | return 0; | |
1689 | } | |
1690 | ||
1691 | void do_sysctl_args(void) | |
1692 | { | |
1693 | char *command_line; | |
1694 | struct vfsmount *proc_mnt = NULL; | |
1695 | ||
1696 | command_line = kstrdup(saved_command_line, GFP_KERNEL); | |
1697 | if (!command_line) | |
1698 | panic("%s: Failed to allocate copy of command line\n", __func__); | |
1699 | ||
1700 | parse_args("Setting sysctl args", command_line, | |
1701 | NULL, 0, -1, -1, &proc_mnt, process_sysctl_arg); | |
1702 | ||
1703 | if (proc_mnt) | |
1704 | kern_unmount(proc_mnt); | |
1705 | ||
1706 | kfree(command_line); | |
1707 | } |