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 <linux/module.h>
13 static const struct dentry_operations proc_sys_dentry_operations;
14 static const struct file_operations proc_sys_file_operations;
15 static const struct inode_operations proc_sys_inode_operations;
16 static const struct file_operations proc_sys_dir_file_operations;
17 static const struct inode_operations proc_sys_dir_operations;
19 void proc_sys_poll_notify(struct ctl_table_poll *poll)
24 atomic_inc(&poll->event);
25 wake_up_interruptible(&poll->wait);
28 static struct ctl_table root_table[] = {
31 .mode = S_IFDIR|S_IRUGO|S_IXUGO,
35 static struct ctl_table_root sysctl_table_root = {
36 .default_set.list = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.header.ctl_entry),
37 .default_set.dir.header = {
40 .ctl_table = root_table,
41 .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
42 .ctl_table_arg = root_table,
43 .root = &sysctl_table_root,
44 .set = &sysctl_table_root.default_set,
48 static DEFINE_SPINLOCK(sysctl_lock);
50 static void drop_sysctl_table(struct ctl_table_header *header);
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);
55 static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table);
57 static void sysctl_print_dir(struct ctl_dir *dir)
59 if (dir->header.parent)
60 sysctl_print_dir(dir->header.parent);
61 printk(KERN_CONT "%s/", dir->header.ctl_table[0].procname);
64 static int namecmp(const char *name1, int len1, const char *name2, int len2)
73 cmp = memcmp(name1, name2, minlen);
79 static struct ctl_table *find_entry(struct ctl_table_header **phead,
80 struct ctl_dir *dir, const char *name, int namelen)
82 struct ctl_table_set *set = dir->header.set;
83 struct ctl_table_header *head;
84 struct ctl_table *entry;
86 list_for_each_entry(head, &set->list, ctl_entry) {
87 if (head->unregistering)
89 if (head->parent != dir)
91 for (entry = head->ctl_table; entry->procname; entry++) {
92 const char *procname = entry->procname;
93 if (namecmp(procname, strlen(procname), name, namelen) == 0) {
102 static void init_header(struct ctl_table_header *head,
103 struct ctl_table_root *root, struct ctl_table_set *set,
104 struct ctl_table *table)
106 head->ctl_table = table;
107 head->ctl_table_arg = table;
108 INIT_LIST_HEAD(&head->ctl_entry);
112 head->unregistering = NULL;
118 static void erase_header(struct ctl_table_header *head)
120 list_del_init(&head->ctl_entry);
123 static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
127 err = sysctl_check_dups(dir, header->ctl_table);
132 header->parent = dir;
133 err = insert_links(header);
136 list_add_tail(&header->ctl_entry, &header->set->list);
139 header->parent = NULL;
140 drop_sysctl_table(&dir->header);
144 /* called under sysctl_lock */
145 static int use_table(struct ctl_table_header *p)
147 if (unlikely(p->unregistering))
153 /* called under sysctl_lock */
154 static void unuse_table(struct ctl_table_header *p)
157 if (unlikely(p->unregistering))
158 complete(p->unregistering);
161 /* called under sysctl_lock, will reacquire if has to wait */
162 static void start_unregistering(struct ctl_table_header *p)
165 * if p->used is 0, nobody will ever touch that entry again;
166 * we'll eliminate all paths to it before dropping sysctl_lock
168 if (unlikely(p->used)) {
169 struct completion wait;
170 init_completion(&wait);
171 p->unregistering = &wait;
172 spin_unlock(&sysctl_lock);
173 wait_for_completion(&wait);
174 spin_lock(&sysctl_lock);
176 /* anything non-NULL; we'll never dereference it */
177 p->unregistering = ERR_PTR(-EINVAL);
180 * do not remove from the list until nobody holds it; walking the
181 * list in do_sysctl() relies on that.
186 static void sysctl_head_get(struct ctl_table_header *head)
188 spin_lock(&sysctl_lock);
190 spin_unlock(&sysctl_lock);
193 void sysctl_head_put(struct ctl_table_header *head)
195 spin_lock(&sysctl_lock);
197 kfree_rcu(head, rcu);
198 spin_unlock(&sysctl_lock);
201 static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
205 spin_lock(&sysctl_lock);
206 if (!use_table(head))
207 head = ERR_PTR(-ENOENT);
208 spin_unlock(&sysctl_lock);
212 static void sysctl_head_finish(struct ctl_table_header *head)
216 spin_lock(&sysctl_lock);
218 spin_unlock(&sysctl_lock);
221 static struct ctl_table_set *
222 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
224 struct ctl_table_set *set = &root->default_set;
226 set = root->lookup(root, namespaces);
230 static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
232 const char *name, int namelen)
234 struct ctl_table_header *head;
235 struct ctl_table *entry;
237 spin_lock(&sysctl_lock);
238 entry = find_entry(&head, dir, name, namelen);
239 if (entry && use_table(head))
243 spin_unlock(&sysctl_lock);
247 static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
248 struct list_head *tmp)
250 struct ctl_table_set *set = dir->header.set;
251 struct ctl_table_header *head;
253 for (tmp = tmp->next; tmp != &set->list; tmp = tmp->next) {
254 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
256 if (head->parent != dir ||
257 !head->ctl_table->procname ||
266 static void first_entry(struct ctl_dir *dir,
267 struct ctl_table_header **phead, struct ctl_table **pentry)
269 struct ctl_table_header *head;
270 struct ctl_table *entry = NULL;
272 spin_lock(&sysctl_lock);
273 head = next_usable_entry(dir, &dir->header.set->list);
274 spin_unlock(&sysctl_lock);
276 entry = head->ctl_table;
281 static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry)
283 struct ctl_table_header *head = *phead;
284 struct ctl_table *entry = *pentry;
287 if (!entry->procname) {
288 spin_lock(&sysctl_lock);
290 head = next_usable_entry(head->parent, &head->ctl_entry);
291 spin_unlock(&sysctl_lock);
293 entry = head->ctl_table;
299 void register_sysctl_root(struct ctl_table_root *root)
304 * sysctl_perm does NOT grant the superuser all rights automatically, because
305 * some sysctl variables are readonly even to root.
308 static int test_perm(int mode, int op)
312 else if (in_egroup_p(0))
314 if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
319 static int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
323 if (root->permissions)
324 mode = root->permissions(root, current->nsproxy, table);
328 return test_perm(mode, op);
331 static struct inode *proc_sys_make_inode(struct super_block *sb,
332 struct ctl_table_header *head, struct ctl_table *table)
335 struct proc_inode *ei;
337 inode = new_inode(sb);
341 inode->i_ino = get_next_ino();
343 sysctl_head_get(head);
346 ei->sysctl_entry = table;
348 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
349 inode->i_mode = table->mode;
350 if (!S_ISDIR(table->mode)) {
351 inode->i_mode |= S_IFREG;
352 inode->i_op = &proc_sys_inode_operations;
353 inode->i_fop = &proc_sys_file_operations;
355 inode->i_mode |= S_IFDIR;
356 inode->i_op = &proc_sys_dir_operations;
357 inode->i_fop = &proc_sys_dir_file_operations;
363 static struct ctl_table_header *grab_header(struct inode *inode)
365 struct ctl_table_header *head = PROC_I(inode)->sysctl;
367 head = &sysctl_table_root.default_set.dir.header;
368 return sysctl_head_grab(head);
371 static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
372 struct nameidata *nd)
374 struct ctl_table_header *head = grab_header(dir);
375 struct ctl_table_header *h = NULL;
376 struct qstr *name = &dentry->d_name;
379 struct dentry *err = ERR_PTR(-ENOENT);
380 struct ctl_dir *ctl_dir;
384 return ERR_CAST(head);
386 ctl_dir = container_of(head, struct ctl_dir, header);
388 p = lookup_entry(&h, ctl_dir, name->name, name->len);
392 ret = sysctl_follow_link(&h, &p, current->nsproxy);
397 err = ERR_PTR(-ENOMEM);
398 inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p);
400 sysctl_head_finish(h);
406 d_set_d_op(dentry, &proc_sys_dentry_operations);
407 d_add(dentry, inode);
410 sysctl_head_finish(head);
414 static ssize_t proc_sys_call_handler(struct file *filp, void __user *buf,
415 size_t count, loff_t *ppos, int write)
417 struct inode *inode = filp->f_path.dentry->d_inode;
418 struct ctl_table_header *head = grab_header(inode);
419 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
424 return PTR_ERR(head);
427 * At this point we know that the sysctl was not unregistered
428 * and won't be until we finish.
431 if (sysctl_perm(head->root, table, write ? MAY_WRITE : MAY_READ))
434 /* if that can happen at all, it should be -EINVAL, not -EISDIR */
436 if (!table->proc_handler)
439 /* careful: calling conventions are nasty here */
441 error = table->proc_handler(table, write, buf, &res, ppos);
445 sysctl_head_finish(head);
450 static ssize_t proc_sys_read(struct file *filp, char __user *buf,
451 size_t count, loff_t *ppos)
453 return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 0);
456 static ssize_t proc_sys_write(struct file *filp, const char __user *buf,
457 size_t count, loff_t *ppos)
459 return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 1);
462 static int proc_sys_open(struct inode *inode, struct file *filp)
464 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
467 filp->private_data = proc_sys_poll_event(table->poll);
472 static unsigned int proc_sys_poll(struct file *filp, poll_table *wait)
474 struct inode *inode = filp->f_path.dentry->d_inode;
475 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
476 unsigned long event = (unsigned long)filp->private_data;
477 unsigned int ret = DEFAULT_POLLMASK;
479 if (!table->proc_handler)
485 poll_wait(filp, &table->poll->wait, wait);
487 if (event != atomic_read(&table->poll->event)) {
488 filp->private_data = proc_sys_poll_event(table->poll);
489 ret = POLLIN | POLLRDNORM | POLLERR | POLLPRI;
496 static int proc_sys_fill_cache(struct file *filp, void *dirent,
498 struct ctl_table_header *head,
499 struct ctl_table *table)
501 struct dentry *child, *dir = filp->f_path.dentry;
505 unsigned type = DT_UNKNOWN;
507 qname.name = table->procname;
508 qname.len = strlen(table->procname);
509 qname.hash = full_name_hash(qname.name, qname.len);
511 child = d_lookup(dir, &qname);
513 child = d_alloc(dir, &qname);
515 inode = proc_sys_make_inode(dir->d_sb, head, table);
520 d_set_d_op(child, &proc_sys_dentry_operations);
527 inode = child->d_inode;
529 type = inode->i_mode >> 12;
531 return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type);
534 static int proc_sys_link_fill_cache(struct file *filp, void *dirent,
536 struct ctl_table_header *head,
537 struct ctl_table *table)
540 head = sysctl_head_grab(head);
542 /* It is not an error if we can not follow the link ignore it */
543 err = sysctl_follow_link(&head, &table, current->nsproxy);
547 ret = proc_sys_fill_cache(filp, dirent, filldir, head, table);
549 sysctl_head_finish(head);
553 static int scan(struct ctl_table_header *head, ctl_table *table,
554 unsigned long *pos, struct file *file,
555 void *dirent, filldir_t filldir)
559 if ((*pos)++ < file->f_pos)
562 if (unlikely(S_ISLNK(table->mode)))
563 res = proc_sys_link_fill_cache(file, dirent, filldir, head, table);
565 res = proc_sys_fill_cache(file, dirent, filldir, head, table);
573 static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
575 struct dentry *dentry = filp->f_path.dentry;
576 struct inode *inode = dentry->d_inode;
577 struct ctl_table_header *head = grab_header(inode);
578 struct ctl_table_header *h = NULL;
579 struct ctl_table *entry;
580 struct ctl_dir *ctl_dir;
585 return PTR_ERR(head);
587 ctl_dir = container_of(head, struct ctl_dir, header);
590 /* Avoid a switch here: arm builds fail with missing __cmpdi2 */
591 if (filp->f_pos == 0) {
592 if (filldir(dirent, ".", 1, filp->f_pos,
593 inode->i_ino, DT_DIR) < 0)
597 if (filp->f_pos == 1) {
598 if (filldir(dirent, "..", 2, filp->f_pos,
599 parent_ino(dentry), DT_DIR) < 0)
605 for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) {
606 ret = scan(h, entry, &pos, filp, dirent, filldir);
608 sysctl_head_finish(h);
614 sysctl_head_finish(head);
618 static int proc_sys_permission(struct inode *inode, int mask)
621 * sysctl entries that are not writeable,
622 * are _NOT_ writeable, capabilities or not.
624 struct ctl_table_header *head;
625 struct ctl_table *table;
628 /* Executable files are not allowed under /proc/sys/ */
629 if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode))
632 head = grab_header(inode);
634 return PTR_ERR(head);
636 table = PROC_I(inode)->sysctl_entry;
637 if (!table) /* global root - r-xr-xr-x */
638 error = mask & MAY_WRITE ? -EACCES : 0;
639 else /* Use the permissions on the sysctl table entry */
640 error = sysctl_perm(head->root, table, mask & ~MAY_NOT_BLOCK);
642 sysctl_head_finish(head);
646 static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr)
648 struct inode *inode = dentry->d_inode;
651 if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID))
654 error = inode_change_ok(inode, attr);
658 if ((attr->ia_valid & ATTR_SIZE) &&
659 attr->ia_size != i_size_read(inode)) {
660 error = vmtruncate(inode, attr->ia_size);
665 setattr_copy(inode, attr);
666 mark_inode_dirty(inode);
670 static int proc_sys_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
672 struct inode *inode = dentry->d_inode;
673 struct ctl_table_header *head = grab_header(inode);
674 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
677 return PTR_ERR(head);
679 generic_fillattr(inode, stat);
681 stat->mode = (stat->mode & S_IFMT) | table->mode;
683 sysctl_head_finish(head);
687 static const struct file_operations proc_sys_file_operations = {
688 .open = proc_sys_open,
689 .poll = proc_sys_poll,
690 .read = proc_sys_read,
691 .write = proc_sys_write,
692 .llseek = default_llseek,
695 static const struct file_operations proc_sys_dir_file_operations = {
696 .read = generic_read_dir,
697 .readdir = proc_sys_readdir,
698 .llseek = generic_file_llseek,
701 static const struct inode_operations proc_sys_inode_operations = {
702 .permission = proc_sys_permission,
703 .setattr = proc_sys_setattr,
704 .getattr = proc_sys_getattr,
707 static const struct inode_operations proc_sys_dir_operations = {
708 .lookup = proc_sys_lookup,
709 .permission = proc_sys_permission,
710 .setattr = proc_sys_setattr,
711 .getattr = proc_sys_getattr,
714 static int proc_sys_revalidate(struct dentry *dentry, struct nameidata *nd)
716 if (nd->flags & LOOKUP_RCU)
718 return !PROC_I(dentry->d_inode)->sysctl->unregistering;
721 static int proc_sys_delete(const struct dentry *dentry)
723 return !!PROC_I(dentry->d_inode)->sysctl->unregistering;
726 static int sysctl_is_seen(struct ctl_table_header *p)
728 struct ctl_table_set *set = p->set;
730 spin_lock(&sysctl_lock);
731 if (p->unregistering)
733 else if (!set->is_seen)
736 res = set->is_seen(set);
737 spin_unlock(&sysctl_lock);
741 static int proc_sys_compare(const struct dentry *parent,
742 const struct inode *pinode,
743 const struct dentry *dentry, const struct inode *inode,
744 unsigned int len, const char *str, const struct qstr *name)
746 struct ctl_table_header *head;
747 /* Although proc doesn't have negative dentries, rcu-walk means
748 * that inode here can be NULL */
749 /* AV: can it, indeed? */
752 if (name->len != len)
754 if (memcmp(name->name, str, len))
756 head = rcu_dereference(PROC_I(inode)->sysctl);
757 return !head || !sysctl_is_seen(head);
760 static const struct dentry_operations proc_sys_dentry_operations = {
761 .d_revalidate = proc_sys_revalidate,
762 .d_delete = proc_sys_delete,
763 .d_compare = proc_sys_compare,
766 static struct ctl_dir *find_subdir(struct ctl_dir *dir,
767 const char *name, int namelen)
769 struct ctl_table_header *head;
770 struct ctl_table *entry;
772 entry = find_entry(&head, dir, name, namelen);
774 return ERR_PTR(-ENOENT);
775 if (S_ISDIR(entry->mode))
776 return container_of(head, struct ctl_dir, header);
777 return ERR_PTR(-ENOTDIR);
780 static struct ctl_dir *new_dir(struct ctl_table_set *set,
781 const char *name, int namelen)
783 struct ctl_table *table;
787 new = kzalloc(sizeof(*new) + sizeof(struct ctl_table)*2 +
788 namelen + 1, GFP_KERNEL);
792 table = (struct ctl_table *)(new + 1);
793 new_name = (char *)(table + 2);
794 memcpy(new_name, name, namelen);
795 new_name[namelen] = '\0';
796 table[0].procname = new_name;
797 table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
798 init_header(&new->header, set->dir.header.root, set, table);
803 static struct ctl_dir *get_subdir(struct ctl_dir *dir,
804 const char *name, int namelen)
806 struct ctl_table_set *set = dir->header.set;
807 struct ctl_dir *subdir, *new = NULL;
809 spin_lock(&sysctl_lock);
810 subdir = find_subdir(dir, name, namelen);
813 if (PTR_ERR(subdir) != -ENOENT)
816 spin_unlock(&sysctl_lock);
817 new = new_dir(set, name, namelen);
818 spin_lock(&sysctl_lock);
819 subdir = ERR_PTR(-ENOMEM);
823 subdir = find_subdir(dir, name, namelen);
826 if (PTR_ERR(subdir) != -ENOENT)
829 if (insert_header(dir, &new->header))
833 subdir->header.nreg++;
835 if (unlikely(IS_ERR(subdir))) {
836 printk(KERN_ERR "sysctl could not get directory: ");
837 sysctl_print_dir(dir);
838 printk(KERN_CONT "/%*.*s %ld\n",
839 namelen, namelen, name, PTR_ERR(subdir));
841 drop_sysctl_table(&dir->header);
843 drop_sysctl_table(&new->header);
844 spin_unlock(&sysctl_lock);
848 static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir)
850 struct ctl_dir *parent;
851 const char *procname;
852 if (!dir->header.parent)
854 parent = xlate_dir(set, dir->header.parent);
857 procname = dir->header.ctl_table[0].procname;
858 return find_subdir(parent, procname, strlen(procname));
861 static int sysctl_follow_link(struct ctl_table_header **phead,
862 struct ctl_table **pentry, struct nsproxy *namespaces)
864 struct ctl_table_header *head;
865 struct ctl_table_root *root;
866 struct ctl_table_set *set;
867 struct ctl_table *entry;
871 /* Get out quickly if not a link */
872 if (!S_ISLNK((*pentry)->mode))
876 spin_lock(&sysctl_lock);
877 root = (*pentry)->data;
878 set = lookup_header_set(root, namespaces);
879 dir = xlate_dir(set, (*phead)->parent);
883 const char *procname = (*pentry)->procname;
885 entry = find_entry(&head, dir, procname, strlen(procname));
887 if (entry && use_table(head)) {
895 spin_unlock(&sysctl_lock);
899 static int sysctl_check_table_dups(struct ctl_dir *dir, struct ctl_table *old,
900 struct ctl_table *table)
902 struct ctl_table *entry, *test;
905 for (entry = old; entry->procname; entry++) {
906 for (test = table; test->procname; test++) {
907 if (strcmp(entry->procname, test->procname) == 0) {
908 printk(KERN_ERR "sysctl duplicate entry: ");
909 sysctl_print_dir(dir);
910 printk(KERN_CONT "/%s\n", test->procname);
918 static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table)
920 struct ctl_table_set *set;
921 struct ctl_table_header *head;
924 set = dir->header.set;
925 list_for_each_entry(head, &set->list, ctl_entry) {
926 if (head->unregistering)
928 if (head->parent != dir)
930 error = sysctl_check_table_dups(dir, head->ctl_table, table);
935 static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...)
937 struct va_format vaf;
944 printk(KERN_ERR "sysctl table check failed: %s/%s %pV\n",
945 path, table->procname, &vaf);
951 static int sysctl_check_table(const char *path, struct ctl_table *table)
954 for (; table->procname; table++) {
956 err = sysctl_err(path, table, "Not a file");
958 if ((table->proc_handler == proc_dostring) ||
959 (table->proc_handler == proc_dointvec) ||
960 (table->proc_handler == proc_dointvec_minmax) ||
961 (table->proc_handler == proc_dointvec_jiffies) ||
962 (table->proc_handler == proc_dointvec_userhz_jiffies) ||
963 (table->proc_handler == proc_dointvec_ms_jiffies) ||
964 (table->proc_handler == proc_doulongvec_minmax) ||
965 (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
967 err = sysctl_err(path, table, "No data");
969 err = sysctl_err(path, table, "No maxlen");
971 if (!table->proc_handler)
972 err = sysctl_err(path, table, "No proc_handler");
974 if ((table->mode & (S_IRUGO|S_IWUGO)) != table->mode)
975 err = sysctl_err(path, table, "bogus .mode 0%o",
981 static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table *table,
982 struct ctl_table_root *link_root)
984 struct ctl_table *link_table, *entry, *link;
985 struct ctl_table_header *links;
987 int nr_entries, name_bytes;
991 for (entry = table; entry->procname; entry++) {
993 name_bytes += strlen(entry->procname) + 1;
996 links = kzalloc(sizeof(struct ctl_table_header) +
997 sizeof(struct ctl_table)*(nr_entries + 1) +
1004 link_table = (struct ctl_table *)(links + 1);
1005 link_name = (char *)&link_table[nr_entries + 1];
1007 for (link = link_table, entry = table; entry->procname; link++, entry++) {
1008 int len = strlen(entry->procname) + 1;
1009 memcpy(link_name, entry->procname, len);
1010 link->procname = link_name;
1011 link->mode = S_IFLNK|S_IRWXUGO;
1012 link->data = link_root;
1015 init_header(links, dir->header.root, dir->header.set, link_table);
1016 links->nreg = nr_entries;
1021 static bool get_links(struct ctl_dir *dir,
1022 struct ctl_table *table, struct ctl_table_root *link_root)
1024 struct ctl_table_header *head;
1025 struct ctl_table *entry, *link;
1027 /* Are there links available for every entry in table? */
1028 for (entry = table; entry->procname; entry++) {
1029 const char *procname = entry->procname;
1030 link = find_entry(&head, dir, procname, strlen(procname));
1033 if (S_ISDIR(link->mode) && S_ISDIR(entry->mode))
1035 if (S_ISLNK(link->mode) && (link->data == link_root))
1040 /* The checks passed. Increase the registration count on the links */
1041 for (entry = table; entry->procname; entry++) {
1042 const char *procname = entry->procname;
1043 link = find_entry(&head, dir, procname, strlen(procname));
1049 static int insert_links(struct ctl_table_header *head)
1051 struct ctl_table_set *root_set = &sysctl_table_root.default_set;
1052 struct ctl_dir *core_parent = NULL;
1053 struct ctl_table_header *links;
1056 if (head->set == root_set)
1059 core_parent = xlate_dir(root_set, head->parent);
1060 if (IS_ERR(core_parent))
1063 if (get_links(core_parent, head->ctl_table, head->root))
1066 core_parent->header.nreg++;
1067 spin_unlock(&sysctl_lock);
1069 links = new_links(core_parent, head->ctl_table, head->root);
1071 spin_lock(&sysctl_lock);
1077 if (get_links(core_parent, head->ctl_table, head->root)) {
1082 err = insert_header(core_parent, links);
1086 drop_sysctl_table(&core_parent->header);
1091 * __register_sysctl_table - register a leaf sysctl table
1092 * @set: Sysctl tree to register on
1093 * @path: The path to the directory the sysctl table is in.
1094 * @table: the top-level table structure
1096 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1097 * array. A completely 0 filled entry terminates the table.
1099 * The members of the &struct ctl_table structure are used as follows:
1101 * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1102 * enter a sysctl file
1104 * data - a pointer to data for use by proc_handler
1106 * maxlen - the maximum size in bytes of the data
1108 * mode - the file permissions for the /proc/sys file
1110 * child - must be %NULL.
1112 * proc_handler - the text handler routine (described below)
1114 * extra1, extra2 - extra pointers usable by the proc handler routines
1116 * Leaf nodes in the sysctl tree will be represented by a single file
1117 * under /proc; non-leaf nodes will be represented by directories.
1119 * There must be a proc_handler routine for any terminal nodes.
1120 * Several default handlers are available to cover common cases -
1122 * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1123 * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1124 * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1126 * It is the handler's job to read the input buffer from user memory
1127 * and process it. The handler should return 0 on success.
1129 * This routine returns %NULL on a failure to register, and a pointer
1130 * to the table header on success.
1132 struct ctl_table_header *__register_sysctl_table(
1133 struct ctl_table_set *set,
1134 const char *path, struct ctl_table *table)
1136 struct ctl_table_root *root = set->dir.header.root;
1137 struct ctl_table_header *links = NULL;
1138 struct ctl_table_header *header;
1139 const char *name, *nextname;
1140 struct ctl_dir *dir;
1142 header = kzalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
1146 init_header(header, root, set, table);
1147 if (sysctl_check_table(path, table))
1150 spin_lock(&sysctl_lock);
1153 spin_unlock(&sysctl_lock);
1155 /* Find the directory for the ctl_table */
1156 for (name = path; name; name = nextname) {
1158 nextname = strchr(name, '/');
1160 namelen = nextname - name;
1163 namelen = strlen(name);
1168 dir = get_subdir(dir, name, namelen);
1173 spin_lock(&sysctl_lock);
1174 if (insert_header(dir, header))
1175 goto fail_put_dir_locked;
1177 drop_sysctl_table(&dir->header);
1178 spin_unlock(&sysctl_lock);
1182 fail_put_dir_locked:
1183 drop_sysctl_table(&dir->header);
1184 spin_unlock(&sysctl_lock);
1192 static char *append_path(const char *path, char *pos, const char *name)
1195 namelen = strlen(name);
1196 if (((pos - path) + namelen + 2) >= PATH_MAX)
1198 memcpy(pos, name, namelen);
1200 pos[namelen + 1] = '\0';
1205 static int count_subheaders(struct ctl_table *table)
1208 int nr_subheaders = 0;
1209 struct ctl_table *entry;
1211 /* special case: no directory and empty directory */
1212 if (!table || !table->procname)
1215 for (entry = table; entry->procname; entry++) {
1217 nr_subheaders += count_subheaders(entry->child);
1221 return nr_subheaders + has_files;
1224 static int register_leaf_sysctl_tables(const char *path, char *pos,
1225 struct ctl_table_header ***subheader, struct ctl_table_set *set,
1226 struct ctl_table *table)
1228 struct ctl_table *ctl_table_arg = NULL;
1229 struct ctl_table *entry, *files;
1234 for (entry = table; entry->procname; entry++) {
1242 /* If there are mixed files and directories we need a new table */
1243 if (nr_dirs && nr_files) {
1244 struct ctl_table *new;
1245 files = kzalloc(sizeof(struct ctl_table) * (nr_files + 1),
1250 ctl_table_arg = files;
1251 for (new = files, entry = table; entry->procname; entry++) {
1259 /* Register everything except a directory full of subdirectories */
1260 if (nr_files || !nr_dirs) {
1261 struct ctl_table_header *header;
1262 header = __register_sysctl_table(set, path, files);
1264 kfree(ctl_table_arg);
1268 /* Remember if we need to free the file table */
1269 header->ctl_table_arg = ctl_table_arg;
1270 **subheader = header;
1274 /* Recurse into the subdirectories. */
1275 for (entry = table; entry->procname; entry++) {
1281 err = -ENAMETOOLONG;
1282 child_pos = append_path(path, pos, entry->procname);
1286 err = register_leaf_sysctl_tables(path, child_pos, subheader,
1294 /* On failure our caller will unregister all registered subheaders */
1299 * __register_sysctl_paths - register a sysctl table hierarchy
1300 * @set: Sysctl tree to register on
1301 * @path: The path to the directory the sysctl table is in.
1302 * @table: the top-level table structure
1304 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1305 * array. A completely 0 filled entry terminates the table.
1307 * See __register_sysctl_table for more details.
1309 struct ctl_table_header *__register_sysctl_paths(
1310 struct ctl_table_set *set,
1311 const struct ctl_path *path, struct ctl_table *table)
1313 struct ctl_table *ctl_table_arg = table;
1314 int nr_subheaders = count_subheaders(table);
1315 struct ctl_table_header *header = NULL, **subheaders, **subheader;
1316 const struct ctl_path *component;
1317 char *new_path, *pos;
1319 pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL);
1324 for (component = path; component->procname; component++) {
1325 pos = append_path(new_path, pos, component->procname);
1329 while (table->procname && table->child && !table[1].procname) {
1330 pos = append_path(new_path, pos, table->procname);
1333 table = table->child;
1335 if (nr_subheaders == 1) {
1336 header = __register_sysctl_table(set, new_path, table);
1338 header->ctl_table_arg = ctl_table_arg;
1340 header = kzalloc(sizeof(*header) +
1341 sizeof(*subheaders)*nr_subheaders, GFP_KERNEL);
1345 subheaders = (struct ctl_table_header **) (header + 1);
1346 subheader = subheaders;
1347 header->ctl_table_arg = ctl_table_arg;
1349 if (register_leaf_sysctl_tables(new_path, pos, &subheader,
1351 goto err_register_leaves;
1358 err_register_leaves:
1359 while (subheader > subheaders) {
1360 struct ctl_table_header *subh = *(--subheader);
1361 struct ctl_table *table = subh->ctl_table_arg;
1362 unregister_sysctl_table(subh);
1371 * register_sysctl_table_path - register a sysctl table hierarchy
1372 * @path: The path to the directory the sysctl table is in.
1373 * @table: the top-level table structure
1375 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1376 * array. A completely 0 filled entry terminates the table.
1378 * See __register_sysctl_paths for more details.
1380 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1381 struct ctl_table *table)
1383 return __register_sysctl_paths(&sysctl_table_root.default_set,
1386 EXPORT_SYMBOL(register_sysctl_paths);
1389 * register_sysctl_table - register a sysctl table hierarchy
1390 * @table: the top-level table structure
1392 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1393 * array. A completely 0 filled entry terminates the table.
1395 * See register_sysctl_paths for more details.
1397 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1399 static const struct ctl_path null_path[] = { {} };
1401 return register_sysctl_paths(null_path, table);
1403 EXPORT_SYMBOL(register_sysctl_table);
1405 static void put_links(struct ctl_table_header *header)
1407 struct ctl_table_set *root_set = &sysctl_table_root.default_set;
1408 struct ctl_table_root *root = header->root;
1409 struct ctl_dir *parent = header->parent;
1410 struct ctl_dir *core_parent;
1411 struct ctl_table *entry;
1413 if (header->set == root_set)
1416 core_parent = xlate_dir(root_set, parent);
1417 if (IS_ERR(core_parent))
1420 for (entry = header->ctl_table; entry->procname; entry++) {
1421 struct ctl_table_header *link_head;
1422 struct ctl_table *link;
1423 const char *name = entry->procname;
1425 link = find_entry(&link_head, core_parent, name, strlen(name));
1427 ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) ||
1428 (S_ISLNK(link->mode) && (link->data == root)))) {
1429 drop_sysctl_table(link_head);
1432 printk(KERN_ERR "sysctl link missing during unregister: ");
1433 sysctl_print_dir(parent);
1434 printk(KERN_CONT "/%s\n", name);
1439 static void drop_sysctl_table(struct ctl_table_header *header)
1441 struct ctl_dir *parent = header->parent;
1447 start_unregistering(header);
1448 if (!--header->count)
1449 kfree_rcu(header, rcu);
1452 drop_sysctl_table(&parent->header);
1456 * unregister_sysctl_table - unregister a sysctl table hierarchy
1457 * @header: the header returned from register_sysctl_table
1459 * Unregisters the sysctl table and all children. proc entries may not
1460 * actually be removed until they are no longer used by anyone.
1462 void unregister_sysctl_table(struct ctl_table_header * header)
1470 nr_subheaders = count_subheaders(header->ctl_table_arg);
1471 if (unlikely(nr_subheaders > 1)) {
1472 struct ctl_table_header **subheaders;
1475 subheaders = (struct ctl_table_header **)(header + 1);
1476 for (i = nr_subheaders -1; i >= 0; i--) {
1477 struct ctl_table_header *subh = subheaders[i];
1478 struct ctl_table *table = subh->ctl_table_arg;
1479 unregister_sysctl_table(subh);
1486 spin_lock(&sysctl_lock);
1487 drop_sysctl_table(header);
1488 spin_unlock(&sysctl_lock);
1490 EXPORT_SYMBOL(unregister_sysctl_table);
1492 void setup_sysctl_set(struct ctl_table_set *set,
1493 struct ctl_table_root *root,
1494 int (*is_seen)(struct ctl_table_set *))
1496 memset(set, sizeof(*set), 0);
1497 INIT_LIST_HEAD(&set->list);
1498 set->is_seen = is_seen;
1499 init_header(&set->dir.header, root, set, root_table);
1502 void retire_sysctl_set(struct ctl_table_set *set)
1504 WARN_ON(!list_empty(&set->list));
1507 int __init proc_sys_init(void)
1509 struct proc_dir_entry *proc_sys_root;
1511 proc_sys_root = proc_mkdir("sys", NULL);
1512 proc_sys_root->proc_iops = &proc_sys_dir_operations;
1513 proc_sys_root->proc_fops = &proc_sys_dir_file_operations;
1514 proc_sys_root->nlink = 0;
1516 return sysctl_init();