Commit | Line | Data |
---|---|---|
77b14db5 EB |
1 | /* |
2 | * /proc/sys support | |
3 | */ | |
1e0edd3f | 4 | #include <linux/init.h> |
77b14db5 | 5 | #include <linux/sysctl.h> |
f1ecf068 | 6 | #include <linux/poll.h> |
77b14db5 EB |
7 | #include <linux/proc_fs.h> |
8 | #include <linux/security.h> | |
40401530 | 9 | #include <linux/sched.h> |
34286d66 | 10 | #include <linux/namei.h> |
40401530 | 11 | #include <linux/mm.h> |
1f87f0b5 | 12 | #include <linux/module.h> |
77b14db5 EB |
13 | #include "internal.h" |
14 | ||
d72f71eb | 15 | static const struct dentry_operations proc_sys_dentry_operations; |
77b14db5 | 16 | static const struct file_operations proc_sys_file_operations; |
03a44825 | 17 | static const struct inode_operations proc_sys_inode_operations; |
9043476f AV |
18 | static const struct file_operations proc_sys_dir_file_operations; |
19 | static const struct inode_operations proc_sys_dir_operations; | |
77b14db5 | 20 | |
f1ecf068 LDM |
21 | void proc_sys_poll_notify(struct ctl_table_poll *poll) |
22 | { | |
23 | if (!poll) | |
24 | return; | |
25 | ||
26 | atomic_inc(&poll->event); | |
27 | wake_up_interruptible(&poll->wait); | |
28 | } | |
29 | ||
a194558e EB |
30 | static struct ctl_table root_table[] = { |
31 | { | |
32 | .procname = "", | |
7ec66d06 | 33 | .mode = S_IFDIR|S_IRUGO|S_IXUGO, |
a194558e EB |
34 | }, |
35 | { } | |
36 | }; | |
0e47c99d | 37 | static struct ctl_table_root sysctl_table_root = { |
0e47c99d | 38 | .default_set.dir.header = { |
7ec66d06 EB |
39 | {{.count = 1, |
40 | .nreg = 1, | |
ac13ac6f | 41 | .ctl_table = root_table }}, |
0e47c99d | 42 | .ctl_table_arg = root_table, |
7ec66d06 EB |
43 | .root = &sysctl_table_root, |
44 | .set = &sysctl_table_root.default_set, | |
45 | }, | |
1f87f0b5 | 46 | }; |
1f87f0b5 EB |
47 | |
48 | static DEFINE_SPINLOCK(sysctl_lock); | |
49 | ||
7ec66d06 | 50 | static void drop_sysctl_table(struct ctl_table_header *header); |
0e47c99d EB |
51 | static int sysctl_follow_link(struct ctl_table_header **phead, |
52 | struct ctl_table **pentry, struct nsproxy *namespaces); | |
53 | static int insert_links(struct ctl_table_header *head); | |
54 | static void put_links(struct ctl_table_header *header); | |
7ec66d06 | 55 | |
6980128f EB |
56 | static void sysctl_print_dir(struct ctl_dir *dir) |
57 | { | |
58 | if (dir->header.parent) | |
59 | sysctl_print_dir(dir->header.parent); | |
60 | printk(KERN_CONT "%s/", dir->header.ctl_table[0].procname); | |
61 | } | |
62 | ||
076c3eed EB |
63 | static int namecmp(const char *name1, int len1, const char *name2, int len2) |
64 | { | |
65 | int minlen; | |
66 | int cmp; | |
67 | ||
68 | minlen = len1; | |
69 | if (minlen > len2) | |
70 | minlen = len2; | |
71 | ||
72 | cmp = memcmp(name1, name2, minlen); | |
73 | if (cmp == 0) | |
74 | cmp = len1 - len2; | |
75 | return cmp; | |
76 | } | |
77 | ||
60f126d9 | 78 | /* Called under sysctl_lock */ |
076c3eed | 79 | static struct ctl_table *find_entry(struct ctl_table_header **phead, |
0e47c99d | 80 | struct ctl_dir *dir, const char *name, int namelen) |
076c3eed EB |
81 | { |
82 | struct ctl_table_header *head; | |
83 | struct ctl_table *entry; | |
ac13ac6f | 84 | struct rb_node *node = dir->root.rb_node; |
076c3eed | 85 | |
ac13ac6f EB |
86 | while (node) |
87 | { | |
88 | struct ctl_node *ctl_node; | |
89 | const char *procname; | |
90 | int cmp; | |
91 | ||
92 | ctl_node = rb_entry(node, struct ctl_node, node); | |
93 | head = ctl_node->header; | |
94 | entry = &head->ctl_table[ctl_node - head->node]; | |
95 | procname = entry->procname; | |
96 | ||
97 | cmp = namecmp(name, namelen, procname, strlen(procname)); | |
98 | if (cmp < 0) | |
99 | node = node->rb_left; | |
100 | else if (cmp > 0) | |
101 | node = node->rb_right; | |
102 | else { | |
103 | *phead = head; | |
104 | return entry; | |
076c3eed EB |
105 | } |
106 | } | |
107 | return NULL; | |
108 | } | |
109 | ||
ac13ac6f EB |
110 | static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry) |
111 | { | |
112 | struct rb_node *node = &head->node[entry - head->ctl_table].node; | |
113 | struct rb_node **p = &head->parent->root.rb_node; | |
114 | struct rb_node *parent = NULL; | |
115 | const char *name = entry->procname; | |
116 | int namelen = strlen(name); | |
117 | ||
118 | while (*p) { | |
119 | struct ctl_table_header *parent_head; | |
120 | struct ctl_table *parent_entry; | |
121 | struct ctl_node *parent_node; | |
122 | const char *parent_name; | |
123 | int cmp; | |
124 | ||
125 | parent = *p; | |
126 | parent_node = rb_entry(parent, struct ctl_node, node); | |
127 | parent_head = parent_node->header; | |
128 | parent_entry = &parent_head->ctl_table[parent_node - parent_head->node]; | |
129 | parent_name = parent_entry->procname; | |
130 | ||
131 | cmp = namecmp(name, namelen, parent_name, strlen(parent_name)); | |
132 | if (cmp < 0) | |
133 | p = &(*p)->rb_left; | |
134 | else if (cmp > 0) | |
135 | p = &(*p)->rb_right; | |
136 | else { | |
137 | printk(KERN_ERR "sysctl duplicate entry: "); | |
138 | sysctl_print_dir(head->parent); | |
139 | printk(KERN_CONT "/%s\n", entry->procname); | |
140 | return -EEXIST; | |
141 | } | |
142 | } | |
143 | ||
144 | rb_link_node(node, parent, p); | |
ea5272f5 | 145 | rb_insert_color(node, &head->parent->root); |
ac13ac6f EB |
146 | return 0; |
147 | } | |
148 | ||
149 | static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry) | |
150 | { | |
151 | struct rb_node *node = &head->node[entry - head->ctl_table].node; | |
152 | ||
153 | rb_erase(node, &head->parent->root); | |
154 | } | |
155 | ||
e0d04529 EB |
156 | static void init_header(struct ctl_table_header *head, |
157 | struct ctl_table_root *root, struct ctl_table_set *set, | |
ac13ac6f | 158 | struct ctl_node *node, struct ctl_table *table) |
e0d04529 | 159 | { |
7ec66d06 | 160 | head->ctl_table = table; |
e0d04529 | 161 | head->ctl_table_arg = table; |
e0d04529 EB |
162 | head->used = 0; |
163 | head->count = 1; | |
164 | head->nreg = 1; | |
165 | head->unregistering = NULL; | |
166 | head->root = root; | |
167 | head->set = set; | |
168 | head->parent = NULL; | |
ac13ac6f EB |
169 | head->node = node; |
170 | if (node) { | |
171 | struct ctl_table *entry; | |
4c199a93 | 172 | for (entry = table; entry->procname; entry++, node++) |
ac13ac6f | 173 | node->header = head; |
ac13ac6f | 174 | } |
e0d04529 EB |
175 | } |
176 | ||
8425d6aa EB |
177 | static void erase_header(struct ctl_table_header *head) |
178 | { | |
ac13ac6f EB |
179 | struct ctl_table *entry; |
180 | for (entry = head->ctl_table; entry->procname; entry++) | |
181 | erase_entry(head, entry); | |
8425d6aa EB |
182 | } |
183 | ||
0e47c99d | 184 | static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header) |
8425d6aa | 185 | { |
ac13ac6f | 186 | struct ctl_table *entry; |
0e47c99d EB |
187 | int err; |
188 | ||
189 | dir->header.nreg++; | |
7ec66d06 | 190 | header->parent = dir; |
0e47c99d EB |
191 | err = insert_links(header); |
192 | if (err) | |
193 | goto fail_links; | |
ac13ac6f EB |
194 | for (entry = header->ctl_table; entry->procname; entry++) { |
195 | err = insert_entry(header, entry); | |
196 | if (err) | |
197 | goto fail; | |
198 | } | |
0e47c99d | 199 | return 0; |
ac13ac6f EB |
200 | fail: |
201 | erase_header(header); | |
202 | put_links(header); | |
0e47c99d EB |
203 | fail_links: |
204 | header->parent = NULL; | |
205 | drop_sysctl_table(&dir->header); | |
206 | return err; | |
8425d6aa EB |
207 | } |
208 | ||
1f87f0b5 EB |
209 | /* called under sysctl_lock */ |
210 | static int use_table(struct ctl_table_header *p) | |
211 | { | |
212 | if (unlikely(p->unregistering)) | |
213 | return 0; | |
214 | p->used++; | |
215 | return 1; | |
216 | } | |
217 | ||
218 | /* called under sysctl_lock */ | |
219 | static void unuse_table(struct ctl_table_header *p) | |
220 | { | |
221 | if (!--p->used) | |
222 | if (unlikely(p->unregistering)) | |
223 | complete(p->unregistering); | |
224 | } | |
225 | ||
226 | /* called under sysctl_lock, will reacquire if has to wait */ | |
227 | static void start_unregistering(struct ctl_table_header *p) | |
228 | { | |
229 | /* | |
230 | * if p->used is 0, nobody will ever touch that entry again; | |
231 | * we'll eliminate all paths to it before dropping sysctl_lock | |
232 | */ | |
233 | if (unlikely(p->used)) { | |
234 | struct completion wait; | |
235 | init_completion(&wait); | |
236 | p->unregistering = &wait; | |
237 | spin_unlock(&sysctl_lock); | |
238 | wait_for_completion(&wait); | |
239 | spin_lock(&sysctl_lock); | |
240 | } else { | |
241 | /* anything non-NULL; we'll never dereference it */ | |
242 | p->unregistering = ERR_PTR(-EINVAL); | |
243 | } | |
244 | /* | |
245 | * do not remove from the list until nobody holds it; walking the | |
246 | * list in do_sysctl() relies on that. | |
247 | */ | |
8425d6aa | 248 | erase_header(p); |
1f87f0b5 EB |
249 | } |
250 | ||
251 | static void sysctl_head_get(struct ctl_table_header *head) | |
252 | { | |
253 | spin_lock(&sysctl_lock); | |
254 | head->count++; | |
255 | spin_unlock(&sysctl_lock); | |
256 | } | |
257 | ||
258 | void sysctl_head_put(struct ctl_table_header *head) | |
259 | { | |
260 | spin_lock(&sysctl_lock); | |
261 | if (!--head->count) | |
262 | kfree_rcu(head, rcu); | |
263 | spin_unlock(&sysctl_lock); | |
264 | } | |
265 | ||
266 | static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head) | |
267 | { | |
ab4a1f24 | 268 | BUG_ON(!head); |
1f87f0b5 EB |
269 | spin_lock(&sysctl_lock); |
270 | if (!use_table(head)) | |
271 | head = ERR_PTR(-ENOENT); | |
272 | spin_unlock(&sysctl_lock); | |
273 | return head; | |
274 | } | |
275 | ||
276 | static void sysctl_head_finish(struct ctl_table_header *head) | |
277 | { | |
278 | if (!head) | |
279 | return; | |
280 | spin_lock(&sysctl_lock); | |
281 | unuse_table(head); | |
282 | spin_unlock(&sysctl_lock); | |
283 | } | |
284 | ||
285 | static struct ctl_table_set * | |
286 | lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces) | |
287 | { | |
288 | struct ctl_table_set *set = &root->default_set; | |
289 | if (root->lookup) | |
290 | set = root->lookup(root, namespaces); | |
291 | return set; | |
292 | } | |
293 | ||
076c3eed | 294 | static struct ctl_table *lookup_entry(struct ctl_table_header **phead, |
7ec66d06 | 295 | struct ctl_dir *dir, |
076c3eed EB |
296 | const char *name, int namelen) |
297 | { | |
298 | struct ctl_table_header *head; | |
299 | struct ctl_table *entry; | |
076c3eed EB |
300 | |
301 | spin_lock(&sysctl_lock); | |
0e47c99d EB |
302 | entry = find_entry(&head, dir, name, namelen); |
303 | if (entry && use_table(head)) | |
304 | *phead = head; | |
305 | else | |
306 | entry = NULL; | |
076c3eed EB |
307 | spin_unlock(&sysctl_lock); |
308 | return entry; | |
309 | } | |
310 | ||
ac13ac6f | 311 | static struct ctl_node *first_usable_entry(struct rb_node *node) |
1f87f0b5 | 312 | { |
ac13ac6f | 313 | struct ctl_node *ctl_node; |
1f87f0b5 | 314 | |
ac13ac6f EB |
315 | for (;node; node = rb_next(node)) { |
316 | ctl_node = rb_entry(node, struct ctl_node, node); | |
317 | if (use_table(ctl_node->header)) | |
318 | return ctl_node; | |
1f87f0b5 | 319 | } |
1f87f0b5 EB |
320 | return NULL; |
321 | } | |
322 | ||
7ec66d06 | 323 | static void first_entry(struct ctl_dir *dir, |
6a75ce16 EB |
324 | struct ctl_table_header **phead, struct ctl_table **pentry) |
325 | { | |
ac13ac6f | 326 | struct ctl_table_header *head = NULL; |
7ec66d06 | 327 | struct ctl_table *entry = NULL; |
ac13ac6f | 328 | struct ctl_node *ctl_node; |
6a75ce16 EB |
329 | |
330 | spin_lock(&sysctl_lock); | |
ac13ac6f | 331 | ctl_node = first_usable_entry(rb_first(&dir->root)); |
6a75ce16 | 332 | spin_unlock(&sysctl_lock); |
ac13ac6f EB |
333 | if (ctl_node) { |
334 | head = ctl_node->header; | |
335 | entry = &head->ctl_table[ctl_node - head->node]; | |
336 | } | |
6a75ce16 EB |
337 | *phead = head; |
338 | *pentry = entry; | |
339 | } | |
340 | ||
7ec66d06 | 341 | static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry) |
1f87f0b5 | 342 | { |
6a75ce16 EB |
343 | struct ctl_table_header *head = *phead; |
344 | struct ctl_table *entry = *pentry; | |
ac13ac6f | 345 | struct ctl_node *ctl_node = &head->node[entry - head->ctl_table]; |
6a75ce16 | 346 | |
ac13ac6f EB |
347 | spin_lock(&sysctl_lock); |
348 | unuse_table(head); | |
349 | ||
350 | ctl_node = first_usable_entry(rb_next(&ctl_node->node)); | |
351 | spin_unlock(&sysctl_lock); | |
352 | head = NULL; | |
353 | if (ctl_node) { | |
354 | head = ctl_node->header; | |
355 | entry = &head->ctl_table[ctl_node - head->node]; | |
6a75ce16 EB |
356 | } |
357 | *phead = head; | |
358 | *pentry = entry; | |
1f87f0b5 EB |
359 | } |
360 | ||
361 | void register_sysctl_root(struct ctl_table_root *root) | |
362 | { | |
1f87f0b5 EB |
363 | } |
364 | ||
365 | /* | |
366 | * sysctl_perm does NOT grant the superuser all rights automatically, because | |
367 | * some sysctl variables are readonly even to root. | |
368 | */ | |
369 | ||
370 | static int test_perm(int mode, int op) | |
371 | { | |
091bd3ea | 372 | if (uid_eq(current_euid(), GLOBAL_ROOT_UID)) |
1f87f0b5 | 373 | mode >>= 6; |
091bd3ea | 374 | else if (in_egroup_p(GLOBAL_ROOT_GID)) |
1f87f0b5 EB |
375 | mode >>= 3; |
376 | if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0) | |
377 | return 0; | |
378 | return -EACCES; | |
379 | } | |
380 | ||
73f7ef43 | 381 | static int sysctl_perm(struct ctl_table_header *head, struct ctl_table *table, int op) |
1f87f0b5 | 382 | { |
73f7ef43 | 383 | struct ctl_table_root *root = head->root; |
1f87f0b5 EB |
384 | int mode; |
385 | ||
386 | if (root->permissions) | |
73f7ef43 | 387 | mode = root->permissions(head, table); |
1f87f0b5 EB |
388 | else |
389 | mode = table->mode; | |
390 | ||
391 | return test_perm(mode, op); | |
392 | } | |
393 | ||
9043476f AV |
394 | static struct inode *proc_sys_make_inode(struct super_block *sb, |
395 | struct ctl_table_header *head, struct ctl_table *table) | |
77b14db5 EB |
396 | { |
397 | struct inode *inode; | |
9043476f | 398 | struct proc_inode *ei; |
77b14db5 | 399 | |
9043476f | 400 | inode = new_inode(sb); |
77b14db5 EB |
401 | if (!inode) |
402 | goto out; | |
403 | ||
85fe4025 CH |
404 | inode->i_ino = get_next_ino(); |
405 | ||
9043476f | 406 | sysctl_head_get(head); |
77b14db5 | 407 | ei = PROC_I(inode); |
9043476f AV |
408 | ei->sysctl = head; |
409 | ei->sysctl_entry = table; | |
410 | ||
77b14db5 | 411 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; |
9043476f | 412 | inode->i_mode = table->mode; |
7ec66d06 | 413 | if (!S_ISDIR(table->mode)) { |
9043476f AV |
414 | inode->i_mode |= S_IFREG; |
415 | inode->i_op = &proc_sys_inode_operations; | |
416 | inode->i_fop = &proc_sys_file_operations; | |
417 | } else { | |
418 | inode->i_mode |= S_IFDIR; | |
9043476f AV |
419 | inode->i_op = &proc_sys_dir_operations; |
420 | inode->i_fop = &proc_sys_dir_file_operations; | |
421 | } | |
77b14db5 EB |
422 | out: |
423 | return inode; | |
424 | } | |
425 | ||
81324364 | 426 | static struct ctl_table_header *grab_header(struct inode *inode) |
77b14db5 | 427 | { |
3cc3e046 EB |
428 | struct ctl_table_header *head = PROC_I(inode)->sysctl; |
429 | if (!head) | |
0e47c99d | 430 | head = &sysctl_table_root.default_set.dir.header; |
3cc3e046 | 431 | return sysctl_head_grab(head); |
9043476f | 432 | } |
77b14db5 | 433 | |
9043476f | 434 | static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry, |
00cd8dd3 | 435 | unsigned int flags) |
9043476f AV |
436 | { |
437 | struct ctl_table_header *head = grab_header(dir); | |
9043476f AV |
438 | struct ctl_table_header *h = NULL; |
439 | struct qstr *name = &dentry->d_name; | |
440 | struct ctl_table *p; | |
441 | struct inode *inode; | |
442 | struct dentry *err = ERR_PTR(-ENOENT); | |
7ec66d06 | 443 | struct ctl_dir *ctl_dir; |
0e47c99d | 444 | int ret; |
77b14db5 | 445 | |
9043476f AV |
446 | if (IS_ERR(head)) |
447 | return ERR_CAST(head); | |
77b14db5 | 448 | |
7ec66d06 | 449 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 | 450 | |
7ec66d06 | 451 | p = lookup_entry(&h, ctl_dir, name->name, name->len); |
9043476f | 452 | if (!p) |
77b14db5 EB |
453 | goto out; |
454 | ||
4e757320 EB |
455 | if (S_ISLNK(p->mode)) { |
456 | ret = sysctl_follow_link(&h, &p, current->nsproxy); | |
457 | err = ERR_PTR(ret); | |
458 | if (ret) | |
459 | goto out; | |
460 | } | |
0e47c99d | 461 | |
77b14db5 | 462 | err = ERR_PTR(-ENOMEM); |
9043476f | 463 | inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p); |
77b14db5 EB |
464 | if (!inode) |
465 | goto out; | |
466 | ||
467 | err = NULL; | |
fb045adb | 468 | d_set_d_op(dentry, &proc_sys_dentry_operations); |
77b14db5 EB |
469 | d_add(dentry, inode); |
470 | ||
471 | out: | |
6bf61045 FR |
472 | if (h) |
473 | sysctl_head_finish(h); | |
77b14db5 EB |
474 | sysctl_head_finish(head); |
475 | return err; | |
476 | } | |
477 | ||
7708bfb1 PE |
478 | static ssize_t proc_sys_call_handler(struct file *filp, void __user *buf, |
479 | size_t count, loff_t *ppos, int write) | |
77b14db5 | 480 | { |
9043476f AV |
481 | struct inode *inode = filp->f_path.dentry->d_inode; |
482 | struct ctl_table_header *head = grab_header(inode); | |
483 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
2a2da53b DH |
484 | ssize_t error; |
485 | size_t res; | |
77b14db5 | 486 | |
9043476f AV |
487 | if (IS_ERR(head)) |
488 | return PTR_ERR(head); | |
77b14db5 EB |
489 | |
490 | /* | |
491 | * At this point we know that the sysctl was not unregistered | |
492 | * and won't be until we finish. | |
493 | */ | |
494 | error = -EPERM; | |
73f7ef43 | 495 | if (sysctl_perm(head, table, write ? MAY_WRITE : MAY_READ)) |
77b14db5 EB |
496 | goto out; |
497 | ||
9043476f AV |
498 | /* if that can happen at all, it should be -EINVAL, not -EISDIR */ |
499 | error = -EINVAL; | |
500 | if (!table->proc_handler) | |
501 | goto out; | |
502 | ||
77b14db5 EB |
503 | /* careful: calling conventions are nasty here */ |
504 | res = count; | |
8d65af78 | 505 | error = table->proc_handler(table, write, buf, &res, ppos); |
77b14db5 EB |
506 | if (!error) |
507 | error = res; | |
508 | out: | |
509 | sysctl_head_finish(head); | |
510 | ||
511 | return error; | |
512 | } | |
513 | ||
7708bfb1 | 514 | static ssize_t proc_sys_read(struct file *filp, char __user *buf, |
77b14db5 EB |
515 | size_t count, loff_t *ppos) |
516 | { | |
7708bfb1 PE |
517 | return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 0); |
518 | } | |
77b14db5 | 519 | |
7708bfb1 PE |
520 | static ssize_t proc_sys_write(struct file *filp, const char __user *buf, |
521 | size_t count, loff_t *ppos) | |
522 | { | |
523 | return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 1); | |
77b14db5 EB |
524 | } |
525 | ||
f1ecf068 LDM |
526 | static int proc_sys_open(struct inode *inode, struct file *filp) |
527 | { | |
4e474a00 | 528 | struct ctl_table_header *head = grab_header(inode); |
f1ecf068 LDM |
529 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; |
530 | ||
4e474a00 LDM |
531 | /* sysctl was unregistered */ |
532 | if (IS_ERR(head)) | |
533 | return PTR_ERR(head); | |
534 | ||
f1ecf068 LDM |
535 | if (table->poll) |
536 | filp->private_data = proc_sys_poll_event(table->poll); | |
537 | ||
4e474a00 LDM |
538 | sysctl_head_finish(head); |
539 | ||
f1ecf068 LDM |
540 | return 0; |
541 | } | |
542 | ||
543 | static unsigned int proc_sys_poll(struct file *filp, poll_table *wait) | |
544 | { | |
545 | struct inode *inode = filp->f_path.dentry->d_inode; | |
4e474a00 | 546 | struct ctl_table_header *head = grab_header(inode); |
f1ecf068 | 547 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; |
f1ecf068 | 548 | unsigned int ret = DEFAULT_POLLMASK; |
4e474a00 LDM |
549 | unsigned long event; |
550 | ||
551 | /* sysctl was unregistered */ | |
552 | if (IS_ERR(head)) | |
553 | return POLLERR | POLLHUP; | |
f1ecf068 LDM |
554 | |
555 | if (!table->proc_handler) | |
556 | goto out; | |
557 | ||
558 | if (!table->poll) | |
559 | goto out; | |
560 | ||
4e474a00 | 561 | event = (unsigned long)filp->private_data; |
f1ecf068 LDM |
562 | poll_wait(filp, &table->poll->wait, wait); |
563 | ||
564 | if (event != atomic_read(&table->poll->event)) { | |
565 | filp->private_data = proc_sys_poll_event(table->poll); | |
566 | ret = POLLIN | POLLRDNORM | POLLERR | POLLPRI; | |
567 | } | |
568 | ||
569 | out: | |
4e474a00 LDM |
570 | sysctl_head_finish(head); |
571 | ||
f1ecf068 LDM |
572 | return ret; |
573 | } | |
77b14db5 EB |
574 | |
575 | static int proc_sys_fill_cache(struct file *filp, void *dirent, | |
9043476f AV |
576 | filldir_t filldir, |
577 | struct ctl_table_header *head, | |
578 | struct ctl_table *table) | |
77b14db5 | 579 | { |
77b14db5 EB |
580 | struct dentry *child, *dir = filp->f_path.dentry; |
581 | struct inode *inode; | |
582 | struct qstr qname; | |
583 | ino_t ino = 0; | |
584 | unsigned type = DT_UNKNOWN; | |
77b14db5 EB |
585 | |
586 | qname.name = table->procname; | |
587 | qname.len = strlen(table->procname); | |
588 | qname.hash = full_name_hash(qname.name, qname.len); | |
589 | ||
77b14db5 EB |
590 | child = d_lookup(dir, &qname); |
591 | if (!child) { | |
9043476f AV |
592 | child = d_alloc(dir, &qname); |
593 | if (child) { | |
594 | inode = proc_sys_make_inode(dir->d_sb, head, table); | |
595 | if (!inode) { | |
596 | dput(child); | |
597 | return -ENOMEM; | |
598 | } else { | |
fb045adb | 599 | d_set_d_op(child, &proc_sys_dentry_operations); |
9043476f | 600 | d_add(child, inode); |
77b14db5 | 601 | } |
9043476f AV |
602 | } else { |
603 | return -ENOMEM; | |
77b14db5 EB |
604 | } |
605 | } | |
77b14db5 | 606 | inode = child->d_inode; |
9043476f AV |
607 | ino = inode->i_ino; |
608 | type = inode->i_mode >> 12; | |
77b14db5 | 609 | dput(child); |
9043476f AV |
610 | return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type); |
611 | } | |
612 | ||
0e47c99d EB |
613 | static int proc_sys_link_fill_cache(struct file *filp, void *dirent, |
614 | filldir_t filldir, | |
615 | struct ctl_table_header *head, | |
616 | struct ctl_table *table) | |
617 | { | |
618 | int err, ret = 0; | |
619 | head = sysctl_head_grab(head); | |
620 | ||
4e757320 EB |
621 | if (S_ISLNK(table->mode)) { |
622 | /* It is not an error if we can not follow the link ignore it */ | |
623 | err = sysctl_follow_link(&head, &table, current->nsproxy); | |
624 | if (err) | |
625 | goto out; | |
626 | } | |
0e47c99d EB |
627 | |
628 | ret = proc_sys_fill_cache(filp, dirent, filldir, head, table); | |
629 | out: | |
630 | sysctl_head_finish(head); | |
631 | return ret; | |
632 | } | |
633 | ||
9043476f AV |
634 | static int scan(struct ctl_table_header *head, ctl_table *table, |
635 | unsigned long *pos, struct file *file, | |
636 | void *dirent, filldir_t filldir) | |
637 | { | |
6a75ce16 | 638 | int res; |
9043476f | 639 | |
6a75ce16 EB |
640 | if ((*pos)++ < file->f_pos) |
641 | return 0; | |
9043476f | 642 | |
0e47c99d EB |
643 | if (unlikely(S_ISLNK(table->mode))) |
644 | res = proc_sys_link_fill_cache(file, dirent, filldir, head, table); | |
645 | else | |
9043476f | 646 | res = proc_sys_fill_cache(file, dirent, filldir, head, table); |
9043476f | 647 | |
6a75ce16 EB |
648 | if (res == 0) |
649 | file->f_pos = *pos; | |
9043476f | 650 | |
6a75ce16 | 651 | return res; |
77b14db5 EB |
652 | } |
653 | ||
654 | static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir) | |
655 | { | |
9043476f | 656 | struct dentry *dentry = filp->f_path.dentry; |
77b14db5 | 657 | struct inode *inode = dentry->d_inode; |
9043476f | 658 | struct ctl_table_header *head = grab_header(inode); |
9043476f | 659 | struct ctl_table_header *h = NULL; |
6a75ce16 | 660 | struct ctl_table *entry; |
7ec66d06 | 661 | struct ctl_dir *ctl_dir; |
77b14db5 | 662 | unsigned long pos; |
9043476f AV |
663 | int ret = -EINVAL; |
664 | ||
665 | if (IS_ERR(head)) | |
666 | return PTR_ERR(head); | |
77b14db5 | 667 | |
7ec66d06 | 668 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 EB |
669 | |
670 | ret = 0; | |
671 | /* Avoid a switch here: arm builds fail with missing __cmpdi2 */ | |
672 | if (filp->f_pos == 0) { | |
673 | if (filldir(dirent, ".", 1, filp->f_pos, | |
674 | inode->i_ino, DT_DIR) < 0) | |
675 | goto out; | |
676 | filp->f_pos++; | |
677 | } | |
678 | if (filp->f_pos == 1) { | |
679 | if (filldir(dirent, "..", 2, filp->f_pos, | |
680 | parent_ino(dentry), DT_DIR) < 0) | |
681 | goto out; | |
682 | filp->f_pos++; | |
683 | } | |
684 | pos = 2; | |
685 | ||
7ec66d06 | 686 | for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) { |
6a75ce16 | 687 | ret = scan(h, entry, &pos, filp, dirent, filldir); |
9043476f AV |
688 | if (ret) { |
689 | sysctl_head_finish(h); | |
690 | break; | |
77b14db5 EB |
691 | } |
692 | } | |
693 | ret = 1; | |
694 | out: | |
695 | sysctl_head_finish(head); | |
696 | return ret; | |
697 | } | |
698 | ||
10556cb2 | 699 | static int proc_sys_permission(struct inode *inode, int mask) |
77b14db5 EB |
700 | { |
701 | /* | |
702 | * sysctl entries that are not writeable, | |
703 | * are _NOT_ writeable, capabilities or not. | |
704 | */ | |
f696a365 MS |
705 | struct ctl_table_header *head; |
706 | struct ctl_table *table; | |
77b14db5 EB |
707 | int error; |
708 | ||
f696a365 MS |
709 | /* Executable files are not allowed under /proc/sys/ */ |
710 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) | |
711 | return -EACCES; | |
712 | ||
713 | head = grab_header(inode); | |
9043476f AV |
714 | if (IS_ERR(head)) |
715 | return PTR_ERR(head); | |
77b14db5 | 716 | |
f696a365 | 717 | table = PROC_I(inode)->sysctl_entry; |
9043476f AV |
718 | if (!table) /* global root - r-xr-xr-x */ |
719 | error = mask & MAY_WRITE ? -EACCES : 0; | |
720 | else /* Use the permissions on the sysctl table entry */ | |
73f7ef43 | 721 | error = sysctl_perm(head, table, mask & ~MAY_NOT_BLOCK); |
77b14db5 | 722 | |
77b14db5 EB |
723 | sysctl_head_finish(head); |
724 | return error; | |
725 | } | |
726 | ||
727 | static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr) | |
728 | { | |
729 | struct inode *inode = dentry->d_inode; | |
730 | int error; | |
731 | ||
732 | if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) | |
733 | return -EPERM; | |
734 | ||
735 | error = inode_change_ok(inode, attr); | |
1025774c CH |
736 | if (error) |
737 | return error; | |
738 | ||
1025774c CH |
739 | setattr_copy(inode, attr); |
740 | mark_inode_dirty(inode); | |
741 | return 0; | |
77b14db5 EB |
742 | } |
743 | ||
9043476f AV |
744 | static int proc_sys_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) |
745 | { | |
746 | struct inode *inode = dentry->d_inode; | |
747 | struct ctl_table_header *head = grab_header(inode); | |
748 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
749 | ||
750 | if (IS_ERR(head)) | |
751 | return PTR_ERR(head); | |
752 | ||
753 | generic_fillattr(inode, stat); | |
754 | if (table) | |
755 | stat->mode = (stat->mode & S_IFMT) | table->mode; | |
756 | ||
757 | sysctl_head_finish(head); | |
758 | return 0; | |
759 | } | |
760 | ||
77b14db5 | 761 | static const struct file_operations proc_sys_file_operations = { |
f1ecf068 LDM |
762 | .open = proc_sys_open, |
763 | .poll = proc_sys_poll, | |
77b14db5 EB |
764 | .read = proc_sys_read, |
765 | .write = proc_sys_write, | |
6038f373 | 766 | .llseek = default_llseek, |
9043476f AV |
767 | }; |
768 | ||
769 | static const struct file_operations proc_sys_dir_file_operations = { | |
887df078 | 770 | .read = generic_read_dir, |
77b14db5 | 771 | .readdir = proc_sys_readdir, |
3222a3e5 | 772 | .llseek = generic_file_llseek, |
77b14db5 EB |
773 | }; |
774 | ||
03a44825 | 775 | static const struct inode_operations proc_sys_inode_operations = { |
9043476f AV |
776 | .permission = proc_sys_permission, |
777 | .setattr = proc_sys_setattr, | |
778 | .getattr = proc_sys_getattr, | |
779 | }; | |
780 | ||
781 | static const struct inode_operations proc_sys_dir_operations = { | |
77b14db5 EB |
782 | .lookup = proc_sys_lookup, |
783 | .permission = proc_sys_permission, | |
784 | .setattr = proc_sys_setattr, | |
9043476f | 785 | .getattr = proc_sys_getattr, |
77b14db5 EB |
786 | }; |
787 | ||
0b728e19 | 788 | static int proc_sys_revalidate(struct dentry *dentry, unsigned int flags) |
77b14db5 | 789 | { |
0b728e19 | 790 | if (flags & LOOKUP_RCU) |
34286d66 | 791 | return -ECHILD; |
9043476f AV |
792 | return !PROC_I(dentry->d_inode)->sysctl->unregistering; |
793 | } | |
794 | ||
fe15ce44 | 795 | static int proc_sys_delete(const struct dentry *dentry) |
9043476f AV |
796 | { |
797 | return !!PROC_I(dentry->d_inode)->sysctl->unregistering; | |
798 | } | |
799 | ||
1f87f0b5 EB |
800 | static int sysctl_is_seen(struct ctl_table_header *p) |
801 | { | |
802 | struct ctl_table_set *set = p->set; | |
803 | int res; | |
804 | spin_lock(&sysctl_lock); | |
805 | if (p->unregistering) | |
806 | res = 0; | |
807 | else if (!set->is_seen) | |
808 | res = 1; | |
809 | else | |
810 | res = set->is_seen(set); | |
811 | spin_unlock(&sysctl_lock); | |
812 | return res; | |
813 | } | |
814 | ||
621e155a NP |
815 | static int proc_sys_compare(const struct dentry *parent, |
816 | const struct inode *pinode, | |
817 | const struct dentry *dentry, const struct inode *inode, | |
818 | unsigned int len, const char *str, const struct qstr *name) | |
9043476f | 819 | { |
dfef6dcd | 820 | struct ctl_table_header *head; |
31e6b01f NP |
821 | /* Although proc doesn't have negative dentries, rcu-walk means |
822 | * that inode here can be NULL */ | |
dfef6dcd | 823 | /* AV: can it, indeed? */ |
31e6b01f | 824 | if (!inode) |
dfef6dcd | 825 | return 1; |
621e155a | 826 | if (name->len != len) |
9043476f | 827 | return 1; |
621e155a | 828 | if (memcmp(name->name, str, len)) |
9043476f | 829 | return 1; |
dfef6dcd AV |
830 | head = rcu_dereference(PROC_I(inode)->sysctl); |
831 | return !head || !sysctl_is_seen(head); | |
77b14db5 EB |
832 | } |
833 | ||
d72f71eb | 834 | static const struct dentry_operations proc_sys_dentry_operations = { |
77b14db5 | 835 | .d_revalidate = proc_sys_revalidate, |
9043476f AV |
836 | .d_delete = proc_sys_delete, |
837 | .d_compare = proc_sys_compare, | |
77b14db5 EB |
838 | }; |
839 | ||
0e47c99d EB |
840 | static struct ctl_dir *find_subdir(struct ctl_dir *dir, |
841 | const char *name, int namelen) | |
1f87f0b5 | 842 | { |
7ec66d06 EB |
843 | struct ctl_table_header *head; |
844 | struct ctl_table *entry; | |
1f87f0b5 | 845 | |
0e47c99d | 846 | entry = find_entry(&head, dir, name, namelen); |
7ec66d06 EB |
847 | if (!entry) |
848 | return ERR_PTR(-ENOENT); | |
51f72f4a EB |
849 | if (!S_ISDIR(entry->mode)) |
850 | return ERR_PTR(-ENOTDIR); | |
851 | return container_of(head, struct ctl_dir, header); | |
7ec66d06 EB |
852 | } |
853 | ||
854 | static struct ctl_dir *new_dir(struct ctl_table_set *set, | |
0e47c99d | 855 | const char *name, int namelen) |
7ec66d06 EB |
856 | { |
857 | struct ctl_table *table; | |
858 | struct ctl_dir *new; | |
ac13ac6f | 859 | struct ctl_node *node; |
7ec66d06 | 860 | char *new_name; |
1f87f0b5 | 861 | |
ac13ac6f EB |
862 | new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) + |
863 | sizeof(struct ctl_table)*2 + namelen + 1, | |
864 | GFP_KERNEL); | |
7ec66d06 | 865 | if (!new) |
1f87f0b5 EB |
866 | return NULL; |
867 | ||
ac13ac6f EB |
868 | node = (struct ctl_node *)(new + 1); |
869 | table = (struct ctl_table *)(node + 1); | |
7ec66d06 EB |
870 | new_name = (char *)(table + 2); |
871 | memcpy(new_name, name, namelen); | |
872 | new_name[namelen] = '\0'; | |
873 | table[0].procname = new_name; | |
874 | table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
ac13ac6f | 875 | init_header(&new->header, set->dir.header.root, set, node, table); |
7ec66d06 EB |
876 | |
877 | return new; | |
1f87f0b5 EB |
878 | } |
879 | ||
60f126d9 EB |
880 | /** |
881 | * get_subdir - find or create a subdir with the specified name. | |
882 | * @dir: Directory to create the subdirectory in | |
883 | * @name: The name of the subdirectory to find or create | |
884 | * @namelen: The length of name | |
885 | * | |
886 | * Takes a directory with an elevated reference count so we know that | |
887 | * if we drop the lock the directory will not go away. Upon success | |
888 | * the reference is moved from @dir to the returned subdirectory. | |
889 | * Upon error an error code is returned and the reference on @dir is | |
890 | * simply dropped. | |
891 | */ | |
0e47c99d EB |
892 | static struct ctl_dir *get_subdir(struct ctl_dir *dir, |
893 | const char *name, int namelen) | |
1f87f0b5 | 894 | { |
0e47c99d | 895 | struct ctl_table_set *set = dir->header.set; |
7ec66d06 | 896 | struct ctl_dir *subdir, *new = NULL; |
0eb97f38 | 897 | int err; |
1f87f0b5 | 898 | |
7ec66d06 | 899 | spin_lock(&sysctl_lock); |
0e47c99d | 900 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
901 | if (!IS_ERR(subdir)) |
902 | goto found; | |
903 | if (PTR_ERR(subdir) != -ENOENT) | |
904 | goto failed; | |
905 | ||
906 | spin_unlock(&sysctl_lock); | |
907 | new = new_dir(set, name, namelen); | |
908 | spin_lock(&sysctl_lock); | |
909 | subdir = ERR_PTR(-ENOMEM); | |
910 | if (!new) | |
911 | goto failed; | |
912 | ||
60f126d9 | 913 | /* Was the subdir added while we dropped the lock? */ |
0e47c99d | 914 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
915 | if (!IS_ERR(subdir)) |
916 | goto found; | |
917 | if (PTR_ERR(subdir) != -ENOENT) | |
918 | goto failed; | |
919 | ||
60f126d9 | 920 | /* Nope. Use the our freshly made directory entry. */ |
0eb97f38 EB |
921 | err = insert_header(dir, &new->header); |
922 | subdir = ERR_PTR(err); | |
923 | if (err) | |
0e47c99d | 924 | goto failed; |
7ec66d06 EB |
925 | subdir = new; |
926 | found: | |
927 | subdir->header.nreg++; | |
928 | failed: | |
929 | if (unlikely(IS_ERR(subdir))) { | |
6980128f EB |
930 | printk(KERN_ERR "sysctl could not get directory: "); |
931 | sysctl_print_dir(dir); | |
932 | printk(KERN_CONT "/%*.*s %ld\n", | |
7ec66d06 | 933 | namelen, namelen, name, PTR_ERR(subdir)); |
1f87f0b5 | 934 | } |
7ec66d06 EB |
935 | drop_sysctl_table(&dir->header); |
936 | if (new) | |
937 | drop_sysctl_table(&new->header); | |
938 | spin_unlock(&sysctl_lock); | |
939 | return subdir; | |
1f87f0b5 EB |
940 | } |
941 | ||
0e47c99d EB |
942 | static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir) |
943 | { | |
944 | struct ctl_dir *parent; | |
945 | const char *procname; | |
946 | if (!dir->header.parent) | |
947 | return &set->dir; | |
948 | parent = xlate_dir(set, dir->header.parent); | |
949 | if (IS_ERR(parent)) | |
950 | return parent; | |
951 | procname = dir->header.ctl_table[0].procname; | |
952 | return find_subdir(parent, procname, strlen(procname)); | |
953 | } | |
954 | ||
955 | static int sysctl_follow_link(struct ctl_table_header **phead, | |
956 | struct ctl_table **pentry, struct nsproxy *namespaces) | |
957 | { | |
958 | struct ctl_table_header *head; | |
959 | struct ctl_table_root *root; | |
960 | struct ctl_table_set *set; | |
961 | struct ctl_table *entry; | |
962 | struct ctl_dir *dir; | |
963 | int ret; | |
964 | ||
0e47c99d EB |
965 | ret = 0; |
966 | spin_lock(&sysctl_lock); | |
967 | root = (*pentry)->data; | |
968 | set = lookup_header_set(root, namespaces); | |
969 | dir = xlate_dir(set, (*phead)->parent); | |
970 | if (IS_ERR(dir)) | |
971 | ret = PTR_ERR(dir); | |
972 | else { | |
973 | const char *procname = (*pentry)->procname; | |
974 | head = NULL; | |
975 | entry = find_entry(&head, dir, procname, strlen(procname)); | |
976 | ret = -ENOENT; | |
977 | if (entry && use_table(head)) { | |
978 | unuse_table(*phead); | |
979 | *phead = head; | |
980 | *pentry = entry; | |
981 | ret = 0; | |
982 | } | |
983 | } | |
984 | ||
985 | spin_unlock(&sysctl_lock); | |
986 | return ret; | |
987 | } | |
988 | ||
7c60c48f | 989 | static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...) |
1f87f0b5 | 990 | { |
7c60c48f EB |
991 | struct va_format vaf; |
992 | va_list args; | |
1f87f0b5 | 993 | |
7c60c48f EB |
994 | va_start(args, fmt); |
995 | vaf.fmt = fmt; | |
996 | vaf.va = &args; | |
997 | ||
998 | printk(KERN_ERR "sysctl table check failed: %s/%s %pV\n", | |
999 | path, table->procname, &vaf); | |
1f87f0b5 | 1000 | |
7c60c48f EB |
1001 | va_end(args); |
1002 | return -EINVAL; | |
1f87f0b5 EB |
1003 | } |
1004 | ||
7c60c48f | 1005 | static int sysctl_check_table(const char *path, struct ctl_table *table) |
1f87f0b5 | 1006 | { |
7c60c48f | 1007 | int err = 0; |
1f87f0b5 | 1008 | for (; table->procname; table++) { |
1f87f0b5 | 1009 | if (table->child) |
7c60c48f EB |
1010 | err = sysctl_err(path, table, "Not a file"); |
1011 | ||
1012 | if ((table->proc_handler == proc_dostring) || | |
1013 | (table->proc_handler == proc_dointvec) || | |
1014 | (table->proc_handler == proc_dointvec_minmax) || | |
1015 | (table->proc_handler == proc_dointvec_jiffies) || | |
1016 | (table->proc_handler == proc_dointvec_userhz_jiffies) || | |
1017 | (table->proc_handler == proc_dointvec_ms_jiffies) || | |
1018 | (table->proc_handler == proc_doulongvec_minmax) || | |
1019 | (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) { | |
1020 | if (!table->data) | |
1021 | err = sysctl_err(path, table, "No data"); | |
1022 | if (!table->maxlen) | |
1023 | err = sysctl_err(path, table, "No maxlen"); | |
1024 | } | |
1025 | if (!table->proc_handler) | |
1026 | err = sysctl_err(path, table, "No proc_handler"); | |
1027 | ||
1028 | if ((table->mode & (S_IRUGO|S_IWUGO)) != table->mode) | |
1029 | err = sysctl_err(path, table, "bogus .mode 0%o", | |
1030 | table->mode); | |
1f87f0b5 | 1031 | } |
7c60c48f | 1032 | return err; |
1f87f0b5 | 1033 | } |
1f87f0b5 | 1034 | |
0e47c99d EB |
1035 | static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table *table, |
1036 | struct ctl_table_root *link_root) | |
1037 | { | |
1038 | struct ctl_table *link_table, *entry, *link; | |
1039 | struct ctl_table_header *links; | |
ac13ac6f | 1040 | struct ctl_node *node; |
0e47c99d EB |
1041 | char *link_name; |
1042 | int nr_entries, name_bytes; | |
1043 | ||
1044 | name_bytes = 0; | |
1045 | nr_entries = 0; | |
1046 | for (entry = table; entry->procname; entry++) { | |
1047 | nr_entries++; | |
1048 | name_bytes += strlen(entry->procname) + 1; | |
1049 | } | |
1050 | ||
1051 | links = kzalloc(sizeof(struct ctl_table_header) + | |
ac13ac6f | 1052 | sizeof(struct ctl_node)*nr_entries + |
0e47c99d EB |
1053 | sizeof(struct ctl_table)*(nr_entries + 1) + |
1054 | name_bytes, | |
1055 | GFP_KERNEL); | |
1056 | ||
1057 | if (!links) | |
1058 | return NULL; | |
1059 | ||
ac13ac6f EB |
1060 | node = (struct ctl_node *)(links + 1); |
1061 | link_table = (struct ctl_table *)(node + nr_entries); | |
0e47c99d EB |
1062 | link_name = (char *)&link_table[nr_entries + 1]; |
1063 | ||
1064 | for (link = link_table, entry = table; entry->procname; link++, entry++) { | |
1065 | int len = strlen(entry->procname) + 1; | |
1066 | memcpy(link_name, entry->procname, len); | |
1067 | link->procname = link_name; | |
1068 | link->mode = S_IFLNK|S_IRWXUGO; | |
1069 | link->data = link_root; | |
1070 | link_name += len; | |
1071 | } | |
ac13ac6f | 1072 | init_header(links, dir->header.root, dir->header.set, node, link_table); |
0e47c99d EB |
1073 | links->nreg = nr_entries; |
1074 | ||
1075 | return links; | |
1076 | } | |
1077 | ||
1078 | static bool get_links(struct ctl_dir *dir, | |
1079 | struct ctl_table *table, struct ctl_table_root *link_root) | |
1080 | { | |
1081 | struct ctl_table_header *head; | |
1082 | struct ctl_table *entry, *link; | |
1083 | ||
1084 | /* Are there links available for every entry in table? */ | |
1085 | for (entry = table; entry->procname; entry++) { | |
1086 | const char *procname = entry->procname; | |
1087 | link = find_entry(&head, dir, procname, strlen(procname)); | |
1088 | if (!link) | |
1089 | return false; | |
1090 | if (S_ISDIR(link->mode) && S_ISDIR(entry->mode)) | |
1091 | continue; | |
1092 | if (S_ISLNK(link->mode) && (link->data == link_root)) | |
1093 | continue; | |
1094 | return false; | |
1095 | } | |
1096 | ||
1097 | /* The checks passed. Increase the registration count on the links */ | |
1098 | for (entry = table; entry->procname; entry++) { | |
1099 | const char *procname = entry->procname; | |
1100 | link = find_entry(&head, dir, procname, strlen(procname)); | |
1101 | head->nreg++; | |
1102 | } | |
1103 | return true; | |
1104 | } | |
1105 | ||
1106 | static int insert_links(struct ctl_table_header *head) | |
1107 | { | |
1108 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
1109 | struct ctl_dir *core_parent = NULL; | |
1110 | struct ctl_table_header *links; | |
1111 | int err; | |
1112 | ||
1113 | if (head->set == root_set) | |
1114 | return 0; | |
1115 | ||
1116 | core_parent = xlate_dir(root_set, head->parent); | |
1117 | if (IS_ERR(core_parent)) | |
1118 | return 0; | |
1119 | ||
1120 | if (get_links(core_parent, head->ctl_table, head->root)) | |
1121 | return 0; | |
1122 | ||
1123 | core_parent->header.nreg++; | |
1124 | spin_unlock(&sysctl_lock); | |
1125 | ||
1126 | links = new_links(core_parent, head->ctl_table, head->root); | |
1127 | ||
1128 | spin_lock(&sysctl_lock); | |
1129 | err = -ENOMEM; | |
1130 | if (!links) | |
1131 | goto out; | |
1132 | ||
1133 | err = 0; | |
1134 | if (get_links(core_parent, head->ctl_table, head->root)) { | |
1135 | kfree(links); | |
1136 | goto out; | |
1137 | } | |
1138 | ||
1139 | err = insert_header(core_parent, links); | |
1140 | if (err) | |
1141 | kfree(links); | |
1142 | out: | |
1143 | drop_sysctl_table(&core_parent->header); | |
1144 | return err; | |
1145 | } | |
1146 | ||
1f87f0b5 | 1147 | /** |
f728019b | 1148 | * __register_sysctl_table - register a leaf sysctl table |
60a47a2e | 1149 | * @set: Sysctl tree to register on |
1f87f0b5 EB |
1150 | * @path: The path to the directory the sysctl table is in. |
1151 | * @table: the top-level table structure | |
1152 | * | |
1153 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1154 | * array. A completely 0 filled entry terminates the table. | |
1155 | * | |
1156 | * The members of the &struct ctl_table structure are used as follows: | |
1157 | * | |
1158 | * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not | |
1159 | * enter a sysctl file | |
1160 | * | |
1161 | * data - a pointer to data for use by proc_handler | |
1162 | * | |
1163 | * maxlen - the maximum size in bytes of the data | |
1164 | * | |
f728019b | 1165 | * mode - the file permissions for the /proc/sys file |
1f87f0b5 | 1166 | * |
f728019b | 1167 | * child - must be %NULL. |
1f87f0b5 EB |
1168 | * |
1169 | * proc_handler - the text handler routine (described below) | |
1170 | * | |
1f87f0b5 EB |
1171 | * extra1, extra2 - extra pointers usable by the proc handler routines |
1172 | * | |
1173 | * Leaf nodes in the sysctl tree will be represented by a single file | |
1174 | * under /proc; non-leaf nodes will be represented by directories. | |
1175 | * | |
f728019b EB |
1176 | * There must be a proc_handler routine for any terminal nodes. |
1177 | * Several default handlers are available to cover common cases - | |
1f87f0b5 EB |
1178 | * |
1179 | * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), | |
1180 | * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), | |
1181 | * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() | |
1182 | * | |
1183 | * It is the handler's job to read the input buffer from user memory | |
1184 | * and process it. The handler should return 0 on success. | |
1185 | * | |
1186 | * This routine returns %NULL on a failure to register, and a pointer | |
1187 | * to the table header on success. | |
1188 | */ | |
6e9d5164 | 1189 | struct ctl_table_header *__register_sysctl_table( |
60a47a2e | 1190 | struct ctl_table_set *set, |
6e9d5164 | 1191 | const char *path, struct ctl_table *table) |
1f87f0b5 | 1192 | { |
60a47a2e | 1193 | struct ctl_table_root *root = set->dir.header.root; |
1f87f0b5 | 1194 | struct ctl_table_header *header; |
6e9d5164 | 1195 | const char *name, *nextname; |
7ec66d06 | 1196 | struct ctl_dir *dir; |
ac13ac6f EB |
1197 | struct ctl_table *entry; |
1198 | struct ctl_node *node; | |
1199 | int nr_entries = 0; | |
1200 | ||
1201 | for (entry = table; entry->procname; entry++) | |
1202 | nr_entries++; | |
1f87f0b5 | 1203 | |
ac13ac6f EB |
1204 | header = kzalloc(sizeof(struct ctl_table_header) + |
1205 | sizeof(struct ctl_node)*nr_entries, GFP_KERNEL); | |
1f87f0b5 EB |
1206 | if (!header) |
1207 | return NULL; | |
1208 | ||
ac13ac6f EB |
1209 | node = (struct ctl_node *)(header + 1); |
1210 | init_header(header, root, set, node, table); | |
7ec66d06 EB |
1211 | if (sysctl_check_table(path, table)) |
1212 | goto fail; | |
1213 | ||
1214 | spin_lock(&sysctl_lock); | |
0e47c99d | 1215 | dir = &set->dir; |
60f126d9 | 1216 | /* Reference moved down the diretory tree get_subdir */ |
7ec66d06 EB |
1217 | dir->header.nreg++; |
1218 | spin_unlock(&sysctl_lock); | |
1f87f0b5 | 1219 | |
7ec66d06 | 1220 | /* Find the directory for the ctl_table */ |
6e9d5164 EB |
1221 | for (name = path; name; name = nextname) { |
1222 | int namelen; | |
1223 | nextname = strchr(name, '/'); | |
1224 | if (nextname) { | |
1225 | namelen = nextname - name; | |
1226 | nextname++; | |
1227 | } else { | |
1228 | namelen = strlen(name); | |
1229 | } | |
1230 | if (namelen == 0) | |
1231 | continue; | |
1f87f0b5 | 1232 | |
0e47c99d | 1233 | dir = get_subdir(dir, name, namelen); |
7ec66d06 EB |
1234 | if (IS_ERR(dir)) |
1235 | goto fail; | |
1f87f0b5 | 1236 | } |
0e47c99d | 1237 | |
1f87f0b5 | 1238 | spin_lock(&sysctl_lock); |
0e47c99d | 1239 | if (insert_header(dir, header)) |
7ec66d06 | 1240 | goto fail_put_dir_locked; |
0e47c99d | 1241 | |
7ec66d06 | 1242 | drop_sysctl_table(&dir->header); |
1f87f0b5 EB |
1243 | spin_unlock(&sysctl_lock); |
1244 | ||
1245 | return header; | |
0e47c99d | 1246 | |
7ec66d06 EB |
1247 | fail_put_dir_locked: |
1248 | drop_sysctl_table(&dir->header); | |
7c60c48f EB |
1249 | spin_unlock(&sysctl_lock); |
1250 | fail: | |
1251 | kfree(header); | |
1252 | dump_stack(); | |
1253 | return NULL; | |
1f87f0b5 EB |
1254 | } |
1255 | ||
fea478d4 EB |
1256 | /** |
1257 | * register_sysctl - register a sysctl table | |
1258 | * @path: The path to the directory the sysctl table is in. | |
1259 | * @table: the table structure | |
1260 | * | |
1261 | * Register a sysctl table. @table should be a filled in ctl_table | |
1262 | * array. A completely 0 filled entry terminates the table. | |
1263 | * | |
1264 | * See __register_sysctl_table for more details. | |
1265 | */ | |
1266 | struct ctl_table_header *register_sysctl(const char *path, struct ctl_table *table) | |
1267 | { | |
1268 | return __register_sysctl_table(&sysctl_table_root.default_set, | |
1269 | path, table); | |
1270 | } | |
1271 | EXPORT_SYMBOL(register_sysctl); | |
1272 | ||
6e9d5164 EB |
1273 | static char *append_path(const char *path, char *pos, const char *name) |
1274 | { | |
1275 | int namelen; | |
1276 | namelen = strlen(name); | |
1277 | if (((pos - path) + namelen + 2) >= PATH_MAX) | |
1278 | return NULL; | |
1279 | memcpy(pos, name, namelen); | |
1280 | pos[namelen] = '/'; | |
1281 | pos[namelen + 1] = '\0'; | |
1282 | pos += namelen + 1; | |
1283 | return pos; | |
1284 | } | |
1285 | ||
f728019b EB |
1286 | static int count_subheaders(struct ctl_table *table) |
1287 | { | |
1288 | int has_files = 0; | |
1289 | int nr_subheaders = 0; | |
1290 | struct ctl_table *entry; | |
1291 | ||
1292 | /* special case: no directory and empty directory */ | |
1293 | if (!table || !table->procname) | |
1294 | return 1; | |
1295 | ||
1296 | for (entry = table; entry->procname; entry++) { | |
1297 | if (entry->child) | |
1298 | nr_subheaders += count_subheaders(entry->child); | |
1299 | else | |
1300 | has_files = 1; | |
1301 | } | |
1302 | return nr_subheaders + has_files; | |
1303 | } | |
1304 | ||
1305 | static int register_leaf_sysctl_tables(const char *path, char *pos, | |
60a47a2e | 1306 | struct ctl_table_header ***subheader, struct ctl_table_set *set, |
f728019b EB |
1307 | struct ctl_table *table) |
1308 | { | |
1309 | struct ctl_table *ctl_table_arg = NULL; | |
1310 | struct ctl_table *entry, *files; | |
1311 | int nr_files = 0; | |
1312 | int nr_dirs = 0; | |
1313 | int err = -ENOMEM; | |
1314 | ||
1315 | for (entry = table; entry->procname; entry++) { | |
1316 | if (entry->child) | |
1317 | nr_dirs++; | |
1318 | else | |
1319 | nr_files++; | |
1320 | } | |
1321 | ||
1322 | files = table; | |
1323 | /* If there are mixed files and directories we need a new table */ | |
1324 | if (nr_dirs && nr_files) { | |
1325 | struct ctl_table *new; | |
1326 | files = kzalloc(sizeof(struct ctl_table) * (nr_files + 1), | |
1327 | GFP_KERNEL); | |
1328 | if (!files) | |
1329 | goto out; | |
1330 | ||
1331 | ctl_table_arg = files; | |
1332 | for (new = files, entry = table; entry->procname; entry++) { | |
1333 | if (entry->child) | |
1334 | continue; | |
1335 | *new = *entry; | |
1336 | new++; | |
1337 | } | |
1338 | } | |
1339 | ||
1340 | /* Register everything except a directory full of subdirectories */ | |
1341 | if (nr_files || !nr_dirs) { | |
1342 | struct ctl_table_header *header; | |
60a47a2e | 1343 | header = __register_sysctl_table(set, path, files); |
f728019b EB |
1344 | if (!header) { |
1345 | kfree(ctl_table_arg); | |
1346 | goto out; | |
1347 | } | |
1348 | ||
1349 | /* Remember if we need to free the file table */ | |
1350 | header->ctl_table_arg = ctl_table_arg; | |
1351 | **subheader = header; | |
1352 | (*subheader)++; | |
1353 | } | |
1354 | ||
1355 | /* Recurse into the subdirectories. */ | |
1356 | for (entry = table; entry->procname; entry++) { | |
1357 | char *child_pos; | |
1358 | ||
1359 | if (!entry->child) | |
1360 | continue; | |
1361 | ||
1362 | err = -ENAMETOOLONG; | |
1363 | child_pos = append_path(path, pos, entry->procname); | |
1364 | if (!child_pos) | |
1365 | goto out; | |
1366 | ||
1367 | err = register_leaf_sysctl_tables(path, child_pos, subheader, | |
60a47a2e | 1368 | set, entry->child); |
f728019b EB |
1369 | pos[0] = '\0'; |
1370 | if (err) | |
1371 | goto out; | |
1372 | } | |
1373 | err = 0; | |
1374 | out: | |
1375 | /* On failure our caller will unregister all registered subheaders */ | |
1376 | return err; | |
1377 | } | |
1378 | ||
6e9d5164 EB |
1379 | /** |
1380 | * __register_sysctl_paths - register a sysctl table hierarchy | |
60a47a2e | 1381 | * @set: Sysctl tree to register on |
6e9d5164 EB |
1382 | * @path: The path to the directory the sysctl table is in. |
1383 | * @table: the top-level table structure | |
1384 | * | |
1385 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1386 | * array. A completely 0 filled entry terminates the table. | |
1387 | * | |
1388 | * See __register_sysctl_table for more details. | |
1389 | */ | |
1390 | struct ctl_table_header *__register_sysctl_paths( | |
60a47a2e | 1391 | struct ctl_table_set *set, |
6e9d5164 EB |
1392 | const struct ctl_path *path, struct ctl_table *table) |
1393 | { | |
ec6a5266 | 1394 | struct ctl_table *ctl_table_arg = table; |
f728019b EB |
1395 | int nr_subheaders = count_subheaders(table); |
1396 | struct ctl_table_header *header = NULL, **subheaders, **subheader; | |
6e9d5164 EB |
1397 | const struct ctl_path *component; |
1398 | char *new_path, *pos; | |
1399 | ||
1400 | pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL); | |
1401 | if (!new_path) | |
1402 | return NULL; | |
1403 | ||
1404 | pos[0] = '\0'; | |
1405 | for (component = path; component->procname; component++) { | |
1406 | pos = append_path(new_path, pos, component->procname); | |
1407 | if (!pos) | |
1408 | goto out; | |
1409 | } | |
ec6a5266 EB |
1410 | while (table->procname && table->child && !table[1].procname) { |
1411 | pos = append_path(new_path, pos, table->procname); | |
1412 | if (!pos) | |
1413 | goto out; | |
1414 | table = table->child; | |
1415 | } | |
f728019b | 1416 | if (nr_subheaders == 1) { |
60a47a2e | 1417 | header = __register_sysctl_table(set, new_path, table); |
f728019b EB |
1418 | if (header) |
1419 | header->ctl_table_arg = ctl_table_arg; | |
1420 | } else { | |
1421 | header = kzalloc(sizeof(*header) + | |
1422 | sizeof(*subheaders)*nr_subheaders, GFP_KERNEL); | |
1423 | if (!header) | |
1424 | goto out; | |
1425 | ||
1426 | subheaders = (struct ctl_table_header **) (header + 1); | |
1427 | subheader = subheaders; | |
ec6a5266 | 1428 | header->ctl_table_arg = ctl_table_arg; |
f728019b EB |
1429 | |
1430 | if (register_leaf_sysctl_tables(new_path, pos, &subheader, | |
60a47a2e | 1431 | set, table)) |
f728019b EB |
1432 | goto err_register_leaves; |
1433 | } | |
1434 | ||
6e9d5164 EB |
1435 | out: |
1436 | kfree(new_path); | |
1437 | return header; | |
f728019b EB |
1438 | |
1439 | err_register_leaves: | |
1440 | while (subheader > subheaders) { | |
1441 | struct ctl_table_header *subh = *(--subheader); | |
1442 | struct ctl_table *table = subh->ctl_table_arg; | |
1443 | unregister_sysctl_table(subh); | |
1444 | kfree(table); | |
1445 | } | |
1446 | kfree(header); | |
1447 | header = NULL; | |
1448 | goto out; | |
6e9d5164 EB |
1449 | } |
1450 | ||
1f87f0b5 EB |
1451 | /** |
1452 | * register_sysctl_table_path - register a sysctl table hierarchy | |
1453 | * @path: The path to the directory the sysctl table is in. | |
1454 | * @table: the top-level table structure | |
1455 | * | |
1456 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1457 | * array. A completely 0 filled entry terminates the table. | |
1458 | * | |
1459 | * See __register_sysctl_paths for more details. | |
1460 | */ | |
1461 | struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path, | |
1462 | struct ctl_table *table) | |
1463 | { | |
60a47a2e | 1464 | return __register_sysctl_paths(&sysctl_table_root.default_set, |
1f87f0b5 EB |
1465 | path, table); |
1466 | } | |
1467 | EXPORT_SYMBOL(register_sysctl_paths); | |
1468 | ||
1469 | /** | |
1470 | * register_sysctl_table - register a sysctl table hierarchy | |
1471 | * @table: the top-level table structure | |
1472 | * | |
1473 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1474 | * array. A completely 0 filled entry terminates the table. | |
1475 | * | |
1476 | * See register_sysctl_paths for more details. | |
1477 | */ | |
1478 | struct ctl_table_header *register_sysctl_table(struct ctl_table *table) | |
1479 | { | |
1480 | static const struct ctl_path null_path[] = { {} }; | |
1481 | ||
1482 | return register_sysctl_paths(null_path, table); | |
1483 | } | |
1484 | EXPORT_SYMBOL(register_sysctl_table); | |
1485 | ||
0e47c99d EB |
1486 | static void put_links(struct ctl_table_header *header) |
1487 | { | |
1488 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
1489 | struct ctl_table_root *root = header->root; | |
1490 | struct ctl_dir *parent = header->parent; | |
1491 | struct ctl_dir *core_parent; | |
1492 | struct ctl_table *entry; | |
1493 | ||
1494 | if (header->set == root_set) | |
1495 | return; | |
1496 | ||
1497 | core_parent = xlate_dir(root_set, parent); | |
1498 | if (IS_ERR(core_parent)) | |
1499 | return; | |
1500 | ||
1501 | for (entry = header->ctl_table; entry->procname; entry++) { | |
1502 | struct ctl_table_header *link_head; | |
1503 | struct ctl_table *link; | |
1504 | const char *name = entry->procname; | |
1505 | ||
1506 | link = find_entry(&link_head, core_parent, name, strlen(name)); | |
1507 | if (link && | |
1508 | ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) || | |
1509 | (S_ISLNK(link->mode) && (link->data == root)))) { | |
1510 | drop_sysctl_table(link_head); | |
1511 | } | |
1512 | else { | |
1513 | printk(KERN_ERR "sysctl link missing during unregister: "); | |
1514 | sysctl_print_dir(parent); | |
1515 | printk(KERN_CONT "/%s\n", name); | |
1516 | } | |
1517 | } | |
1518 | } | |
1519 | ||
938aaa4f EB |
1520 | static void drop_sysctl_table(struct ctl_table_header *header) |
1521 | { | |
7ec66d06 EB |
1522 | struct ctl_dir *parent = header->parent; |
1523 | ||
938aaa4f EB |
1524 | if (--header->nreg) |
1525 | return; | |
1526 | ||
0e47c99d | 1527 | put_links(header); |
938aaa4f | 1528 | start_unregistering(header); |
938aaa4f EB |
1529 | if (!--header->count) |
1530 | kfree_rcu(header, rcu); | |
7ec66d06 EB |
1531 | |
1532 | if (parent) | |
1533 | drop_sysctl_table(&parent->header); | |
938aaa4f EB |
1534 | } |
1535 | ||
1f87f0b5 EB |
1536 | /** |
1537 | * unregister_sysctl_table - unregister a sysctl table hierarchy | |
1538 | * @header: the header returned from register_sysctl_table | |
1539 | * | |
1540 | * Unregisters the sysctl table and all children. proc entries may not | |
1541 | * actually be removed until they are no longer used by anyone. | |
1542 | */ | |
1543 | void unregister_sysctl_table(struct ctl_table_header * header) | |
1544 | { | |
f728019b | 1545 | int nr_subheaders; |
1f87f0b5 EB |
1546 | might_sleep(); |
1547 | ||
1548 | if (header == NULL) | |
1549 | return; | |
1550 | ||
f728019b EB |
1551 | nr_subheaders = count_subheaders(header->ctl_table_arg); |
1552 | if (unlikely(nr_subheaders > 1)) { | |
1553 | struct ctl_table_header **subheaders; | |
1554 | int i; | |
1555 | ||
1556 | subheaders = (struct ctl_table_header **)(header + 1); | |
1557 | for (i = nr_subheaders -1; i >= 0; i--) { | |
1558 | struct ctl_table_header *subh = subheaders[i]; | |
1559 | struct ctl_table *table = subh->ctl_table_arg; | |
1560 | unregister_sysctl_table(subh); | |
1561 | kfree(table); | |
1562 | } | |
1563 | kfree(header); | |
1564 | return; | |
1565 | } | |
1566 | ||
1f87f0b5 | 1567 | spin_lock(&sysctl_lock); |
938aaa4f | 1568 | drop_sysctl_table(header); |
1f87f0b5 EB |
1569 | spin_unlock(&sysctl_lock); |
1570 | } | |
1571 | EXPORT_SYMBOL(unregister_sysctl_table); | |
1572 | ||
0e47c99d | 1573 | void setup_sysctl_set(struct ctl_table_set *set, |
9eb47c26 | 1574 | struct ctl_table_root *root, |
1f87f0b5 EB |
1575 | int (*is_seen)(struct ctl_table_set *)) |
1576 | { | |
1347440d | 1577 | memset(set, 0, sizeof(*set)); |
0e47c99d | 1578 | set->is_seen = is_seen; |
ac13ac6f | 1579 | init_header(&set->dir.header, root, set, NULL, root_table); |
1f87f0b5 EB |
1580 | } |
1581 | ||
97324cd8 EB |
1582 | void retire_sysctl_set(struct ctl_table_set *set) |
1583 | { | |
ac13ac6f | 1584 | WARN_ON(!RB_EMPTY_ROOT(&set->dir.root)); |
97324cd8 | 1585 | } |
1f87f0b5 | 1586 | |
1e0edd3f | 1587 | int __init proc_sys_init(void) |
77b14db5 | 1588 | { |
e1675231 AD |
1589 | struct proc_dir_entry *proc_sys_root; |
1590 | ||
77b14db5 | 1591 | proc_sys_root = proc_mkdir("sys", NULL); |
9043476f AV |
1592 | proc_sys_root->proc_iops = &proc_sys_dir_operations; |
1593 | proc_sys_root->proc_fops = &proc_sys_dir_file_operations; | |
77b14db5 | 1594 | proc_sys_root->nlink = 0; |
de4e83bd EB |
1595 | |
1596 | return sysctl_init(); | |
77b14db5 | 1597 | } |