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>
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)
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)
31 struct proc_inode *ei;
33 inode = new_inode(sb);
37 inode->i_ino = get_next_ino();
39 sysctl_head_get(head);
42 ei->sysctl_entry = table;
44 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
45 inode->i_mode = table->mode;
47 inode->i_mode |= S_IFREG;
48 inode->i_op = &proc_sys_inode_operations;
49 inode->i_fop = &proc_sys_file_operations;
51 inode->i_mode |= S_IFDIR;
53 inode->i_op = &proc_sys_dir_operations;
54 inode->i_fop = &proc_sys_dir_file_operations;
60 static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
63 for ( ; p->procname; p++) {
68 len = strlen(p->procname);
72 if (memcmp(p->procname, name->name, len) != 0)
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);
86 return sysctl_head_next(NULL);
89 static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
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;
98 struct dentry *err = ERR_PTR(-ENOENT);
101 return ERR_CAST(head);
103 if (table && !table->child) {
108 table = table ? table->child : head->ctl_table;
110 p = find_in_table(table, name);
112 for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
113 if (h->attached_to != table)
115 p = find_in_table(h->attached_by, name);
124 err = ERR_PTR(-ENOMEM);
125 inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p);
130 d_set_d_op(dentry, &proc_sys_dentry_operations);
131 d_add(dentry, inode);
135 sysctl_head_finish(h);
136 sysctl_head_finish(head);
140 static ssize_t proc_sys_call_handler(struct file *filp, void __user *buf,
141 size_t count, loff_t *ppos, int write)
143 struct inode *inode = filp->f_path.dentry->d_inode;
144 struct ctl_table_header *head = grab_header(inode);
145 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
150 return PTR_ERR(head);
153 * At this point we know that the sysctl was not unregistered
154 * and won't be until we finish.
157 if (sysctl_perm(head->root, table, write ? MAY_WRITE : MAY_READ))
160 /* if that can happen at all, it should be -EINVAL, not -EISDIR */
162 if (!table->proc_handler)
165 /* careful: calling conventions are nasty here */
167 error = table->proc_handler(table, write, buf, &res, ppos);
171 sysctl_head_finish(head);
176 static ssize_t proc_sys_read(struct file *filp, char __user *buf,
177 size_t count, loff_t *ppos)
179 return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 0);
182 static ssize_t proc_sys_write(struct file *filp, const char __user *buf,
183 size_t count, loff_t *ppos)
185 return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 1);
188 static int proc_sys_open(struct inode *inode, struct file *filp)
190 struct ctl_table_header *head = grab_header(inode);
191 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
193 /* sysctl was unregistered */
195 return PTR_ERR(head);
198 filp->private_data = proc_sys_poll_event(table->poll);
200 sysctl_head_finish(head);
205 static unsigned int proc_sys_poll(struct file *filp, poll_table *wait)
207 struct inode *inode = filp->f_path.dentry->d_inode;
208 struct ctl_table_header *head = grab_header(inode);
209 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
210 unsigned int ret = DEFAULT_POLLMASK;
213 /* sysctl was unregistered */
215 return POLLERR | POLLHUP;
217 if (!table->proc_handler)
223 event = (unsigned long)filp->private_data;
224 poll_wait(filp, &table->poll->wait, wait);
226 if (event != atomic_read(&table->poll->event)) {
227 filp->private_data = proc_sys_poll_event(table->poll);
228 ret = POLLIN | POLLRDNORM | POLLERR | POLLPRI;
232 sysctl_head_finish(head);
237 static int proc_sys_fill_cache(struct file *filp, void *dirent,
239 struct ctl_table_header *head,
240 struct ctl_table *table)
242 struct dentry *child, *dir = filp->f_path.dentry;
246 unsigned type = DT_UNKNOWN;
248 qname.name = table->procname;
249 qname.len = strlen(table->procname);
250 qname.hash = full_name_hash(qname.name, qname.len);
252 child = d_lookup(dir, &qname);
254 child = d_alloc(dir, &qname);
256 inode = proc_sys_make_inode(dir->d_sb, head, table);
261 d_set_d_op(child, &proc_sys_dentry_operations);
268 inode = child->d_inode;
270 type = inode->i_mode >> 12;
272 return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type);
275 static int scan(struct ctl_table_header *head, ctl_table *table,
276 unsigned long *pos, struct file *file,
277 void *dirent, filldir_t filldir)
280 for (; table->procname; table++, (*pos)++) {
283 /* Can't do anything without a proc name */
284 if (!table->procname)
287 if (*pos < file->f_pos)
290 res = proc_sys_fill_cache(file, dirent, filldir, head, table);
294 file->f_pos = *pos + 1;
299 static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
301 struct dentry *dentry = filp->f_path.dentry;
302 struct inode *inode = dentry->d_inode;
303 struct ctl_table_header *head = grab_header(inode);
304 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
305 struct ctl_table_header *h = NULL;
310 return PTR_ERR(head);
312 if (table && !table->child) {
317 table = table ? table->child : head->ctl_table;
320 /* Avoid a switch here: arm builds fail with missing __cmpdi2 */
321 if (filp->f_pos == 0) {
322 if (filldir(dirent, ".", 1, filp->f_pos,
323 inode->i_ino, DT_DIR) < 0)
327 if (filp->f_pos == 1) {
328 if (filldir(dirent, "..", 2, filp->f_pos,
329 parent_ino(dentry), DT_DIR) < 0)
335 ret = scan(head, table, &pos, filp, dirent, filldir);
339 for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
340 if (h->attached_to != table)
342 ret = scan(h, h->attached_by, &pos, filp, dirent, filldir);
344 sysctl_head_finish(h);
350 sysctl_head_finish(head);
354 static int proc_sys_permission(struct inode *inode, int mask)
357 * sysctl entries that are not writeable,
358 * are _NOT_ writeable, capabilities or not.
360 struct ctl_table_header *head;
361 struct ctl_table *table;
364 /* Executable files are not allowed under /proc/sys/ */
365 if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode))
368 head = grab_header(inode);
370 return PTR_ERR(head);
372 table = PROC_I(inode)->sysctl_entry;
373 if (!table) /* global root - r-xr-xr-x */
374 error = mask & MAY_WRITE ? -EACCES : 0;
375 else /* Use the permissions on the sysctl table entry */
376 error = sysctl_perm(head->root, table, mask & ~MAY_NOT_BLOCK);
378 sysctl_head_finish(head);
382 static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr)
384 struct inode *inode = dentry->d_inode;
387 if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID))
390 error = setattr_prepare(dentry, attr);
394 if ((attr->ia_valid & ATTR_SIZE) &&
395 attr->ia_size != i_size_read(inode)) {
396 error = vmtruncate(inode, attr->ia_size);
401 setattr_copy(inode, attr);
402 mark_inode_dirty(inode);
406 static int proc_sys_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
408 struct inode *inode = dentry->d_inode;
409 struct ctl_table_header *head = grab_header(inode);
410 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
413 return PTR_ERR(head);
415 generic_fillattr(inode, stat);
417 stat->mode = (stat->mode & S_IFMT) | table->mode;
419 sysctl_head_finish(head);
423 static const struct file_operations proc_sys_file_operations = {
424 .open = proc_sys_open,
425 .poll = proc_sys_poll,
426 .read = proc_sys_read,
427 .write = proc_sys_write,
428 .llseek = default_llseek,
431 static const struct file_operations proc_sys_dir_file_operations = {
432 .read = generic_read_dir,
433 .readdir = proc_sys_readdir,
434 .llseek = generic_file_llseek,
437 static const struct inode_operations proc_sys_inode_operations = {
438 .permission = proc_sys_permission,
439 .setattr = proc_sys_setattr,
440 .getattr = proc_sys_getattr,
443 static const struct inode_operations proc_sys_dir_operations = {
444 .lookup = proc_sys_lookup,
445 .permission = proc_sys_permission,
446 .setattr = proc_sys_setattr,
447 .getattr = proc_sys_getattr,
450 static int proc_sys_revalidate(struct dentry *dentry, struct nameidata *nd)
452 if (nd->flags & LOOKUP_RCU)
454 return !PROC_I(dentry->d_inode)->sysctl->unregistering;
457 static int proc_sys_delete(const struct dentry *dentry)
459 return !!PROC_I(dentry->d_inode)->sysctl->unregistering;
462 static int proc_sys_compare(const struct dentry *parent,
463 const struct inode *pinode,
464 const struct dentry *dentry, const struct inode *inode,
465 unsigned int len, const char *str, const struct qstr *name)
467 struct ctl_table_header *head;
468 /* Although proc doesn't have negative dentries, rcu-walk means
469 * that inode here can be NULL */
470 /* AV: can it, indeed? */
473 if (name->len != len)
475 if (memcmp(name->name, str, len))
477 head = rcu_dereference(PROC_I(inode)->sysctl);
478 return !head || !sysctl_is_seen(head);
481 static const struct dentry_operations proc_sys_dentry_operations = {
482 .d_revalidate = proc_sys_revalidate,
483 .d_delete = proc_sys_delete,
484 .d_compare = proc_sys_compare,
487 int __init proc_sys_init(void)
489 struct proc_dir_entry *proc_sys_root;
491 proc_sys_root = proc_mkdir("sys", NULL);
492 proc_sys_root->proc_iops = &proc_sys_dir_operations;
493 proc_sys_root->proc_fops = &proc_sys_dir_file_operations;
494 proc_sys_root->nlink = 0;