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 set_nlink(inode, de->nlink);
144 generic_fillattr(inode, stat);
148 static const struct inode_operations proc_file_inode_operations = {
149 .setattr = proc_notify_change,
153 * This function parses a name such as "tty/driver/serial", and
154 * returns the struct proc_dir_entry for "/proc/tty/driver", and
155 * returns "serial" in residual.
157 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
158 const char **residual)
160 const char *cp = name, *next;
161 struct proc_dir_entry *de;
169 next = strchr(cp, '/');
174 de = pde_subdir_find(de, cp, len);
176 WARN(1, "name '%s'\n", name);
186 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
187 const char **residual)
191 read_lock(&proc_subdir_lock);
192 rv = __xlate_proc_name(name, ret, residual);
193 read_unlock(&proc_subdir_lock);
197 static DEFINE_IDA(proc_inum_ida);
199 #define PROC_DYNAMIC_FIRST 0xF0000000U
202 * Return an inode number between PROC_DYNAMIC_FIRST and
203 * 0xffffffff, or zero on failure.
205 int proc_alloc_inum(unsigned int *inum)
209 i = ida_simple_get(&proc_inum_ida, 0, UINT_MAX - PROC_DYNAMIC_FIRST + 1,
214 *inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
218 void proc_free_inum(unsigned int inum)
220 ida_simple_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
223 static int proc_misc_d_revalidate(struct dentry *dentry, unsigned int flags)
225 if (flags & LOOKUP_RCU)
228 if (atomic_read(&PDE(d_inode(dentry))->in_use) < 0)
229 return 0; /* revalidate */
233 static int proc_misc_d_delete(const struct dentry *dentry)
235 return atomic_read(&PDE(d_inode(dentry))->in_use) < 0;
238 static const struct dentry_operations proc_misc_dentry_ops = {
239 .d_revalidate = proc_misc_d_revalidate,
240 .d_delete = proc_misc_d_delete,
244 * Don't create negative dentries here, return -ENOENT by hand
247 struct dentry *proc_lookup_de(struct inode *dir, struct dentry *dentry,
248 struct proc_dir_entry *de)
252 read_lock(&proc_subdir_lock);
253 de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
256 read_unlock(&proc_subdir_lock);
257 inode = proc_get_inode(dir->i_sb, de);
259 return ERR_PTR(-ENOMEM);
260 d_set_d_op(dentry, de->proc_dops);
261 return d_splice_alias(inode, dentry);
263 read_unlock(&proc_subdir_lock);
264 return ERR_PTR(-ENOENT);
267 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
270 return proc_lookup_de(dir, dentry, PDE(dir));
274 * This returns non-zero if at EOF, so that the /proc
275 * root directory can use this and check if it should
276 * continue with the <pid> entries..
278 * Note that the VFS-layer doesn't care about the return
279 * value of the readdir() call, as long as it's non-negative
282 int proc_readdir_de(struct file *file, struct dir_context *ctx,
283 struct proc_dir_entry *de)
287 if (!dir_emit_dots(file, ctx))
291 read_lock(&proc_subdir_lock);
292 de = pde_subdir_first(de);
295 read_unlock(&proc_subdir_lock);
300 de = pde_subdir_next(de);
305 struct proc_dir_entry *next;
307 read_unlock(&proc_subdir_lock);
308 if (!dir_emit(ctx, de->name, de->namelen,
309 de->low_ino, de->mode >> 12)) {
314 read_lock(&proc_subdir_lock);
315 next = pde_subdir_next(de);
319 read_unlock(&proc_subdir_lock);
323 int proc_readdir(struct file *file, struct dir_context *ctx)
325 struct inode *inode = file_inode(file);
327 return proc_readdir_de(file, ctx, PDE(inode));
331 * These are the generic /proc directory operations. They
332 * use the in-memory "struct proc_dir_entry" tree to parse
333 * the /proc directory.
335 static const struct file_operations proc_dir_operations = {
336 .llseek = generic_file_llseek,
337 .read = generic_read_dir,
338 .iterate_shared = proc_readdir,
342 * proc directories can do almost nothing..
344 static const struct inode_operations proc_dir_inode_operations = {
345 .lookup = proc_lookup,
346 .getattr = proc_getattr,
347 .setattr = proc_notify_change,
350 /* returns the registered entry, or frees dp and returns NULL on failure */
351 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
352 struct proc_dir_entry *dp)
354 if (proc_alloc_inum(&dp->low_ino))
357 write_lock(&proc_subdir_lock);
359 if (pde_subdir_insert(dir, dp) == false) {
360 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
361 dir->name, dp->name);
362 write_unlock(&proc_subdir_lock);
365 write_unlock(&proc_subdir_lock);
369 proc_free_inum(dp->low_ino);
375 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
380 struct proc_dir_entry *ent = NULL;
384 if (xlate_proc_name(name, parent, &fn) != 0)
387 qstr.len = strlen(fn);
388 if (qstr.len == 0 || qstr.len >= 256) {
389 WARN(1, "name len %u\n", qstr.len);
392 if (qstr.len == 1 && fn[0] == '.') {
393 WARN(1, "name '.'\n");
396 if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
397 WARN(1, "name '..'\n");
400 if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
401 WARN(1, "create '/proc/%s' by hand\n", qstr.name);
404 if (is_empty_pde(*parent)) {
405 WARN(1, "attempt to add to permanently empty directory");
409 ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
413 if (qstr.len + 1 <= SIZEOF_PDE_INLINE_NAME) {
414 ent->name = ent->inline_name;
416 ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
423 memcpy(ent->name, fn, qstr.len + 1);
424 ent->namelen = qstr.len;
427 ent->subdir = RB_ROOT;
428 refcount_set(&ent->refcnt, 1);
429 spin_lock_init(&ent->pde_unload_lock);
430 INIT_LIST_HEAD(&ent->pde_openers);
431 proc_set_user(ent, (*parent)->uid, (*parent)->gid);
433 ent->proc_dops = &proc_misc_dentry_ops;
439 struct proc_dir_entry *proc_symlink(const char *name,
440 struct proc_dir_entry *parent, const char *dest)
442 struct proc_dir_entry *ent;
444 ent = __proc_create(&parent, name,
445 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
448 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
450 strcpy((char*)ent->data,dest);
451 ent->proc_iops = &proc_link_inode_operations;
452 ent = proc_register(parent, ent);
460 EXPORT_SYMBOL(proc_symlink);
462 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
463 struct proc_dir_entry *parent, void *data)
465 struct proc_dir_entry *ent;
468 mode = S_IRUGO | S_IXUGO;
470 ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
473 ent->proc_fops = &proc_dir_operations;
474 ent->proc_iops = &proc_dir_inode_operations;
476 ent = proc_register(parent, ent);
482 EXPORT_SYMBOL_GPL(proc_mkdir_data);
484 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
485 struct proc_dir_entry *parent)
487 return proc_mkdir_data(name, mode, parent, NULL);
489 EXPORT_SYMBOL(proc_mkdir_mode);
491 struct proc_dir_entry *proc_mkdir(const char *name,
492 struct proc_dir_entry *parent)
494 return proc_mkdir_data(name, 0, parent, NULL);
496 EXPORT_SYMBOL(proc_mkdir);
498 struct proc_dir_entry *proc_create_mount_point(const char *name)
500 umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
501 struct proc_dir_entry *ent, *parent = NULL;
503 ent = __proc_create(&parent, name, mode, 2);
506 ent->proc_fops = NULL;
507 ent->proc_iops = NULL;
509 ent = proc_register(parent, ent);
515 EXPORT_SYMBOL(proc_create_mount_point);
517 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
518 struct proc_dir_entry **parent, void *data)
520 struct proc_dir_entry *p;
522 if ((mode & S_IFMT) == 0)
524 if ((mode & S_IALLUGO) == 0)
526 if (WARN_ON_ONCE(!S_ISREG(mode)))
529 p = __proc_create(parent, name, mode, 1);
531 p->proc_iops = &proc_file_inode_operations;
537 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
538 struct proc_dir_entry *parent,
539 const struct file_operations *proc_fops, void *data)
541 struct proc_dir_entry *p;
543 BUG_ON(proc_fops == NULL);
545 p = proc_create_reg(name, mode, &parent, data);
548 p->proc_fops = proc_fops;
549 return proc_register(parent, p);
551 EXPORT_SYMBOL(proc_create_data);
553 struct proc_dir_entry *proc_create(const char *name, umode_t mode,
554 struct proc_dir_entry *parent,
555 const struct file_operations *proc_fops)
557 return proc_create_data(name, mode, parent, proc_fops, NULL);
559 EXPORT_SYMBOL(proc_create);
561 static int proc_seq_open(struct inode *inode, struct file *file)
563 struct proc_dir_entry *de = PDE(inode);
566 return seq_open_private(file, de->seq_ops, de->state_size);
567 return seq_open(file, de->seq_ops);
570 static int proc_seq_release(struct inode *inode, struct file *file)
572 struct proc_dir_entry *de = PDE(inode);
575 return seq_release_private(inode, file);
576 return seq_release(inode, file);
579 static const struct file_operations proc_seq_fops = {
580 .open = proc_seq_open,
583 .release = proc_seq_release,
586 struct proc_dir_entry *proc_create_seq_private(const char *name, umode_t mode,
587 struct proc_dir_entry *parent, const struct seq_operations *ops,
588 unsigned int state_size, void *data)
590 struct proc_dir_entry *p;
592 p = proc_create_reg(name, mode, &parent, data);
595 p->proc_fops = &proc_seq_fops;
597 p->state_size = state_size;
598 return proc_register(parent, p);
600 EXPORT_SYMBOL(proc_create_seq_private);
602 static int proc_single_open(struct inode *inode, struct file *file)
604 struct proc_dir_entry *de = PDE(inode);
606 return single_open(file, de->single_show, de->data);
609 static const struct file_operations proc_single_fops = {
610 .open = proc_single_open,
613 .release = single_release,
616 struct proc_dir_entry *proc_create_single_data(const char *name, umode_t mode,
617 struct proc_dir_entry *parent,
618 int (*show)(struct seq_file *, void *), void *data)
620 struct proc_dir_entry *p;
622 p = proc_create_reg(name, mode, &parent, data);
625 p->proc_fops = &proc_single_fops;
626 p->single_show = show;
627 return proc_register(parent, p);
629 EXPORT_SYMBOL(proc_create_single_data);
631 void proc_set_size(struct proc_dir_entry *de, loff_t size)
635 EXPORT_SYMBOL(proc_set_size);
637 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
642 EXPORT_SYMBOL(proc_set_user);
644 void pde_put(struct proc_dir_entry *pde)
646 if (refcount_dec_and_test(&pde->refcnt)) {
647 proc_free_inum(pde->low_ino);
653 * Remove a /proc entry and free it if it's not currently in use.
655 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
657 struct proc_dir_entry *de = NULL;
658 const char *fn = name;
661 write_lock(&proc_subdir_lock);
662 if (__xlate_proc_name(name, &parent, &fn) != 0) {
663 write_unlock(&proc_subdir_lock);
668 de = pde_subdir_find(parent, fn, len);
670 rb_erase(&de->subdir_node, &parent->subdir);
671 write_unlock(&proc_subdir_lock);
673 WARN(1, "name '%s'\n", name);
677 proc_entry_rundown(de);
679 if (S_ISDIR(de->mode))
682 WARN(pde_subdir_first(de),
683 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
684 __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
687 EXPORT_SYMBOL(remove_proc_entry);
689 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
691 struct proc_dir_entry *root = NULL, *de, *next;
692 const char *fn = name;
695 write_lock(&proc_subdir_lock);
696 if (__xlate_proc_name(name, &parent, &fn) != 0) {
697 write_unlock(&proc_subdir_lock);
702 root = pde_subdir_find(parent, fn, len);
704 write_unlock(&proc_subdir_lock);
707 rb_erase(&root->subdir_node, &parent->subdir);
711 next = pde_subdir_first(de);
713 rb_erase(&next->subdir_node, &de->subdir);
717 write_unlock(&proc_subdir_lock);
719 proc_entry_rundown(de);
721 if (S_ISDIR(de->mode))
728 write_lock(&proc_subdir_lock);
734 EXPORT_SYMBOL(remove_proc_subtree);
736 void *proc_get_parent_data(const struct inode *inode)
738 struct proc_dir_entry *de = PDE(inode);
739 return de->parent->data;
741 EXPORT_SYMBOL_GPL(proc_get_parent_data);
743 void proc_remove(struct proc_dir_entry *de)
746 remove_proc_subtree(de->name, de->parent);
748 EXPORT_SYMBOL(proc_remove);
750 void *PDE_DATA(const struct inode *inode)
752 return __PDE_DATA(inode);
754 EXPORT_SYMBOL(PDE_DATA);
757 * Pull a user buffer into memory and pass it to the file's write handler if
758 * one is supplied. The ->write() method is permitted to modify the
759 * kernel-side buffer.
761 ssize_t proc_simple_write(struct file *f, const char __user *ubuf, size_t size,
764 struct proc_dir_entry *pde = PDE(file_inode(f));
770 if (size == 0 || size > PAGE_SIZE - 1)
772 buf = memdup_user_nul(ubuf, size);
775 ret = pde->write(f, buf, size);
777 return ret == 0 ? size : ret;