]>
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> | |
34286d66 | 9 | #include <linux/namei.h> |
1f87f0b5 | 10 | #include <linux/module.h> |
77b14db5 EB |
11 | #include "internal.h" |
12 | ||
d72f71eb | 13 | static const struct dentry_operations proc_sys_dentry_operations; |
77b14db5 | 14 | static const struct file_operations proc_sys_file_operations; |
03a44825 | 15 | static const struct inode_operations proc_sys_inode_operations; |
9043476f AV |
16 | static const struct file_operations proc_sys_dir_file_operations; |
17 | static const struct inode_operations proc_sys_dir_operations; | |
77b14db5 | 18 | |
f1ecf068 LDM |
19 | void proc_sys_poll_notify(struct ctl_table_poll *poll) |
20 | { | |
21 | if (!poll) | |
22 | return; | |
23 | ||
24 | atomic_inc(&poll->event); | |
25 | wake_up_interruptible(&poll->wait); | |
26 | } | |
27 | ||
a194558e EB |
28 | static struct ctl_table root_table[] = { |
29 | { | |
30 | .procname = "", | |
7ec66d06 | 31 | .mode = S_IFDIR|S_IRUGO|S_IXUGO, |
a194558e EB |
32 | }, |
33 | { } | |
34 | }; | |
0e47c99d | 35 | static struct ctl_table_root sysctl_table_root = { |
9e3d47df | 36 | .default_set.dir.list = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.list), |
0e47c99d | 37 | .default_set.dir.header = { |
7ec66d06 EB |
38 | {{.count = 1, |
39 | .nreg = 1, | |
40 | .ctl_table = root_table, | |
9e3d47df | 41 | .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.header.ctl_entry),}}, |
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); | |
e54012ce | 55 | static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table); |
7ec66d06 | 56 | |
6980128f EB |
57 | static void sysctl_print_dir(struct ctl_dir *dir) |
58 | { | |
59 | if (dir->header.parent) | |
60 | sysctl_print_dir(dir->header.parent); | |
61 | printk(KERN_CONT "%s/", dir->header.ctl_table[0].procname); | |
62 | } | |
63 | ||
076c3eed EB |
64 | static int namecmp(const char *name1, int len1, const char *name2, int len2) |
65 | { | |
66 | int minlen; | |
67 | int cmp; | |
68 | ||
69 | minlen = len1; | |
70 | if (minlen > len2) | |
71 | minlen = len2; | |
72 | ||
73 | cmp = memcmp(name1, name2, minlen); | |
74 | if (cmp == 0) | |
75 | cmp = len1 - len2; | |
76 | return cmp; | |
77 | } | |
78 | ||
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; | |
84 | ||
9e3d47df | 85 | list_for_each_entry(head, &dir->list, ctl_entry) { |
076c3eed EB |
86 | if (head->unregistering) |
87 | continue; | |
7ec66d06 | 88 | for (entry = head->ctl_table; entry->procname; entry++) { |
076c3eed EB |
89 | const char *procname = entry->procname; |
90 | if (namecmp(procname, strlen(procname), name, namelen) == 0) { | |
91 | *phead = head; | |
92 | return entry; | |
93 | } | |
94 | } | |
95 | } | |
96 | return NULL; | |
97 | } | |
98 | ||
e0d04529 EB |
99 | static void init_header(struct ctl_table_header *head, |
100 | struct ctl_table_root *root, struct ctl_table_set *set, | |
101 | struct ctl_table *table) | |
102 | { | |
7ec66d06 | 103 | head->ctl_table = table; |
e0d04529 EB |
104 | head->ctl_table_arg = table; |
105 | INIT_LIST_HEAD(&head->ctl_entry); | |
106 | head->used = 0; | |
107 | head->count = 1; | |
108 | head->nreg = 1; | |
109 | head->unregistering = NULL; | |
110 | head->root = root; | |
111 | head->set = set; | |
112 | head->parent = NULL; | |
113 | } | |
114 | ||
8425d6aa EB |
115 | static void erase_header(struct ctl_table_header *head) |
116 | { | |
117 | list_del_init(&head->ctl_entry); | |
118 | } | |
119 | ||
0e47c99d | 120 | static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header) |
8425d6aa | 121 | { |
0e47c99d EB |
122 | int err; |
123 | ||
e54012ce EB |
124 | err = sysctl_check_dups(dir, header->ctl_table); |
125 | if (err) | |
126 | return err; | |
127 | ||
0e47c99d | 128 | dir->header.nreg++; |
7ec66d06 | 129 | header->parent = dir; |
0e47c99d EB |
130 | err = insert_links(header); |
131 | if (err) | |
132 | goto fail_links; | |
9e3d47df | 133 | list_add_tail(&header->ctl_entry, &header->parent->list); |
0e47c99d EB |
134 | return 0; |
135 | fail_links: | |
136 | header->parent = NULL; | |
137 | drop_sysctl_table(&dir->header); | |
138 | return err; | |
8425d6aa EB |
139 | } |
140 | ||
1f87f0b5 EB |
141 | /* called under sysctl_lock */ |
142 | static int use_table(struct ctl_table_header *p) | |
143 | { | |
144 | if (unlikely(p->unregistering)) | |
145 | return 0; | |
146 | p->used++; | |
147 | return 1; | |
148 | } | |
149 | ||
150 | /* called under sysctl_lock */ | |
151 | static void unuse_table(struct ctl_table_header *p) | |
152 | { | |
153 | if (!--p->used) | |
154 | if (unlikely(p->unregistering)) | |
155 | complete(p->unregistering); | |
156 | } | |
157 | ||
158 | /* called under sysctl_lock, will reacquire if has to wait */ | |
159 | static void start_unregistering(struct ctl_table_header *p) | |
160 | { | |
161 | /* | |
162 | * if p->used is 0, nobody will ever touch that entry again; | |
163 | * we'll eliminate all paths to it before dropping sysctl_lock | |
164 | */ | |
165 | if (unlikely(p->used)) { | |
166 | struct completion wait; | |
167 | init_completion(&wait); | |
168 | p->unregistering = &wait; | |
169 | spin_unlock(&sysctl_lock); | |
170 | wait_for_completion(&wait); | |
171 | spin_lock(&sysctl_lock); | |
172 | } else { | |
173 | /* anything non-NULL; we'll never dereference it */ | |
174 | p->unregistering = ERR_PTR(-EINVAL); | |
175 | } | |
176 | /* | |
177 | * do not remove from the list until nobody holds it; walking the | |
178 | * list in do_sysctl() relies on that. | |
179 | */ | |
8425d6aa | 180 | erase_header(p); |
1f87f0b5 EB |
181 | } |
182 | ||
183 | static void sysctl_head_get(struct ctl_table_header *head) | |
184 | { | |
185 | spin_lock(&sysctl_lock); | |
186 | head->count++; | |
187 | spin_unlock(&sysctl_lock); | |
188 | } | |
189 | ||
190 | void sysctl_head_put(struct ctl_table_header *head) | |
191 | { | |
192 | spin_lock(&sysctl_lock); | |
193 | if (!--head->count) | |
194 | kfree_rcu(head, rcu); | |
195 | spin_unlock(&sysctl_lock); | |
196 | } | |
197 | ||
198 | static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head) | |
199 | { | |
200 | if (!head) | |
201 | BUG(); | |
202 | spin_lock(&sysctl_lock); | |
203 | if (!use_table(head)) | |
204 | head = ERR_PTR(-ENOENT); | |
205 | spin_unlock(&sysctl_lock); | |
206 | return head; | |
207 | } | |
208 | ||
209 | static void sysctl_head_finish(struct ctl_table_header *head) | |
210 | { | |
211 | if (!head) | |
212 | return; | |
213 | spin_lock(&sysctl_lock); | |
214 | unuse_table(head); | |
215 | spin_unlock(&sysctl_lock); | |
216 | } | |
217 | ||
218 | static struct ctl_table_set * | |
219 | lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces) | |
220 | { | |
221 | struct ctl_table_set *set = &root->default_set; | |
222 | if (root->lookup) | |
223 | set = root->lookup(root, namespaces); | |
224 | return set; | |
225 | } | |
226 | ||
076c3eed | 227 | static struct ctl_table *lookup_entry(struct ctl_table_header **phead, |
7ec66d06 | 228 | struct ctl_dir *dir, |
076c3eed EB |
229 | const char *name, int namelen) |
230 | { | |
231 | struct ctl_table_header *head; | |
232 | struct ctl_table *entry; | |
076c3eed EB |
233 | |
234 | spin_lock(&sysctl_lock); | |
0e47c99d EB |
235 | entry = find_entry(&head, dir, name, namelen); |
236 | if (entry && use_table(head)) | |
237 | *phead = head; | |
238 | else | |
239 | entry = NULL; | |
076c3eed EB |
240 | spin_unlock(&sysctl_lock); |
241 | return entry; | |
242 | } | |
243 | ||
7ec66d06 | 244 | static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir, |
0e47c99d | 245 | struct list_head *tmp) |
1f87f0b5 | 246 | { |
1f87f0b5 | 247 | struct ctl_table_header *head; |
1f87f0b5 | 248 | |
9e3d47df | 249 | for (tmp = tmp->next; tmp != &dir->list; tmp = tmp->next) { |
1f87f0b5 EB |
250 | head = list_entry(tmp, struct ctl_table_header, ctl_entry); |
251 | ||
9e3d47df | 252 | if (!head->ctl_table->procname || |
6a75ce16 | 253 | !use_table(head)) |
1f87f0b5 EB |
254 | continue; |
255 | ||
0e47c99d | 256 | return head; |
1f87f0b5 | 257 | } |
1f87f0b5 EB |
258 | return NULL; |
259 | } | |
260 | ||
7ec66d06 | 261 | static void first_entry(struct ctl_dir *dir, |
6a75ce16 EB |
262 | struct ctl_table_header **phead, struct ctl_table **pentry) |
263 | { | |
7ec66d06 EB |
264 | struct ctl_table_header *head; |
265 | struct ctl_table *entry = NULL; | |
6a75ce16 EB |
266 | |
267 | spin_lock(&sysctl_lock); | |
9e3d47df | 268 | head = next_usable_entry(dir, &dir->list); |
6a75ce16 | 269 | spin_unlock(&sysctl_lock); |
7ec66d06 EB |
270 | if (head) |
271 | entry = head->ctl_table; | |
6a75ce16 EB |
272 | *phead = head; |
273 | *pentry = entry; | |
274 | } | |
275 | ||
7ec66d06 | 276 | static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry) |
1f87f0b5 | 277 | { |
6a75ce16 EB |
278 | struct ctl_table_header *head = *phead; |
279 | struct ctl_table *entry = *pentry; | |
280 | ||
281 | entry++; | |
282 | if (!entry->procname) { | |
6a75ce16 EB |
283 | spin_lock(&sysctl_lock); |
284 | unuse_table(head); | |
0e47c99d | 285 | head = next_usable_entry(head->parent, &head->ctl_entry); |
6a75ce16 EB |
286 | spin_unlock(&sysctl_lock); |
287 | if (head) | |
7ec66d06 | 288 | entry = head->ctl_table; |
6a75ce16 EB |
289 | } |
290 | *phead = head; | |
291 | *pentry = entry; | |
1f87f0b5 EB |
292 | } |
293 | ||
294 | void register_sysctl_root(struct ctl_table_root *root) | |
295 | { | |
1f87f0b5 EB |
296 | } |
297 | ||
298 | /* | |
299 | * sysctl_perm does NOT grant the superuser all rights automatically, because | |
300 | * some sysctl variables are readonly even to root. | |
301 | */ | |
302 | ||
303 | static int test_perm(int mode, int op) | |
304 | { | |
305 | if (!current_euid()) | |
306 | mode >>= 6; | |
307 | else if (in_egroup_p(0)) | |
308 | mode >>= 3; | |
309 | if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0) | |
310 | return 0; | |
311 | return -EACCES; | |
312 | } | |
313 | ||
314 | static int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op) | |
315 | { | |
316 | int mode; | |
317 | ||
318 | if (root->permissions) | |
319 | mode = root->permissions(root, current->nsproxy, table); | |
320 | else | |
321 | mode = table->mode; | |
322 | ||
323 | return test_perm(mode, op); | |
324 | } | |
325 | ||
9043476f AV |
326 | static struct inode *proc_sys_make_inode(struct super_block *sb, |
327 | struct ctl_table_header *head, struct ctl_table *table) | |
77b14db5 EB |
328 | { |
329 | struct inode *inode; | |
9043476f | 330 | struct proc_inode *ei; |
77b14db5 | 331 | |
9043476f | 332 | inode = new_inode(sb); |
77b14db5 EB |
333 | if (!inode) |
334 | goto out; | |
335 | ||
85fe4025 CH |
336 | inode->i_ino = get_next_ino(); |
337 | ||
9043476f | 338 | sysctl_head_get(head); |
77b14db5 | 339 | ei = PROC_I(inode); |
9043476f AV |
340 | ei->sysctl = head; |
341 | ei->sysctl_entry = table; | |
342 | ||
77b14db5 | 343 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; |
9043476f | 344 | inode->i_mode = table->mode; |
7ec66d06 | 345 | if (!S_ISDIR(table->mode)) { |
9043476f AV |
346 | inode->i_mode |= S_IFREG; |
347 | inode->i_op = &proc_sys_inode_operations; | |
348 | inode->i_fop = &proc_sys_file_operations; | |
349 | } else { | |
350 | inode->i_mode |= S_IFDIR; | |
9043476f AV |
351 | inode->i_op = &proc_sys_dir_operations; |
352 | inode->i_fop = &proc_sys_dir_file_operations; | |
353 | } | |
77b14db5 EB |
354 | out: |
355 | return inode; | |
356 | } | |
357 | ||
81324364 | 358 | static struct ctl_table_header *grab_header(struct inode *inode) |
77b14db5 | 359 | { |
3cc3e046 EB |
360 | struct ctl_table_header *head = PROC_I(inode)->sysctl; |
361 | if (!head) | |
0e47c99d | 362 | head = &sysctl_table_root.default_set.dir.header; |
3cc3e046 | 363 | return sysctl_head_grab(head); |
9043476f | 364 | } |
77b14db5 | 365 | |
9043476f AV |
366 | static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry, |
367 | struct nameidata *nd) | |
368 | { | |
369 | struct ctl_table_header *head = grab_header(dir); | |
9043476f AV |
370 | struct ctl_table_header *h = NULL; |
371 | struct qstr *name = &dentry->d_name; | |
372 | struct ctl_table *p; | |
373 | struct inode *inode; | |
374 | struct dentry *err = ERR_PTR(-ENOENT); | |
7ec66d06 | 375 | struct ctl_dir *ctl_dir; |
0e47c99d | 376 | int ret; |
77b14db5 | 377 | |
9043476f AV |
378 | if (IS_ERR(head)) |
379 | return ERR_CAST(head); | |
77b14db5 | 380 | |
7ec66d06 | 381 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 | 382 | |
7ec66d06 | 383 | p = lookup_entry(&h, ctl_dir, name->name, name->len); |
9043476f | 384 | if (!p) |
77b14db5 EB |
385 | goto out; |
386 | ||
0e47c99d EB |
387 | ret = sysctl_follow_link(&h, &p, current->nsproxy); |
388 | err = ERR_PTR(ret); | |
389 | if (ret) | |
390 | goto out; | |
391 | ||
77b14db5 | 392 | err = ERR_PTR(-ENOMEM); |
9043476f AV |
393 | inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p); |
394 | if (h) | |
395 | sysctl_head_finish(h); | |
396 | ||
77b14db5 EB |
397 | if (!inode) |
398 | goto out; | |
399 | ||
400 | err = NULL; | |
fb045adb | 401 | d_set_d_op(dentry, &proc_sys_dentry_operations); |
77b14db5 EB |
402 | d_add(dentry, inode); |
403 | ||
404 | out: | |
405 | sysctl_head_finish(head); | |
406 | return err; | |
407 | } | |
408 | ||
7708bfb1 PE |
409 | static ssize_t proc_sys_call_handler(struct file *filp, void __user *buf, |
410 | size_t count, loff_t *ppos, int write) | |
77b14db5 | 411 | { |
9043476f AV |
412 | struct inode *inode = filp->f_path.dentry->d_inode; |
413 | struct ctl_table_header *head = grab_header(inode); | |
414 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
2a2da53b DH |
415 | ssize_t error; |
416 | size_t res; | |
77b14db5 | 417 | |
9043476f AV |
418 | if (IS_ERR(head)) |
419 | return PTR_ERR(head); | |
77b14db5 EB |
420 | |
421 | /* | |
422 | * At this point we know that the sysctl was not unregistered | |
423 | * and won't be until we finish. | |
424 | */ | |
425 | error = -EPERM; | |
d7321cd6 | 426 | if (sysctl_perm(head->root, table, write ? MAY_WRITE : MAY_READ)) |
77b14db5 EB |
427 | goto out; |
428 | ||
9043476f AV |
429 | /* if that can happen at all, it should be -EINVAL, not -EISDIR */ |
430 | error = -EINVAL; | |
431 | if (!table->proc_handler) | |
432 | goto out; | |
433 | ||
77b14db5 EB |
434 | /* careful: calling conventions are nasty here */ |
435 | res = count; | |
8d65af78 | 436 | error = table->proc_handler(table, write, buf, &res, ppos); |
77b14db5 EB |
437 | if (!error) |
438 | error = res; | |
439 | out: | |
440 | sysctl_head_finish(head); | |
441 | ||
442 | return error; | |
443 | } | |
444 | ||
7708bfb1 | 445 | static ssize_t proc_sys_read(struct file *filp, char __user *buf, |
77b14db5 EB |
446 | size_t count, loff_t *ppos) |
447 | { | |
7708bfb1 PE |
448 | return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 0); |
449 | } | |
77b14db5 | 450 | |
7708bfb1 PE |
451 | static ssize_t proc_sys_write(struct file *filp, const char __user *buf, |
452 | size_t count, loff_t *ppos) | |
453 | { | |
454 | return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 1); | |
77b14db5 EB |
455 | } |
456 | ||
f1ecf068 LDM |
457 | static int proc_sys_open(struct inode *inode, struct file *filp) |
458 | { | |
459 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
460 | ||
461 | if (table->poll) | |
462 | filp->private_data = proc_sys_poll_event(table->poll); | |
463 | ||
464 | return 0; | |
465 | } | |
466 | ||
467 | static unsigned int proc_sys_poll(struct file *filp, poll_table *wait) | |
468 | { | |
469 | struct inode *inode = filp->f_path.dentry->d_inode; | |
470 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
471 | unsigned long event = (unsigned long)filp->private_data; | |
472 | unsigned int ret = DEFAULT_POLLMASK; | |
473 | ||
474 | if (!table->proc_handler) | |
475 | goto out; | |
476 | ||
477 | if (!table->poll) | |
478 | goto out; | |
479 | ||
480 | poll_wait(filp, &table->poll->wait, wait); | |
481 | ||
482 | if (event != atomic_read(&table->poll->event)) { | |
483 | filp->private_data = proc_sys_poll_event(table->poll); | |
484 | ret = POLLIN | POLLRDNORM | POLLERR | POLLPRI; | |
485 | } | |
486 | ||
487 | out: | |
488 | return ret; | |
489 | } | |
77b14db5 EB |
490 | |
491 | static int proc_sys_fill_cache(struct file *filp, void *dirent, | |
9043476f AV |
492 | filldir_t filldir, |
493 | struct ctl_table_header *head, | |
494 | struct ctl_table *table) | |
77b14db5 | 495 | { |
77b14db5 EB |
496 | struct dentry *child, *dir = filp->f_path.dentry; |
497 | struct inode *inode; | |
498 | struct qstr qname; | |
499 | ino_t ino = 0; | |
500 | unsigned type = DT_UNKNOWN; | |
77b14db5 EB |
501 | |
502 | qname.name = table->procname; | |
503 | qname.len = strlen(table->procname); | |
504 | qname.hash = full_name_hash(qname.name, qname.len); | |
505 | ||
77b14db5 EB |
506 | child = d_lookup(dir, &qname); |
507 | if (!child) { | |
9043476f AV |
508 | child = d_alloc(dir, &qname); |
509 | if (child) { | |
510 | inode = proc_sys_make_inode(dir->d_sb, head, table); | |
511 | if (!inode) { | |
512 | dput(child); | |
513 | return -ENOMEM; | |
514 | } else { | |
fb045adb | 515 | d_set_d_op(child, &proc_sys_dentry_operations); |
9043476f | 516 | d_add(child, inode); |
77b14db5 | 517 | } |
9043476f AV |
518 | } else { |
519 | return -ENOMEM; | |
77b14db5 EB |
520 | } |
521 | } | |
77b14db5 | 522 | inode = child->d_inode; |
9043476f AV |
523 | ino = inode->i_ino; |
524 | type = inode->i_mode >> 12; | |
77b14db5 | 525 | dput(child); |
9043476f AV |
526 | return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type); |
527 | } | |
528 | ||
0e47c99d EB |
529 | static int proc_sys_link_fill_cache(struct file *filp, void *dirent, |
530 | filldir_t filldir, | |
531 | struct ctl_table_header *head, | |
532 | struct ctl_table *table) | |
533 | { | |
534 | int err, ret = 0; | |
535 | head = sysctl_head_grab(head); | |
536 | ||
537 | /* It is not an error if we can not follow the link ignore it */ | |
538 | err = sysctl_follow_link(&head, &table, current->nsproxy); | |
539 | if (err) | |
540 | goto out; | |
541 | ||
542 | ret = proc_sys_fill_cache(filp, dirent, filldir, head, table); | |
543 | out: | |
544 | sysctl_head_finish(head); | |
545 | return ret; | |
546 | } | |
547 | ||
9043476f AV |
548 | static int scan(struct ctl_table_header *head, ctl_table *table, |
549 | unsigned long *pos, struct file *file, | |
550 | void *dirent, filldir_t filldir) | |
551 | { | |
6a75ce16 | 552 | int res; |
9043476f | 553 | |
6a75ce16 EB |
554 | if ((*pos)++ < file->f_pos) |
555 | return 0; | |
9043476f | 556 | |
0e47c99d EB |
557 | if (unlikely(S_ISLNK(table->mode))) |
558 | res = proc_sys_link_fill_cache(file, dirent, filldir, head, table); | |
559 | else | |
560 | res = proc_sys_fill_cache(file, dirent, filldir, head, table); | |
9043476f | 561 | |
6a75ce16 EB |
562 | if (res == 0) |
563 | file->f_pos = *pos; | |
9043476f | 564 | |
6a75ce16 | 565 | return res; |
77b14db5 EB |
566 | } |
567 | ||
568 | static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir) | |
569 | { | |
9043476f | 570 | struct dentry *dentry = filp->f_path.dentry; |
77b14db5 | 571 | struct inode *inode = dentry->d_inode; |
9043476f | 572 | struct ctl_table_header *head = grab_header(inode); |
9043476f | 573 | struct ctl_table_header *h = NULL; |
6a75ce16 | 574 | struct ctl_table *entry; |
7ec66d06 | 575 | struct ctl_dir *ctl_dir; |
77b14db5 | 576 | unsigned long pos; |
9043476f AV |
577 | int ret = -EINVAL; |
578 | ||
579 | if (IS_ERR(head)) | |
580 | return PTR_ERR(head); | |
77b14db5 | 581 | |
7ec66d06 | 582 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 EB |
583 | |
584 | ret = 0; | |
585 | /* Avoid a switch here: arm builds fail with missing __cmpdi2 */ | |
586 | if (filp->f_pos == 0) { | |
587 | if (filldir(dirent, ".", 1, filp->f_pos, | |
588 | inode->i_ino, DT_DIR) < 0) | |
589 | goto out; | |
590 | filp->f_pos++; | |
591 | } | |
592 | if (filp->f_pos == 1) { | |
593 | if (filldir(dirent, "..", 2, filp->f_pos, | |
594 | parent_ino(dentry), DT_DIR) < 0) | |
595 | goto out; | |
596 | filp->f_pos++; | |
597 | } | |
598 | pos = 2; | |
599 | ||
7ec66d06 | 600 | for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) { |
6a75ce16 | 601 | ret = scan(h, entry, &pos, filp, dirent, filldir); |
9043476f AV |
602 | if (ret) { |
603 | sysctl_head_finish(h); | |
604 | break; | |
77b14db5 EB |
605 | } |
606 | } | |
607 | ret = 1; | |
608 | out: | |
609 | sysctl_head_finish(head); | |
610 | return ret; | |
611 | } | |
612 | ||
10556cb2 | 613 | static int proc_sys_permission(struct inode *inode, int mask) |
77b14db5 EB |
614 | { |
615 | /* | |
616 | * sysctl entries that are not writeable, | |
617 | * are _NOT_ writeable, capabilities or not. | |
618 | */ | |
f696a365 MS |
619 | struct ctl_table_header *head; |
620 | struct ctl_table *table; | |
77b14db5 EB |
621 | int error; |
622 | ||
f696a365 MS |
623 | /* Executable files are not allowed under /proc/sys/ */ |
624 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) | |
625 | return -EACCES; | |
626 | ||
627 | head = grab_header(inode); | |
9043476f AV |
628 | if (IS_ERR(head)) |
629 | return PTR_ERR(head); | |
77b14db5 | 630 | |
f696a365 | 631 | table = PROC_I(inode)->sysctl_entry; |
9043476f AV |
632 | if (!table) /* global root - r-xr-xr-x */ |
633 | error = mask & MAY_WRITE ? -EACCES : 0; | |
634 | else /* Use the permissions on the sysctl table entry */ | |
1fc0f78c | 635 | error = sysctl_perm(head->root, table, mask & ~MAY_NOT_BLOCK); |
77b14db5 | 636 | |
77b14db5 EB |
637 | sysctl_head_finish(head); |
638 | return error; | |
639 | } | |
640 | ||
641 | static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr) | |
642 | { | |
643 | struct inode *inode = dentry->d_inode; | |
644 | int error; | |
645 | ||
646 | if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) | |
647 | return -EPERM; | |
648 | ||
649 | error = inode_change_ok(inode, attr); | |
1025774c CH |
650 | if (error) |
651 | return error; | |
652 | ||
653 | if ((attr->ia_valid & ATTR_SIZE) && | |
654 | attr->ia_size != i_size_read(inode)) { | |
655 | error = vmtruncate(inode, attr->ia_size); | |
656 | if (error) | |
657 | return error; | |
658 | } | |
77b14db5 | 659 | |
1025774c CH |
660 | setattr_copy(inode, attr); |
661 | mark_inode_dirty(inode); | |
662 | return 0; | |
77b14db5 EB |
663 | } |
664 | ||
9043476f AV |
665 | static int proc_sys_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) |
666 | { | |
667 | struct inode *inode = dentry->d_inode; | |
668 | struct ctl_table_header *head = grab_header(inode); | |
669 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
670 | ||
671 | if (IS_ERR(head)) | |
672 | return PTR_ERR(head); | |
673 | ||
674 | generic_fillattr(inode, stat); | |
675 | if (table) | |
676 | stat->mode = (stat->mode & S_IFMT) | table->mode; | |
677 | ||
678 | sysctl_head_finish(head); | |
679 | return 0; | |
680 | } | |
681 | ||
77b14db5 | 682 | static const struct file_operations proc_sys_file_operations = { |
f1ecf068 LDM |
683 | .open = proc_sys_open, |
684 | .poll = proc_sys_poll, | |
77b14db5 EB |
685 | .read = proc_sys_read, |
686 | .write = proc_sys_write, | |
6038f373 | 687 | .llseek = default_llseek, |
9043476f AV |
688 | }; |
689 | ||
690 | static const struct file_operations proc_sys_dir_file_operations = { | |
887df078 | 691 | .read = generic_read_dir, |
77b14db5 | 692 | .readdir = proc_sys_readdir, |
3222a3e5 | 693 | .llseek = generic_file_llseek, |
77b14db5 EB |
694 | }; |
695 | ||
03a44825 | 696 | static const struct inode_operations proc_sys_inode_operations = { |
9043476f AV |
697 | .permission = proc_sys_permission, |
698 | .setattr = proc_sys_setattr, | |
699 | .getattr = proc_sys_getattr, | |
700 | }; | |
701 | ||
702 | static const struct inode_operations proc_sys_dir_operations = { | |
77b14db5 EB |
703 | .lookup = proc_sys_lookup, |
704 | .permission = proc_sys_permission, | |
705 | .setattr = proc_sys_setattr, | |
9043476f | 706 | .getattr = proc_sys_getattr, |
77b14db5 EB |
707 | }; |
708 | ||
709 | static int proc_sys_revalidate(struct dentry *dentry, struct nameidata *nd) | |
710 | { | |
34286d66 NP |
711 | if (nd->flags & LOOKUP_RCU) |
712 | return -ECHILD; | |
9043476f AV |
713 | return !PROC_I(dentry->d_inode)->sysctl->unregistering; |
714 | } | |
715 | ||
fe15ce44 | 716 | static int proc_sys_delete(const struct dentry *dentry) |
9043476f AV |
717 | { |
718 | return !!PROC_I(dentry->d_inode)->sysctl->unregistering; | |
719 | } | |
720 | ||
1f87f0b5 EB |
721 | static int sysctl_is_seen(struct ctl_table_header *p) |
722 | { | |
723 | struct ctl_table_set *set = p->set; | |
724 | int res; | |
725 | spin_lock(&sysctl_lock); | |
726 | if (p->unregistering) | |
727 | res = 0; | |
728 | else if (!set->is_seen) | |
729 | res = 1; | |
730 | else | |
731 | res = set->is_seen(set); | |
732 | spin_unlock(&sysctl_lock); | |
733 | return res; | |
734 | } | |
735 | ||
621e155a NP |
736 | static int proc_sys_compare(const struct dentry *parent, |
737 | const struct inode *pinode, | |
738 | const struct dentry *dentry, const struct inode *inode, | |
739 | unsigned int len, const char *str, const struct qstr *name) | |
9043476f | 740 | { |
dfef6dcd | 741 | struct ctl_table_header *head; |
31e6b01f NP |
742 | /* Although proc doesn't have negative dentries, rcu-walk means |
743 | * that inode here can be NULL */ | |
dfef6dcd | 744 | /* AV: can it, indeed? */ |
31e6b01f | 745 | if (!inode) |
dfef6dcd | 746 | return 1; |
621e155a | 747 | if (name->len != len) |
9043476f | 748 | return 1; |
621e155a | 749 | if (memcmp(name->name, str, len)) |
9043476f | 750 | return 1; |
dfef6dcd AV |
751 | head = rcu_dereference(PROC_I(inode)->sysctl); |
752 | return !head || !sysctl_is_seen(head); | |
77b14db5 EB |
753 | } |
754 | ||
d72f71eb | 755 | static const struct dentry_operations proc_sys_dentry_operations = { |
77b14db5 | 756 | .d_revalidate = proc_sys_revalidate, |
9043476f AV |
757 | .d_delete = proc_sys_delete, |
758 | .d_compare = proc_sys_compare, | |
77b14db5 EB |
759 | }; |
760 | ||
0e47c99d EB |
761 | static struct ctl_dir *find_subdir(struct ctl_dir *dir, |
762 | const char *name, int namelen) | |
1f87f0b5 | 763 | { |
7ec66d06 EB |
764 | struct ctl_table_header *head; |
765 | struct ctl_table *entry; | |
1f87f0b5 | 766 | |
0e47c99d | 767 | entry = find_entry(&head, dir, name, namelen); |
7ec66d06 EB |
768 | if (!entry) |
769 | return ERR_PTR(-ENOENT); | |
770 | if (S_ISDIR(entry->mode)) | |
771 | return container_of(head, struct ctl_dir, header); | |
772 | return ERR_PTR(-ENOTDIR); | |
773 | } | |
774 | ||
775 | static struct ctl_dir *new_dir(struct ctl_table_set *set, | |
0e47c99d | 776 | const char *name, int namelen) |
7ec66d06 EB |
777 | { |
778 | struct ctl_table *table; | |
779 | struct ctl_dir *new; | |
780 | char *new_name; | |
1f87f0b5 | 781 | |
7ec66d06 EB |
782 | new = kzalloc(sizeof(*new) + sizeof(struct ctl_table)*2 + |
783 | namelen + 1, GFP_KERNEL); | |
784 | if (!new) | |
1f87f0b5 EB |
785 | return NULL; |
786 | ||
7ec66d06 EB |
787 | table = (struct ctl_table *)(new + 1); |
788 | new_name = (char *)(table + 2); | |
789 | memcpy(new_name, name, namelen); | |
790 | new_name[namelen] = '\0'; | |
9e3d47df | 791 | INIT_LIST_HEAD(&new->list); |
7ec66d06 EB |
792 | table[0].procname = new_name; |
793 | table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
0e47c99d | 794 | init_header(&new->header, set->dir.header.root, set, table); |
7ec66d06 EB |
795 | |
796 | return new; | |
1f87f0b5 EB |
797 | } |
798 | ||
0e47c99d EB |
799 | static struct ctl_dir *get_subdir(struct ctl_dir *dir, |
800 | const char *name, int namelen) | |
1f87f0b5 | 801 | { |
0e47c99d | 802 | struct ctl_table_set *set = dir->header.set; |
7ec66d06 | 803 | struct ctl_dir *subdir, *new = NULL; |
1f87f0b5 | 804 | |
7ec66d06 | 805 | spin_lock(&sysctl_lock); |
0e47c99d | 806 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
807 | if (!IS_ERR(subdir)) |
808 | goto found; | |
809 | if (PTR_ERR(subdir) != -ENOENT) | |
810 | goto failed; | |
811 | ||
812 | spin_unlock(&sysctl_lock); | |
813 | new = new_dir(set, name, namelen); | |
814 | spin_lock(&sysctl_lock); | |
815 | subdir = ERR_PTR(-ENOMEM); | |
816 | if (!new) | |
817 | goto failed; | |
818 | ||
0e47c99d | 819 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
820 | if (!IS_ERR(subdir)) |
821 | goto found; | |
822 | if (PTR_ERR(subdir) != -ENOENT) | |
823 | goto failed; | |
824 | ||
0e47c99d EB |
825 | if (insert_header(dir, &new->header)) |
826 | goto failed; | |
7ec66d06 EB |
827 | subdir = new; |
828 | found: | |
829 | subdir->header.nreg++; | |
830 | failed: | |
831 | if (unlikely(IS_ERR(subdir))) { | |
6980128f EB |
832 | printk(KERN_ERR "sysctl could not get directory: "); |
833 | sysctl_print_dir(dir); | |
834 | printk(KERN_CONT "/%*.*s %ld\n", | |
7ec66d06 | 835 | namelen, namelen, name, PTR_ERR(subdir)); |
1f87f0b5 | 836 | } |
7ec66d06 EB |
837 | drop_sysctl_table(&dir->header); |
838 | if (new) | |
839 | drop_sysctl_table(&new->header); | |
840 | spin_unlock(&sysctl_lock); | |
841 | return subdir; | |
1f87f0b5 EB |
842 | } |
843 | ||
0e47c99d EB |
844 | static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir) |
845 | { | |
846 | struct ctl_dir *parent; | |
847 | const char *procname; | |
848 | if (!dir->header.parent) | |
849 | return &set->dir; | |
850 | parent = xlate_dir(set, dir->header.parent); | |
851 | if (IS_ERR(parent)) | |
852 | return parent; | |
853 | procname = dir->header.ctl_table[0].procname; | |
854 | return find_subdir(parent, procname, strlen(procname)); | |
855 | } | |
856 | ||
857 | static int sysctl_follow_link(struct ctl_table_header **phead, | |
858 | struct ctl_table **pentry, struct nsproxy *namespaces) | |
859 | { | |
860 | struct ctl_table_header *head; | |
861 | struct ctl_table_root *root; | |
862 | struct ctl_table_set *set; | |
863 | struct ctl_table *entry; | |
864 | struct ctl_dir *dir; | |
865 | int ret; | |
866 | ||
867 | /* Get out quickly if not a link */ | |
868 | if (!S_ISLNK((*pentry)->mode)) | |
869 | return 0; | |
870 | ||
871 | ret = 0; | |
872 | spin_lock(&sysctl_lock); | |
873 | root = (*pentry)->data; | |
874 | set = lookup_header_set(root, namespaces); | |
875 | dir = xlate_dir(set, (*phead)->parent); | |
876 | if (IS_ERR(dir)) | |
877 | ret = PTR_ERR(dir); | |
878 | else { | |
879 | const char *procname = (*pentry)->procname; | |
880 | head = NULL; | |
881 | entry = find_entry(&head, dir, procname, strlen(procname)); | |
882 | ret = -ENOENT; | |
883 | if (entry && use_table(head)) { | |
884 | unuse_table(*phead); | |
885 | *phead = head; | |
886 | *pentry = entry; | |
887 | ret = 0; | |
888 | } | |
889 | } | |
890 | ||
891 | spin_unlock(&sysctl_lock); | |
892 | return ret; | |
893 | } | |
894 | ||
e54012ce | 895 | static int sysctl_check_table_dups(struct ctl_dir *dir, struct ctl_table *old, |
7c60c48f | 896 | struct ctl_table *table) |
1f87f0b5 | 897 | { |
7c60c48f EB |
898 | struct ctl_table *entry, *test; |
899 | int error = 0; | |
1f87f0b5 | 900 | |
7c60c48f EB |
901 | for (entry = old; entry->procname; entry++) { |
902 | for (test = table; test->procname; test++) { | |
903 | if (strcmp(entry->procname, test->procname) == 0) { | |
e54012ce EB |
904 | printk(KERN_ERR "sysctl duplicate entry: "); |
905 | sysctl_print_dir(dir); | |
906 | printk(KERN_CONT "/%s\n", test->procname); | |
7c60c48f EB |
907 | error = -EEXIST; |
908 | } | |
909 | } | |
910 | } | |
911 | return error; | |
1f87f0b5 EB |
912 | } |
913 | ||
e54012ce | 914 | static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table) |
1f87f0b5 | 915 | { |
7ec66d06 | 916 | struct ctl_table_header *head; |
7c60c48f | 917 | int error = 0; |
1f87f0b5 | 918 | |
9e3d47df | 919 | list_for_each_entry(head, &dir->list, ctl_entry) { |
0e47c99d EB |
920 | if (head->unregistering) |
921 | continue; | |
922 | if (head->parent != dir) | |
923 | continue; | |
e54012ce | 924 | error = sysctl_check_table_dups(dir, head->ctl_table, table); |
0e47c99d | 925 | } |
7c60c48f | 926 | return error; |
1f87f0b5 EB |
927 | } |
928 | ||
7c60c48f | 929 | static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...) |
1f87f0b5 | 930 | { |
7c60c48f EB |
931 | struct va_format vaf; |
932 | va_list args; | |
1f87f0b5 | 933 | |
7c60c48f EB |
934 | va_start(args, fmt); |
935 | vaf.fmt = fmt; | |
936 | vaf.va = &args; | |
937 | ||
938 | printk(KERN_ERR "sysctl table check failed: %s/%s %pV\n", | |
939 | path, table->procname, &vaf); | |
1f87f0b5 | 940 | |
7c60c48f EB |
941 | va_end(args); |
942 | return -EINVAL; | |
1f87f0b5 EB |
943 | } |
944 | ||
7c60c48f | 945 | static int sysctl_check_table(const char *path, struct ctl_table *table) |
1f87f0b5 | 946 | { |
7c60c48f | 947 | int err = 0; |
1f87f0b5 | 948 | for (; table->procname; table++) { |
1f87f0b5 | 949 | if (table->child) |
7c60c48f EB |
950 | err = sysctl_err(path, table, "Not a file"); |
951 | ||
952 | if ((table->proc_handler == proc_dostring) || | |
953 | (table->proc_handler == proc_dointvec) || | |
954 | (table->proc_handler == proc_dointvec_minmax) || | |
955 | (table->proc_handler == proc_dointvec_jiffies) || | |
956 | (table->proc_handler == proc_dointvec_userhz_jiffies) || | |
957 | (table->proc_handler == proc_dointvec_ms_jiffies) || | |
958 | (table->proc_handler == proc_doulongvec_minmax) || | |
959 | (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) { | |
960 | if (!table->data) | |
961 | err = sysctl_err(path, table, "No data"); | |
962 | if (!table->maxlen) | |
963 | err = sysctl_err(path, table, "No maxlen"); | |
964 | } | |
965 | if (!table->proc_handler) | |
966 | err = sysctl_err(path, table, "No proc_handler"); | |
967 | ||
968 | if ((table->mode & (S_IRUGO|S_IWUGO)) != table->mode) | |
969 | err = sysctl_err(path, table, "bogus .mode 0%o", | |
970 | table->mode); | |
1f87f0b5 | 971 | } |
7c60c48f | 972 | return err; |
1f87f0b5 | 973 | } |
1f87f0b5 | 974 | |
0e47c99d EB |
975 | static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table *table, |
976 | struct ctl_table_root *link_root) | |
977 | { | |
978 | struct ctl_table *link_table, *entry, *link; | |
979 | struct ctl_table_header *links; | |
980 | char *link_name; | |
981 | int nr_entries, name_bytes; | |
982 | ||
983 | name_bytes = 0; | |
984 | nr_entries = 0; | |
985 | for (entry = table; entry->procname; entry++) { | |
986 | nr_entries++; | |
987 | name_bytes += strlen(entry->procname) + 1; | |
988 | } | |
989 | ||
990 | links = kzalloc(sizeof(struct ctl_table_header) + | |
991 | sizeof(struct ctl_table)*(nr_entries + 1) + | |
992 | name_bytes, | |
993 | GFP_KERNEL); | |
994 | ||
995 | if (!links) | |
996 | return NULL; | |
997 | ||
998 | link_table = (struct ctl_table *)(links + 1); | |
999 | link_name = (char *)&link_table[nr_entries + 1]; | |
1000 | ||
1001 | for (link = link_table, entry = table; entry->procname; link++, entry++) { | |
1002 | int len = strlen(entry->procname) + 1; | |
1003 | memcpy(link_name, entry->procname, len); | |
1004 | link->procname = link_name; | |
1005 | link->mode = S_IFLNK|S_IRWXUGO; | |
1006 | link->data = link_root; | |
1007 | link_name += len; | |
1008 | } | |
1009 | init_header(links, dir->header.root, dir->header.set, link_table); | |
1010 | links->nreg = nr_entries; | |
1011 | ||
1012 | return links; | |
1013 | } | |
1014 | ||
1015 | static bool get_links(struct ctl_dir *dir, | |
1016 | struct ctl_table *table, struct ctl_table_root *link_root) | |
1017 | { | |
1018 | struct ctl_table_header *head; | |
1019 | struct ctl_table *entry, *link; | |
1020 | ||
1021 | /* Are there links available for every entry in table? */ | |
1022 | for (entry = table; entry->procname; entry++) { | |
1023 | const char *procname = entry->procname; | |
1024 | link = find_entry(&head, dir, procname, strlen(procname)); | |
1025 | if (!link) | |
1026 | return false; | |
1027 | if (S_ISDIR(link->mode) && S_ISDIR(entry->mode)) | |
1028 | continue; | |
1029 | if (S_ISLNK(link->mode) && (link->data == link_root)) | |
1030 | continue; | |
1031 | return false; | |
1032 | } | |
1033 | ||
1034 | /* The checks passed. Increase the registration count on the links */ | |
1035 | for (entry = table; entry->procname; entry++) { | |
1036 | const char *procname = entry->procname; | |
1037 | link = find_entry(&head, dir, procname, strlen(procname)); | |
1038 | head->nreg++; | |
1039 | } | |
1040 | return true; | |
1041 | } | |
1042 | ||
1043 | static int insert_links(struct ctl_table_header *head) | |
1044 | { | |
1045 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
1046 | struct ctl_dir *core_parent = NULL; | |
1047 | struct ctl_table_header *links; | |
1048 | int err; | |
1049 | ||
1050 | if (head->set == root_set) | |
1051 | return 0; | |
1052 | ||
1053 | core_parent = xlate_dir(root_set, head->parent); | |
1054 | if (IS_ERR(core_parent)) | |
1055 | return 0; | |
1056 | ||
1057 | if (get_links(core_parent, head->ctl_table, head->root)) | |
1058 | return 0; | |
1059 | ||
1060 | core_parent->header.nreg++; | |
1061 | spin_unlock(&sysctl_lock); | |
1062 | ||
1063 | links = new_links(core_parent, head->ctl_table, head->root); | |
1064 | ||
1065 | spin_lock(&sysctl_lock); | |
1066 | err = -ENOMEM; | |
1067 | if (!links) | |
1068 | goto out; | |
1069 | ||
1070 | err = 0; | |
1071 | if (get_links(core_parent, head->ctl_table, head->root)) { | |
1072 | kfree(links); | |
1073 | goto out; | |
1074 | } | |
1075 | ||
1076 | err = insert_header(core_parent, links); | |
1077 | if (err) | |
1078 | kfree(links); | |
1079 | out: | |
1080 | drop_sysctl_table(&core_parent->header); | |
1081 | return err; | |
1082 | } | |
1083 | ||
1f87f0b5 | 1084 | /** |
f728019b | 1085 | * __register_sysctl_table - register a leaf sysctl table |
60a47a2e | 1086 | * @set: Sysctl tree to register on |
1f87f0b5 EB |
1087 | * @path: The path to the directory the sysctl table is in. |
1088 | * @table: the top-level table structure | |
1089 | * | |
1090 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1091 | * array. A completely 0 filled entry terminates the table. | |
1092 | * | |
1093 | * The members of the &struct ctl_table structure are used as follows: | |
1094 | * | |
1095 | * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not | |
1096 | * enter a sysctl file | |
1097 | * | |
1098 | * data - a pointer to data for use by proc_handler | |
1099 | * | |
1100 | * maxlen - the maximum size in bytes of the data | |
1101 | * | |
f728019b | 1102 | * mode - the file permissions for the /proc/sys file |
1f87f0b5 | 1103 | * |
f728019b | 1104 | * child - must be %NULL. |
1f87f0b5 EB |
1105 | * |
1106 | * proc_handler - the text handler routine (described below) | |
1107 | * | |
1f87f0b5 EB |
1108 | * extra1, extra2 - extra pointers usable by the proc handler routines |
1109 | * | |
1110 | * Leaf nodes in the sysctl tree will be represented by a single file | |
1111 | * under /proc; non-leaf nodes will be represented by directories. | |
1112 | * | |
f728019b EB |
1113 | * There must be a proc_handler routine for any terminal nodes. |
1114 | * Several default handlers are available to cover common cases - | |
1f87f0b5 EB |
1115 | * |
1116 | * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), | |
1117 | * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), | |
1118 | * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() | |
1119 | * | |
1120 | * It is the handler's job to read the input buffer from user memory | |
1121 | * and process it. The handler should return 0 on success. | |
1122 | * | |
1123 | * This routine returns %NULL on a failure to register, and a pointer | |
1124 | * to the table header on success. | |
1125 | */ | |
6e9d5164 | 1126 | struct ctl_table_header *__register_sysctl_table( |
60a47a2e | 1127 | struct ctl_table_set *set, |
6e9d5164 | 1128 | const char *path, struct ctl_table *table) |
1f87f0b5 | 1129 | { |
60a47a2e | 1130 | struct ctl_table_root *root = set->dir.header.root; |
0e47c99d | 1131 | struct ctl_table_header *links = NULL; |
1f87f0b5 | 1132 | struct ctl_table_header *header; |
6e9d5164 | 1133 | const char *name, *nextname; |
7ec66d06 | 1134 | struct ctl_dir *dir; |
1f87f0b5 | 1135 | |
7ec66d06 | 1136 | header = kzalloc(sizeof(struct ctl_table_header), GFP_KERNEL); |
1f87f0b5 EB |
1137 | if (!header) |
1138 | return NULL; | |
1139 | ||
60a47a2e | 1140 | init_header(header, root, set, table); |
7ec66d06 EB |
1141 | if (sysctl_check_table(path, table)) |
1142 | goto fail; | |
1143 | ||
1144 | spin_lock(&sysctl_lock); | |
0e47c99d | 1145 | dir = &set->dir; |
7ec66d06 EB |
1146 | dir->header.nreg++; |
1147 | spin_unlock(&sysctl_lock); | |
1f87f0b5 | 1148 | |
7ec66d06 | 1149 | /* Find the directory for the ctl_table */ |
6e9d5164 EB |
1150 | for (name = path; name; name = nextname) { |
1151 | int namelen; | |
1152 | nextname = strchr(name, '/'); | |
1153 | if (nextname) { | |
1154 | namelen = nextname - name; | |
1155 | nextname++; | |
1156 | } else { | |
1157 | namelen = strlen(name); | |
1158 | } | |
1159 | if (namelen == 0) | |
1160 | continue; | |
1f87f0b5 | 1161 | |
0e47c99d | 1162 | dir = get_subdir(dir, name, namelen); |
7ec66d06 EB |
1163 | if (IS_ERR(dir)) |
1164 | goto fail; | |
1f87f0b5 | 1165 | } |
0e47c99d | 1166 | |
1f87f0b5 | 1167 | spin_lock(&sysctl_lock); |
0e47c99d | 1168 | if (insert_header(dir, header)) |
7ec66d06 | 1169 | goto fail_put_dir_locked; |
0e47c99d | 1170 | |
7ec66d06 | 1171 | drop_sysctl_table(&dir->header); |
1f87f0b5 EB |
1172 | spin_unlock(&sysctl_lock); |
1173 | ||
1174 | return header; | |
0e47c99d | 1175 | |
7ec66d06 EB |
1176 | fail_put_dir_locked: |
1177 | drop_sysctl_table(&dir->header); | |
7c60c48f EB |
1178 | spin_unlock(&sysctl_lock); |
1179 | fail: | |
0e47c99d | 1180 | kfree(links); |
7c60c48f EB |
1181 | kfree(header); |
1182 | dump_stack(); | |
1183 | return NULL; | |
1f87f0b5 EB |
1184 | } |
1185 | ||
6e9d5164 EB |
1186 | static char *append_path(const char *path, char *pos, const char *name) |
1187 | { | |
1188 | int namelen; | |
1189 | namelen = strlen(name); | |
1190 | if (((pos - path) + namelen + 2) >= PATH_MAX) | |
1191 | return NULL; | |
1192 | memcpy(pos, name, namelen); | |
1193 | pos[namelen] = '/'; | |
1194 | pos[namelen + 1] = '\0'; | |
1195 | pos += namelen + 1; | |
1196 | return pos; | |
1197 | } | |
1198 | ||
f728019b EB |
1199 | static int count_subheaders(struct ctl_table *table) |
1200 | { | |
1201 | int has_files = 0; | |
1202 | int nr_subheaders = 0; | |
1203 | struct ctl_table *entry; | |
1204 | ||
1205 | /* special case: no directory and empty directory */ | |
1206 | if (!table || !table->procname) | |
1207 | return 1; | |
1208 | ||
1209 | for (entry = table; entry->procname; entry++) { | |
1210 | if (entry->child) | |
1211 | nr_subheaders += count_subheaders(entry->child); | |
1212 | else | |
1213 | has_files = 1; | |
1214 | } | |
1215 | return nr_subheaders + has_files; | |
1216 | } | |
1217 | ||
1218 | static int register_leaf_sysctl_tables(const char *path, char *pos, | |
60a47a2e | 1219 | struct ctl_table_header ***subheader, struct ctl_table_set *set, |
f728019b EB |
1220 | struct ctl_table *table) |
1221 | { | |
1222 | struct ctl_table *ctl_table_arg = NULL; | |
1223 | struct ctl_table *entry, *files; | |
1224 | int nr_files = 0; | |
1225 | int nr_dirs = 0; | |
1226 | int err = -ENOMEM; | |
1227 | ||
1228 | for (entry = table; entry->procname; entry++) { | |
1229 | if (entry->child) | |
1230 | nr_dirs++; | |
1231 | else | |
1232 | nr_files++; | |
1233 | } | |
1234 | ||
1235 | files = table; | |
1236 | /* If there are mixed files and directories we need a new table */ | |
1237 | if (nr_dirs && nr_files) { | |
1238 | struct ctl_table *new; | |
1239 | files = kzalloc(sizeof(struct ctl_table) * (nr_files + 1), | |
1240 | GFP_KERNEL); | |
1241 | if (!files) | |
1242 | goto out; | |
1243 | ||
1244 | ctl_table_arg = files; | |
1245 | for (new = files, entry = table; entry->procname; entry++) { | |
1246 | if (entry->child) | |
1247 | continue; | |
1248 | *new = *entry; | |
1249 | new++; | |
1250 | } | |
1251 | } | |
1252 | ||
1253 | /* Register everything except a directory full of subdirectories */ | |
1254 | if (nr_files || !nr_dirs) { | |
1255 | struct ctl_table_header *header; | |
60a47a2e | 1256 | header = __register_sysctl_table(set, path, files); |
f728019b EB |
1257 | if (!header) { |
1258 | kfree(ctl_table_arg); | |
1259 | goto out; | |
1260 | } | |
1261 | ||
1262 | /* Remember if we need to free the file table */ | |
1263 | header->ctl_table_arg = ctl_table_arg; | |
1264 | **subheader = header; | |
1265 | (*subheader)++; | |
1266 | } | |
1267 | ||
1268 | /* Recurse into the subdirectories. */ | |
1269 | for (entry = table; entry->procname; entry++) { | |
1270 | char *child_pos; | |
1271 | ||
1272 | if (!entry->child) | |
1273 | continue; | |
1274 | ||
1275 | err = -ENAMETOOLONG; | |
1276 | child_pos = append_path(path, pos, entry->procname); | |
1277 | if (!child_pos) | |
1278 | goto out; | |
1279 | ||
1280 | err = register_leaf_sysctl_tables(path, child_pos, subheader, | |
60a47a2e | 1281 | set, entry->child); |
f728019b EB |
1282 | pos[0] = '\0'; |
1283 | if (err) | |
1284 | goto out; | |
1285 | } | |
1286 | err = 0; | |
1287 | out: | |
1288 | /* On failure our caller will unregister all registered subheaders */ | |
1289 | return err; | |
1290 | } | |
1291 | ||
6e9d5164 EB |
1292 | /** |
1293 | * __register_sysctl_paths - register a sysctl table hierarchy | |
60a47a2e | 1294 | * @set: Sysctl tree to register on |
6e9d5164 EB |
1295 | * @path: The path to the directory the sysctl table is in. |
1296 | * @table: the top-level table structure | |
1297 | * | |
1298 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1299 | * array. A completely 0 filled entry terminates the table. | |
1300 | * | |
1301 | * See __register_sysctl_table for more details. | |
1302 | */ | |
1303 | struct ctl_table_header *__register_sysctl_paths( | |
60a47a2e | 1304 | struct ctl_table_set *set, |
6e9d5164 EB |
1305 | const struct ctl_path *path, struct ctl_table *table) |
1306 | { | |
ec6a5266 | 1307 | struct ctl_table *ctl_table_arg = table; |
f728019b EB |
1308 | int nr_subheaders = count_subheaders(table); |
1309 | struct ctl_table_header *header = NULL, **subheaders, **subheader; | |
6e9d5164 EB |
1310 | const struct ctl_path *component; |
1311 | char *new_path, *pos; | |
1312 | ||
1313 | pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL); | |
1314 | if (!new_path) | |
1315 | return NULL; | |
1316 | ||
1317 | pos[0] = '\0'; | |
1318 | for (component = path; component->procname; component++) { | |
1319 | pos = append_path(new_path, pos, component->procname); | |
1320 | if (!pos) | |
1321 | goto out; | |
1322 | } | |
ec6a5266 EB |
1323 | while (table->procname && table->child && !table[1].procname) { |
1324 | pos = append_path(new_path, pos, table->procname); | |
1325 | if (!pos) | |
1326 | goto out; | |
1327 | table = table->child; | |
1328 | } | |
f728019b | 1329 | if (nr_subheaders == 1) { |
60a47a2e | 1330 | header = __register_sysctl_table(set, new_path, table); |
f728019b EB |
1331 | if (header) |
1332 | header->ctl_table_arg = ctl_table_arg; | |
1333 | } else { | |
1334 | header = kzalloc(sizeof(*header) + | |
1335 | sizeof(*subheaders)*nr_subheaders, GFP_KERNEL); | |
1336 | if (!header) | |
1337 | goto out; | |
1338 | ||
1339 | subheaders = (struct ctl_table_header **) (header + 1); | |
1340 | subheader = subheaders; | |
ec6a5266 | 1341 | header->ctl_table_arg = ctl_table_arg; |
f728019b EB |
1342 | |
1343 | if (register_leaf_sysctl_tables(new_path, pos, &subheader, | |
60a47a2e | 1344 | set, table)) |
f728019b EB |
1345 | goto err_register_leaves; |
1346 | } | |
1347 | ||
6e9d5164 EB |
1348 | out: |
1349 | kfree(new_path); | |
1350 | return header; | |
f728019b EB |
1351 | |
1352 | err_register_leaves: | |
1353 | while (subheader > subheaders) { | |
1354 | struct ctl_table_header *subh = *(--subheader); | |
1355 | struct ctl_table *table = subh->ctl_table_arg; | |
1356 | unregister_sysctl_table(subh); | |
1357 | kfree(table); | |
1358 | } | |
1359 | kfree(header); | |
1360 | header = NULL; | |
1361 | goto out; | |
6e9d5164 EB |
1362 | } |
1363 | ||
1f87f0b5 EB |
1364 | /** |
1365 | * register_sysctl_table_path - register a sysctl table hierarchy | |
1366 | * @path: The path to the directory the sysctl table is in. | |
1367 | * @table: the top-level table structure | |
1368 | * | |
1369 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1370 | * array. A completely 0 filled entry terminates the table. | |
1371 | * | |
1372 | * See __register_sysctl_paths for more details. | |
1373 | */ | |
1374 | struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path, | |
1375 | struct ctl_table *table) | |
1376 | { | |
60a47a2e | 1377 | return __register_sysctl_paths(&sysctl_table_root.default_set, |
1f87f0b5 EB |
1378 | path, table); |
1379 | } | |
1380 | EXPORT_SYMBOL(register_sysctl_paths); | |
1381 | ||
1382 | /** | |
1383 | * register_sysctl_table - register a sysctl table hierarchy | |
1384 | * @table: the top-level table structure | |
1385 | * | |
1386 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1387 | * array. A completely 0 filled entry terminates the table. | |
1388 | * | |
1389 | * See register_sysctl_paths for more details. | |
1390 | */ | |
1391 | struct ctl_table_header *register_sysctl_table(struct ctl_table *table) | |
1392 | { | |
1393 | static const struct ctl_path null_path[] = { {} }; | |
1394 | ||
1395 | return register_sysctl_paths(null_path, table); | |
1396 | } | |
1397 | EXPORT_SYMBOL(register_sysctl_table); | |
1398 | ||
0e47c99d EB |
1399 | static void put_links(struct ctl_table_header *header) |
1400 | { | |
1401 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
1402 | struct ctl_table_root *root = header->root; | |
1403 | struct ctl_dir *parent = header->parent; | |
1404 | struct ctl_dir *core_parent; | |
1405 | struct ctl_table *entry; | |
1406 | ||
1407 | if (header->set == root_set) | |
1408 | return; | |
1409 | ||
1410 | core_parent = xlate_dir(root_set, parent); | |
1411 | if (IS_ERR(core_parent)) | |
1412 | return; | |
1413 | ||
1414 | for (entry = header->ctl_table; entry->procname; entry++) { | |
1415 | struct ctl_table_header *link_head; | |
1416 | struct ctl_table *link; | |
1417 | const char *name = entry->procname; | |
1418 | ||
1419 | link = find_entry(&link_head, core_parent, name, strlen(name)); | |
1420 | if (link && | |
1421 | ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) || | |
1422 | (S_ISLNK(link->mode) && (link->data == root)))) { | |
1423 | drop_sysctl_table(link_head); | |
1424 | } | |
1425 | else { | |
1426 | printk(KERN_ERR "sysctl link missing during unregister: "); | |
1427 | sysctl_print_dir(parent); | |
1428 | printk(KERN_CONT "/%s\n", name); | |
1429 | } | |
1430 | } | |
1431 | } | |
1432 | ||
938aaa4f EB |
1433 | static void drop_sysctl_table(struct ctl_table_header *header) |
1434 | { | |
7ec66d06 EB |
1435 | struct ctl_dir *parent = header->parent; |
1436 | ||
938aaa4f EB |
1437 | if (--header->nreg) |
1438 | return; | |
1439 | ||
0e47c99d | 1440 | put_links(header); |
938aaa4f | 1441 | start_unregistering(header); |
938aaa4f EB |
1442 | if (!--header->count) |
1443 | kfree_rcu(header, rcu); | |
7ec66d06 EB |
1444 | |
1445 | if (parent) | |
1446 | drop_sysctl_table(&parent->header); | |
938aaa4f EB |
1447 | } |
1448 | ||
1f87f0b5 EB |
1449 | /** |
1450 | * unregister_sysctl_table - unregister a sysctl table hierarchy | |
1451 | * @header: the header returned from register_sysctl_table | |
1452 | * | |
1453 | * Unregisters the sysctl table and all children. proc entries may not | |
1454 | * actually be removed until they are no longer used by anyone. | |
1455 | */ | |
1456 | void unregister_sysctl_table(struct ctl_table_header * header) | |
1457 | { | |
f728019b | 1458 | int nr_subheaders; |
1f87f0b5 EB |
1459 | might_sleep(); |
1460 | ||
1461 | if (header == NULL) | |
1462 | return; | |
1463 | ||
f728019b EB |
1464 | nr_subheaders = count_subheaders(header->ctl_table_arg); |
1465 | if (unlikely(nr_subheaders > 1)) { | |
1466 | struct ctl_table_header **subheaders; | |
1467 | int i; | |
1468 | ||
1469 | subheaders = (struct ctl_table_header **)(header + 1); | |
1470 | for (i = nr_subheaders -1; i >= 0; i--) { | |
1471 | struct ctl_table_header *subh = subheaders[i]; | |
1472 | struct ctl_table *table = subh->ctl_table_arg; | |
1473 | unregister_sysctl_table(subh); | |
1474 | kfree(table); | |
1475 | } | |
1476 | kfree(header); | |
1477 | return; | |
1478 | } | |
1479 | ||
1f87f0b5 | 1480 | spin_lock(&sysctl_lock); |
938aaa4f | 1481 | drop_sysctl_table(header); |
1f87f0b5 EB |
1482 | spin_unlock(&sysctl_lock); |
1483 | } | |
1484 | EXPORT_SYMBOL(unregister_sysctl_table); | |
1485 | ||
0e47c99d | 1486 | void setup_sysctl_set(struct ctl_table_set *set, |
9eb47c26 | 1487 | struct ctl_table_root *root, |
1f87f0b5 EB |
1488 | int (*is_seen)(struct ctl_table_set *)) |
1489 | { | |
0e47c99d | 1490 | memset(set, sizeof(*set), 0); |
0e47c99d | 1491 | set->is_seen = is_seen; |
9e3d47df | 1492 | INIT_LIST_HEAD(&set->dir.list); |
0e47c99d | 1493 | init_header(&set->dir.header, root, set, root_table); |
1f87f0b5 EB |
1494 | } |
1495 | ||
97324cd8 EB |
1496 | void retire_sysctl_set(struct ctl_table_set *set) |
1497 | { | |
9e3d47df | 1498 | WARN_ON(!list_empty(&set->dir.list)); |
97324cd8 | 1499 | } |
1f87f0b5 | 1500 | |
1e0edd3f | 1501 | int __init proc_sys_init(void) |
77b14db5 | 1502 | { |
e1675231 AD |
1503 | struct proc_dir_entry *proc_sys_root; |
1504 | ||
77b14db5 | 1505 | proc_sys_root = proc_mkdir("sys", NULL); |
9043476f AV |
1506 | proc_sys_root->proc_iops = &proc_sys_dir_operations; |
1507 | proc_sys_root->proc_fops = &proc_sys_dir_file_operations; | |
77b14db5 | 1508 | proc_sys_root->nlink = 0; |
de4e83bd EB |
1509 | |
1510 | return sysctl_init(); | |
77b14db5 | 1511 | } |