2 * proc/fs/generic.c --- generic routines for the proc-fs
4 * This file contains generic proc-fs routines for handling
5 * directories and files.
7 * Copyright (C) 1991, 1992 Linus Torvalds.
8 * Copyright (C) 1997 Theodore Ts'o
11 #include <linux/cache.h>
12 #include <linux/errno.h>
13 #include <linux/time.h>
14 #include <linux/proc_fs.h>
15 #include <linux/stat.h>
17 #include <linux/module.h>
18 #include <linux/namei.h>
19 #include <linux/slab.h>
20 #include <linux/printk.h>
21 #include <linux/mount.h>
22 #include <linux/init.h>
23 #include <linux/idr.h>
24 #include <linux/bitops.h>
25 #include <linux/spinlock.h>
26 #include <linux/completion.h>
27 #include <linux/uaccess.h>
28 #include <linux/seq_file.h>
32 static DEFINE_RWLOCK(proc_subdir_lock);
34 struct kmem_cache *proc_dir_entry_cache __ro_after_init;
36 void pde_free(struct proc_dir_entry *pde)
38 if (S_ISLNK(pde->mode))
40 if (pde->name != pde->inline_name)
42 kmem_cache_free(proc_dir_entry_cache, pde);
45 static int proc_match(const char *name, struct proc_dir_entry *de, unsigned int len)
47 if (len < de->namelen)
49 if (len > de->namelen)
52 return memcmp(name, de->name, len);
55 static struct proc_dir_entry *pde_subdir_first(struct proc_dir_entry *dir)
57 return rb_entry_safe(rb_first(&dir->subdir), struct proc_dir_entry,
61 static struct proc_dir_entry *pde_subdir_next(struct proc_dir_entry *dir)
63 return rb_entry_safe(rb_next(&dir->subdir_node), struct proc_dir_entry,
67 static struct proc_dir_entry *pde_subdir_find(struct proc_dir_entry *dir,
71 struct rb_node *node = dir->subdir.rb_node;
74 struct proc_dir_entry *de = rb_entry(node,
75 struct proc_dir_entry,
77 int result = proc_match(name, de, len);
82 node = node->rb_right;
89 static bool pde_subdir_insert(struct proc_dir_entry *dir,
90 struct proc_dir_entry *de)
92 struct rb_root *root = &dir->subdir;
93 struct rb_node **new = &root->rb_node, *parent = NULL;
95 /* Figure out where to put new node */
97 struct proc_dir_entry *this = rb_entry(*new,
98 struct proc_dir_entry,
100 int result = proc_match(de->name, this, de->namelen);
104 new = &(*new)->rb_left;
106 new = &(*new)->rb_right;
111 /* Add new node and rebalance tree. */
112 rb_link_node(&de->subdir_node, parent, new);
113 rb_insert_color(&de->subdir_node, root);
117 static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
119 struct inode *inode = d_inode(dentry);
120 struct proc_dir_entry *de = PDE(inode);
123 error = setattr_prepare(dentry, iattr);
127 setattr_copy(inode, iattr);
128 mark_inode_dirty(inode);
130 proc_set_user(de, inode->i_uid, inode->i_gid);
131 de->mode = inode->i_mode;
135 static int proc_getattr(const struct path *path, struct kstat *stat,
136 u32 request_mask, unsigned int query_flags)
138 struct inode *inode = d_inode(path->dentry);
139 struct proc_dir_entry *de = PDE(inode);
141 nlink_t nlink = READ_ONCE(de->nlink);
143 set_nlink(inode, nlink);
147 generic_fillattr(inode, stat);
151 static const struct inode_operations proc_file_inode_operations = {
152 .setattr = proc_notify_change,
156 * This function parses a name such as "tty/driver/serial", and
157 * returns the struct proc_dir_entry for "/proc/tty/driver", and
158 * returns "serial" in residual.
160 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
161 const char **residual)
163 const char *cp = name, *next;
164 struct proc_dir_entry *de;
172 next = strchr(cp, '/');
177 de = pde_subdir_find(de, cp, len);
179 WARN(1, "name '%s'\n", name);
189 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
190 const char **residual)
194 read_lock(&proc_subdir_lock);
195 rv = __xlate_proc_name(name, ret, residual);
196 read_unlock(&proc_subdir_lock);
200 static DEFINE_IDA(proc_inum_ida);
202 #define PROC_DYNAMIC_FIRST 0xF0000000U
205 * Return an inode number between PROC_DYNAMIC_FIRST and
206 * 0xffffffff, or zero on failure.
208 int proc_alloc_inum(unsigned int *inum)
212 i = ida_simple_get(&proc_inum_ida, 0, UINT_MAX - PROC_DYNAMIC_FIRST + 1,
217 *inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
221 void proc_free_inum(unsigned int inum)
223 ida_simple_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
226 static int proc_misc_d_revalidate(struct dentry *dentry, unsigned int flags)
228 if (flags & LOOKUP_RCU)
231 if (atomic_read(&PDE(d_inode(dentry))->in_use) < 0)
232 return 0; /* revalidate */
236 static int proc_misc_d_delete(const struct dentry *dentry)
238 return atomic_read(&PDE(d_inode(dentry))->in_use) < 0;
241 static const struct dentry_operations proc_misc_dentry_ops = {
242 .d_revalidate = proc_misc_d_revalidate,
243 .d_delete = proc_misc_d_delete,
247 * Don't create negative dentries here, return -ENOENT by hand
250 struct dentry *proc_lookup_de(struct inode *dir, struct dentry *dentry,
251 struct proc_dir_entry *de)
255 read_lock(&proc_subdir_lock);
256 de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
259 read_unlock(&proc_subdir_lock);
260 inode = proc_get_inode(dir->i_sb, de);
262 return ERR_PTR(-ENOMEM);
263 d_set_d_op(dentry, de->proc_dops);
264 return d_splice_alias(inode, dentry);
266 read_unlock(&proc_subdir_lock);
267 return ERR_PTR(-ENOENT);
270 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
273 return proc_lookup_de(dir, dentry, PDE(dir));
277 * This returns non-zero if at EOF, so that the /proc
278 * root directory can use this and check if it should
279 * continue with the <pid> entries..
281 * Note that the VFS-layer doesn't care about the return
282 * value of the readdir() call, as long as it's non-negative
285 int proc_readdir_de(struct file *file, struct dir_context *ctx,
286 struct proc_dir_entry *de)
290 if (!dir_emit_dots(file, ctx))
294 read_lock(&proc_subdir_lock);
295 de = pde_subdir_first(de);
298 read_unlock(&proc_subdir_lock);
303 de = pde_subdir_next(de);
308 struct proc_dir_entry *next;
310 read_unlock(&proc_subdir_lock);
311 if (!dir_emit(ctx, de->name, de->namelen,
312 de->low_ino, de->mode >> 12)) {
317 read_lock(&proc_subdir_lock);
318 next = pde_subdir_next(de);
322 read_unlock(&proc_subdir_lock);
326 int proc_readdir(struct file *file, struct dir_context *ctx)
328 struct inode *inode = file_inode(file);
330 return proc_readdir_de(file, ctx, PDE(inode));
334 * These are the generic /proc directory operations. They
335 * use the in-memory "struct proc_dir_entry" tree to parse
336 * the /proc directory.
338 static const struct file_operations proc_dir_operations = {
339 .llseek = generic_file_llseek,
340 .read = generic_read_dir,
341 .iterate_shared = proc_readdir,
344 static int proc_net_d_revalidate(struct dentry *dentry, unsigned int flags)
349 const struct dentry_operations proc_net_dentry_ops = {
350 .d_revalidate = proc_net_d_revalidate,
351 .d_delete = always_delete_dentry,
355 * proc directories can do almost nothing..
357 static const struct inode_operations proc_dir_inode_operations = {
358 .lookup = proc_lookup,
359 .getattr = proc_getattr,
360 .setattr = proc_notify_change,
363 /* returns the registered entry, or frees dp and returns NULL on failure */
364 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
365 struct proc_dir_entry *dp)
367 if (proc_alloc_inum(&dp->low_ino))
370 write_lock(&proc_subdir_lock);
372 if (pde_subdir_insert(dir, dp) == false) {
373 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
374 dir->name, dp->name);
375 write_unlock(&proc_subdir_lock);
379 write_unlock(&proc_subdir_lock);
383 proc_free_inum(dp->low_ino);
389 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
394 struct proc_dir_entry *ent = NULL;
398 if (xlate_proc_name(name, parent, &fn) != 0)
401 qstr.len = strlen(fn);
402 if (qstr.len == 0 || qstr.len >= 256) {
403 WARN(1, "name len %u\n", qstr.len);
406 if (qstr.len == 1 && fn[0] == '.') {
407 WARN(1, "name '.'\n");
410 if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
411 WARN(1, "name '..'\n");
414 if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
415 WARN(1, "create '/proc/%s' by hand\n", qstr.name);
418 if (is_empty_pde(*parent)) {
419 WARN(1, "attempt to add to permanently empty directory");
423 ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
427 if (qstr.len + 1 <= SIZEOF_PDE_INLINE_NAME) {
428 ent->name = ent->inline_name;
430 ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
437 memcpy(ent->name, fn, qstr.len + 1);
438 ent->namelen = qstr.len;
441 ent->subdir = RB_ROOT;
442 refcount_set(&ent->refcnt, 1);
443 spin_lock_init(&ent->pde_unload_lock);
444 INIT_LIST_HEAD(&ent->pde_openers);
445 proc_set_user(ent, (*parent)->uid, (*parent)->gid);
447 ent->proc_dops = &proc_misc_dentry_ops;
453 struct proc_dir_entry *proc_symlink(const char *name,
454 struct proc_dir_entry *parent, const char *dest)
456 struct proc_dir_entry *ent;
458 ent = __proc_create(&parent, name,
459 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
462 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
464 strcpy((char*)ent->data,dest);
465 ent->proc_iops = &proc_link_inode_operations;
466 ent = proc_register(parent, ent);
474 EXPORT_SYMBOL(proc_symlink);
476 struct proc_dir_entry *_proc_mkdir(const char *name, umode_t mode,
477 struct proc_dir_entry *parent, void *data, bool force_lookup)
479 struct proc_dir_entry *ent;
482 mode = S_IRUGO | S_IXUGO;
484 ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
487 ent->proc_fops = &proc_dir_operations;
488 ent->proc_iops = &proc_dir_inode_operations;
490 pde_force_lookup(ent);
492 ent = proc_register(parent, ent);
496 EXPORT_SYMBOL_GPL(_proc_mkdir);
498 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
499 struct proc_dir_entry *parent, void *data)
501 return _proc_mkdir(name, mode, parent, data, false);
503 EXPORT_SYMBOL_GPL(proc_mkdir_data);
505 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
506 struct proc_dir_entry *parent)
508 return proc_mkdir_data(name, mode, parent, NULL);
510 EXPORT_SYMBOL(proc_mkdir_mode);
512 struct proc_dir_entry *proc_mkdir(const char *name,
513 struct proc_dir_entry *parent)
515 return proc_mkdir_data(name, 0, parent, NULL);
517 EXPORT_SYMBOL(proc_mkdir);
519 struct proc_dir_entry *proc_create_mount_point(const char *name)
521 umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
522 struct proc_dir_entry *ent, *parent = NULL;
524 ent = __proc_create(&parent, name, mode, 2);
527 ent->proc_fops = NULL;
528 ent->proc_iops = NULL;
529 ent = proc_register(parent, ent);
533 EXPORT_SYMBOL(proc_create_mount_point);
535 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
536 struct proc_dir_entry **parent, void *data)
538 struct proc_dir_entry *p;
540 if ((mode & S_IFMT) == 0)
542 if ((mode & S_IALLUGO) == 0)
544 if (WARN_ON_ONCE(!S_ISREG(mode)))
547 p = __proc_create(parent, name, mode, 1);
549 p->proc_iops = &proc_file_inode_operations;
555 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
556 struct proc_dir_entry *parent,
557 const struct file_operations *proc_fops, void *data)
559 struct proc_dir_entry *p;
561 BUG_ON(proc_fops == NULL);
563 p = proc_create_reg(name, mode, &parent, data);
566 p->proc_fops = proc_fops;
567 return proc_register(parent, p);
569 EXPORT_SYMBOL(proc_create_data);
571 struct proc_dir_entry *proc_create(const char *name, umode_t mode,
572 struct proc_dir_entry *parent,
573 const struct file_operations *proc_fops)
575 return proc_create_data(name, mode, parent, proc_fops, NULL);
577 EXPORT_SYMBOL(proc_create);
579 static int proc_seq_open(struct inode *inode, struct file *file)
581 struct proc_dir_entry *de = PDE(inode);
584 return seq_open_private(file, de->seq_ops, de->state_size);
585 return seq_open(file, de->seq_ops);
588 static int proc_seq_release(struct inode *inode, struct file *file)
590 struct proc_dir_entry *de = PDE(inode);
593 return seq_release_private(inode, file);
594 return seq_release(inode, file);
597 static const struct file_operations proc_seq_fops = {
598 .open = proc_seq_open,
601 .release = proc_seq_release,
604 struct proc_dir_entry *proc_create_seq_private(const char *name, umode_t mode,
605 struct proc_dir_entry *parent, const struct seq_operations *ops,
606 unsigned int state_size, void *data)
608 struct proc_dir_entry *p;
610 p = proc_create_reg(name, mode, &parent, data);
613 p->proc_fops = &proc_seq_fops;
615 p->state_size = state_size;
616 return proc_register(parent, p);
618 EXPORT_SYMBOL(proc_create_seq_private);
620 static int proc_single_open(struct inode *inode, struct file *file)
622 struct proc_dir_entry *de = PDE(inode);
624 return single_open(file, de->single_show, de->data);
627 static const struct file_operations proc_single_fops = {
628 .open = proc_single_open,
631 .release = single_release,
634 struct proc_dir_entry *proc_create_single_data(const char *name, umode_t mode,
635 struct proc_dir_entry *parent,
636 int (*show)(struct seq_file *, void *), void *data)
638 struct proc_dir_entry *p;
640 p = proc_create_reg(name, mode, &parent, data);
643 p->proc_fops = &proc_single_fops;
644 p->single_show = show;
645 return proc_register(parent, p);
647 EXPORT_SYMBOL(proc_create_single_data);
649 void proc_set_size(struct proc_dir_entry *de, loff_t size)
653 EXPORT_SYMBOL(proc_set_size);
655 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
660 EXPORT_SYMBOL(proc_set_user);
662 void pde_put(struct proc_dir_entry *pde)
664 if (refcount_dec_and_test(&pde->refcnt)) {
665 proc_free_inum(pde->low_ino);
671 * Remove a /proc entry and free it if it's not currently in use.
673 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
675 struct proc_dir_entry *de = NULL;
676 const char *fn = name;
679 write_lock(&proc_subdir_lock);
680 if (__xlate_proc_name(name, &parent, &fn) != 0) {
681 write_unlock(&proc_subdir_lock);
686 de = pde_subdir_find(parent, fn, len);
688 rb_erase(&de->subdir_node, &parent->subdir);
689 if (S_ISDIR(de->mode)) {
693 write_unlock(&proc_subdir_lock);
695 WARN(1, "name '%s'\n", name);
699 proc_entry_rundown(de);
701 WARN(pde_subdir_first(de),
702 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
703 __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
706 EXPORT_SYMBOL(remove_proc_entry);
708 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
710 struct proc_dir_entry *root = NULL, *de, *next;
711 const char *fn = name;
714 write_lock(&proc_subdir_lock);
715 if (__xlate_proc_name(name, &parent, &fn) != 0) {
716 write_unlock(&proc_subdir_lock);
721 root = pde_subdir_find(parent, fn, len);
723 write_unlock(&proc_subdir_lock);
726 rb_erase(&root->subdir_node, &parent->subdir);
730 next = pde_subdir_first(de);
732 rb_erase(&next->subdir_node, &de->subdir);
737 if (S_ISDIR(de->mode))
739 write_unlock(&proc_subdir_lock);
741 proc_entry_rundown(de);
746 write_lock(&proc_subdir_lock);
752 EXPORT_SYMBOL(remove_proc_subtree);
754 void *proc_get_parent_data(const struct inode *inode)
756 struct proc_dir_entry *de = PDE(inode);
757 return de->parent->data;
759 EXPORT_SYMBOL_GPL(proc_get_parent_data);
761 void proc_remove(struct proc_dir_entry *de)
764 remove_proc_subtree(de->name, de->parent);
766 EXPORT_SYMBOL(proc_remove);
768 void *PDE_DATA(const struct inode *inode)
770 return __PDE_DATA(inode);
772 EXPORT_SYMBOL(PDE_DATA);
775 * Pull a user buffer into memory and pass it to the file's write handler if
776 * one is supplied. The ->write() method is permitted to modify the
777 * kernel-side buffer.
779 ssize_t proc_simple_write(struct file *f, const char __user *ubuf, size_t size,
782 struct proc_dir_entry *pde = PDE(file_inode(f));
788 if (size == 0 || size > PAGE_SIZE - 1)
790 buf = memdup_user_nul(ubuf, size);
793 ret = pde->write(f, buf, size);
795 return ret == 0 ? size : ret;