1 // SPDX-License-Identifier: GPL-2.0-only
3 * proc/fs/generic.c --- generic routines for the proc-fs
5 * This file contains generic proc-fs routines for handling
6 * directories and files.
8 * Copyright (C) 1991, 1992 Linus Torvalds.
9 * Copyright (C) 1997 Theodore Ts'o
12 #include <linux/cache.h>
13 #include <linux/errno.h>
14 #include <linux/time.h>
15 #include <linux/proc_fs.h>
16 #include <linux/stat.h>
18 #include <linux/module.h>
19 #include <linux/namei.h>
20 #include <linux/slab.h>
21 #include <linux/printk.h>
22 #include <linux/mount.h>
23 #include <linux/init.h>
24 #include <linux/idr.h>
25 #include <linux/bitops.h>
26 #include <linux/spinlock.h>
27 #include <linux/completion.h>
28 #include <linux/uaccess.h>
29 #include <linux/seq_file.h>
33 static DEFINE_RWLOCK(proc_subdir_lock
);
35 struct kmem_cache
*proc_dir_entry_cache __ro_after_init
;
37 void pde_free(struct proc_dir_entry
*pde
)
39 if (S_ISLNK(pde
->mode
))
41 if (pde
->name
!= pde
->inline_name
)
43 kmem_cache_free(proc_dir_entry_cache
, pde
);
46 static int proc_match(const char *name
, struct proc_dir_entry
*de
, unsigned int len
)
48 if (len
< de
->namelen
)
50 if (len
> de
->namelen
)
53 return memcmp(name
, de
->name
, len
);
56 static struct proc_dir_entry
*pde_subdir_first(struct proc_dir_entry
*dir
)
58 return rb_entry_safe(rb_first(&dir
->subdir
), struct proc_dir_entry
,
62 static struct proc_dir_entry
*pde_subdir_next(struct proc_dir_entry
*dir
)
64 return rb_entry_safe(rb_next(&dir
->subdir_node
), struct proc_dir_entry
,
68 static struct proc_dir_entry
*pde_subdir_find(struct proc_dir_entry
*dir
,
72 struct rb_node
*node
= dir
->subdir
.rb_node
;
75 struct proc_dir_entry
*de
= rb_entry(node
,
76 struct proc_dir_entry
,
78 int result
= proc_match(name
, de
, len
);
83 node
= node
->rb_right
;
90 static bool pde_subdir_insert(struct proc_dir_entry
*dir
,
91 struct proc_dir_entry
*de
)
93 struct rb_root
*root
= &dir
->subdir
;
94 struct rb_node
**new = &root
->rb_node
, *parent
= NULL
;
96 /* Figure out where to put new node */
98 struct proc_dir_entry
*this = rb_entry(*new,
99 struct proc_dir_entry
,
101 int result
= proc_match(de
->name
, this, de
->namelen
);
105 new = &(*new)->rb_left
;
107 new = &(*new)->rb_right
;
112 /* Add new node and rebalance tree. */
113 rb_link_node(&de
->subdir_node
, parent
, new);
114 rb_insert_color(&de
->subdir_node
, root
);
118 static int proc_notify_change(struct mnt_idmap
*idmap
,
119 struct dentry
*dentry
, struct iattr
*iattr
)
121 struct inode
*inode
= d_inode(dentry
);
122 struct proc_dir_entry
*de
= PDE(inode
);
125 error
= setattr_prepare(&nop_mnt_idmap
, dentry
, iattr
);
129 setattr_copy(&nop_mnt_idmap
, inode
, iattr
);
131 proc_set_user(de
, inode
->i_uid
, inode
->i_gid
);
132 de
->mode
= inode
->i_mode
;
136 static int proc_getattr(struct mnt_idmap
*idmap
,
137 const struct path
*path
, struct kstat
*stat
,
138 u32 request_mask
, unsigned int query_flags
)
140 struct inode
*inode
= d_inode(path
->dentry
);
141 struct proc_dir_entry
*de
= PDE(inode
);
143 nlink_t nlink
= READ_ONCE(de
->nlink
);
145 set_nlink(inode
, nlink
);
149 generic_fillattr(&nop_mnt_idmap
, request_mask
, inode
, stat
);
153 static const struct inode_operations proc_file_inode_operations
= {
154 .setattr
= proc_notify_change
,
158 * This function parses a name such as "tty/driver/serial", and
159 * returns the struct proc_dir_entry for "/proc/tty/driver", and
160 * returns "serial" in residual.
162 static int __xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
163 const char **residual
)
165 const char *cp
= name
, *next
;
166 struct proc_dir_entry
*de
;
168 de
= *ret
?: &proc_root
;
169 while ((next
= strchr(cp
, '/')) != NULL
) {
170 de
= pde_subdir_find(de
, cp
, next
- cp
);
172 WARN(1, "name '%s'\n", name
);
182 static int xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
183 const char **residual
)
187 read_lock(&proc_subdir_lock
);
188 rv
= __xlate_proc_name(name
, ret
, residual
);
189 read_unlock(&proc_subdir_lock
);
193 static DEFINE_IDA(proc_inum_ida
);
195 #define PROC_DYNAMIC_FIRST 0xF0000000U
198 * Return an inode number between PROC_DYNAMIC_FIRST and
199 * 0xffffffff, or zero on failure.
201 int proc_alloc_inum(unsigned int *inum
)
205 i
= ida_alloc_max(&proc_inum_ida
, UINT_MAX
- PROC_DYNAMIC_FIRST
,
210 *inum
= PROC_DYNAMIC_FIRST
+ (unsigned int)i
;
214 void proc_free_inum(unsigned int inum
)
216 ida_free(&proc_inum_ida
, inum
- PROC_DYNAMIC_FIRST
);
219 static int proc_misc_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
221 if (flags
& LOOKUP_RCU
)
224 if (atomic_read(&PDE(d_inode(dentry
))->in_use
) < 0)
225 return 0; /* revalidate */
229 static int proc_misc_d_delete(const struct dentry
*dentry
)
231 return atomic_read(&PDE(d_inode(dentry
))->in_use
) < 0;
234 static const struct dentry_operations proc_misc_dentry_ops
= {
235 .d_revalidate
= proc_misc_d_revalidate
,
236 .d_delete
= proc_misc_d_delete
,
240 * Don't create negative dentries here, return -ENOENT by hand
243 struct dentry
*proc_lookup_de(struct inode
*dir
, struct dentry
*dentry
,
244 struct proc_dir_entry
*de
)
248 read_lock(&proc_subdir_lock
);
249 de
= pde_subdir_find(de
, dentry
->d_name
.name
, dentry
->d_name
.len
);
252 read_unlock(&proc_subdir_lock
);
253 inode
= proc_get_inode(dir
->i_sb
, de
);
255 return ERR_PTR(-ENOMEM
);
256 d_set_d_op(dentry
, de
->proc_dops
);
257 return d_splice_alias(inode
, dentry
);
259 read_unlock(&proc_subdir_lock
);
260 return ERR_PTR(-ENOENT
);
263 struct dentry
*proc_lookup(struct inode
*dir
, struct dentry
*dentry
,
266 struct proc_fs_info
*fs_info
= proc_sb_info(dir
->i_sb
);
268 if (fs_info
->pidonly
== PROC_PIDONLY_ON
)
269 return ERR_PTR(-ENOENT
);
271 return proc_lookup_de(dir
, dentry
, PDE(dir
));
275 * This returns non-zero if at EOF, so that the /proc
276 * root directory can use this and check if it should
277 * continue with the <pid> entries..
279 * Note that the VFS-layer doesn't care about the return
280 * value of the readdir() call, as long as it's non-negative
283 int proc_readdir_de(struct file
*file
, struct dir_context
*ctx
,
284 struct proc_dir_entry
*de
)
288 if (!dir_emit_dots(file
, ctx
))
292 read_lock(&proc_subdir_lock
);
293 de
= pde_subdir_first(de
);
296 read_unlock(&proc_subdir_lock
);
301 de
= pde_subdir_next(de
);
306 struct proc_dir_entry
*next
;
308 read_unlock(&proc_subdir_lock
);
309 if (!dir_emit(ctx
, de
->name
, de
->namelen
,
310 de
->low_ino
, de
->mode
>> 12)) {
315 read_lock(&proc_subdir_lock
);
316 next
= pde_subdir_next(de
);
320 read_unlock(&proc_subdir_lock
);
324 int proc_readdir(struct file
*file
, struct dir_context
*ctx
)
326 struct inode
*inode
= file_inode(file
);
327 struct proc_fs_info
*fs_info
= proc_sb_info(inode
->i_sb
);
329 if (fs_info
->pidonly
== PROC_PIDONLY_ON
)
332 return proc_readdir_de(file
, ctx
, PDE(inode
));
336 * These are the generic /proc directory operations. They
337 * use the in-memory "struct proc_dir_entry" tree to parse
338 * the /proc directory.
340 static const struct file_operations proc_dir_operations
= {
341 .llseek
= generic_file_llseek
,
342 .read
= generic_read_dir
,
343 .iterate_shared
= proc_readdir
,
346 static int proc_net_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
351 const struct dentry_operations proc_net_dentry_ops
= {
352 .d_revalidate
= proc_net_d_revalidate
,
353 .d_delete
= always_delete_dentry
,
357 * proc directories can do almost nothing..
359 static const struct inode_operations proc_dir_inode_operations
= {
360 .lookup
= proc_lookup
,
361 .getattr
= proc_getattr
,
362 .setattr
= proc_notify_change
,
365 /* returns the registered entry, or frees dp and returns NULL on failure */
366 struct proc_dir_entry
*proc_register(struct proc_dir_entry
*dir
,
367 struct proc_dir_entry
*dp
)
369 if (proc_alloc_inum(&dp
->low_ino
))
372 write_lock(&proc_subdir_lock
);
374 if (pde_subdir_insert(dir
, dp
) == false) {
375 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
376 dir
->name
, dp
->name
);
377 write_unlock(&proc_subdir_lock
);
381 write_unlock(&proc_subdir_lock
);
385 proc_free_inum(dp
->low_ino
);
391 static struct proc_dir_entry
*__proc_create(struct proc_dir_entry
**parent
,
396 struct proc_dir_entry
*ent
= NULL
;
400 if (xlate_proc_name(name
, parent
, &fn
) != 0)
403 qstr
.len
= strlen(fn
);
404 if (qstr
.len
== 0 || qstr
.len
>= 256) {
405 WARN(1, "name len %u\n", qstr
.len
);
408 if (qstr
.len
== 1 && fn
[0] == '.') {
409 WARN(1, "name '.'\n");
412 if (qstr
.len
== 2 && fn
[0] == '.' && fn
[1] == '.') {
413 WARN(1, "name '..'\n");
416 if (*parent
== &proc_root
&& name_to_int(&qstr
) != ~0U) {
417 WARN(1, "create '/proc/%s' by hand\n", qstr
.name
);
420 if (is_empty_pde(*parent
)) {
421 WARN(1, "attempt to add to permanently empty directory");
425 ent
= kmem_cache_zalloc(proc_dir_entry_cache
, GFP_KERNEL
);
429 if (qstr
.len
+ 1 <= SIZEOF_PDE_INLINE_NAME
) {
430 ent
->name
= ent
->inline_name
;
432 ent
->name
= kmalloc(qstr
.len
+ 1, GFP_KERNEL
);
439 memcpy(ent
->name
, fn
, qstr
.len
+ 1);
440 ent
->namelen
= qstr
.len
;
443 ent
->subdir
= RB_ROOT
;
444 refcount_set(&ent
->refcnt
, 1);
445 spin_lock_init(&ent
->pde_unload_lock
);
446 INIT_LIST_HEAD(&ent
->pde_openers
);
447 proc_set_user(ent
, (*parent
)->uid
, (*parent
)->gid
);
449 ent
->proc_dops
= &proc_misc_dentry_ops
;
450 /* Revalidate everything under /proc/${pid}/net */
451 if ((*parent
)->proc_dops
== &proc_net_dentry_ops
)
452 pde_force_lookup(ent
);
458 struct proc_dir_entry
*proc_symlink(const char *name
,
459 struct proc_dir_entry
*parent
, const char *dest
)
461 struct proc_dir_entry
*ent
;
463 ent
= __proc_create(&parent
, name
,
464 (S_IFLNK
| S_IRUGO
| S_IWUGO
| S_IXUGO
),1);
467 ent
->size
= strlen(dest
);
468 ent
->data
= kmemdup(dest
, ent
->size
+ 1, GFP_KERNEL
);
470 ent
->proc_iops
= &proc_link_inode_operations
;
471 ent
= proc_register(parent
, ent
);
479 EXPORT_SYMBOL(proc_symlink
);
481 struct proc_dir_entry
*_proc_mkdir(const char *name
, umode_t mode
,
482 struct proc_dir_entry
*parent
, void *data
, bool force_lookup
)
484 struct proc_dir_entry
*ent
;
487 mode
= S_IRUGO
| S_IXUGO
;
489 ent
= __proc_create(&parent
, name
, S_IFDIR
| mode
, 2);
492 ent
->proc_dir_ops
= &proc_dir_operations
;
493 ent
->proc_iops
= &proc_dir_inode_operations
;
495 pde_force_lookup(ent
);
497 ent
= proc_register(parent
, ent
);
501 EXPORT_SYMBOL_GPL(_proc_mkdir
);
503 struct proc_dir_entry
*proc_mkdir_data(const char *name
, umode_t mode
,
504 struct proc_dir_entry
*parent
, void *data
)
506 return _proc_mkdir(name
, mode
, parent
, data
, false);
508 EXPORT_SYMBOL_GPL(proc_mkdir_data
);
510 struct proc_dir_entry
*proc_mkdir_mode(const char *name
, umode_t mode
,
511 struct proc_dir_entry
*parent
)
513 return proc_mkdir_data(name
, mode
, parent
, NULL
);
515 EXPORT_SYMBOL(proc_mkdir_mode
);
517 struct proc_dir_entry
*proc_mkdir(const char *name
,
518 struct proc_dir_entry
*parent
)
520 return proc_mkdir_data(name
, 0, parent
, NULL
);
522 EXPORT_SYMBOL(proc_mkdir
);
524 struct proc_dir_entry
*proc_create_mount_point(const char *name
)
526 umode_t mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
527 struct proc_dir_entry
*ent
, *parent
= NULL
;
529 ent
= __proc_create(&parent
, name
, mode
, 2);
532 ent
->proc_dir_ops
= NULL
;
533 ent
->proc_iops
= NULL
;
534 ent
= proc_register(parent
, ent
);
538 EXPORT_SYMBOL(proc_create_mount_point
);
540 struct proc_dir_entry
*proc_create_reg(const char *name
, umode_t mode
,
541 struct proc_dir_entry
**parent
, void *data
)
543 struct proc_dir_entry
*p
;
545 if ((mode
& S_IFMT
) == 0)
547 if ((mode
& S_IALLUGO
) == 0)
549 if (WARN_ON_ONCE(!S_ISREG(mode
)))
552 p
= __proc_create(parent
, name
, mode
, 1);
554 p
->proc_iops
= &proc_file_inode_operations
;
560 static inline void pde_set_flags(struct proc_dir_entry
*pde
)
562 if (pde
->proc_ops
->proc_flags
& PROC_ENTRY_PERMANENT
)
563 pde
->flags
|= PROC_ENTRY_PERMANENT
;
566 struct proc_dir_entry
*proc_create_data(const char *name
, umode_t mode
,
567 struct proc_dir_entry
*parent
,
568 const struct proc_ops
*proc_ops
, void *data
)
570 struct proc_dir_entry
*p
;
572 p
= proc_create_reg(name
, mode
, &parent
, data
);
575 p
->proc_ops
= proc_ops
;
577 return proc_register(parent
, p
);
579 EXPORT_SYMBOL(proc_create_data
);
581 struct proc_dir_entry
*proc_create(const char *name
, umode_t mode
,
582 struct proc_dir_entry
*parent
,
583 const struct proc_ops
*proc_ops
)
585 return proc_create_data(name
, mode
, parent
, proc_ops
, NULL
);
587 EXPORT_SYMBOL(proc_create
);
589 static int proc_seq_open(struct inode
*inode
, struct file
*file
)
591 struct proc_dir_entry
*de
= PDE(inode
);
594 return seq_open_private(file
, de
->seq_ops
, de
->state_size
);
595 return seq_open(file
, de
->seq_ops
);
598 static int proc_seq_release(struct inode
*inode
, struct file
*file
)
600 struct proc_dir_entry
*de
= PDE(inode
);
603 return seq_release_private(inode
, file
);
604 return seq_release(inode
, file
);
607 static const struct proc_ops proc_seq_ops
= {
608 /* not permanent -- can call into arbitrary seq_operations */
609 .proc_open
= proc_seq_open
,
610 .proc_read_iter
= seq_read_iter
,
611 .proc_lseek
= seq_lseek
,
612 .proc_release
= proc_seq_release
,
615 struct proc_dir_entry
*proc_create_seq_private(const char *name
, umode_t mode
,
616 struct proc_dir_entry
*parent
, const struct seq_operations
*ops
,
617 unsigned int state_size
, void *data
)
619 struct proc_dir_entry
*p
;
621 p
= proc_create_reg(name
, mode
, &parent
, data
);
624 p
->proc_ops
= &proc_seq_ops
;
626 p
->state_size
= state_size
;
627 return proc_register(parent
, p
);
629 EXPORT_SYMBOL(proc_create_seq_private
);
631 static int proc_single_open(struct inode
*inode
, struct file
*file
)
633 struct proc_dir_entry
*de
= PDE(inode
);
635 return single_open(file
, de
->single_show
, de
->data
);
638 static const struct proc_ops proc_single_ops
= {
639 /* not permanent -- can call into arbitrary ->single_show */
640 .proc_open
= proc_single_open
,
641 .proc_read_iter
= seq_read_iter
,
642 .proc_lseek
= seq_lseek
,
643 .proc_release
= single_release
,
646 struct proc_dir_entry
*proc_create_single_data(const char *name
, umode_t mode
,
647 struct proc_dir_entry
*parent
,
648 int (*show
)(struct seq_file
*, void *), void *data
)
650 struct proc_dir_entry
*p
;
652 p
= proc_create_reg(name
, mode
, &parent
, data
);
655 p
->proc_ops
= &proc_single_ops
;
656 p
->single_show
= show
;
657 return proc_register(parent
, p
);
659 EXPORT_SYMBOL(proc_create_single_data
);
661 void proc_set_size(struct proc_dir_entry
*de
, loff_t size
)
665 EXPORT_SYMBOL(proc_set_size
);
667 void proc_set_user(struct proc_dir_entry
*de
, kuid_t uid
, kgid_t gid
)
672 EXPORT_SYMBOL(proc_set_user
);
674 void pde_put(struct proc_dir_entry
*pde
)
676 if (refcount_dec_and_test(&pde
->refcnt
)) {
677 proc_free_inum(pde
->low_ino
);
683 * Remove a /proc entry and free it if it's not currently in use.
685 void remove_proc_entry(const char *name
, struct proc_dir_entry
*parent
)
687 struct proc_dir_entry
*de
= NULL
;
688 const char *fn
= name
;
691 write_lock(&proc_subdir_lock
);
692 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
693 write_unlock(&proc_subdir_lock
);
698 de
= pde_subdir_find(parent
, fn
, len
);
700 if (unlikely(pde_is_permanent(de
))) {
701 WARN(1, "removing permanent /proc entry '%s'", de
->name
);
704 rb_erase(&de
->subdir_node
, &parent
->subdir
);
705 if (S_ISDIR(de
->mode
))
709 write_unlock(&proc_subdir_lock
);
711 WARN(1, "name '%s'\n", name
);
715 proc_entry_rundown(de
);
717 WARN(pde_subdir_first(de
),
718 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
719 __func__
, de
->parent
->name
, de
->name
, pde_subdir_first(de
)->name
);
722 EXPORT_SYMBOL(remove_proc_entry
);
724 int remove_proc_subtree(const char *name
, struct proc_dir_entry
*parent
)
726 struct proc_dir_entry
*root
= NULL
, *de
, *next
;
727 const char *fn
= name
;
730 write_lock(&proc_subdir_lock
);
731 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
732 write_unlock(&proc_subdir_lock
);
737 root
= pde_subdir_find(parent
, fn
, len
);
739 write_unlock(&proc_subdir_lock
);
742 if (unlikely(pde_is_permanent(root
))) {
743 write_unlock(&proc_subdir_lock
);
744 WARN(1, "removing permanent /proc entry '%s/%s'",
745 root
->parent
->name
, root
->name
);
748 rb_erase(&root
->subdir_node
, &parent
->subdir
);
752 next
= pde_subdir_first(de
);
754 if (unlikely(pde_is_permanent(next
))) {
755 write_unlock(&proc_subdir_lock
);
756 WARN(1, "removing permanent /proc entry '%s/%s'",
757 next
->parent
->name
, next
->name
);
760 rb_erase(&next
->subdir_node
, &de
->subdir
);
765 if (S_ISDIR(de
->mode
))
767 write_unlock(&proc_subdir_lock
);
769 proc_entry_rundown(de
);
774 write_lock(&proc_subdir_lock
);
780 EXPORT_SYMBOL(remove_proc_subtree
);
782 void *proc_get_parent_data(const struct inode
*inode
)
784 struct proc_dir_entry
*de
= PDE(inode
);
785 return de
->parent
->data
;
787 EXPORT_SYMBOL_GPL(proc_get_parent_data
);
789 void proc_remove(struct proc_dir_entry
*de
)
792 remove_proc_subtree(de
->name
, de
->parent
);
794 EXPORT_SYMBOL(proc_remove
);
797 * Pull a user buffer into memory and pass it to the file's write handler if
798 * one is supplied. The ->write() method is permitted to modify the
799 * kernel-side buffer.
801 ssize_t
proc_simple_write(struct file
*f
, const char __user
*ubuf
, size_t size
,
804 struct proc_dir_entry
*pde
= PDE(file_inode(f
));
810 if (size
== 0 || size
> PAGE_SIZE
- 1)
812 buf
= memdup_user_nul(ubuf
, size
);
815 ret
= pde
->write(f
, buf
, size
);
817 return ret
== 0 ? size
: ret
;