4 #include <linux/init.h>
5 #include <linux/sysctl.h>
6 #include <linux/poll.h>
7 #include <linux/proc_fs.h>
8 #include <linux/printk.h>
9 #include <linux/security.h>
10 #include <linux/sched.h>
11 #include <linux/namei.h>
13 #include <linux/module.h>
16 static const struct dentry_operations proc_sys_dentry_operations
;
17 static const struct file_operations proc_sys_file_operations
;
18 static const struct inode_operations proc_sys_inode_operations
;
19 static const struct file_operations proc_sys_dir_file_operations
;
20 static const struct inode_operations proc_sys_dir_operations
;
22 void proc_sys_poll_notify(struct ctl_table_poll
*poll
)
27 atomic_inc(&poll
->event
);
28 wake_up_interruptible(&poll
->wait
);
31 static struct ctl_table root_table
[] = {
34 .mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
,
38 static struct ctl_table_root sysctl_table_root
= {
39 .default_set
.dir
.header
= {
42 .ctl_table
= root_table
}},
43 .ctl_table_arg
= root_table
,
44 .root
= &sysctl_table_root
,
45 .set
= &sysctl_table_root
.default_set
,
49 static DEFINE_SPINLOCK(sysctl_lock
);
51 static void drop_sysctl_table(struct ctl_table_header
*header
);
52 static int sysctl_follow_link(struct ctl_table_header
**phead
,
53 struct ctl_table
**pentry
, struct nsproxy
*namespaces
);
54 static int insert_links(struct ctl_table_header
*head
);
55 static void put_links(struct ctl_table_header
*header
);
57 static void sysctl_print_dir(struct ctl_dir
*dir
)
59 if (dir
->header
.parent
)
60 sysctl_print_dir(dir
->header
.parent
);
61 pr_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 /* Called under sysctl_lock */
80 static struct ctl_table
*find_entry(struct ctl_table_header
**phead
,
81 struct ctl_dir
*dir
, const char *name
, int namelen
)
83 struct ctl_table_header
*head
;
84 struct ctl_table
*entry
;
85 struct rb_node
*node
= dir
->root
.rb_node
;
89 struct ctl_node
*ctl_node
;
93 ctl_node
= rb_entry(node
, struct ctl_node
, node
);
94 head
= ctl_node
->header
;
95 entry
= &head
->ctl_table
[ctl_node
- head
->node
];
96 procname
= entry
->procname
;
98 cmp
= namecmp(name
, namelen
, procname
, strlen(procname
));
100 node
= node
->rb_left
;
102 node
= node
->rb_right
;
111 static int insert_entry(struct ctl_table_header
*head
, struct ctl_table
*entry
)
113 struct rb_node
*node
= &head
->node
[entry
- head
->ctl_table
].node
;
114 struct rb_node
**p
= &head
->parent
->root
.rb_node
;
115 struct rb_node
*parent
= NULL
;
116 const char *name
= entry
->procname
;
117 int namelen
= strlen(name
);
120 struct ctl_table_header
*parent_head
;
121 struct ctl_table
*parent_entry
;
122 struct ctl_node
*parent_node
;
123 const char *parent_name
;
127 parent_node
= rb_entry(parent
, struct ctl_node
, node
);
128 parent_head
= parent_node
->header
;
129 parent_entry
= &parent_head
->ctl_table
[parent_node
- parent_head
->node
];
130 parent_name
= parent_entry
->procname
;
132 cmp
= namecmp(name
, namelen
, parent_name
, strlen(parent_name
));
138 pr_err("sysctl duplicate entry: ");
139 sysctl_print_dir(head
->parent
);
140 pr_cont("/%s\n", entry
->procname
);
145 rb_link_node(node
, parent
, p
);
146 rb_insert_color(node
, &head
->parent
->root
);
150 static void erase_entry(struct ctl_table_header
*head
, struct ctl_table
*entry
)
152 struct rb_node
*node
= &head
->node
[entry
- head
->ctl_table
].node
;
154 rb_erase(node
, &head
->parent
->root
);
157 static void init_header(struct ctl_table_header
*head
,
158 struct ctl_table_root
*root
, struct ctl_table_set
*set
,
159 struct ctl_node
*node
, struct ctl_table
*table
)
161 head
->ctl_table
= table
;
162 head
->ctl_table_arg
= table
;
166 head
->unregistering
= NULL
;
172 struct ctl_table
*entry
;
173 for (entry
= table
; entry
->procname
; entry
++, node
++)
178 static void erase_header(struct ctl_table_header
*head
)
180 struct ctl_table
*entry
;
181 for (entry
= head
->ctl_table
; entry
->procname
; entry
++)
182 erase_entry(head
, entry
);
185 static int insert_header(struct ctl_dir
*dir
, struct ctl_table_header
*header
)
187 struct ctl_table
*entry
;
191 header
->parent
= dir
;
192 err
= insert_links(header
);
195 for (entry
= header
->ctl_table
; entry
->procname
; entry
++) {
196 err
= insert_entry(header
, entry
);
202 erase_header(header
);
205 header
->parent
= NULL
;
206 drop_sysctl_table(&dir
->header
);
210 /* called under sysctl_lock */
211 static int use_table(struct ctl_table_header
*p
)
213 if (unlikely(p
->unregistering
))
219 /* called under sysctl_lock */
220 static void unuse_table(struct ctl_table_header
*p
)
223 if (unlikely(p
->unregistering
))
224 complete(p
->unregistering
);
227 /* called under sysctl_lock, will reacquire if has to wait */
228 static void start_unregistering(struct ctl_table_header
*p
)
231 * if p->used is 0, nobody will ever touch that entry again;
232 * we'll eliminate all paths to it before dropping sysctl_lock
234 if (unlikely(p
->used
)) {
235 struct completion wait
;
236 init_completion(&wait
);
237 p
->unregistering
= &wait
;
238 spin_unlock(&sysctl_lock
);
239 wait_for_completion(&wait
);
240 spin_lock(&sysctl_lock
);
242 /* anything non-NULL; we'll never dereference it */
243 p
->unregistering
= ERR_PTR(-EINVAL
);
246 * do not remove from the list until nobody holds it; walking the
247 * list in do_sysctl() relies on that.
252 static void sysctl_head_get(struct ctl_table_header
*head
)
254 spin_lock(&sysctl_lock
);
256 spin_unlock(&sysctl_lock
);
259 void sysctl_head_put(struct ctl_table_header
*head
)
261 spin_lock(&sysctl_lock
);
263 kfree_rcu(head
, rcu
);
264 spin_unlock(&sysctl_lock
);
267 static struct ctl_table_header
*sysctl_head_grab(struct ctl_table_header
*head
)
270 spin_lock(&sysctl_lock
);
271 if (!use_table(head
))
272 head
= ERR_PTR(-ENOENT
);
273 spin_unlock(&sysctl_lock
);
277 static void sysctl_head_finish(struct ctl_table_header
*head
)
281 spin_lock(&sysctl_lock
);
283 spin_unlock(&sysctl_lock
);
286 static struct ctl_table_set
*
287 lookup_header_set(struct ctl_table_root
*root
, struct nsproxy
*namespaces
)
289 struct ctl_table_set
*set
= &root
->default_set
;
291 set
= root
->lookup(root
, namespaces
);
295 static struct ctl_table
*lookup_entry(struct ctl_table_header
**phead
,
297 const char *name
, int namelen
)
299 struct ctl_table_header
*head
;
300 struct ctl_table
*entry
;
302 spin_lock(&sysctl_lock
);
303 entry
= find_entry(&head
, dir
, name
, namelen
);
304 if (entry
&& use_table(head
))
308 spin_unlock(&sysctl_lock
);
312 static struct ctl_node
*first_usable_entry(struct rb_node
*node
)
314 struct ctl_node
*ctl_node
;
316 for (;node
; node
= rb_next(node
)) {
317 ctl_node
= rb_entry(node
, struct ctl_node
, node
);
318 if (use_table(ctl_node
->header
))
324 static void first_entry(struct ctl_dir
*dir
,
325 struct ctl_table_header
**phead
, struct ctl_table
**pentry
)
327 struct ctl_table_header
*head
= NULL
;
328 struct ctl_table
*entry
= NULL
;
329 struct ctl_node
*ctl_node
;
331 spin_lock(&sysctl_lock
);
332 ctl_node
= first_usable_entry(rb_first(&dir
->root
));
333 spin_unlock(&sysctl_lock
);
335 head
= ctl_node
->header
;
336 entry
= &head
->ctl_table
[ctl_node
- head
->node
];
342 static void next_entry(struct ctl_table_header
**phead
, struct ctl_table
**pentry
)
344 struct ctl_table_header
*head
= *phead
;
345 struct ctl_table
*entry
= *pentry
;
346 struct ctl_node
*ctl_node
= &head
->node
[entry
- head
->ctl_table
];
348 spin_lock(&sysctl_lock
);
351 ctl_node
= first_usable_entry(rb_next(&ctl_node
->node
));
352 spin_unlock(&sysctl_lock
);
355 head
= ctl_node
->header
;
356 entry
= &head
->ctl_table
[ctl_node
- head
->node
];
362 void register_sysctl_root(struct ctl_table_root
*root
)
367 * sysctl_perm does NOT grant the superuser all rights automatically, because
368 * some sysctl variables are readonly even to root.
371 static int test_perm(int mode
, int op
)
373 if (uid_eq(current_euid(), GLOBAL_ROOT_UID
))
375 else if (in_egroup_p(GLOBAL_ROOT_GID
))
377 if ((op
& ~mode
& (MAY_READ
|MAY_WRITE
|MAY_EXEC
)) == 0)
382 static int sysctl_perm(struct ctl_table_header
*head
, struct ctl_table
*table
, int op
)
384 struct ctl_table_root
*root
= head
->root
;
387 if (root
->permissions
)
388 mode
= root
->permissions(head
, table
);
392 return test_perm(mode
, op
);
395 static struct inode
*proc_sys_make_inode(struct super_block
*sb
,
396 struct ctl_table_header
*head
, struct ctl_table
*table
)
399 struct proc_inode
*ei
;
401 inode
= new_inode(sb
);
405 inode
->i_ino
= get_next_ino();
407 sysctl_head_get(head
);
410 ei
->sysctl_entry
= table
;
412 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
413 inode
->i_mode
= table
->mode
;
414 if (!S_ISDIR(table
->mode
)) {
415 inode
->i_mode
|= S_IFREG
;
416 inode
->i_op
= &proc_sys_inode_operations
;
417 inode
->i_fop
= &proc_sys_file_operations
;
419 inode
->i_mode
|= S_IFDIR
;
420 inode
->i_op
= &proc_sys_dir_operations
;
421 inode
->i_fop
= &proc_sys_dir_file_operations
;
427 static struct ctl_table_header
*grab_header(struct inode
*inode
)
429 struct ctl_table_header
*head
= PROC_I(inode
)->sysctl
;
431 head
= &sysctl_table_root
.default_set
.dir
.header
;
432 return sysctl_head_grab(head
);
435 static struct dentry
*proc_sys_lookup(struct inode
*dir
, struct dentry
*dentry
,
438 struct ctl_table_header
*head
= grab_header(dir
);
439 struct ctl_table_header
*h
= NULL
;
440 struct qstr
*name
= &dentry
->d_name
;
443 struct dentry
*err
= ERR_PTR(-ENOENT
);
444 struct ctl_dir
*ctl_dir
;
448 return ERR_CAST(head
);
450 ctl_dir
= container_of(head
, struct ctl_dir
, header
);
452 p
= lookup_entry(&h
, ctl_dir
, name
->name
, name
->len
);
456 if (S_ISLNK(p
->mode
)) {
457 ret
= sysctl_follow_link(&h
, &p
, current
->nsproxy
);
463 err
= ERR_PTR(-ENOMEM
);
464 inode
= proc_sys_make_inode(dir
->i_sb
, h
? h
: head
, p
);
469 d_set_d_op(dentry
, &proc_sys_dentry_operations
);
470 d_add(dentry
, inode
);
474 sysctl_head_finish(h
);
475 sysctl_head_finish(head
);
479 static ssize_t
proc_sys_call_handler(struct file
*filp
, void __user
*buf
,
480 size_t count
, loff_t
*ppos
, int write
)
482 struct inode
*inode
= file_inode(filp
);
483 struct ctl_table_header
*head
= grab_header(inode
);
484 struct ctl_table
*table
= PROC_I(inode
)->sysctl_entry
;
489 return PTR_ERR(head
);
492 * At this point we know that the sysctl was not unregistered
493 * and won't be until we finish.
496 if (sysctl_perm(head
, table
, write
? MAY_WRITE
: MAY_READ
))
499 /* if that can happen at all, it should be -EINVAL, not -EISDIR */
501 if (!table
->proc_handler
)
504 /* careful: calling conventions are nasty here */
506 error
= table
->proc_handler(table
, write
, buf
, &res
, ppos
);
510 sysctl_head_finish(head
);
515 static ssize_t
proc_sys_read(struct file
*filp
, char __user
*buf
,
516 size_t count
, loff_t
*ppos
)
518 return proc_sys_call_handler(filp
, (void __user
*)buf
, count
, ppos
, 0);
521 static ssize_t
proc_sys_write(struct file
*filp
, const char __user
*buf
,
522 size_t count
, loff_t
*ppos
)
524 return proc_sys_call_handler(filp
, (void __user
*)buf
, count
, ppos
, 1);
527 static int proc_sys_open(struct inode
*inode
, struct file
*filp
)
529 struct ctl_table_header
*head
= grab_header(inode
);
530 struct ctl_table
*table
= PROC_I(inode
)->sysctl_entry
;
532 /* sysctl was unregistered */
534 return PTR_ERR(head
);
537 filp
->private_data
= proc_sys_poll_event(table
->poll
);
539 sysctl_head_finish(head
);
544 static unsigned int proc_sys_poll(struct file
*filp
, poll_table
*wait
)
546 struct inode
*inode
= file_inode(filp
);
547 struct ctl_table_header
*head
= grab_header(inode
);
548 struct ctl_table
*table
= PROC_I(inode
)->sysctl_entry
;
549 unsigned int ret
= DEFAULT_POLLMASK
;
552 /* sysctl was unregistered */
554 return POLLERR
| POLLHUP
;
556 if (!table
->proc_handler
)
562 event
= (unsigned long)filp
->private_data
;
563 poll_wait(filp
, &table
->poll
->wait
, wait
);
565 if (event
!= atomic_read(&table
->poll
->event
)) {
566 filp
->private_data
= proc_sys_poll_event(table
->poll
);
567 ret
= POLLIN
| POLLRDNORM
| POLLERR
| POLLPRI
;
571 sysctl_head_finish(head
);
576 static int proc_sys_fill_cache(struct file
*filp
, void *dirent
,
578 struct ctl_table_header
*head
,
579 struct ctl_table
*table
)
581 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
585 unsigned type
= DT_UNKNOWN
;
587 qname
.name
= table
->procname
;
588 qname
.len
= strlen(table
->procname
);
589 qname
.hash
= full_name_hash(qname
.name
, qname
.len
);
591 child
= d_lookup(dir
, &qname
);
593 child
= d_alloc(dir
, &qname
);
595 inode
= proc_sys_make_inode(dir
->d_sb
, head
, table
);
600 d_set_d_op(child
, &proc_sys_dentry_operations
);
607 inode
= child
->d_inode
;
609 type
= inode
->i_mode
>> 12;
611 return !!filldir(dirent
, qname
.name
, qname
.len
, filp
->f_pos
, ino
, type
);
614 static int proc_sys_link_fill_cache(struct file
*filp
, void *dirent
,
616 struct ctl_table_header
*head
,
617 struct ctl_table
*table
)
620 head
= sysctl_head_grab(head
);
622 if (S_ISLNK(table
->mode
)) {
623 /* It is not an error if we can not follow the link ignore it */
624 err
= sysctl_follow_link(&head
, &table
, current
->nsproxy
);
629 ret
= proc_sys_fill_cache(filp
, dirent
, filldir
, head
, table
);
631 sysctl_head_finish(head
);
635 static int scan(struct ctl_table_header
*head
, ctl_table
*table
,
636 unsigned long *pos
, struct file
*file
,
637 void *dirent
, filldir_t filldir
)
641 if ((*pos
)++ < file
->f_pos
)
644 if (unlikely(S_ISLNK(table
->mode
)))
645 res
= proc_sys_link_fill_cache(file
, dirent
, filldir
, head
, table
);
647 res
= proc_sys_fill_cache(file
, dirent
, filldir
, head
, table
);
655 static int proc_sys_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
657 struct dentry
*dentry
= filp
->f_path
.dentry
;
658 struct inode
*inode
= dentry
->d_inode
;
659 struct ctl_table_header
*head
= grab_header(inode
);
660 struct ctl_table_header
*h
= NULL
;
661 struct ctl_table
*entry
;
662 struct ctl_dir
*ctl_dir
;
667 return PTR_ERR(head
);
669 ctl_dir
= container_of(head
, struct ctl_dir
, header
);
672 /* Avoid a switch here: arm builds fail with missing __cmpdi2 */
673 if (filp
->f_pos
== 0) {
674 if (filldir(dirent
, ".", 1, filp
->f_pos
,
675 inode
->i_ino
, DT_DIR
) < 0)
679 if (filp
->f_pos
== 1) {
680 if (filldir(dirent
, "..", 2, filp
->f_pos
,
681 parent_ino(dentry
), DT_DIR
) < 0)
687 for (first_entry(ctl_dir
, &h
, &entry
); h
; next_entry(&h
, &entry
)) {
688 ret
= scan(h
, entry
, &pos
, filp
, dirent
, filldir
);
690 sysctl_head_finish(h
);
696 sysctl_head_finish(head
);
700 static int proc_sys_permission(struct inode
*inode
, int mask
)
703 * sysctl entries that are not writeable,
704 * are _NOT_ writeable, capabilities or not.
706 struct ctl_table_header
*head
;
707 struct ctl_table
*table
;
710 /* Executable files are not allowed under /proc/sys/ */
711 if ((mask
& MAY_EXEC
) && S_ISREG(inode
->i_mode
))
714 head
= grab_header(inode
);
716 return PTR_ERR(head
);
718 table
= PROC_I(inode
)->sysctl_entry
;
719 if (!table
) /* global root - r-xr-xr-x */
720 error
= mask
& MAY_WRITE
? -EACCES
: 0;
721 else /* Use the permissions on the sysctl table entry */
722 error
= sysctl_perm(head
, table
, mask
& ~MAY_NOT_BLOCK
);
724 sysctl_head_finish(head
);
728 static int proc_sys_setattr(struct dentry
*dentry
, struct iattr
*attr
)
730 struct inode
*inode
= dentry
->d_inode
;
733 if (attr
->ia_valid
& (ATTR_MODE
| ATTR_UID
| ATTR_GID
))
736 error
= inode_change_ok(inode
, attr
);
740 setattr_copy(inode
, attr
);
741 mark_inode_dirty(inode
);
745 static int proc_sys_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
747 struct inode
*inode
= dentry
->d_inode
;
748 struct ctl_table_header
*head
= grab_header(inode
);
749 struct ctl_table
*table
= PROC_I(inode
)->sysctl_entry
;
752 return PTR_ERR(head
);
754 generic_fillattr(inode
, stat
);
756 stat
->mode
= (stat
->mode
& S_IFMT
) | table
->mode
;
758 sysctl_head_finish(head
);
762 static const struct file_operations proc_sys_file_operations
= {
763 .open
= proc_sys_open
,
764 .poll
= proc_sys_poll
,
765 .read
= proc_sys_read
,
766 .write
= proc_sys_write
,
767 .llseek
= default_llseek
,
770 static const struct file_operations proc_sys_dir_file_operations
= {
771 .read
= generic_read_dir
,
772 .readdir
= proc_sys_readdir
,
773 .llseek
= generic_file_llseek
,
776 static const struct inode_operations proc_sys_inode_operations
= {
777 .permission
= proc_sys_permission
,
778 .setattr
= proc_sys_setattr
,
779 .getattr
= proc_sys_getattr
,
782 static const struct inode_operations proc_sys_dir_operations
= {
783 .lookup
= proc_sys_lookup
,
784 .permission
= proc_sys_permission
,
785 .setattr
= proc_sys_setattr
,
786 .getattr
= proc_sys_getattr
,
789 static int proc_sys_revalidate(struct dentry
*dentry
, unsigned int flags
)
791 if (flags
& LOOKUP_RCU
)
793 return !PROC_I(dentry
->d_inode
)->sysctl
->unregistering
;
796 static int proc_sys_delete(const struct dentry
*dentry
)
798 return !!PROC_I(dentry
->d_inode
)->sysctl
->unregistering
;
801 static int sysctl_is_seen(struct ctl_table_header
*p
)
803 struct ctl_table_set
*set
= p
->set
;
805 spin_lock(&sysctl_lock
);
806 if (p
->unregistering
)
808 else if (!set
->is_seen
)
811 res
= set
->is_seen(set
);
812 spin_unlock(&sysctl_lock
);
816 static int proc_sys_compare(const struct dentry
*parent
,
817 const struct inode
*pinode
,
818 const struct dentry
*dentry
, const struct inode
*inode
,
819 unsigned int len
, const char *str
, const struct qstr
*name
)
821 struct ctl_table_header
*head
;
822 /* Although proc doesn't have negative dentries, rcu-walk means
823 * that inode here can be NULL */
824 /* AV: can it, indeed? */
827 if (name
->len
!= len
)
829 if (memcmp(name
->name
, str
, len
))
831 head
= rcu_dereference(PROC_I(inode
)->sysctl
);
832 return !head
|| !sysctl_is_seen(head
);
835 static const struct dentry_operations proc_sys_dentry_operations
= {
836 .d_revalidate
= proc_sys_revalidate
,
837 .d_delete
= proc_sys_delete
,
838 .d_compare
= proc_sys_compare
,
841 static struct ctl_dir
*find_subdir(struct ctl_dir
*dir
,
842 const char *name
, int namelen
)
844 struct ctl_table_header
*head
;
845 struct ctl_table
*entry
;
847 entry
= find_entry(&head
, dir
, name
, namelen
);
849 return ERR_PTR(-ENOENT
);
850 if (!S_ISDIR(entry
->mode
))
851 return ERR_PTR(-ENOTDIR
);
852 return container_of(head
, struct ctl_dir
, header
);
855 static struct ctl_dir
*new_dir(struct ctl_table_set
*set
,
856 const char *name
, int namelen
)
858 struct ctl_table
*table
;
860 struct ctl_node
*node
;
863 new = kzalloc(sizeof(*new) + sizeof(struct ctl_node
) +
864 sizeof(struct ctl_table
)*2 + namelen
+ 1,
869 node
= (struct ctl_node
*)(new + 1);
870 table
= (struct ctl_table
*)(node
+ 1);
871 new_name
= (char *)(table
+ 2);
872 memcpy(new_name
, name
, namelen
);
873 new_name
[namelen
] = '\0';
874 table
[0].procname
= new_name
;
875 table
[0].mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
876 init_header(&new->header
, set
->dir
.header
.root
, set
, node
, table
);
882 * get_subdir - find or create a subdir with the specified name.
883 * @dir: Directory to create the subdirectory in
884 * @name: The name of the subdirectory to find or create
885 * @namelen: The length of name
887 * Takes a directory with an elevated reference count so we know that
888 * if we drop the lock the directory will not go away. Upon success
889 * the reference is moved from @dir to the returned subdirectory.
890 * Upon error an error code is returned and the reference on @dir is
893 static struct ctl_dir
*get_subdir(struct ctl_dir
*dir
,
894 const char *name
, int namelen
)
896 struct ctl_table_set
*set
= dir
->header
.set
;
897 struct ctl_dir
*subdir
, *new = NULL
;
900 spin_lock(&sysctl_lock
);
901 subdir
= find_subdir(dir
, name
, namelen
);
904 if (PTR_ERR(subdir
) != -ENOENT
)
907 spin_unlock(&sysctl_lock
);
908 new = new_dir(set
, name
, namelen
);
909 spin_lock(&sysctl_lock
);
910 subdir
= ERR_PTR(-ENOMEM
);
914 /* Was the subdir added while we dropped the lock? */
915 subdir
= find_subdir(dir
, name
, namelen
);
918 if (PTR_ERR(subdir
) != -ENOENT
)
921 /* Nope. Use the our freshly made directory entry. */
922 err
= insert_header(dir
, &new->header
);
923 subdir
= ERR_PTR(err
);
928 subdir
->header
.nreg
++;
930 if (unlikely(IS_ERR(subdir
))) {
931 pr_err("sysctl could not get directory: ");
932 sysctl_print_dir(dir
);
933 pr_cont("/%*.*s %ld\n",
934 namelen
, namelen
, name
, PTR_ERR(subdir
));
936 drop_sysctl_table(&dir
->header
);
938 drop_sysctl_table(&new->header
);
939 spin_unlock(&sysctl_lock
);
943 static struct ctl_dir
*xlate_dir(struct ctl_table_set
*set
, struct ctl_dir
*dir
)
945 struct ctl_dir
*parent
;
946 const char *procname
;
947 if (!dir
->header
.parent
)
949 parent
= xlate_dir(set
, dir
->header
.parent
);
952 procname
= dir
->header
.ctl_table
[0].procname
;
953 return find_subdir(parent
, procname
, strlen(procname
));
956 static int sysctl_follow_link(struct ctl_table_header
**phead
,
957 struct ctl_table
**pentry
, struct nsproxy
*namespaces
)
959 struct ctl_table_header
*head
;
960 struct ctl_table_root
*root
;
961 struct ctl_table_set
*set
;
962 struct ctl_table
*entry
;
967 spin_lock(&sysctl_lock
);
968 root
= (*pentry
)->data
;
969 set
= lookup_header_set(root
, namespaces
);
970 dir
= xlate_dir(set
, (*phead
)->parent
);
974 const char *procname
= (*pentry
)->procname
;
976 entry
= find_entry(&head
, dir
, procname
, strlen(procname
));
978 if (entry
&& use_table(head
)) {
986 spin_unlock(&sysctl_lock
);
990 static int sysctl_err(const char *path
, struct ctl_table
*table
, char *fmt
, ...)
992 struct va_format vaf
;
999 pr_err("sysctl table check failed: %s/%s %pV\n",
1000 path
, table
->procname
, &vaf
);
1006 static int sysctl_check_table(const char *path
, struct ctl_table
*table
)
1009 for (; table
->procname
; table
++) {
1011 err
= sysctl_err(path
, table
, "Not a file");
1013 if ((table
->proc_handler
== proc_dostring
) ||
1014 (table
->proc_handler
== proc_dointvec
) ||
1015 (table
->proc_handler
== proc_dointvec_minmax
) ||
1016 (table
->proc_handler
== proc_dointvec_jiffies
) ||
1017 (table
->proc_handler
== proc_dointvec_userhz_jiffies
) ||
1018 (table
->proc_handler
== proc_dointvec_ms_jiffies
) ||
1019 (table
->proc_handler
== proc_doulongvec_minmax
) ||
1020 (table
->proc_handler
== proc_doulongvec_ms_jiffies_minmax
)) {
1022 err
= sysctl_err(path
, table
, "No data");
1024 err
= sysctl_err(path
, table
, "No maxlen");
1026 if (!table
->proc_handler
)
1027 err
= sysctl_err(path
, table
, "No proc_handler");
1029 if ((table
->mode
& (S_IRUGO
|S_IWUGO
)) != table
->mode
)
1030 err
= sysctl_err(path
, table
, "bogus .mode 0%o",
1036 static struct ctl_table_header
*new_links(struct ctl_dir
*dir
, struct ctl_table
*table
,
1037 struct ctl_table_root
*link_root
)
1039 struct ctl_table
*link_table
, *entry
, *link
;
1040 struct ctl_table_header
*links
;
1041 struct ctl_node
*node
;
1043 int nr_entries
, name_bytes
;
1047 for (entry
= table
; entry
->procname
; entry
++) {
1049 name_bytes
+= strlen(entry
->procname
) + 1;
1052 links
= kzalloc(sizeof(struct ctl_table_header
) +
1053 sizeof(struct ctl_node
)*nr_entries
+
1054 sizeof(struct ctl_table
)*(nr_entries
+ 1) +
1061 node
= (struct ctl_node
*)(links
+ 1);
1062 link_table
= (struct ctl_table
*)(node
+ nr_entries
);
1063 link_name
= (char *)&link_table
[nr_entries
+ 1];
1065 for (link
= link_table
, entry
= table
; entry
->procname
; link
++, entry
++) {
1066 int len
= strlen(entry
->procname
) + 1;
1067 memcpy(link_name
, entry
->procname
, len
);
1068 link
->procname
= link_name
;
1069 link
->mode
= S_IFLNK
|S_IRWXUGO
;
1070 link
->data
= link_root
;
1073 init_header(links
, dir
->header
.root
, dir
->header
.set
, node
, link_table
);
1074 links
->nreg
= nr_entries
;
1079 static bool get_links(struct ctl_dir
*dir
,
1080 struct ctl_table
*table
, struct ctl_table_root
*link_root
)
1082 struct ctl_table_header
*head
;
1083 struct ctl_table
*entry
, *link
;
1085 /* Are there links available for every entry in table? */
1086 for (entry
= table
; entry
->procname
; entry
++) {
1087 const char *procname
= entry
->procname
;
1088 link
= find_entry(&head
, dir
, procname
, strlen(procname
));
1091 if (S_ISDIR(link
->mode
) && S_ISDIR(entry
->mode
))
1093 if (S_ISLNK(link
->mode
) && (link
->data
== link_root
))
1098 /* The checks passed. Increase the registration count on the links */
1099 for (entry
= table
; entry
->procname
; entry
++) {
1100 const char *procname
= entry
->procname
;
1101 link
= find_entry(&head
, dir
, procname
, strlen(procname
));
1107 static int insert_links(struct ctl_table_header
*head
)
1109 struct ctl_table_set
*root_set
= &sysctl_table_root
.default_set
;
1110 struct ctl_dir
*core_parent
= NULL
;
1111 struct ctl_table_header
*links
;
1114 if (head
->set
== root_set
)
1117 core_parent
= xlate_dir(root_set
, head
->parent
);
1118 if (IS_ERR(core_parent
))
1121 if (get_links(core_parent
, head
->ctl_table
, head
->root
))
1124 core_parent
->header
.nreg
++;
1125 spin_unlock(&sysctl_lock
);
1127 links
= new_links(core_parent
, head
->ctl_table
, head
->root
);
1129 spin_lock(&sysctl_lock
);
1135 if (get_links(core_parent
, head
->ctl_table
, head
->root
)) {
1140 err
= insert_header(core_parent
, links
);
1144 drop_sysctl_table(&core_parent
->header
);
1149 * __register_sysctl_table - register a leaf sysctl table
1150 * @set: Sysctl tree to register on
1151 * @path: The path to the directory the sysctl table is in.
1152 * @table: the top-level table structure
1154 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1155 * array. A completely 0 filled entry terminates the table.
1157 * The members of the &struct ctl_table structure are used as follows:
1159 * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1160 * enter a sysctl file
1162 * data - a pointer to data for use by proc_handler
1164 * maxlen - the maximum size in bytes of the data
1166 * mode - the file permissions for the /proc/sys file
1168 * child - must be %NULL.
1170 * proc_handler - the text handler routine (described below)
1172 * extra1, extra2 - extra pointers usable by the proc handler routines
1174 * Leaf nodes in the sysctl tree will be represented by a single file
1175 * under /proc; non-leaf nodes will be represented by directories.
1177 * There must be a proc_handler routine for any terminal nodes.
1178 * Several default handlers are available to cover common cases -
1180 * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1181 * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1182 * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1184 * It is the handler's job to read the input buffer from user memory
1185 * and process it. The handler should return 0 on success.
1187 * This routine returns %NULL on a failure to register, and a pointer
1188 * to the table header on success.
1190 struct ctl_table_header
*__register_sysctl_table(
1191 struct ctl_table_set
*set
,
1192 const char *path
, struct ctl_table
*table
)
1194 struct ctl_table_root
*root
= set
->dir
.header
.root
;
1195 struct ctl_table_header
*header
;
1196 const char *name
, *nextname
;
1197 struct ctl_dir
*dir
;
1198 struct ctl_table
*entry
;
1199 struct ctl_node
*node
;
1202 for (entry
= table
; entry
->procname
; entry
++)
1205 header
= kzalloc(sizeof(struct ctl_table_header
) +
1206 sizeof(struct ctl_node
)*nr_entries
, GFP_KERNEL
);
1210 node
= (struct ctl_node
*)(header
+ 1);
1211 init_header(header
, root
, set
, node
, table
);
1212 if (sysctl_check_table(path
, table
))
1215 spin_lock(&sysctl_lock
);
1217 /* Reference moved down the diretory tree get_subdir */
1219 spin_unlock(&sysctl_lock
);
1221 /* Find the directory for the ctl_table */
1222 for (name
= path
; name
; name
= nextname
) {
1224 nextname
= strchr(name
, '/');
1226 namelen
= nextname
- name
;
1229 namelen
= strlen(name
);
1234 dir
= get_subdir(dir
, name
, namelen
);
1239 spin_lock(&sysctl_lock
);
1240 if (insert_header(dir
, header
))
1241 goto fail_put_dir_locked
;
1243 drop_sysctl_table(&dir
->header
);
1244 spin_unlock(&sysctl_lock
);
1248 fail_put_dir_locked
:
1249 drop_sysctl_table(&dir
->header
);
1250 spin_unlock(&sysctl_lock
);
1258 * register_sysctl - register a sysctl table
1259 * @path: The path to the directory the sysctl table is in.
1260 * @table: the table structure
1262 * Register a sysctl table. @table should be a filled in ctl_table
1263 * array. A completely 0 filled entry terminates the table.
1265 * See __register_sysctl_table for more details.
1267 struct ctl_table_header
*register_sysctl(const char *path
, struct ctl_table
*table
)
1269 return __register_sysctl_table(&sysctl_table_root
.default_set
,
1272 EXPORT_SYMBOL(register_sysctl
);
1274 static char *append_path(const char *path
, char *pos
, const char *name
)
1277 namelen
= strlen(name
);
1278 if (((pos
- path
) + namelen
+ 2) >= PATH_MAX
)
1280 memcpy(pos
, name
, namelen
);
1282 pos
[namelen
+ 1] = '\0';
1287 static int count_subheaders(struct ctl_table
*table
)
1290 int nr_subheaders
= 0;
1291 struct ctl_table
*entry
;
1293 /* special case: no directory and empty directory */
1294 if (!table
|| !table
->procname
)
1297 for (entry
= table
; entry
->procname
; entry
++) {
1299 nr_subheaders
+= count_subheaders(entry
->child
);
1303 return nr_subheaders
+ has_files
;
1306 static int register_leaf_sysctl_tables(const char *path
, char *pos
,
1307 struct ctl_table_header
***subheader
, struct ctl_table_set
*set
,
1308 struct ctl_table
*table
)
1310 struct ctl_table
*ctl_table_arg
= NULL
;
1311 struct ctl_table
*entry
, *files
;
1316 for (entry
= table
; entry
->procname
; entry
++) {
1324 /* If there are mixed files and directories we need a new table */
1325 if (nr_dirs
&& nr_files
) {
1326 struct ctl_table
*new;
1327 files
= kzalloc(sizeof(struct ctl_table
) * (nr_files
+ 1),
1332 ctl_table_arg
= files
;
1333 for (new = files
, entry
= table
; entry
->procname
; entry
++) {
1341 /* Register everything except a directory full of subdirectories */
1342 if (nr_files
|| !nr_dirs
) {
1343 struct ctl_table_header
*header
;
1344 header
= __register_sysctl_table(set
, path
, files
);
1346 kfree(ctl_table_arg
);
1350 /* Remember if we need to free the file table */
1351 header
->ctl_table_arg
= ctl_table_arg
;
1352 **subheader
= header
;
1356 /* Recurse into the subdirectories. */
1357 for (entry
= table
; entry
->procname
; entry
++) {
1363 err
= -ENAMETOOLONG
;
1364 child_pos
= append_path(path
, pos
, entry
->procname
);
1368 err
= register_leaf_sysctl_tables(path
, child_pos
, subheader
,
1376 /* On failure our caller will unregister all registered subheaders */
1381 * __register_sysctl_paths - register a sysctl table hierarchy
1382 * @set: Sysctl tree to register on
1383 * @path: The path to the directory the sysctl table is in.
1384 * @table: the top-level table structure
1386 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1387 * array. A completely 0 filled entry terminates the table.
1389 * See __register_sysctl_table for more details.
1391 struct ctl_table_header
*__register_sysctl_paths(
1392 struct ctl_table_set
*set
,
1393 const struct ctl_path
*path
, struct ctl_table
*table
)
1395 struct ctl_table
*ctl_table_arg
= table
;
1396 int nr_subheaders
= count_subheaders(table
);
1397 struct ctl_table_header
*header
= NULL
, **subheaders
, **subheader
;
1398 const struct ctl_path
*component
;
1399 char *new_path
, *pos
;
1401 pos
= new_path
= kmalloc(PATH_MAX
, GFP_KERNEL
);
1406 for (component
= path
; component
->procname
; component
++) {
1407 pos
= append_path(new_path
, pos
, component
->procname
);
1411 while (table
->procname
&& table
->child
&& !table
[1].procname
) {
1412 pos
= append_path(new_path
, pos
, table
->procname
);
1415 table
= table
->child
;
1417 if (nr_subheaders
== 1) {
1418 header
= __register_sysctl_table(set
, new_path
, table
);
1420 header
->ctl_table_arg
= ctl_table_arg
;
1422 header
= kzalloc(sizeof(*header
) +
1423 sizeof(*subheaders
)*nr_subheaders
, GFP_KERNEL
);
1427 subheaders
= (struct ctl_table_header
**) (header
+ 1);
1428 subheader
= subheaders
;
1429 header
->ctl_table_arg
= ctl_table_arg
;
1431 if (register_leaf_sysctl_tables(new_path
, pos
, &subheader
,
1433 goto err_register_leaves
;
1440 err_register_leaves
:
1441 while (subheader
> subheaders
) {
1442 struct ctl_table_header
*subh
= *(--subheader
);
1443 struct ctl_table
*table
= subh
->ctl_table_arg
;
1444 unregister_sysctl_table(subh
);
1453 * register_sysctl_table_path - register a sysctl table hierarchy
1454 * @path: The path to the directory the sysctl table is in.
1455 * @table: the top-level table structure
1457 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1458 * array. A completely 0 filled entry terminates the table.
1460 * See __register_sysctl_paths for more details.
1462 struct ctl_table_header
*register_sysctl_paths(const struct ctl_path
*path
,
1463 struct ctl_table
*table
)
1465 return __register_sysctl_paths(&sysctl_table_root
.default_set
,
1468 EXPORT_SYMBOL(register_sysctl_paths
);
1471 * register_sysctl_table - register a sysctl table hierarchy
1472 * @table: the top-level table structure
1474 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1475 * array. A completely 0 filled entry terminates the table.
1477 * See register_sysctl_paths for more details.
1479 struct ctl_table_header
*register_sysctl_table(struct ctl_table
*table
)
1481 static const struct ctl_path null_path
[] = { {} };
1483 return register_sysctl_paths(null_path
, table
);
1485 EXPORT_SYMBOL(register_sysctl_table
);
1487 static void put_links(struct ctl_table_header
*header
)
1489 struct ctl_table_set
*root_set
= &sysctl_table_root
.default_set
;
1490 struct ctl_table_root
*root
= header
->root
;
1491 struct ctl_dir
*parent
= header
->parent
;
1492 struct ctl_dir
*core_parent
;
1493 struct ctl_table
*entry
;
1495 if (header
->set
== root_set
)
1498 core_parent
= xlate_dir(root_set
, parent
);
1499 if (IS_ERR(core_parent
))
1502 for (entry
= header
->ctl_table
; entry
->procname
; entry
++) {
1503 struct ctl_table_header
*link_head
;
1504 struct ctl_table
*link
;
1505 const char *name
= entry
->procname
;
1507 link
= find_entry(&link_head
, core_parent
, name
, strlen(name
));
1509 ((S_ISDIR(link
->mode
) && S_ISDIR(entry
->mode
)) ||
1510 (S_ISLNK(link
->mode
) && (link
->data
== root
)))) {
1511 drop_sysctl_table(link_head
);
1514 pr_err("sysctl link missing during unregister: ");
1515 sysctl_print_dir(parent
);
1516 pr_cont("/%s\n", name
);
1521 static void drop_sysctl_table(struct ctl_table_header
*header
)
1523 struct ctl_dir
*parent
= header
->parent
;
1529 start_unregistering(header
);
1530 if (!--header
->count
)
1531 kfree_rcu(header
, rcu
);
1534 drop_sysctl_table(&parent
->header
);
1538 * unregister_sysctl_table - unregister a sysctl table hierarchy
1539 * @header: the header returned from register_sysctl_table
1541 * Unregisters the sysctl table and all children. proc entries may not
1542 * actually be removed until they are no longer used by anyone.
1544 void unregister_sysctl_table(struct ctl_table_header
* header
)
1552 nr_subheaders
= count_subheaders(header
->ctl_table_arg
);
1553 if (unlikely(nr_subheaders
> 1)) {
1554 struct ctl_table_header
**subheaders
;
1557 subheaders
= (struct ctl_table_header
**)(header
+ 1);
1558 for (i
= nr_subheaders
-1; i
>= 0; i
--) {
1559 struct ctl_table_header
*subh
= subheaders
[i
];
1560 struct ctl_table
*table
= subh
->ctl_table_arg
;
1561 unregister_sysctl_table(subh
);
1568 spin_lock(&sysctl_lock
);
1569 drop_sysctl_table(header
);
1570 spin_unlock(&sysctl_lock
);
1572 EXPORT_SYMBOL(unregister_sysctl_table
);
1574 void setup_sysctl_set(struct ctl_table_set
*set
,
1575 struct ctl_table_root
*root
,
1576 int (*is_seen
)(struct ctl_table_set
*))
1578 memset(set
, 0, sizeof(*set
));
1579 set
->is_seen
= is_seen
;
1580 init_header(&set
->dir
.header
, root
, set
, NULL
, root_table
);
1583 void retire_sysctl_set(struct ctl_table_set
*set
)
1585 WARN_ON(!RB_EMPTY_ROOT(&set
->dir
.root
));
1588 int __init
proc_sys_init(void)
1590 struct proc_dir_entry
*proc_sys_root
;
1592 proc_sys_root
= proc_mkdir("sys", NULL
);
1593 proc_sys_root
->proc_iops
= &proc_sys_dir_operations
;
1594 proc_sys_root
->proc_fops
= &proc_sys_dir_file_operations
;
1595 proc_sys_root
->nlink
= 0;
1597 return sysctl_init();