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 dentry *dentry, struct iattr *iattr)
120 struct inode *inode = d_inode(dentry);
121 struct proc_dir_entry *de = PDE(inode);
124 error = setattr_prepare(dentry, iattr);
128 setattr_copy(inode, iattr);
129 mark_inode_dirty(inode);
131 proc_set_user(de, inode->i_uid, inode->i_gid);
132 de->mode = inode->i_mode;
136 static int proc_getattr(const struct path *path, struct kstat *stat,
137 u32 request_mask, unsigned int query_flags)
139 struct inode *inode = d_inode(path->dentry);
140 struct proc_dir_entry *de = PDE(inode);
142 nlink_t nlink = READ_ONCE(de->nlink);
144 set_nlink(inode, nlink);
148 generic_fillattr(inode, stat);
152 static const struct inode_operations proc_file_inode_operations = {
153 .setattr = proc_notify_change,
157 * This function parses a name such as "tty/driver/serial", and
158 * returns the struct proc_dir_entry for "/proc/tty/driver", and
159 * returns "serial" in residual.
161 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
162 const char **residual)
164 const char *cp = name, *next;
165 struct proc_dir_entry *de;
172 next = strchr(cp, '/');
176 de = pde_subdir_find(de, cp, next - cp);
178 WARN(1, "name '%s'\n", name);
188 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
189 const char **residual)
193 read_lock(&proc_subdir_lock);
194 rv = __xlate_proc_name(name, ret, residual);
195 read_unlock(&proc_subdir_lock);
199 static DEFINE_IDA(proc_inum_ida);
201 #define PROC_DYNAMIC_FIRST 0xF0000000U
204 * Return an inode number between PROC_DYNAMIC_FIRST and
205 * 0xffffffff, or zero on failure.
207 int proc_alloc_inum(unsigned int *inum)
211 i = ida_simple_get(&proc_inum_ida, 0, UINT_MAX - PROC_DYNAMIC_FIRST + 1,
216 *inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
220 void proc_free_inum(unsigned int inum)
222 ida_simple_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
225 static int proc_misc_d_revalidate(struct dentry *dentry, unsigned int flags)
227 if (flags & LOOKUP_RCU)
230 if (atomic_read(&PDE(d_inode(dentry))->in_use) < 0)
231 return 0; /* revalidate */
235 static int proc_misc_d_delete(const struct dentry *dentry)
237 return atomic_read(&PDE(d_inode(dentry))->in_use) < 0;
240 static const struct dentry_operations proc_misc_dentry_ops = {
241 .d_revalidate = proc_misc_d_revalidate,
242 .d_delete = proc_misc_d_delete,
246 * Don't create negative dentries here, return -ENOENT by hand
249 struct dentry *proc_lookup_de(struct inode *dir, struct dentry *dentry,
250 struct proc_dir_entry *de)
254 read_lock(&proc_subdir_lock);
255 de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
258 read_unlock(&proc_subdir_lock);
259 inode = proc_get_inode(dir->i_sb, de);
261 return ERR_PTR(-ENOMEM);
262 d_set_d_op(dentry, de->proc_dops);
263 return d_splice_alias(inode, dentry);
265 read_unlock(&proc_subdir_lock);
266 return ERR_PTR(-ENOENT);
269 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
272 struct proc_fs_info *fs_info = proc_sb_info(dir->i_sb);
274 if (fs_info->pidonly == PROC_PIDONLY_ON)
275 return ERR_PTR(-ENOENT);
277 return proc_lookup_de(dir, dentry, PDE(dir));
281 * This returns non-zero if at EOF, so that the /proc
282 * root directory can use this and check if it should
283 * continue with the <pid> entries..
285 * Note that the VFS-layer doesn't care about the return
286 * value of the readdir() call, as long as it's non-negative
289 int proc_readdir_de(struct file *file, struct dir_context *ctx,
290 struct proc_dir_entry *de)
294 if (!dir_emit_dots(file, ctx))
298 read_lock(&proc_subdir_lock);
299 de = pde_subdir_first(de);
302 read_unlock(&proc_subdir_lock);
307 de = pde_subdir_next(de);
312 struct proc_dir_entry *next;
314 read_unlock(&proc_subdir_lock);
315 if (!dir_emit(ctx, de->name, de->namelen,
316 de->low_ino, de->mode >> 12)) {
321 read_lock(&proc_subdir_lock);
322 next = pde_subdir_next(de);
326 read_unlock(&proc_subdir_lock);
330 int proc_readdir(struct file *file, struct dir_context *ctx)
332 struct inode *inode = file_inode(file);
333 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
335 if (fs_info->pidonly == PROC_PIDONLY_ON)
338 return proc_readdir_de(file, ctx, PDE(inode));
342 * These are the generic /proc directory operations. They
343 * use the in-memory "struct proc_dir_entry" tree to parse
344 * the /proc directory.
346 static const struct file_operations proc_dir_operations = {
347 .llseek = generic_file_llseek,
348 .read = generic_read_dir,
349 .iterate_shared = proc_readdir,
352 static int proc_net_d_revalidate(struct dentry *dentry, unsigned int flags)
357 const struct dentry_operations proc_net_dentry_ops = {
358 .d_revalidate = proc_net_d_revalidate,
359 .d_delete = always_delete_dentry,
363 * proc directories can do almost nothing..
365 static const struct inode_operations proc_dir_inode_operations = {
366 .lookup = proc_lookup,
367 .getattr = proc_getattr,
368 .setattr = proc_notify_change,
371 /* returns the registered entry, or frees dp and returns NULL on failure */
372 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
373 struct proc_dir_entry *dp)
375 if (proc_alloc_inum(&dp->low_ino))
378 write_lock(&proc_subdir_lock);
380 if (pde_subdir_insert(dir, dp) == false) {
381 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
382 dir->name, dp->name);
383 write_unlock(&proc_subdir_lock);
387 write_unlock(&proc_subdir_lock);
391 proc_free_inum(dp->low_ino);
397 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
402 struct proc_dir_entry *ent = NULL;
406 if (xlate_proc_name(name, parent, &fn) != 0)
409 qstr.len = strlen(fn);
410 if (qstr.len == 0 || qstr.len >= 256) {
411 WARN(1, "name len %u\n", qstr.len);
414 if (qstr.len == 1 && fn[0] == '.') {
415 WARN(1, "name '.'\n");
418 if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
419 WARN(1, "name '..'\n");
422 if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
423 WARN(1, "create '/proc/%s' by hand\n", qstr.name);
426 if (is_empty_pde(*parent)) {
427 WARN(1, "attempt to add to permanently empty directory");
431 ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
435 if (qstr.len + 1 <= SIZEOF_PDE_INLINE_NAME) {
436 ent->name = ent->inline_name;
438 ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
445 memcpy(ent->name, fn, qstr.len + 1);
446 ent->namelen = qstr.len;
449 ent->subdir = RB_ROOT;
450 refcount_set(&ent->refcnt, 1);
451 spin_lock_init(&ent->pde_unload_lock);
452 INIT_LIST_HEAD(&ent->pde_openers);
453 proc_set_user(ent, (*parent)->uid, (*parent)->gid);
455 ent->proc_dops = &proc_misc_dentry_ops;
456 /* Revalidate everything under /proc/${pid}/net */
457 if ((*parent)->proc_dops == &proc_net_dentry_ops)
458 pde_force_lookup(ent);
464 struct proc_dir_entry *proc_symlink(const char *name,
465 struct proc_dir_entry *parent, const char *dest)
467 struct proc_dir_entry *ent;
469 ent = __proc_create(&parent, name,
470 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
473 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
475 strcpy((char*)ent->data,dest);
476 ent->proc_iops = &proc_link_inode_operations;
477 ent = proc_register(parent, ent);
485 EXPORT_SYMBOL(proc_symlink);
487 struct proc_dir_entry *_proc_mkdir(const char *name, umode_t mode,
488 struct proc_dir_entry *parent, void *data, bool force_lookup)
490 struct proc_dir_entry *ent;
493 mode = S_IRUGO | S_IXUGO;
495 ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
498 ent->proc_dir_ops = &proc_dir_operations;
499 ent->proc_iops = &proc_dir_inode_operations;
501 pde_force_lookup(ent);
503 ent = proc_register(parent, ent);
507 EXPORT_SYMBOL_GPL(_proc_mkdir);
509 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
510 struct proc_dir_entry *parent, void *data)
512 return _proc_mkdir(name, mode, parent, data, false);
514 EXPORT_SYMBOL_GPL(proc_mkdir_data);
516 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
517 struct proc_dir_entry *parent)
519 return proc_mkdir_data(name, mode, parent, NULL);
521 EXPORT_SYMBOL(proc_mkdir_mode);
523 struct proc_dir_entry *proc_mkdir(const char *name,
524 struct proc_dir_entry *parent)
526 return proc_mkdir_data(name, 0, parent, NULL);
528 EXPORT_SYMBOL(proc_mkdir);
530 struct proc_dir_entry *proc_create_mount_point(const char *name)
532 umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
533 struct proc_dir_entry *ent, *parent = NULL;
535 ent = __proc_create(&parent, name, mode, 2);
538 ent->proc_dir_ops = NULL;
539 ent->proc_iops = NULL;
540 ent = proc_register(parent, ent);
544 EXPORT_SYMBOL(proc_create_mount_point);
546 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
547 struct proc_dir_entry **parent, void *data)
549 struct proc_dir_entry *p;
551 if ((mode & S_IFMT) == 0)
553 if ((mode & S_IALLUGO) == 0)
555 if (WARN_ON_ONCE(!S_ISREG(mode)))
558 p = __proc_create(parent, name, mode, 1);
560 p->proc_iops = &proc_file_inode_operations;
566 static inline void pde_set_flags(struct proc_dir_entry *pde)
568 if (pde->proc_ops->proc_flags & PROC_ENTRY_PERMANENT)
569 pde->flags |= PROC_ENTRY_PERMANENT;
572 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
573 struct proc_dir_entry *parent,
574 const struct proc_ops *proc_ops, void *data)
576 struct proc_dir_entry *p;
578 p = proc_create_reg(name, mode, &parent, data);
581 p->proc_ops = proc_ops;
583 return proc_register(parent, p);
585 EXPORT_SYMBOL(proc_create_data);
587 struct proc_dir_entry *proc_create(const char *name, umode_t mode,
588 struct proc_dir_entry *parent,
589 const struct proc_ops *proc_ops)
591 return proc_create_data(name, mode, parent, proc_ops, NULL);
593 EXPORT_SYMBOL(proc_create);
595 static int proc_seq_open(struct inode *inode, struct file *file)
597 struct proc_dir_entry *de = PDE(inode);
600 return seq_open_private(file, de->seq_ops, de->state_size);
601 return seq_open(file, de->seq_ops);
604 static int proc_seq_release(struct inode *inode, struct file *file)
606 struct proc_dir_entry *de = PDE(inode);
609 return seq_release_private(inode, file);
610 return seq_release(inode, file);
613 static const struct proc_ops proc_seq_ops = {
614 /* not permanent -- can call into arbitrary seq_operations */
615 .proc_open = proc_seq_open,
616 .proc_read_iter = seq_read_iter,
617 .proc_lseek = seq_lseek,
618 .proc_release = proc_seq_release,
621 struct proc_dir_entry *proc_create_seq_private(const char *name, umode_t mode,
622 struct proc_dir_entry *parent, const struct seq_operations *ops,
623 unsigned int state_size, void *data)
625 struct proc_dir_entry *p;
627 p = proc_create_reg(name, mode, &parent, data);
630 p->proc_ops = &proc_seq_ops;
632 p->state_size = state_size;
633 return proc_register(parent, p);
635 EXPORT_SYMBOL(proc_create_seq_private);
637 static int proc_single_open(struct inode *inode, struct file *file)
639 struct proc_dir_entry *de = PDE(inode);
641 return single_open(file, de->single_show, de->data);
644 static const struct proc_ops proc_single_ops = {
645 /* not permanent -- can call into arbitrary ->single_show */
646 .proc_open = proc_single_open,
647 .proc_read_iter = seq_read_iter,
648 .proc_lseek = seq_lseek,
649 .proc_release = single_release,
652 struct proc_dir_entry *proc_create_single_data(const char *name, umode_t mode,
653 struct proc_dir_entry *parent,
654 int (*show)(struct seq_file *, void *), void *data)
656 struct proc_dir_entry *p;
658 p = proc_create_reg(name, mode, &parent, data);
661 p->proc_ops = &proc_single_ops;
662 p->single_show = show;
663 return proc_register(parent, p);
665 EXPORT_SYMBOL(proc_create_single_data);
667 void proc_set_size(struct proc_dir_entry *de, loff_t size)
671 EXPORT_SYMBOL(proc_set_size);
673 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
678 EXPORT_SYMBOL(proc_set_user);
680 void pde_put(struct proc_dir_entry *pde)
682 if (refcount_dec_and_test(&pde->refcnt)) {
683 proc_free_inum(pde->low_ino);
689 * Remove a /proc entry and free it if it's not currently in use.
691 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
693 struct proc_dir_entry *de = NULL;
694 const char *fn = name;
697 write_lock(&proc_subdir_lock);
698 if (__xlate_proc_name(name, &parent, &fn) != 0) {
699 write_unlock(&proc_subdir_lock);
704 de = pde_subdir_find(parent, fn, len);
706 if (unlikely(pde_is_permanent(de))) {
707 WARN(1, "removing permanent /proc entry '%s'", de->name);
710 rb_erase(&de->subdir_node, &parent->subdir);
711 if (S_ISDIR(de->mode))
715 write_unlock(&proc_subdir_lock);
717 WARN(1, "name '%s'\n", name);
721 proc_entry_rundown(de);
723 WARN(pde_subdir_first(de),
724 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
725 __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
728 EXPORT_SYMBOL(remove_proc_entry);
730 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
732 struct proc_dir_entry *root = NULL, *de, *next;
733 const char *fn = name;
736 write_lock(&proc_subdir_lock);
737 if (__xlate_proc_name(name, &parent, &fn) != 0) {
738 write_unlock(&proc_subdir_lock);
743 root = pde_subdir_find(parent, fn, len);
745 write_unlock(&proc_subdir_lock);
748 if (unlikely(pde_is_permanent(root))) {
749 write_unlock(&proc_subdir_lock);
750 WARN(1, "removing permanent /proc entry '%s/%s'",
751 root->parent->name, root->name);
754 rb_erase(&root->subdir_node, &parent->subdir);
758 next = pde_subdir_first(de);
760 if (unlikely(pde_is_permanent(next))) {
761 write_unlock(&proc_subdir_lock);
762 WARN(1, "removing permanent /proc entry '%s/%s'",
763 next->parent->name, next->name);
766 rb_erase(&next->subdir_node, &de->subdir);
771 if (S_ISDIR(de->mode))
773 write_unlock(&proc_subdir_lock);
775 proc_entry_rundown(de);
780 write_lock(&proc_subdir_lock);
786 EXPORT_SYMBOL(remove_proc_subtree);
788 void *proc_get_parent_data(const struct inode *inode)
790 struct proc_dir_entry *de = PDE(inode);
791 return de->parent->data;
793 EXPORT_SYMBOL_GPL(proc_get_parent_data);
795 void proc_remove(struct proc_dir_entry *de)
798 remove_proc_subtree(de->name, de->parent);
800 EXPORT_SYMBOL(proc_remove);
802 void *PDE_DATA(const struct inode *inode)
804 return __PDE_DATA(inode);
806 EXPORT_SYMBOL(PDE_DATA);
809 * Pull a user buffer into memory and pass it to the file's write handler if
810 * one is supplied. The ->write() method is permitted to modify the
811 * kernel-side buffer.
813 ssize_t proc_simple_write(struct file *f, const char __user *ubuf, size_t size,
816 struct proc_dir_entry *pde = PDE(file_inode(f));
822 if (size == 0 || size > PAGE_SIZE - 1)
824 buf = memdup_user_nul(ubuf, size);
827 ret = pde->write(f, buf, size);
829 return ret == 0 ? size : ret;