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 */ |
a35dd3a7 | 33 | static struct ctl_table sysctl_mount_point[] = { }; |
f9bd6733 | 34 | |
ee9efac4 LC |
35 | /** |
36 | * register_sysctl_mount_point() - registers a sysctl mount point | |
37 | * @path: path for the mount point | |
38 | * | |
39 | * Used to create a permanently empty directory to serve as mount point. | |
40 | * There are some subtle but important permission checks this allows in the | |
41 | * case of unprivileged mounts. | |
42 | */ | |
43 | struct ctl_table_header *register_sysctl_mount_point(const char *path) | |
44 | { | |
31555231 | 45 | return register_sysctl(path, sysctl_mount_point); |
ee9efac4 LC |
46 | } |
47 | EXPORT_SYMBOL(register_sysctl_mount_point); | |
48 | ||
2f2665c1 | 49 | #define sysctl_is_perm_empty_ctl_header(hptr) \ |
4a7b29f6 | 50 | (hptr->type == SYSCTL_TABLE_TYPE_PERMANENTLY_EMPTY) |
2f2665c1 | 51 | #define sysctl_set_perm_empty_ctl_header(hptr) \ |
4a7b29f6 | 52 | (hptr->type = SYSCTL_TABLE_TYPE_PERMANENTLY_EMPTY) |
2f2665c1 | 53 | #define sysctl_clear_perm_empty_ctl_header(hptr) \ |
4a7b29f6 | 54 | (hptr->type = SYSCTL_TABLE_TYPE_DEFAULT) |
f9bd6733 | 55 | |
f1ecf068 LDM |
56 | void proc_sys_poll_notify(struct ctl_table_poll *poll) |
57 | { | |
58 | if (!poll) | |
59 | return; | |
60 | ||
61 | atomic_inc(&poll->event); | |
62 | wake_up_interruptible(&poll->wait); | |
63 | } | |
64 | ||
a194558e EB |
65 | static struct ctl_table root_table[] = { |
66 | { | |
67 | .procname = "", | |
7ec66d06 | 68 | .mode = S_IFDIR|S_IRUGO|S_IXUGO, |
a194558e | 69 | }, |
a194558e | 70 | }; |
0e47c99d | 71 | static struct ctl_table_root sysctl_table_root = { |
0e47c99d | 72 | .default_set.dir.header = { |
7ec66d06 EB |
73 | {{.count = 1, |
74 | .nreg = 1, | |
ac13ac6f | 75 | .ctl_table = root_table }}, |
0e47c99d | 76 | .ctl_table_arg = root_table, |
7ec66d06 EB |
77 | .root = &sysctl_table_root, |
78 | .set = &sysctl_table_root.default_set, | |
79 | }, | |
1f87f0b5 | 80 | }; |
1f87f0b5 EB |
81 | |
82 | static DEFINE_SPINLOCK(sysctl_lock); | |
83 | ||
7ec66d06 | 84 | static void drop_sysctl_table(struct ctl_table_header *header); |
0e47c99d | 85 | static int sysctl_follow_link(struct ctl_table_header **phead, |
13bcc6a2 | 86 | struct ctl_table **pentry); |
0e47c99d EB |
87 | static int insert_links(struct ctl_table_header *head); |
88 | static void put_links(struct ctl_table_header *header); | |
7ec66d06 | 89 | |
6980128f EB |
90 | static void sysctl_print_dir(struct ctl_dir *dir) |
91 | { | |
92 | if (dir->header.parent) | |
93 | sysctl_print_dir(dir->header.parent); | |
87ebdc00 | 94 | pr_cont("%s/", dir->header.ctl_table[0].procname); |
6980128f EB |
95 | } |
96 | ||
076c3eed EB |
97 | static int namecmp(const char *name1, int len1, const char *name2, int len2) |
98 | { | |
076c3eed EB |
99 | int cmp; |
100 | ||
d8fc9b66 | 101 | cmp = memcmp(name1, name2, min(len1, len2)); |
076c3eed EB |
102 | if (cmp == 0) |
103 | cmp = len1 - len2; | |
104 | return cmp; | |
105 | } | |
106 | ||
60f126d9 | 107 | /* Called under sysctl_lock */ |
076c3eed | 108 | static struct ctl_table *find_entry(struct ctl_table_header **phead, |
0e47c99d | 109 | struct ctl_dir *dir, const char *name, int namelen) |
076c3eed EB |
110 | { |
111 | struct ctl_table_header *head; | |
112 | struct ctl_table *entry; | |
ac13ac6f | 113 | struct rb_node *node = dir->root.rb_node; |
076c3eed | 114 | |
ac13ac6f EB |
115 | while (node) |
116 | { | |
117 | struct ctl_node *ctl_node; | |
118 | const char *procname; | |
119 | int cmp; | |
120 | ||
121 | ctl_node = rb_entry(node, struct ctl_node, node); | |
122 | head = ctl_node->header; | |
123 | entry = &head->ctl_table[ctl_node - head->node]; | |
124 | procname = entry->procname; | |
125 | ||
126 | cmp = namecmp(name, namelen, procname, strlen(procname)); | |
127 | if (cmp < 0) | |
128 | node = node->rb_left; | |
129 | else if (cmp > 0) | |
130 | node = node->rb_right; | |
131 | else { | |
132 | *phead = head; | |
133 | return entry; | |
076c3eed EB |
134 | } |
135 | } | |
136 | return NULL; | |
137 | } | |
138 | ||
ac13ac6f EB |
139 | static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry) |
140 | { | |
141 | struct rb_node *node = &head->node[entry - head->ctl_table].node; | |
142 | struct rb_node **p = &head->parent->root.rb_node; | |
143 | struct rb_node *parent = NULL; | |
144 | const char *name = entry->procname; | |
145 | int namelen = strlen(name); | |
146 | ||
147 | while (*p) { | |
148 | struct ctl_table_header *parent_head; | |
149 | struct ctl_table *parent_entry; | |
150 | struct ctl_node *parent_node; | |
151 | const char *parent_name; | |
152 | int cmp; | |
153 | ||
154 | parent = *p; | |
155 | parent_node = rb_entry(parent, struct ctl_node, node); | |
156 | parent_head = parent_node->header; | |
157 | parent_entry = &parent_head->ctl_table[parent_node - parent_head->node]; | |
158 | parent_name = parent_entry->procname; | |
159 | ||
160 | cmp = namecmp(name, namelen, parent_name, strlen(parent_name)); | |
161 | if (cmp < 0) | |
162 | p = &(*p)->rb_left; | |
163 | else if (cmp > 0) | |
164 | p = &(*p)->rb_right; | |
165 | else { | |
87ebdc00 | 166 | pr_err("sysctl duplicate entry: "); |
ac13ac6f | 167 | sysctl_print_dir(head->parent); |
153ee1c4 | 168 | pr_cont("%s\n", entry->procname); |
ac13ac6f EB |
169 | return -EEXIST; |
170 | } | |
171 | } | |
172 | ||
173 | rb_link_node(node, parent, p); | |
ea5272f5 | 174 | rb_insert_color(node, &head->parent->root); |
ac13ac6f EB |
175 | return 0; |
176 | } | |
177 | ||
178 | static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry) | |
179 | { | |
180 | struct rb_node *node = &head->node[entry - head->ctl_table].node; | |
181 | ||
182 | rb_erase(node, &head->parent->root); | |
183 | } | |
184 | ||
e0d04529 EB |
185 | static void init_header(struct ctl_table_header *head, |
186 | struct ctl_table_root *root, struct ctl_table_set *set, | |
b1f01e2b | 187 | struct ctl_node *node, struct ctl_table *table, size_t table_size) |
e0d04529 | 188 | { |
7ec66d06 | 189 | head->ctl_table = table; |
b1f01e2b | 190 | head->ctl_table_size = table_size; |
e0d04529 | 191 | head->ctl_table_arg = table; |
e0d04529 EB |
192 | head->used = 0; |
193 | head->count = 1; | |
194 | head->nreg = 1; | |
195 | head->unregistering = NULL; | |
196 | head->root = root; | |
197 | head->set = set; | |
198 | head->parent = NULL; | |
ac13ac6f | 199 | head->node = node; |
2fd1d2c4 | 200 | INIT_HLIST_HEAD(&head->inodes); |
ac13ac6f EB |
201 | if (node) { |
202 | struct ctl_table *entry; | |
cb55f27a | 203 | |
18d4b42e | 204 | list_for_each_table_entry(entry, head) { |
ac13ac6f | 205 | node->header = head; |
cb55f27a MT |
206 | node++; |
207 | } | |
ac13ac6f | 208 | } |
4a7b29f6 TW |
209 | if (table == sysctl_mount_point) |
210 | sysctl_set_perm_empty_ctl_header(head); | |
e0d04529 EB |
211 | } |
212 | ||
8425d6aa EB |
213 | static void erase_header(struct ctl_table_header *head) |
214 | { | |
ac13ac6f | 215 | struct ctl_table *entry; |
cb55f27a | 216 | |
18d4b42e | 217 | list_for_each_table_entry(entry, head) |
ac13ac6f | 218 | erase_entry(head, entry); |
8425d6aa EB |
219 | } |
220 | ||
0e47c99d | 221 | static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header) |
8425d6aa | 222 | { |
ac13ac6f | 223 | struct ctl_table *entry; |
2f2665c1 | 224 | struct ctl_table_header *dir_h = &dir->header; |
0e47c99d EB |
225 | int err; |
226 | ||
2f2665c1 | 227 | |
f9bd6733 | 228 | /* Is this a permanently empty directory? */ |
2f2665c1 | 229 | if (sysctl_is_perm_empty_ctl_header(dir_h)) |
f9bd6733 EB |
230 | return -EROFS; |
231 | ||
232 | /* Am I creating a permanently empty directory? */ | |
a35dd3a7 | 233 | if (sysctl_is_perm_empty_ctl_header(header)) { |
f9bd6733 EB |
234 | if (!RB_EMPTY_ROOT(&dir->root)) |
235 | return -EINVAL; | |
2f2665c1 | 236 | sysctl_set_perm_empty_ctl_header(dir_h); |
f9bd6733 EB |
237 | } |
238 | ||
2f2665c1 | 239 | dir_h->nreg++; |
7ec66d06 | 240 | header->parent = dir; |
0e47c99d EB |
241 | err = insert_links(header); |
242 | if (err) | |
243 | goto fail_links; | |
18d4b42e | 244 | list_for_each_table_entry(entry, header) { |
ac13ac6f EB |
245 | err = insert_entry(header, entry); |
246 | if (err) | |
247 | goto fail; | |
248 | } | |
0e47c99d | 249 | return 0; |
ac13ac6f EB |
250 | fail: |
251 | erase_header(header); | |
252 | put_links(header); | |
0e47c99d | 253 | fail_links: |
f9bd6733 | 254 | if (header->ctl_table == sysctl_mount_point) |
2f2665c1 | 255 | sysctl_clear_perm_empty_ctl_header(dir_h); |
0e47c99d | 256 | header->parent = NULL; |
2f2665c1 | 257 | drop_sysctl_table(dir_h); |
0e47c99d | 258 | return err; |
8425d6aa EB |
259 | } |
260 | ||
1f87f0b5 EB |
261 | /* called under sysctl_lock */ |
262 | static int use_table(struct ctl_table_header *p) | |
263 | { | |
264 | if (unlikely(p->unregistering)) | |
265 | return 0; | |
266 | p->used++; | |
267 | return 1; | |
268 | } | |
269 | ||
270 | /* called under sysctl_lock */ | |
271 | static void unuse_table(struct ctl_table_header *p) | |
272 | { | |
273 | if (!--p->used) | |
274 | if (unlikely(p->unregistering)) | |
275 | complete(p->unregistering); | |
276 | } | |
277 | ||
f90f3caf | 278 | static void proc_sys_invalidate_dcache(struct ctl_table_header *head) |
d6cffbbe | 279 | { |
f90f3caf | 280 | proc_invalidate_siblings_dcache(&head->inodes, &sysctl_lock); |
d6cffbbe KK |
281 | } |
282 | ||
1f87f0b5 EB |
283 | /* called under sysctl_lock, will reacquire if has to wait */ |
284 | static void start_unregistering(struct ctl_table_header *p) | |
285 | { | |
286 | /* | |
287 | * if p->used is 0, nobody will ever touch that entry again; | |
288 | * we'll eliminate all paths to it before dropping sysctl_lock | |
289 | */ | |
290 | if (unlikely(p->used)) { | |
291 | struct completion wait; | |
292 | init_completion(&wait); | |
293 | p->unregistering = &wait; | |
294 | spin_unlock(&sysctl_lock); | |
295 | wait_for_completion(&wait); | |
1f87f0b5 EB |
296 | } else { |
297 | /* anything non-NULL; we'll never dereference it */ | |
298 | p->unregistering = ERR_PTR(-EINVAL); | |
ace0c791 | 299 | spin_unlock(&sysctl_lock); |
1f87f0b5 | 300 | } |
d6cffbbe | 301 | /* |
f90f3caf | 302 | * Invalidate dentries for unregistered sysctls: namespaced sysctls |
d6cffbbe KK |
303 | * can have duplicate names and contaminate dcache very badly. |
304 | */ | |
f90f3caf | 305 | proc_sys_invalidate_dcache(p); |
1f87f0b5 EB |
306 | /* |
307 | * do not remove from the list until nobody holds it; walking the | |
308 | * list in do_sysctl() relies on that. | |
309 | */ | |
ace0c791 | 310 | spin_lock(&sysctl_lock); |
8425d6aa | 311 | erase_header(p); |
1f87f0b5 EB |
312 | } |
313 | ||
1f87f0b5 EB |
314 | static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head) |
315 | { | |
ab4a1f24 | 316 | BUG_ON(!head); |
1f87f0b5 EB |
317 | spin_lock(&sysctl_lock); |
318 | if (!use_table(head)) | |
319 | head = ERR_PTR(-ENOENT); | |
320 | spin_unlock(&sysctl_lock); | |
321 | return head; | |
322 | } | |
323 | ||
324 | static void sysctl_head_finish(struct ctl_table_header *head) | |
325 | { | |
326 | if (!head) | |
327 | return; | |
328 | spin_lock(&sysctl_lock); | |
329 | unuse_table(head); | |
330 | spin_unlock(&sysctl_lock); | |
331 | } | |
332 | ||
333 | static struct ctl_table_set * | |
13bcc6a2 | 334 | lookup_header_set(struct ctl_table_root *root) |
1f87f0b5 EB |
335 | { |
336 | struct ctl_table_set *set = &root->default_set; | |
337 | if (root->lookup) | |
13bcc6a2 | 338 | set = root->lookup(root); |
1f87f0b5 EB |
339 | return set; |
340 | } | |
341 | ||
076c3eed | 342 | static struct ctl_table *lookup_entry(struct ctl_table_header **phead, |
7ec66d06 | 343 | struct ctl_dir *dir, |
076c3eed EB |
344 | const char *name, int namelen) |
345 | { | |
346 | struct ctl_table_header *head; | |
347 | struct ctl_table *entry; | |
076c3eed EB |
348 | |
349 | spin_lock(&sysctl_lock); | |
0e47c99d EB |
350 | entry = find_entry(&head, dir, name, namelen); |
351 | if (entry && use_table(head)) | |
352 | *phead = head; | |
353 | else | |
354 | entry = NULL; | |
076c3eed EB |
355 | spin_unlock(&sysctl_lock); |
356 | return entry; | |
357 | } | |
358 | ||
ac13ac6f | 359 | static struct ctl_node *first_usable_entry(struct rb_node *node) |
1f87f0b5 | 360 | { |
ac13ac6f | 361 | struct ctl_node *ctl_node; |
1f87f0b5 | 362 | |
ac13ac6f EB |
363 | for (;node; node = rb_next(node)) { |
364 | ctl_node = rb_entry(node, struct ctl_node, node); | |
365 | if (use_table(ctl_node->header)) | |
366 | return ctl_node; | |
1f87f0b5 | 367 | } |
1f87f0b5 EB |
368 | return NULL; |
369 | } | |
370 | ||
7ec66d06 | 371 | static void first_entry(struct ctl_dir *dir, |
6a75ce16 EB |
372 | struct ctl_table_header **phead, struct ctl_table **pentry) |
373 | { | |
ac13ac6f | 374 | struct ctl_table_header *head = NULL; |
7ec66d06 | 375 | struct ctl_table *entry = NULL; |
ac13ac6f | 376 | struct ctl_node *ctl_node; |
6a75ce16 EB |
377 | |
378 | spin_lock(&sysctl_lock); | |
ac13ac6f | 379 | ctl_node = first_usable_entry(rb_first(&dir->root)); |
6a75ce16 | 380 | spin_unlock(&sysctl_lock); |
ac13ac6f EB |
381 | if (ctl_node) { |
382 | head = ctl_node->header; | |
383 | entry = &head->ctl_table[ctl_node - head->node]; | |
384 | } | |
6a75ce16 EB |
385 | *phead = head; |
386 | *pentry = entry; | |
387 | } | |
388 | ||
7ec66d06 | 389 | static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry) |
1f87f0b5 | 390 | { |
6a75ce16 EB |
391 | struct ctl_table_header *head = *phead; |
392 | struct ctl_table *entry = *pentry; | |
ac13ac6f | 393 | struct ctl_node *ctl_node = &head->node[entry - head->ctl_table]; |
6a75ce16 | 394 | |
ac13ac6f EB |
395 | spin_lock(&sysctl_lock); |
396 | unuse_table(head); | |
397 | ||
398 | ctl_node = first_usable_entry(rb_next(&ctl_node->node)); | |
399 | spin_unlock(&sysctl_lock); | |
400 | head = NULL; | |
401 | if (ctl_node) { | |
402 | head = ctl_node->header; | |
403 | entry = &head->ctl_table[ctl_node - head->node]; | |
6a75ce16 EB |
404 | } |
405 | *phead = head; | |
406 | *pentry = entry; | |
1f87f0b5 EB |
407 | } |
408 | ||
1f87f0b5 EB |
409 | /* |
410 | * sysctl_perm does NOT grant the superuser all rights automatically, because | |
411 | * some sysctl variables are readonly even to root. | |
412 | */ | |
413 | ||
414 | static int test_perm(int mode, int op) | |
415 | { | |
091bd3ea | 416 | if (uid_eq(current_euid(), GLOBAL_ROOT_UID)) |
1f87f0b5 | 417 | mode >>= 6; |
091bd3ea | 418 | else if (in_egroup_p(GLOBAL_ROOT_GID)) |
1f87f0b5 EB |
419 | mode >>= 3; |
420 | if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0) | |
421 | return 0; | |
422 | return -EACCES; | |
423 | } | |
424 | ||
73f7ef43 | 425 | static int sysctl_perm(struct ctl_table_header *head, struct ctl_table *table, int op) |
1f87f0b5 | 426 | { |
73f7ef43 | 427 | struct ctl_table_root *root = head->root; |
1f87f0b5 EB |
428 | int mode; |
429 | ||
430 | if (root->permissions) | |
73f7ef43 | 431 | mode = root->permissions(head, table); |
1f87f0b5 EB |
432 | else |
433 | mode = table->mode; | |
434 | ||
435 | return test_perm(mode, op); | |
436 | } | |
437 | ||
9043476f AV |
438 | static struct inode *proc_sys_make_inode(struct super_block *sb, |
439 | struct ctl_table_header *head, struct ctl_table *table) | |
77b14db5 | 440 | { |
e79c6a4f | 441 | struct ctl_table_root *root = head->root; |
77b14db5 | 442 | struct inode *inode; |
9043476f | 443 | struct proc_inode *ei; |
77b14db5 | 444 | |
9043476f | 445 | inode = new_inode(sb); |
77b14db5 | 446 | if (!inode) |
ea5751cc | 447 | return ERR_PTR(-ENOMEM); |
77b14db5 | 448 | |
85fe4025 CH |
449 | inode->i_ino = get_next_ino(); |
450 | ||
77b14db5 | 451 | ei = PROC_I(inode); |
9043476f | 452 | |
d6cffbbe | 453 | spin_lock(&sysctl_lock); |
ace0c791 EB |
454 | if (unlikely(head->unregistering)) { |
455 | spin_unlock(&sysctl_lock); | |
456 | iput(inode); | |
ea5751cc | 457 | return ERR_PTR(-ENOENT); |
ace0c791 EB |
458 | } |
459 | ei->sysctl = head; | |
460 | ei->sysctl_entry = table; | |
0afa5ca8 | 461 | hlist_add_head_rcu(&ei->sibling_inodes, &head->inodes); |
d6cffbbe KK |
462 | head->count++; |
463 | spin_unlock(&sysctl_lock); | |
464 | ||
200d9421 | 465 | simple_inode_init_ts(inode); |
9043476f | 466 | inode->i_mode = table->mode; |
7ec66d06 | 467 | if (!S_ISDIR(table->mode)) { |
9043476f AV |
468 | inode->i_mode |= S_IFREG; |
469 | inode->i_op = &proc_sys_inode_operations; | |
470 | inode->i_fop = &proc_sys_file_operations; | |
471 | } else { | |
472 | inode->i_mode |= S_IFDIR; | |
9043476f AV |
473 | inode->i_op = &proc_sys_dir_operations; |
474 | inode->i_fop = &proc_sys_dir_file_operations; | |
2f2665c1 | 475 | if (sysctl_is_perm_empty_ctl_header(head)) |
f9bd6733 | 476 | make_empty_dir_inode(inode); |
9043476f | 477 | } |
e79c6a4f DT |
478 | |
479 | if (root->set_ownership) | |
520713a9 | 480 | root->set_ownership(head, &inode->i_uid, &inode->i_gid); |
5ec27ec7 RB |
481 | else { |
482 | inode->i_uid = GLOBAL_ROOT_UID; | |
483 | inode->i_gid = GLOBAL_ROOT_GID; | |
484 | } | |
e79c6a4f | 485 | |
77b14db5 EB |
486 | return inode; |
487 | } | |
488 | ||
d6cffbbe KK |
489 | void proc_sys_evict_inode(struct inode *inode, struct ctl_table_header *head) |
490 | { | |
491 | spin_lock(&sysctl_lock); | |
0afa5ca8 | 492 | hlist_del_init_rcu(&PROC_I(inode)->sibling_inodes); |
d6cffbbe KK |
493 | if (!--head->count) |
494 | kfree_rcu(head, rcu); | |
495 | spin_unlock(&sysctl_lock); | |
496 | } | |
497 | ||
81324364 | 498 | static struct ctl_table_header *grab_header(struct inode *inode) |
77b14db5 | 499 | { |
3cc3e046 EB |
500 | struct ctl_table_header *head = PROC_I(inode)->sysctl; |
501 | if (!head) | |
0e47c99d | 502 | head = &sysctl_table_root.default_set.dir.header; |
3cc3e046 | 503 | return sysctl_head_grab(head); |
9043476f | 504 | } |
77b14db5 | 505 | |
9043476f | 506 | static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry, |
00cd8dd3 | 507 | unsigned int flags) |
9043476f AV |
508 | { |
509 | struct ctl_table_header *head = grab_header(dir); | |
9043476f | 510 | struct ctl_table_header *h = NULL; |
dc12e909 | 511 | const struct qstr *name = &dentry->d_name; |
9043476f AV |
512 | struct ctl_table *p; |
513 | struct inode *inode; | |
514 | struct dentry *err = ERR_PTR(-ENOENT); | |
7ec66d06 | 515 | struct ctl_dir *ctl_dir; |
0e47c99d | 516 | int ret; |
77b14db5 | 517 | |
9043476f AV |
518 | if (IS_ERR(head)) |
519 | return ERR_CAST(head); | |
77b14db5 | 520 | |
7ec66d06 | 521 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 | 522 | |
7ec66d06 | 523 | p = lookup_entry(&h, ctl_dir, name->name, name->len); |
9043476f | 524 | if (!p) |
77b14db5 EB |
525 | goto out; |
526 | ||
4e757320 | 527 | if (S_ISLNK(p->mode)) { |
13bcc6a2 | 528 | ret = sysctl_follow_link(&h, &p); |
4e757320 EB |
529 | err = ERR_PTR(ret); |
530 | if (ret) | |
531 | goto out; | |
532 | } | |
0e47c99d | 533 | |
fb045adb | 534 | d_set_d_op(dentry, &proc_sys_dentry_operations); |
1eae9a47 | 535 | inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p); |
888e2b03 | 536 | err = d_splice_alias(inode, dentry); |
77b14db5 EB |
537 | |
538 | out: | |
6bf61045 FR |
539 | if (h) |
540 | sysctl_head_finish(h); | |
77b14db5 EB |
541 | sysctl_head_finish(head); |
542 | return err; | |
543 | } | |
544 | ||
4bd6a735 MWO |
545 | static ssize_t proc_sys_call_handler(struct kiocb *iocb, struct iov_iter *iter, |
546 | int write) | |
77b14db5 | 547 | { |
4bd6a735 | 548 | struct inode *inode = file_inode(iocb->ki_filp); |
9043476f AV |
549 | struct ctl_table_header *head = grab_header(inode); |
550 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
4bd6a735 MWO |
551 | size_t count = iov_iter_count(iter); |
552 | char *kbuf; | |
2a2da53b | 553 | ssize_t error; |
77b14db5 | 554 | |
9043476f AV |
555 | if (IS_ERR(head)) |
556 | return PTR_ERR(head); | |
77b14db5 EB |
557 | |
558 | /* | |
559 | * At this point we know that the sysctl was not unregistered | |
560 | * and won't be until we finish. | |
561 | */ | |
562 | error = -EPERM; | |
73f7ef43 | 563 | if (sysctl_perm(head, table, write ? MAY_WRITE : MAY_READ)) |
77b14db5 EB |
564 | goto out; |
565 | ||
9043476f AV |
566 | /* if that can happen at all, it should be -EINVAL, not -EISDIR */ |
567 | error = -EINVAL; | |
568 | if (!table->proc_handler) | |
569 | goto out; | |
570 | ||
ef9d965b | 571 | /* don't even try if the size is too large */ |
d4d80e69 MWO |
572 | error = -ENOMEM; |
573 | if (count >= KMALLOC_MAX_SIZE) | |
574 | goto out; | |
45089437 | 575 | kbuf = kvzalloc(count + 1, GFP_KERNEL); |
4bd6a735 MWO |
576 | if (!kbuf) |
577 | goto out; | |
ef9d965b | 578 | |
32927393 | 579 | if (write) { |
4bd6a735 MWO |
580 | error = -EFAULT; |
581 | if (!copy_from_iter_full(kbuf, count, iter)) | |
582 | goto out_free_buf; | |
583 | kbuf[count] = '\0'; | |
32927393 CH |
584 | } |
585 | ||
586 | error = BPF_CGROUP_RUN_PROG_SYSCTL(head, table, write, &kbuf, &count, | |
4bd6a735 | 587 | &iocb->ki_pos); |
7b146ceb | 588 | if (error) |
32927393 | 589 | goto out_free_buf; |
7b146ceb | 590 | |
77b14db5 | 591 | /* careful: calling conventions are nasty here */ |
4bd6a735 | 592 | error = table->proc_handler(table, write, kbuf, &count, &iocb->ki_pos); |
32927393 CH |
593 | if (error) |
594 | goto out_free_buf; | |
595 | ||
596 | if (!write) { | |
597 | error = -EFAULT; | |
4bd6a735 | 598 | if (copy_to_iter(kbuf, count, iter) < count) |
32927393 | 599 | goto out_free_buf; |
4e63acdf AI |
600 | } |
601 | ||
32927393 CH |
602 | error = count; |
603 | out_free_buf: | |
45089437 | 604 | kvfree(kbuf); |
77b14db5 EB |
605 | out: |
606 | sysctl_head_finish(head); | |
607 | ||
608 | return error; | |
609 | } | |
610 | ||
4bd6a735 | 611 | static ssize_t proc_sys_read(struct kiocb *iocb, struct iov_iter *iter) |
77b14db5 | 612 | { |
4bd6a735 | 613 | return proc_sys_call_handler(iocb, iter, 0); |
7708bfb1 | 614 | } |
77b14db5 | 615 | |
4bd6a735 | 616 | static ssize_t proc_sys_write(struct kiocb *iocb, struct iov_iter *iter) |
7708bfb1 | 617 | { |
4bd6a735 | 618 | return proc_sys_call_handler(iocb, iter, 1); |
77b14db5 EB |
619 | } |
620 | ||
f1ecf068 LDM |
621 | static int proc_sys_open(struct inode *inode, struct file *filp) |
622 | { | |
4e474a00 | 623 | struct ctl_table_header *head = grab_header(inode); |
f1ecf068 LDM |
624 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; |
625 | ||
4e474a00 LDM |
626 | /* sysctl was unregistered */ |
627 | if (IS_ERR(head)) | |
628 | return PTR_ERR(head); | |
629 | ||
f1ecf068 LDM |
630 | if (table->poll) |
631 | filp->private_data = proc_sys_poll_event(table->poll); | |
632 | ||
4e474a00 LDM |
633 | sysctl_head_finish(head); |
634 | ||
f1ecf068 LDM |
635 | return 0; |
636 | } | |
637 | ||
076ccb76 | 638 | static __poll_t proc_sys_poll(struct file *filp, poll_table *wait) |
f1ecf068 | 639 | { |
496ad9aa | 640 | struct inode *inode = file_inode(filp); |
4e474a00 | 641 | struct ctl_table_header *head = grab_header(inode); |
f1ecf068 | 642 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; |
076ccb76 | 643 | __poll_t ret = DEFAULT_POLLMASK; |
4e474a00 LDM |
644 | unsigned long event; |
645 | ||
646 | /* sysctl was unregistered */ | |
647 | if (IS_ERR(head)) | |
a9a08845 | 648 | return EPOLLERR | EPOLLHUP; |
f1ecf068 LDM |
649 | |
650 | if (!table->proc_handler) | |
651 | goto out; | |
652 | ||
653 | if (!table->poll) | |
654 | goto out; | |
655 | ||
4e474a00 | 656 | event = (unsigned long)filp->private_data; |
f1ecf068 LDM |
657 | poll_wait(filp, &table->poll->wait, wait); |
658 | ||
659 | if (event != atomic_read(&table->poll->event)) { | |
660 | filp->private_data = proc_sys_poll_event(table->poll); | |
a9a08845 | 661 | ret = EPOLLIN | EPOLLRDNORM | EPOLLERR | EPOLLPRI; |
f1ecf068 LDM |
662 | } |
663 | ||
664 | out: | |
4e474a00 LDM |
665 | sysctl_head_finish(head); |
666 | ||
f1ecf068 LDM |
667 | return ret; |
668 | } | |
77b14db5 | 669 | |
f0c3b509 AV |
670 | static bool proc_sys_fill_cache(struct file *file, |
671 | struct dir_context *ctx, | |
9043476f AV |
672 | struct ctl_table_header *head, |
673 | struct ctl_table *table) | |
77b14db5 | 674 | { |
f0c3b509 | 675 | struct dentry *child, *dir = file->f_path.dentry; |
77b14db5 EB |
676 | struct inode *inode; |
677 | struct qstr qname; | |
678 | ino_t ino = 0; | |
679 | unsigned type = DT_UNKNOWN; | |
77b14db5 EB |
680 | |
681 | qname.name = table->procname; | |
682 | qname.len = strlen(table->procname); | |
8387ff25 | 683 | qname.hash = full_name_hash(dir, qname.name, qname.len); |
77b14db5 | 684 | |
77b14db5 EB |
685 | child = d_lookup(dir, &qname); |
686 | if (!child) { | |
76aab3ab AV |
687 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
688 | child = d_alloc_parallel(dir, &qname, &wq); | |
689 | if (IS_ERR(child)) | |
690 | return false; | |
691 | if (d_in_lookup(child)) { | |
888e2b03 | 692 | struct dentry *res; |
76aab3ab | 693 | d_set_d_op(child, &proc_sys_dentry_operations); |
1eae9a47 | 694 | inode = proc_sys_make_inode(dir->d_sb, head, table); |
888e2b03 AV |
695 | res = d_splice_alias(inode, child); |
696 | d_lookup_done(child); | |
697 | if (unlikely(res)) { | |
698 | if (IS_ERR(res)) { | |
699 | dput(child); | |
700 | return false; | |
701 | } | |
702 | dput(child); | |
703 | child = res; | |
704 | } | |
77b14db5 EB |
705 | } |
706 | } | |
2b0143b5 | 707 | inode = d_inode(child); |
9043476f AV |
708 | ino = inode->i_ino; |
709 | type = inode->i_mode >> 12; | |
77b14db5 | 710 | dput(child); |
f0c3b509 | 711 | return dir_emit(ctx, qname.name, qname.len, ino, type); |
9043476f AV |
712 | } |
713 | ||
f0c3b509 AV |
714 | static bool proc_sys_link_fill_cache(struct file *file, |
715 | struct dir_context *ctx, | |
0e47c99d EB |
716 | struct ctl_table_header *head, |
717 | struct ctl_table *table) | |
718 | { | |
f0c3b509 | 719 | bool ret = true; |
a0b0d1c3 | 720 | |
0e47c99d | 721 | head = sysctl_head_grab(head); |
a0b0d1c3 DK |
722 | if (IS_ERR(head)) |
723 | return false; | |
0e47c99d | 724 | |
835b94e0 DK |
725 | /* It is not an error if we can not follow the link ignore it */ |
726 | if (sysctl_follow_link(&head, &table)) | |
727 | goto out; | |
0e47c99d | 728 | |
f0c3b509 | 729 | ret = proc_sys_fill_cache(file, ctx, head, table); |
0e47c99d EB |
730 | out: |
731 | sysctl_head_finish(head); | |
732 | return ret; | |
733 | } | |
734 | ||
e5eea098 | 735 | static int scan(struct ctl_table_header *head, struct ctl_table *table, |
9043476f | 736 | unsigned long *pos, struct file *file, |
f0c3b509 | 737 | struct dir_context *ctx) |
9043476f | 738 | { |
f0c3b509 | 739 | bool res; |
9043476f | 740 | |
f0c3b509 AV |
741 | if ((*pos)++ < ctx->pos) |
742 | return true; | |
9043476f | 743 | |
0e47c99d | 744 | if (unlikely(S_ISLNK(table->mode))) |
f0c3b509 | 745 | res = proc_sys_link_fill_cache(file, ctx, head, table); |
0e47c99d | 746 | else |
f0c3b509 | 747 | res = proc_sys_fill_cache(file, ctx, head, table); |
9043476f | 748 | |
f0c3b509 AV |
749 | if (res) |
750 | ctx->pos = *pos; | |
9043476f | 751 | |
6a75ce16 | 752 | return res; |
77b14db5 EB |
753 | } |
754 | ||
f0c3b509 | 755 | static int proc_sys_readdir(struct file *file, struct dir_context *ctx) |
77b14db5 | 756 | { |
f0c3b509 | 757 | struct ctl_table_header *head = grab_header(file_inode(file)); |
9043476f | 758 | struct ctl_table_header *h = NULL; |
6a75ce16 | 759 | struct ctl_table *entry; |
7ec66d06 | 760 | struct ctl_dir *ctl_dir; |
77b14db5 | 761 | unsigned long pos; |
9043476f AV |
762 | |
763 | if (IS_ERR(head)) | |
764 | return PTR_ERR(head); | |
77b14db5 | 765 | |
7ec66d06 | 766 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 | 767 | |
f0c3b509 | 768 | if (!dir_emit_dots(file, ctx)) |
93362fa4 | 769 | goto out; |
f0c3b509 | 770 | |
77b14db5 EB |
771 | pos = 2; |
772 | ||
7ec66d06 | 773 | for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) { |
f0c3b509 | 774 | if (!scan(h, entry, &pos, file, ctx)) { |
9043476f AV |
775 | sysctl_head_finish(h); |
776 | break; | |
77b14db5 EB |
777 | } |
778 | } | |
93362fa4 | 779 | out: |
77b14db5 | 780 | sysctl_head_finish(head); |
f0c3b509 | 781 | return 0; |
77b14db5 EB |
782 | } |
783 | ||
4609e1f1 | 784 | static int proc_sys_permission(struct mnt_idmap *idmap, |
549c7297 | 785 | struct inode *inode, int mask) |
77b14db5 EB |
786 | { |
787 | /* | |
788 | * sysctl entries that are not writeable, | |
789 | * are _NOT_ writeable, capabilities or not. | |
790 | */ | |
f696a365 MS |
791 | struct ctl_table_header *head; |
792 | struct ctl_table *table; | |
77b14db5 EB |
793 | int error; |
794 | ||
f696a365 MS |
795 | /* Executable files are not allowed under /proc/sys/ */ |
796 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) | |
797 | return -EACCES; | |
798 | ||
799 | head = grab_header(inode); | |
9043476f AV |
800 | if (IS_ERR(head)) |
801 | return PTR_ERR(head); | |
77b14db5 | 802 | |
f696a365 | 803 | table = PROC_I(inode)->sysctl_entry; |
9043476f AV |
804 | if (!table) /* global root - r-xr-xr-x */ |
805 | error = mask & MAY_WRITE ? -EACCES : 0; | |
806 | else /* Use the permissions on the sysctl table entry */ | |
73f7ef43 | 807 | error = sysctl_perm(head, table, mask & ~MAY_NOT_BLOCK); |
77b14db5 | 808 | |
77b14db5 EB |
809 | sysctl_head_finish(head); |
810 | return error; | |
811 | } | |
812 | ||
c1632a0f | 813 | static int proc_sys_setattr(struct mnt_idmap *idmap, |
549c7297 | 814 | struct dentry *dentry, struct iattr *attr) |
77b14db5 | 815 | { |
2b0143b5 | 816 | struct inode *inode = d_inode(dentry); |
77b14db5 EB |
817 | int error; |
818 | ||
819 | if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) | |
820 | return -EPERM; | |
821 | ||
c1632a0f | 822 | error = setattr_prepare(&nop_mnt_idmap, dentry, attr); |
1025774c CH |
823 | if (error) |
824 | return error; | |
825 | ||
c1632a0f | 826 | setattr_copy(&nop_mnt_idmap, inode, attr); |
1025774c | 827 | return 0; |
77b14db5 EB |
828 | } |
829 | ||
b74d24f7 | 830 | static int proc_sys_getattr(struct mnt_idmap *idmap, |
549c7297 | 831 | const struct path *path, struct kstat *stat, |
a528d35e | 832 | u32 request_mask, unsigned int query_flags) |
9043476f | 833 | { |
a528d35e | 834 | struct inode *inode = d_inode(path->dentry); |
9043476f AV |
835 | struct ctl_table_header *head = grab_header(inode); |
836 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
837 | ||
838 | if (IS_ERR(head)) | |
839 | return PTR_ERR(head); | |
840 | ||
0d72b928 | 841 | generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat); |
9043476f AV |
842 | if (table) |
843 | stat->mode = (stat->mode & S_IFMT) | table->mode; | |
844 | ||
845 | sysctl_head_finish(head); | |
846 | return 0; | |
847 | } | |
848 | ||
77b14db5 | 849 | static const struct file_operations proc_sys_file_operations = { |
f1ecf068 LDM |
850 | .open = proc_sys_open, |
851 | .poll = proc_sys_poll, | |
4bd6a735 MWO |
852 | .read_iter = proc_sys_read, |
853 | .write_iter = proc_sys_write, | |
b0072734 | 854 | .splice_read = copy_splice_read, |
4bd6a735 | 855 | .splice_write = iter_file_splice_write, |
6038f373 | 856 | .llseek = default_llseek, |
9043476f AV |
857 | }; |
858 | ||
859 | static const struct file_operations proc_sys_dir_file_operations = { | |
887df078 | 860 | .read = generic_read_dir, |
f50752ea | 861 | .iterate_shared = proc_sys_readdir, |
3222a3e5 | 862 | .llseek = generic_file_llseek, |
77b14db5 EB |
863 | }; |
864 | ||
03a44825 | 865 | static const struct inode_operations proc_sys_inode_operations = { |
9043476f AV |
866 | .permission = proc_sys_permission, |
867 | .setattr = proc_sys_setattr, | |
868 | .getattr = proc_sys_getattr, | |
869 | }; | |
870 | ||
871 | static const struct inode_operations proc_sys_dir_operations = { | |
77b14db5 EB |
872 | .lookup = proc_sys_lookup, |
873 | .permission = proc_sys_permission, | |
874 | .setattr = proc_sys_setattr, | |
9043476f | 875 | .getattr = proc_sys_getattr, |
77b14db5 EB |
876 | }; |
877 | ||
0b728e19 | 878 | static int proc_sys_revalidate(struct dentry *dentry, unsigned int flags) |
77b14db5 | 879 | { |
0b728e19 | 880 | if (flags & LOOKUP_RCU) |
34286d66 | 881 | return -ECHILD; |
2b0143b5 | 882 | return !PROC_I(d_inode(dentry))->sysctl->unregistering; |
9043476f AV |
883 | } |
884 | ||
fe15ce44 | 885 | static int proc_sys_delete(const struct dentry *dentry) |
9043476f | 886 | { |
2b0143b5 | 887 | return !!PROC_I(d_inode(dentry))->sysctl->unregistering; |
9043476f AV |
888 | } |
889 | ||
1f87f0b5 EB |
890 | static int sysctl_is_seen(struct ctl_table_header *p) |
891 | { | |
892 | struct ctl_table_set *set = p->set; | |
893 | int res; | |
894 | spin_lock(&sysctl_lock); | |
895 | if (p->unregistering) | |
896 | res = 0; | |
897 | else if (!set->is_seen) | |
898 | res = 1; | |
899 | else | |
900 | res = set->is_seen(set); | |
901 | spin_unlock(&sysctl_lock); | |
902 | return res; | |
903 | } | |
904 | ||
6fa67e70 | 905 | static int proc_sys_compare(const struct dentry *dentry, |
621e155a | 906 | unsigned int len, const char *str, const struct qstr *name) |
9043476f | 907 | { |
dfef6dcd | 908 | struct ctl_table_header *head; |
da53be12 LT |
909 | struct inode *inode; |
910 | ||
31e6b01f NP |
911 | /* Although proc doesn't have negative dentries, rcu-walk means |
912 | * that inode here can be NULL */ | |
dfef6dcd | 913 | /* AV: can it, indeed? */ |
2b0143b5 | 914 | inode = d_inode_rcu(dentry); |
31e6b01f | 915 | if (!inode) |
dfef6dcd | 916 | return 1; |
621e155a | 917 | if (name->len != len) |
9043476f | 918 | return 1; |
621e155a | 919 | if (memcmp(name->name, str, len)) |
9043476f | 920 | return 1; |
dfef6dcd AV |
921 | head = rcu_dereference(PROC_I(inode)->sysctl); |
922 | return !head || !sysctl_is_seen(head); | |
77b14db5 EB |
923 | } |
924 | ||
d72f71eb | 925 | static const struct dentry_operations proc_sys_dentry_operations = { |
77b14db5 | 926 | .d_revalidate = proc_sys_revalidate, |
9043476f AV |
927 | .d_delete = proc_sys_delete, |
928 | .d_compare = proc_sys_compare, | |
77b14db5 EB |
929 | }; |
930 | ||
0e47c99d EB |
931 | static struct ctl_dir *find_subdir(struct ctl_dir *dir, |
932 | const char *name, int namelen) | |
1f87f0b5 | 933 | { |
7ec66d06 EB |
934 | struct ctl_table_header *head; |
935 | struct ctl_table *entry; | |
1f87f0b5 | 936 | |
0e47c99d | 937 | entry = find_entry(&head, dir, name, namelen); |
7ec66d06 EB |
938 | if (!entry) |
939 | return ERR_PTR(-ENOENT); | |
51f72f4a EB |
940 | if (!S_ISDIR(entry->mode)) |
941 | return ERR_PTR(-ENOTDIR); | |
942 | return container_of(head, struct ctl_dir, header); | |
7ec66d06 EB |
943 | } |
944 | ||
945 | static struct ctl_dir *new_dir(struct ctl_table_set *set, | |
0e47c99d | 946 | const char *name, int namelen) |
7ec66d06 EB |
947 | { |
948 | struct ctl_table *table; | |
949 | struct ctl_dir *new; | |
ac13ac6f | 950 | struct ctl_node *node; |
7ec66d06 | 951 | char *new_name; |
1f87f0b5 | 952 | |
ac13ac6f EB |
953 | new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) + |
954 | sizeof(struct ctl_table)*2 + namelen + 1, | |
955 | GFP_KERNEL); | |
7ec66d06 | 956 | if (!new) |
1f87f0b5 EB |
957 | return NULL; |
958 | ||
ac13ac6f EB |
959 | node = (struct ctl_node *)(new + 1); |
960 | table = (struct ctl_table *)(node + 1); | |
7ec66d06 EB |
961 | new_name = (char *)(table + 2); |
962 | memcpy(new_name, name, namelen); | |
7ec66d06 EB |
963 | table[0].procname = new_name; |
964 | table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
b1f01e2b | 965 | init_header(&new->header, set->dir.header.root, set, node, table, 1); |
7ec66d06 EB |
966 | |
967 | return new; | |
1f87f0b5 EB |
968 | } |
969 | ||
60f126d9 EB |
970 | /** |
971 | * get_subdir - find or create a subdir with the specified name. | |
972 | * @dir: Directory to create the subdirectory in | |
973 | * @name: The name of the subdirectory to find or create | |
974 | * @namelen: The length of name | |
975 | * | |
976 | * Takes a directory with an elevated reference count so we know that | |
977 | * if we drop the lock the directory will not go away. Upon success | |
978 | * the reference is moved from @dir to the returned subdirectory. | |
979 | * Upon error an error code is returned and the reference on @dir is | |
980 | * simply dropped. | |
981 | */ | |
0e47c99d EB |
982 | static struct ctl_dir *get_subdir(struct ctl_dir *dir, |
983 | const char *name, int namelen) | |
1f87f0b5 | 984 | { |
0e47c99d | 985 | struct ctl_table_set *set = dir->header.set; |
7ec66d06 | 986 | struct ctl_dir *subdir, *new = NULL; |
0eb97f38 | 987 | int err; |
1f87f0b5 | 988 | |
7ec66d06 | 989 | spin_lock(&sysctl_lock); |
0e47c99d | 990 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
991 | if (!IS_ERR(subdir)) |
992 | goto found; | |
993 | if (PTR_ERR(subdir) != -ENOENT) | |
994 | goto failed; | |
995 | ||
996 | spin_unlock(&sysctl_lock); | |
997 | new = new_dir(set, name, namelen); | |
998 | spin_lock(&sysctl_lock); | |
999 | subdir = ERR_PTR(-ENOMEM); | |
1000 | if (!new) | |
1001 | goto failed; | |
1002 | ||
60f126d9 | 1003 | /* Was the subdir added while we dropped the lock? */ |
0e47c99d | 1004 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
1005 | if (!IS_ERR(subdir)) |
1006 | goto found; | |
1007 | if (PTR_ERR(subdir) != -ENOENT) | |
1008 | goto failed; | |
1009 | ||
60f126d9 | 1010 | /* Nope. Use the our freshly made directory entry. */ |
0eb97f38 EB |
1011 | err = insert_header(dir, &new->header); |
1012 | subdir = ERR_PTR(err); | |
1013 | if (err) | |
0e47c99d | 1014 | goto failed; |
7ec66d06 EB |
1015 | subdir = new; |
1016 | found: | |
1017 | subdir->header.nreg++; | |
1018 | failed: | |
a1c83681 | 1019 | if (IS_ERR(subdir)) { |
87ebdc00 | 1020 | pr_err("sysctl could not get directory: "); |
6980128f | 1021 | sysctl_print_dir(dir); |
153ee1c4 GU |
1022 | pr_cont("%*.*s %ld\n", namelen, namelen, name, |
1023 | PTR_ERR(subdir)); | |
1f87f0b5 | 1024 | } |
7ec66d06 EB |
1025 | drop_sysctl_table(&dir->header); |
1026 | if (new) | |
1027 | drop_sysctl_table(&new->header); | |
1028 | spin_unlock(&sysctl_lock); | |
1029 | return subdir; | |
1f87f0b5 EB |
1030 | } |
1031 | ||
0e47c99d EB |
1032 | static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir) |
1033 | { | |
1034 | struct ctl_dir *parent; | |
1035 | const char *procname; | |
1036 | if (!dir->header.parent) | |
1037 | return &set->dir; | |
1038 | parent = xlate_dir(set, dir->header.parent); | |
1039 | if (IS_ERR(parent)) | |
1040 | return parent; | |
1041 | procname = dir->header.ctl_table[0].procname; | |
1042 | return find_subdir(parent, procname, strlen(procname)); | |
1043 | } | |
1044 | ||
1045 | static int sysctl_follow_link(struct ctl_table_header **phead, | |
13bcc6a2 | 1046 | struct ctl_table **pentry) |
0e47c99d EB |
1047 | { |
1048 | struct ctl_table_header *head; | |
1049 | struct ctl_table_root *root; | |
1050 | struct ctl_table_set *set; | |
1051 | struct ctl_table *entry; | |
1052 | struct ctl_dir *dir; | |
1053 | int ret; | |
1054 | ||
0e47c99d EB |
1055 | spin_lock(&sysctl_lock); |
1056 | root = (*pentry)->data; | |
13bcc6a2 | 1057 | set = lookup_header_set(root); |
0e47c99d EB |
1058 | dir = xlate_dir(set, (*phead)->parent); |
1059 | if (IS_ERR(dir)) | |
1060 | ret = PTR_ERR(dir); | |
1061 | else { | |
1062 | const char *procname = (*pentry)->procname; | |
1063 | head = NULL; | |
1064 | entry = find_entry(&head, dir, procname, strlen(procname)); | |
1065 | ret = -ENOENT; | |
1066 | if (entry && use_table(head)) { | |
1067 | unuse_table(*phead); | |
1068 | *phead = head; | |
1069 | *pentry = entry; | |
1070 | ret = 0; | |
1071 | } | |
1072 | } | |
1073 | ||
1074 | spin_unlock(&sysctl_lock); | |
1075 | return ret; | |
1076 | } | |
1077 | ||
7c60c48f | 1078 | static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...) |
1f87f0b5 | 1079 | { |
7c60c48f EB |
1080 | struct va_format vaf; |
1081 | va_list args; | |
1f87f0b5 | 1082 | |
7c60c48f EB |
1083 | va_start(args, fmt); |
1084 | vaf.fmt = fmt; | |
1085 | vaf.va = &args; | |
1086 | ||
87ebdc00 AM |
1087 | pr_err("sysctl table check failed: %s/%s %pV\n", |
1088 | path, table->procname, &vaf); | |
1f87f0b5 | 1089 | |
7c60c48f EB |
1090 | va_end(args); |
1091 | return -EINVAL; | |
1f87f0b5 EB |
1092 | } |
1093 | ||
4f2fec00 LR |
1094 | static int sysctl_check_table_array(const char *path, struct ctl_table *table) |
1095 | { | |
1096 | int err = 0; | |
1097 | ||
61d9b56a LR |
1098 | if ((table->proc_handler == proc_douintvec) || |
1099 | (table->proc_handler == proc_douintvec_minmax)) { | |
4f2fec00 | 1100 | if (table->maxlen != sizeof(unsigned int)) |
64a11f3d | 1101 | err |= sysctl_err(path, table, "array not allowed"); |
4f2fec00 LR |
1102 | } |
1103 | ||
cb944413 ED |
1104 | if (table->proc_handler == proc_dou8vec_minmax) { |
1105 | if (table->maxlen != sizeof(u8)) | |
1106 | err |= sysctl_err(path, table, "array not allowed"); | |
1107 | } | |
1108 | ||
f1aa2eb5 OM |
1109 | if (table->proc_handler == proc_dobool) { |
1110 | if (table->maxlen != sizeof(bool)) | |
1111 | err |= sysctl_err(path, table, "array not allowed"); | |
1112 | } | |
1113 | ||
4f2fec00 LR |
1114 | return err; |
1115 | } | |
1116 | ||
cc9f7ee0 | 1117 | static int sysctl_check_table(const char *path, struct ctl_table_header *header) |
1f87f0b5 | 1118 | { |
cb55f27a | 1119 | struct ctl_table *entry; |
7c60c48f | 1120 | int err = 0; |
18d4b42e | 1121 | list_for_each_table_entry(entry, header) { |
cb55f27a | 1122 | if ((entry->proc_handler == proc_dostring) || |
f1aa2eb5 | 1123 | (entry->proc_handler == proc_dobool) || |
cb55f27a MT |
1124 | (entry->proc_handler == proc_dointvec) || |
1125 | (entry->proc_handler == proc_douintvec) || | |
1126 | (entry->proc_handler == proc_douintvec_minmax) || | |
1127 | (entry->proc_handler == proc_dointvec_minmax) || | |
1128 | (entry->proc_handler == proc_dou8vec_minmax) || | |
1129 | (entry->proc_handler == proc_dointvec_jiffies) || | |
1130 | (entry->proc_handler == proc_dointvec_userhz_jiffies) || | |
1131 | (entry->proc_handler == proc_dointvec_ms_jiffies) || | |
1132 | (entry->proc_handler == proc_doulongvec_minmax) || | |
1133 | (entry->proc_handler == proc_doulongvec_ms_jiffies_minmax)) { | |
1134 | if (!entry->data) | |
1135 | err |= sysctl_err(path, entry, "No data"); | |
1136 | if (!entry->maxlen) | |
1137 | err |= sysctl_err(path, entry, "No maxlen"); | |
4f2fec00 | 1138 | else |
cb55f27a | 1139 | err |= sysctl_check_table_array(path, entry); |
7c60c48f | 1140 | } |
cb55f27a MT |
1141 | if (!entry->proc_handler) |
1142 | err |= sysctl_err(path, entry, "No proc_handler"); | |
7c60c48f | 1143 | |
cb55f27a MT |
1144 | if ((entry->mode & (S_IRUGO|S_IWUGO)) != entry->mode) |
1145 | err |= sysctl_err(path, entry, "bogus .mode 0%o", | |
1146 | entry->mode); | |
1f87f0b5 | 1147 | } |
7c60c48f | 1148 | return err; |
1f87f0b5 | 1149 | } |
1f87f0b5 | 1150 | |
cc9f7ee0 | 1151 | static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table_header *head) |
0e47c99d EB |
1152 | { |
1153 | struct ctl_table *link_table, *entry, *link; | |
1154 | struct ctl_table_header *links; | |
ac13ac6f | 1155 | struct ctl_node *node; |
0e47c99d EB |
1156 | char *link_name; |
1157 | int nr_entries, name_bytes; | |
1158 | ||
1159 | name_bytes = 0; | |
1160 | nr_entries = 0; | |
18d4b42e | 1161 | list_for_each_table_entry(entry, head) { |
0e47c99d EB |
1162 | nr_entries++; |
1163 | name_bytes += strlen(entry->procname) + 1; | |
1164 | } | |
1165 | ||
1166 | links = kzalloc(sizeof(struct ctl_table_header) + | |
ac13ac6f | 1167 | sizeof(struct ctl_node)*nr_entries + |
0e47c99d EB |
1168 | sizeof(struct ctl_table)*(nr_entries + 1) + |
1169 | name_bytes, | |
1170 | GFP_KERNEL); | |
1171 | ||
1172 | if (!links) | |
1173 | return NULL; | |
1174 | ||
ac13ac6f EB |
1175 | node = (struct ctl_node *)(links + 1); |
1176 | link_table = (struct ctl_table *)(node + nr_entries); | |
0e47c99d | 1177 | link_name = (char *)&link_table[nr_entries + 1]; |
cb55f27a | 1178 | link = link_table; |
0e47c99d | 1179 | |
18d4b42e | 1180 | list_for_each_table_entry(entry, head) { |
0e47c99d EB |
1181 | int len = strlen(entry->procname) + 1; |
1182 | memcpy(link_name, entry->procname, len); | |
1183 | link->procname = link_name; | |
1184 | link->mode = S_IFLNK|S_IRWXUGO; | |
cc9f7ee0 | 1185 | link->data = head->root; |
0e47c99d | 1186 | link_name += len; |
cb55f27a | 1187 | link++; |
0e47c99d | 1188 | } |
b1f01e2b JG |
1189 | init_header(links, dir->header.root, dir->header.set, node, link_table, |
1190 | head->ctl_table_size); | |
0e47c99d EB |
1191 | links->nreg = nr_entries; |
1192 | ||
1193 | return links; | |
1194 | } | |
1195 | ||
1196 | static bool get_links(struct ctl_dir *dir, | |
cc9f7ee0 JG |
1197 | struct ctl_table_header *header, |
1198 | struct ctl_table_root *link_root) | |
0e47c99d | 1199 | { |
cc9f7ee0 | 1200 | struct ctl_table_header *tmp_head; |
0e47c99d EB |
1201 | struct ctl_table *entry, *link; |
1202 | ||
31555231 | 1203 | if (header->ctl_table_size == 0 || |
eb32d3ad | 1204 | sysctl_is_perm_empty_ctl_header(header)) |
31555231 JG |
1205 | return true; |
1206 | ||
0e47c99d | 1207 | /* Are there links available for every entry in table? */ |
18d4b42e | 1208 | list_for_each_table_entry(entry, header) { |
0e47c99d | 1209 | const char *procname = entry->procname; |
cc9f7ee0 | 1210 | link = find_entry(&tmp_head, dir, procname, strlen(procname)); |
0e47c99d EB |
1211 | if (!link) |
1212 | return false; | |
1213 | if (S_ISDIR(link->mode) && S_ISDIR(entry->mode)) | |
1214 | continue; | |
1215 | if (S_ISLNK(link->mode) && (link->data == link_root)) | |
1216 | continue; | |
1217 | return false; | |
1218 | } | |
1219 | ||
1220 | /* The checks passed. Increase the registration count on the links */ | |
18d4b42e | 1221 | list_for_each_table_entry(entry, header) { |
0e47c99d | 1222 | const char *procname = entry->procname; |
cc9f7ee0 JG |
1223 | link = find_entry(&tmp_head, dir, procname, strlen(procname)); |
1224 | tmp_head->nreg++; | |
0e47c99d EB |
1225 | } |
1226 | return true; | |
1227 | } | |
1228 | ||
1229 | static int insert_links(struct ctl_table_header *head) | |
1230 | { | |
1231 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
9a521359 | 1232 | struct ctl_dir *core_parent; |
0e47c99d EB |
1233 | struct ctl_table_header *links; |
1234 | int err; | |
1235 | ||
1236 | if (head->set == root_set) | |
1237 | return 0; | |
1238 | ||
1239 | core_parent = xlate_dir(root_set, head->parent); | |
1240 | if (IS_ERR(core_parent)) | |
1241 | return 0; | |
1242 | ||
cc9f7ee0 | 1243 | if (get_links(core_parent, head, head->root)) |
0e47c99d EB |
1244 | return 0; |
1245 | ||
1246 | core_parent->header.nreg++; | |
1247 | spin_unlock(&sysctl_lock); | |
1248 | ||
cc9f7ee0 | 1249 | links = new_links(core_parent, head); |
0e47c99d EB |
1250 | |
1251 | spin_lock(&sysctl_lock); | |
1252 | err = -ENOMEM; | |
1253 | if (!links) | |
1254 | goto out; | |
1255 | ||
1256 | err = 0; | |
cc9f7ee0 | 1257 | if (get_links(core_parent, head, head->root)) { |
0e47c99d EB |
1258 | kfree(links); |
1259 | goto out; | |
1260 | } | |
1261 | ||
1262 | err = insert_header(core_parent, links); | |
1263 | if (err) | |
1264 | kfree(links); | |
1265 | out: | |
1266 | drop_sysctl_table(&core_parent->header); | |
1267 | return err; | |
1268 | } | |
1269 | ||
b2f56e55 LC |
1270 | /* Find the directory for the ctl_table. If one is not found create it. */ |
1271 | static struct ctl_dir *sysctl_mkdir_p(struct ctl_dir *dir, const char *path) | |
1272 | { | |
1273 | const char *name, *nextname; | |
1274 | ||
1275 | for (name = path; name; name = nextname) { | |
1276 | int namelen; | |
1277 | nextname = strchr(name, '/'); | |
1278 | if (nextname) { | |
1279 | namelen = nextname - name; | |
1280 | nextname++; | |
1281 | } else { | |
1282 | namelen = strlen(name); | |
1283 | } | |
1284 | if (namelen == 0) | |
1285 | continue; | |
1286 | ||
1287 | /* | |
1288 | * namelen ensures if name is "foo/bar/yay" only foo is | |
1289 | * registered first. We traverse as if using mkdir -p and | |
1290 | * return a ctl_dir for the last directory entry. | |
1291 | */ | |
1292 | dir = get_subdir(dir, name, namelen); | |
1293 | if (IS_ERR(dir)) | |
1294 | break; | |
1295 | } | |
1296 | return dir; | |
1297 | } | |
1298 | ||
1f87f0b5 | 1299 | /** |
f728019b | 1300 | * __register_sysctl_table - register a leaf sysctl table |
60a47a2e | 1301 | * @set: Sysctl tree to register on |
1f87f0b5 | 1302 | * @path: The path to the directory the sysctl table is in. |
1dc8689e LC |
1303 | * @table: the top-level table structure without any child. This table |
1304 | * should not be free'd after registration. So it should not be | |
1305 | * used on stack. It can either be a global or dynamically allocated | |
1306 | * by the caller and free'd later after sysctl unregistration. | |
bff97cf1 | 1307 | * @table_size : The number of elements in table |
1f87f0b5 EB |
1308 | * |
1309 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1310 | * array. A completely 0 filled entry terminates the table. | |
1311 | * | |
1312 | * The members of the &struct ctl_table structure are used as follows: | |
1313 | * | |
1314 | * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not | |
1315 | * enter a sysctl file | |
1316 | * | |
1317 | * data - a pointer to data for use by proc_handler | |
1318 | * | |
1319 | * maxlen - the maximum size in bytes of the data | |
1320 | * | |
f728019b | 1321 | * mode - the file permissions for the /proc/sys file |
1f87f0b5 | 1322 | * |
f728019b | 1323 | * child - must be %NULL. |
1f87f0b5 EB |
1324 | * |
1325 | * proc_handler - the text handler routine (described below) | |
1326 | * | |
1f87f0b5 | 1327 | * extra1, extra2 - extra pointers usable by the proc handler routines |
67ff3228 LC |
1328 | * XXX: we should eventually modify these to use long min / max [0] |
1329 | * [0] https://lkml.kernel.org/87zgpte9o4.fsf@email.froward.int.ebiederm.org | |
1f87f0b5 EB |
1330 | * |
1331 | * Leaf nodes in the sysctl tree will be represented by a single file | |
67ff3228 LC |
1332 | * under /proc; non-leaf nodes (where child is not NULL) are not allowed, |
1333 | * sysctl_check_table() verifies this. | |
1f87f0b5 | 1334 | * |
f728019b EB |
1335 | * There must be a proc_handler routine for any terminal nodes. |
1336 | * Several default handlers are available to cover common cases - | |
1f87f0b5 EB |
1337 | * |
1338 | * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), | |
1339 | * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), | |
1340 | * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() | |
1341 | * | |
1342 | * It is the handler's job to read the input buffer from user memory | |
1343 | * and process it. The handler should return 0 on success. | |
1344 | * | |
1345 | * This routine returns %NULL on a failure to register, and a pointer | |
1346 | * to the table header on success. | |
1347 | */ | |
6e9d5164 | 1348 | struct ctl_table_header *__register_sysctl_table( |
60a47a2e | 1349 | struct ctl_table_set *set, |
bff97cf1 | 1350 | const char *path, struct ctl_table *table, size_t table_size) |
1f87f0b5 | 1351 | { |
60a47a2e | 1352 | struct ctl_table_root *root = set->dir.header.root; |
1f87f0b5 | 1353 | struct ctl_table_header *header; |
7ec66d06 | 1354 | struct ctl_dir *dir; |
ac13ac6f | 1355 | struct ctl_node *node; |
1f87f0b5 | 1356 | |
ac13ac6f | 1357 | header = kzalloc(sizeof(struct ctl_table_header) + |
bff97cf1 | 1358 | sizeof(struct ctl_node)*table_size, GFP_KERNEL_ACCOUNT); |
1f87f0b5 EB |
1359 | if (!header) |
1360 | return NULL; | |
1361 | ||
ac13ac6f | 1362 | node = (struct ctl_node *)(header + 1); |
bff97cf1 | 1363 | init_header(header, root, set, node, table, table_size); |
cc9f7ee0 | 1364 | if (sysctl_check_table(path, header)) |
7ec66d06 EB |
1365 | goto fail; |
1366 | ||
1367 | spin_lock(&sysctl_lock); | |
0e47c99d | 1368 | dir = &set->dir; |
67ff3228 | 1369 | /* Reference moved down the directory tree get_subdir */ |
7ec66d06 EB |
1370 | dir->header.nreg++; |
1371 | spin_unlock(&sysctl_lock); | |
1f87f0b5 | 1372 | |
b2f56e55 LC |
1373 | dir = sysctl_mkdir_p(dir, path); |
1374 | if (IS_ERR(dir)) | |
1375 | goto fail; | |
1f87f0b5 | 1376 | spin_lock(&sysctl_lock); |
0e47c99d | 1377 | if (insert_header(dir, header)) |
7ec66d06 | 1378 | goto fail_put_dir_locked; |
0e47c99d | 1379 | |
7ec66d06 | 1380 | drop_sysctl_table(&dir->header); |
1f87f0b5 EB |
1381 | spin_unlock(&sysctl_lock); |
1382 | ||
1383 | return header; | |
0e47c99d | 1384 | |
7ec66d06 EB |
1385 | fail_put_dir_locked: |
1386 | drop_sysctl_table(&dir->header); | |
7c60c48f EB |
1387 | spin_unlock(&sysctl_lock); |
1388 | fail: | |
1389 | kfree(header); | |
7c60c48f | 1390 | return NULL; |
1f87f0b5 EB |
1391 | } |
1392 | ||
fea478d4 | 1393 | /** |
9edbfe92 | 1394 | * register_sysctl_sz - register a sysctl table |
1dc8689e LC |
1395 | * @path: The path to the directory the sysctl table is in. If the path |
1396 | * doesn't exist we will create it for you. | |
1397 | * @table: the table structure. The calller must ensure the life of the @table | |
1398 | * will be kept during the lifetime use of the syctl. It must not be freed | |
1399 | * until unregister_sysctl_table() is called with the given returned table | |
1400 | * with this registration. If your code is non modular then you don't need | |
1401 | * to call unregister_sysctl_table() and can instead use something like | |
1402 | * register_sysctl_init() which does not care for the result of the syctl | |
1403 | * registration. | |
9edbfe92 | 1404 | * @table_size: The number of elements in table. |
fea478d4 EB |
1405 | * |
1406 | * Register a sysctl table. @table should be a filled in ctl_table | |
1407 | * array. A completely 0 filled entry terminates the table. | |
1408 | * | |
1409 | * See __register_sysctl_table for more details. | |
1410 | */ | |
9edbfe92 JG |
1411 | struct ctl_table_header *register_sysctl_sz(const char *path, struct ctl_table *table, |
1412 | size_t table_size) | |
fea478d4 EB |
1413 | { |
1414 | return __register_sysctl_table(&sysctl_table_root.default_set, | |
9edbfe92 | 1415 | path, table, table_size); |
fea478d4 | 1416 | } |
9edbfe92 | 1417 | EXPORT_SYMBOL(register_sysctl_sz); |
fea478d4 | 1418 | |
3ddd9a80 XN |
1419 | /** |
1420 | * __register_sysctl_init() - register sysctl table to path | |
1dc8689e LC |
1421 | * @path: path name for sysctl base. If that path doesn't exist we will create |
1422 | * it for you. | |
1423 | * @table: This is the sysctl table that needs to be registered to the path. | |
1424 | * The caller must ensure the life of the @table will be kept during the | |
1425 | * lifetime use of the sysctl. | |
3ddd9a80 XN |
1426 | * @table_name: The name of sysctl table, only used for log printing when |
1427 | * registration fails | |
3bc269cf | 1428 | * @table_size: The number of elements in table |
3ddd9a80 XN |
1429 | * |
1430 | * The sysctl interface is used by userspace to query or modify at runtime | |
1431 | * a predefined value set on a variable. These variables however have default | |
1432 | * values pre-set. Code which depends on these variables will always work even | |
1433 | * if register_sysctl() fails. If register_sysctl() fails you'd just loose the | |
1434 | * ability to query or modify the sysctls dynamically at run time. Chances of | |
1435 | * register_sysctl() failing on init are extremely low, and so for both reasons | |
1436 | * this function does not return any error as it is used by initialization code. | |
1437 | * | |
228b09de | 1438 | * Context: if your base directory does not exist it will be created for you. |
3ddd9a80 XN |
1439 | */ |
1440 | void __init __register_sysctl_init(const char *path, struct ctl_table *table, | |
3bc269cf | 1441 | const char *table_name, size_t table_size) |
3ddd9a80 | 1442 | { |
3bc269cf | 1443 | struct ctl_table_header *hdr = register_sysctl_sz(path, table, table_size); |
3ddd9a80 XN |
1444 | |
1445 | if (unlikely(!hdr)) { | |
9edbfe92 | 1446 | pr_err("failed when register_sysctl_sz %s to %s\n", table_name, path); |
3ddd9a80 XN |
1447 | return; |
1448 | } | |
1449 | kmemleak_not_leak(hdr); | |
1450 | } | |
1451 | ||
0e47c99d EB |
1452 | static void put_links(struct ctl_table_header *header) |
1453 | { | |
1454 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
1455 | struct ctl_table_root *root = header->root; | |
1456 | struct ctl_dir *parent = header->parent; | |
1457 | struct ctl_dir *core_parent; | |
1458 | struct ctl_table *entry; | |
1459 | ||
1460 | if (header->set == root_set) | |
1461 | return; | |
1462 | ||
1463 | core_parent = xlate_dir(root_set, parent); | |
1464 | if (IS_ERR(core_parent)) | |
1465 | return; | |
1466 | ||
18d4b42e | 1467 | list_for_each_table_entry(entry, header) { |
0e47c99d EB |
1468 | struct ctl_table_header *link_head; |
1469 | struct ctl_table *link; | |
1470 | const char *name = entry->procname; | |
1471 | ||
1472 | link = find_entry(&link_head, core_parent, name, strlen(name)); | |
1473 | if (link && | |
1474 | ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) || | |
1475 | (S_ISLNK(link->mode) && (link->data == root)))) { | |
1476 | drop_sysctl_table(link_head); | |
1477 | } | |
1478 | else { | |
87ebdc00 | 1479 | pr_err("sysctl link missing during unregister: "); |
0e47c99d | 1480 | sysctl_print_dir(parent); |
153ee1c4 | 1481 | pr_cont("%s\n", name); |
0e47c99d EB |
1482 | } |
1483 | } | |
1484 | } | |
1485 | ||
938aaa4f EB |
1486 | static void drop_sysctl_table(struct ctl_table_header *header) |
1487 | { | |
7ec66d06 EB |
1488 | struct ctl_dir *parent = header->parent; |
1489 | ||
938aaa4f EB |
1490 | if (--header->nreg) |
1491 | return; | |
1492 | ||
89189557 | 1493 | if (parent) { |
23da9588 | 1494 | put_links(header); |
89189557 Y |
1495 | start_unregistering(header); |
1496 | } | |
1497 | ||
938aaa4f EB |
1498 | if (!--header->count) |
1499 | kfree_rcu(header, rcu); | |
7ec66d06 EB |
1500 | |
1501 | if (parent) | |
1502 | drop_sysctl_table(&parent->header); | |
938aaa4f EB |
1503 | } |
1504 | ||
1f87f0b5 EB |
1505 | /** |
1506 | * unregister_sysctl_table - unregister a sysctl table hierarchy | |
19c4e618 | 1507 | * @header: the header returned from register_sysctl or __register_sysctl_table |
1f87f0b5 EB |
1508 | * |
1509 | * Unregisters the sysctl table and all children. proc entries may not | |
1510 | * actually be removed until they are no longer used by anyone. | |
1511 | */ | |
1512 | void unregister_sysctl_table(struct ctl_table_header * header) | |
1513 | { | |
1514 | might_sleep(); | |
1515 | ||
1516 | if (header == NULL) | |
1517 | return; | |
1518 | ||
1519 | spin_lock(&sysctl_lock); | |
938aaa4f | 1520 | drop_sysctl_table(header); |
1f87f0b5 EB |
1521 | spin_unlock(&sysctl_lock); |
1522 | } | |
1523 | EXPORT_SYMBOL(unregister_sysctl_table); | |
1524 | ||
0e47c99d | 1525 | void setup_sysctl_set(struct ctl_table_set *set, |
9eb47c26 | 1526 | struct ctl_table_root *root, |
1f87f0b5 EB |
1527 | int (*is_seen)(struct ctl_table_set *)) |
1528 | { | |
1347440d | 1529 | memset(set, 0, sizeof(*set)); |
0e47c99d | 1530 | set->is_seen = is_seen; |
b1f01e2b | 1531 | init_header(&set->dir.header, root, set, NULL, root_table, 1); |
1f87f0b5 EB |
1532 | } |
1533 | ||
97324cd8 EB |
1534 | void retire_sysctl_set(struct ctl_table_set *set) |
1535 | { | |
ac13ac6f | 1536 | WARN_ON(!RB_EMPTY_ROOT(&set->dir.root)); |
97324cd8 | 1537 | } |
1f87f0b5 | 1538 | |
1e0edd3f | 1539 | int __init proc_sys_init(void) |
77b14db5 | 1540 | { |
e1675231 AD |
1541 | struct proc_dir_entry *proc_sys_root; |
1542 | ||
77b14db5 | 1543 | proc_sys_root = proc_mkdir("sys", NULL); |
9043476f | 1544 | proc_sys_root->proc_iops = &proc_sys_dir_operations; |
d56c0d45 | 1545 | proc_sys_root->proc_dir_ops = &proc_sys_dir_file_operations; |
77b14db5 | 1546 | proc_sys_root->nlink = 0; |
de4e83bd | 1547 | |
d8c0418a | 1548 | return sysctl_init_bases(); |
77b14db5 | 1549 | } |
3db978d4 | 1550 | |
0a477e1a VB |
1551 | struct sysctl_alias { |
1552 | const char *kernel_param; | |
1553 | const char *sysctl_param; | |
1554 | }; | |
1555 | ||
1556 | /* | |
1557 | * Historically some settings had both sysctl and a command line parameter. | |
1558 | * With the generic sysctl. parameter support, we can handle them at a single | |
1559 | * place and only keep the historical name for compatibility. This is not meant | |
1560 | * to add brand new aliases. When adding existing aliases, consider whether | |
1561 | * the possibly different moment of changing the value (e.g. from early_param | |
1562 | * to the moment do_sysctl_args() is called) is an issue for the specific | |
1563 | * parameter. | |
1564 | */ | |
1565 | static const struct sysctl_alias sysctl_aliases[] = { | |
f117955a GP |
1566 | {"hardlockup_all_cpu_backtrace", "kernel.hardlockup_all_cpu_backtrace" }, |
1567 | {"hung_task_panic", "kernel.hung_task_panic" }, | |
1568 | {"numa_zonelist_order", "vm.numa_zonelist_order" }, | |
1569 | {"softlockup_all_cpu_backtrace", "kernel.softlockup_all_cpu_backtrace" }, | |
0a477e1a VB |
1570 | { } |
1571 | }; | |
1572 | ||
1573 | static const char *sysctl_find_alias(char *param) | |
1574 | { | |
1575 | const struct sysctl_alias *alias; | |
1576 | ||
1577 | for (alias = &sysctl_aliases[0]; alias->kernel_param != NULL; alias++) { | |
1578 | if (strcmp(alias->kernel_param, param) == 0) | |
1579 | return alias->sysctl_param; | |
1580 | } | |
1581 | ||
1582 | return NULL; | |
1583 | } | |
1584 | ||
8001f493 KJ |
1585 | bool sysctl_is_alias(char *param) |
1586 | { | |
1587 | const char *alias = sysctl_find_alias(param); | |
1588 | ||
1589 | return alias != NULL; | |
1590 | } | |
1591 | ||
3db978d4 VB |
1592 | /* Set sysctl value passed on kernel command line. */ |
1593 | static int process_sysctl_arg(char *param, char *val, | |
1594 | const char *unused, void *arg) | |
1595 | { | |
1596 | char *path; | |
1597 | struct vfsmount **proc_mnt = arg; | |
1598 | struct file_system_type *proc_fs_type; | |
1599 | struct file *file; | |
1600 | int len; | |
1601 | int err; | |
1602 | loff_t pos = 0; | |
1603 | ssize_t wret; | |
1604 | ||
0a477e1a VB |
1605 | if (strncmp(param, "sysctl", sizeof("sysctl") - 1) == 0) { |
1606 | param += sizeof("sysctl") - 1; | |
3db978d4 | 1607 | |
0a477e1a VB |
1608 | if (param[0] != '/' && param[0] != '.') |
1609 | return 0; | |
3db978d4 | 1610 | |
0a477e1a VB |
1611 | param++; |
1612 | } else { | |
1613 | param = (char *) sysctl_find_alias(param); | |
1614 | if (!param) | |
1615 | return 0; | |
1616 | } | |
3db978d4 | 1617 | |
697edcb0 XN |
1618 | if (!val) |
1619 | return -EINVAL; | |
1620 | len = strlen(val); | |
1621 | if (len == 0) | |
1622 | return -EINVAL; | |
1623 | ||
3db978d4 VB |
1624 | /* |
1625 | * To set sysctl options, we use a temporary mount of proc, look up the | |
1626 | * respective sys/ file and write to it. To avoid mounting it when no | |
1627 | * options were given, we mount it only when the first sysctl option is | |
1628 | * found. Why not a persistent mount? There are problems with a | |
1629 | * persistent mount of proc in that it forces userspace not to use any | |
1630 | * proc mount options. | |
1631 | */ | |
1632 | if (!*proc_mnt) { | |
1633 | proc_fs_type = get_fs_type("proc"); | |
1634 | if (!proc_fs_type) { | |
1635 | pr_err("Failed to find procfs to set sysctl from command line\n"); | |
1636 | return 0; | |
1637 | } | |
1638 | *proc_mnt = kern_mount(proc_fs_type); | |
1639 | put_filesystem(proc_fs_type); | |
1640 | if (IS_ERR(*proc_mnt)) { | |
1641 | pr_err("Failed to mount procfs to set sysctl from command line\n"); | |
1642 | return 0; | |
1643 | } | |
1644 | } | |
1645 | ||
1646 | path = kasprintf(GFP_KERNEL, "sys/%s", param); | |
1647 | if (!path) | |
1648 | panic("%s: Failed to allocate path for %s\n", __func__, param); | |
1649 | strreplace(path, '.', '/'); | |
1650 | ||
ffb37ca3 | 1651 | file = file_open_root_mnt(*proc_mnt, path, O_WRONLY, 0); |
3db978d4 VB |
1652 | if (IS_ERR(file)) { |
1653 | err = PTR_ERR(file); | |
1654 | if (err == -ENOENT) | |
1655 | pr_err("Failed to set sysctl parameter '%s=%s': parameter not found\n", | |
1656 | param, val); | |
1657 | else if (err == -EACCES) | |
1658 | pr_err("Failed to set sysctl parameter '%s=%s': permission denied (read-only?)\n", | |
1659 | param, val); | |
1660 | else | |
1661 | pr_err("Error %pe opening proc file to set sysctl parameter '%s=%s'\n", | |
1662 | file, param, val); | |
1663 | goto out; | |
1664 | } | |
3db978d4 VB |
1665 | wret = kernel_write(file, val, len, &pos); |
1666 | if (wret < 0) { | |
1667 | err = wret; | |
1668 | if (err == -EINVAL) | |
1669 | pr_err("Failed to set sysctl parameter '%s=%s': invalid value\n", | |
1670 | param, val); | |
1671 | else | |
1672 | pr_err("Error %pe writing to proc file to set sysctl parameter '%s=%s'\n", | |
1673 | ERR_PTR(err), param, val); | |
1674 | } else if (wret != len) { | |
1675 | pr_err("Wrote only %zd bytes of %d writing to proc file %s to set sysctl parameter '%s=%s\n", | |
1676 | wret, len, path, param, val); | |
1677 | } | |
1678 | ||
1679 | err = filp_close(file, NULL); | |
1680 | if (err) | |
1681 | pr_err("Error %pe closing proc file to set sysctl parameter '%s=%s\n", | |
1682 | ERR_PTR(err), param, val); | |
1683 | out: | |
1684 | kfree(path); | |
1685 | return 0; | |
1686 | } | |
1687 | ||
1688 | void do_sysctl_args(void) | |
1689 | { | |
1690 | char *command_line; | |
1691 | struct vfsmount *proc_mnt = NULL; | |
1692 | ||
1693 | command_line = kstrdup(saved_command_line, GFP_KERNEL); | |
1694 | if (!command_line) | |
1695 | panic("%s: Failed to allocate copy of command line\n", __func__); | |
1696 | ||
1697 | parse_args("Setting sysctl args", command_line, | |
1698 | NULL, 0, -1, -1, &proc_mnt, process_sysctl_arg); | |
1699 | ||
1700 | if (proc_mnt) | |
1701 | kern_unmount(proc_mnt); | |
1702 | ||
1703 | kfree(command_line); | |
1704 | } |