Avoid beyond bounds copy while caching ACL
[zen-stable.git] / fs / proc / proc_sysctl.c
blob53c3bcea56cf4e554f333cbcc01d2b48861a7391
1 /*
2 * /proc/sys support
3 */
4 #include <linux/init.h>
5 #include <linux/sysctl.h>
6 #include <linux/poll.h>
7 #include <linux/proc_fs.h>
8 #include <linux/security.h>
9 #include <linux/namei.h>
10 #include "internal.h"
12 static const struct dentry_operations proc_sys_dentry_operations;
13 static const struct file_operations proc_sys_file_operations;
14 static const struct inode_operations proc_sys_inode_operations;
15 static const struct file_operations proc_sys_dir_file_operations;
16 static const struct inode_operations proc_sys_dir_operations;
18 void proc_sys_poll_notify(struct ctl_table_poll *poll)
20 if (!poll)
21 return;
23 atomic_inc(&poll->event);
24 wake_up_interruptible(&poll->wait);
27 static struct inode *proc_sys_make_inode(struct super_block *sb,
28 struct ctl_table_header *head, struct ctl_table *table)
30 struct inode *inode;
31 struct proc_inode *ei;
33 inode = new_inode(sb);
34 if (!inode)
35 goto out;
37 inode->i_ino = get_next_ino();
39 sysctl_head_get(head);
40 ei = PROC_I(inode);
41 ei->sysctl = head;
42 ei->sysctl_entry = table;
44 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
45 inode->i_mode = table->mode;
46 if (!table->child) {
47 inode->i_mode |= S_IFREG;
48 inode->i_op = &proc_sys_inode_operations;
49 inode->i_fop = &proc_sys_file_operations;
50 } else {
51 inode->i_mode |= S_IFDIR;
52 clear_nlink(inode);
53 inode->i_op = &proc_sys_dir_operations;
54 inode->i_fop = &proc_sys_dir_file_operations;
56 out:
57 return inode;
60 static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
62 int len;
63 for ( ; p->procname; p++) {
65 if (!p->procname)
66 continue;
68 len = strlen(p->procname);
69 if (len != name->len)
70 continue;
72 if (memcmp(p->procname, name->name, len) != 0)
73 continue;
75 /* I have a match */
76 return p;
78 return NULL;
81 static struct ctl_table_header *grab_header(struct inode *inode)
83 if (PROC_I(inode)->sysctl)
84 return sysctl_head_grab(PROC_I(inode)->sysctl);
85 else
86 return sysctl_head_next(NULL);
89 static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
90 struct nameidata *nd)
92 struct ctl_table_header *head = grab_header(dir);
93 struct ctl_table *table = PROC_I(dir)->sysctl_entry;
94 struct ctl_table_header *h = NULL;
95 struct qstr *name = &dentry->d_name;
96 struct ctl_table *p;
97 struct inode *inode;
98 struct dentry *err = ERR_PTR(-ENOENT);
100 if (IS_ERR(head))
101 return ERR_CAST(head);
103 if (table && !table->child) {
104 WARN_ON(1);
105 goto out;
108 table = table ? table->child : head->ctl_table;
110 p = find_in_table(table, name);
111 if (!p) {
112 for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
113 if (h->attached_to != table)
114 continue;
115 p = find_in_table(h->attached_by, name);
116 if (p)
117 break;
121 if (!p)
122 goto out;
124 err = ERR_PTR(-ENOMEM);
125 inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p);
126 if (h)
127 sysctl_head_finish(h);
129 if (!inode)
130 goto out;
132 err = NULL;
133 d_set_d_op(dentry, &proc_sys_dentry_operations);
134 d_add(dentry, inode);
136 out:
137 sysctl_head_finish(head);
138 return err;
141 static ssize_t proc_sys_call_handler(struct file *filp, void __user *buf,
142 size_t count, loff_t *ppos, int write)
144 struct inode *inode = filp->f_path.dentry->d_inode;
145 struct ctl_table_header *head = grab_header(inode);
146 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
147 ssize_t error;
148 size_t res;
150 if (IS_ERR(head))
151 return PTR_ERR(head);
154 * At this point we know that the sysctl was not unregistered
155 * and won't be until we finish.
157 error = -EPERM;
158 if (sysctl_perm(head->root, table, write ? MAY_WRITE : MAY_READ))
159 goto out;
161 /* if that can happen at all, it should be -EINVAL, not -EISDIR */
162 error = -EINVAL;
163 if (!table->proc_handler)
164 goto out;
166 /* careful: calling conventions are nasty here */
167 res = count;
168 error = table->proc_handler(table, write, buf, &res, ppos);
169 if (!error)
170 error = res;
171 out:
172 sysctl_head_finish(head);
174 return error;
177 static ssize_t proc_sys_read(struct file *filp, char __user *buf,
178 size_t count, loff_t *ppos)
180 return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 0);
183 static ssize_t proc_sys_write(struct file *filp, const char __user *buf,
184 size_t count, loff_t *ppos)
186 return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 1);
189 static int proc_sys_open(struct inode *inode, struct file *filp)
191 struct ctl_table_header *head = grab_header(inode);
192 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
194 /* sysctl was unregistered */
195 if (IS_ERR(head))
196 return PTR_ERR(head);
198 if (table->poll)
199 filp->private_data = proc_sys_poll_event(table->poll);
201 sysctl_head_finish(head);
203 return 0;
206 static unsigned int proc_sys_poll(struct file *filp, poll_table *wait)
208 struct inode *inode = filp->f_path.dentry->d_inode;
209 struct ctl_table_header *head = grab_header(inode);
210 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
211 unsigned int ret = DEFAULT_POLLMASK;
212 unsigned long event;
214 /* sysctl was unregistered */
215 if (IS_ERR(head))
216 return POLLERR | POLLHUP;
218 if (!table->proc_handler)
219 goto out;
221 if (!table->poll)
222 goto out;
224 event = (unsigned long)filp->private_data;
225 poll_wait(filp, &table->poll->wait, wait);
227 if (event != atomic_read(&table->poll->event)) {
228 filp->private_data = proc_sys_poll_event(table->poll);
229 ret = POLLIN | POLLRDNORM | POLLERR | POLLPRI;
232 out:
233 sysctl_head_finish(head);
235 return ret;
238 static int proc_sys_fill_cache(struct file *filp, void *dirent,
239 filldir_t filldir,
240 struct ctl_table_header *head,
241 struct ctl_table *table)
243 struct dentry *child, *dir = filp->f_path.dentry;
244 struct inode *inode;
245 struct qstr qname;
246 ino_t ino = 0;
247 unsigned type = DT_UNKNOWN;
249 qname.name = table->procname;
250 qname.len = strlen(table->procname);
251 qname.hash = full_name_hash(qname.name, qname.len);
253 child = d_lookup(dir, &qname);
254 if (!child) {
255 child = d_alloc(dir, &qname);
256 if (child) {
257 inode = proc_sys_make_inode(dir->d_sb, head, table);
258 if (!inode) {
259 dput(child);
260 return -ENOMEM;
261 } else {
262 d_set_d_op(child, &proc_sys_dentry_operations);
263 d_add(child, inode);
265 } else {
266 return -ENOMEM;
269 inode = child->d_inode;
270 ino = inode->i_ino;
271 type = inode->i_mode >> 12;
272 dput(child);
273 return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type);
276 static int scan(struct ctl_table_header *head, ctl_table *table,
277 unsigned long *pos, struct file *file,
278 void *dirent, filldir_t filldir)
281 for (; table->procname; table++, (*pos)++) {
282 int res;
284 /* Can't do anything without a proc name */
285 if (!table->procname)
286 continue;
288 if (*pos < file->f_pos)
289 continue;
291 res = proc_sys_fill_cache(file, dirent, filldir, head, table);
292 if (res)
293 return res;
295 file->f_pos = *pos + 1;
297 return 0;
300 static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
302 struct dentry *dentry = filp->f_path.dentry;
303 struct inode *inode = dentry->d_inode;
304 struct ctl_table_header *head = grab_header(inode);
305 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
306 struct ctl_table_header *h = NULL;
307 unsigned long pos;
308 int ret = -EINVAL;
310 if (IS_ERR(head))
311 return PTR_ERR(head);
313 if (table && !table->child) {
314 WARN_ON(1);
315 goto out;
318 table = table ? table->child : head->ctl_table;
320 ret = 0;
321 /* Avoid a switch here: arm builds fail with missing __cmpdi2 */
322 if (filp->f_pos == 0) {
323 if (filldir(dirent, ".", 1, filp->f_pos,
324 inode->i_ino, DT_DIR) < 0)
325 goto out;
326 filp->f_pos++;
328 if (filp->f_pos == 1) {
329 if (filldir(dirent, "..", 2, filp->f_pos,
330 parent_ino(dentry), DT_DIR) < 0)
331 goto out;
332 filp->f_pos++;
334 pos = 2;
336 ret = scan(head, table, &pos, filp, dirent, filldir);
337 if (ret)
338 goto out;
340 for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
341 if (h->attached_to != table)
342 continue;
343 ret = scan(h, h->attached_by, &pos, filp, dirent, filldir);
344 if (ret) {
345 sysctl_head_finish(h);
346 break;
349 ret = 1;
350 out:
351 sysctl_head_finish(head);
352 return ret;
355 static int proc_sys_permission(struct inode *inode, int mask)
358 * sysctl entries that are not writeable,
359 * are _NOT_ writeable, capabilities or not.
361 struct ctl_table_header *head;
362 struct ctl_table *table;
363 int error;
365 /* Executable files are not allowed under /proc/sys/ */
366 if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode))
367 return -EACCES;
369 head = grab_header(inode);
370 if (IS_ERR(head))
371 return PTR_ERR(head);
373 table = PROC_I(inode)->sysctl_entry;
374 if (!table) /* global root - r-xr-xr-x */
375 error = mask & MAY_WRITE ? -EACCES : 0;
376 else /* Use the permissions on the sysctl table entry */
377 error = sysctl_perm(head->root, table, mask & ~MAY_NOT_BLOCK);
379 sysctl_head_finish(head);
380 return error;
383 static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr)
385 struct inode *inode = dentry->d_inode;
386 int error;
388 if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID))
389 return -EPERM;
391 error = inode_change_ok(inode, attr);
392 if (error)
393 return error;
395 if ((attr->ia_valid & ATTR_SIZE) &&
396 attr->ia_size != i_size_read(inode)) {
397 error = vmtruncate(inode, attr->ia_size);
398 if (error)
399 return error;
402 setattr_copy(inode, attr);
403 mark_inode_dirty(inode);
404 return 0;
407 static int proc_sys_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
409 struct inode *inode = dentry->d_inode;
410 struct ctl_table_header *head = grab_header(inode);
411 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
413 if (IS_ERR(head))
414 return PTR_ERR(head);
416 generic_fillattr(inode, stat);
417 if (table)
418 stat->mode = (stat->mode & S_IFMT) | table->mode;
420 sysctl_head_finish(head);
421 return 0;
424 static const struct file_operations proc_sys_file_operations = {
425 .open = proc_sys_open,
426 .poll = proc_sys_poll,
427 .read = proc_sys_read,
428 .write = proc_sys_write,
429 .llseek = default_llseek,
432 static const struct file_operations proc_sys_dir_file_operations = {
433 .read = generic_read_dir,
434 .readdir = proc_sys_readdir,
435 .llseek = generic_file_llseek,
438 static const struct inode_operations proc_sys_inode_operations = {
439 .permission = proc_sys_permission,
440 .setattr = proc_sys_setattr,
441 .getattr = proc_sys_getattr,
444 static const struct inode_operations proc_sys_dir_operations = {
445 .lookup = proc_sys_lookup,
446 .permission = proc_sys_permission,
447 .setattr = proc_sys_setattr,
448 .getattr = proc_sys_getattr,
451 static int proc_sys_revalidate(struct dentry *dentry, struct nameidata *nd)
453 if (nd->flags & LOOKUP_RCU)
454 return -ECHILD;
455 return !PROC_I(dentry->d_inode)->sysctl->unregistering;
458 static int proc_sys_delete(const struct dentry *dentry)
460 return !!PROC_I(dentry->d_inode)->sysctl->unregistering;
463 static int proc_sys_compare(const struct dentry *parent,
464 const struct inode *pinode,
465 const struct dentry *dentry, const struct inode *inode,
466 unsigned int len, const char *str, const struct qstr *name)
468 struct ctl_table_header *head;
469 /* Although proc doesn't have negative dentries, rcu-walk means
470 * that inode here can be NULL */
471 /* AV: can it, indeed? */
472 if (!inode)
473 return 1;
474 if (name->len != len)
475 return 1;
476 if (memcmp(name->name, str, len))
477 return 1;
478 head = rcu_dereference(PROC_I(inode)->sysctl);
479 return !head || !sysctl_is_seen(head);
482 static const struct dentry_operations proc_sys_dentry_operations = {
483 .d_revalidate = proc_sys_revalidate,
484 .d_delete = proc_sys_delete,
485 .d_compare = proc_sys_compare,
488 int __init proc_sys_init(void)
490 struct proc_dir_entry *proc_sys_root;
492 proc_sys_root = proc_mkdir("sys", NULL);
493 proc_sys_root->proc_iops = &proc_sys_dir_operations;
494 proc_sys_root->proc_fops = &proc_sys_dir_file_operations;
495 proc_sys_root->nlink = 0;
496 return 0;