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 set_nlink(inode, de->nlink);
143 generic_fillattr(inode, stat);
147 static const struct inode_operations proc_file_inode_operations = {
148 .setattr = proc_notify_change,
152 * This function parses a name such as "tty/driver/serial", and
153 * returns the struct proc_dir_entry for "/proc/tty/driver", and
154 * returns "serial" in residual.
156 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
157 const char **residual)
159 const char *cp = name, *next;
160 struct proc_dir_entry *de;
168 next = strchr(cp, '/');
173 de = pde_subdir_find(de, cp, len);
175 WARN(1, "name '%s'\n", name);
185 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
186 const char **residual)
190 read_lock(&proc_subdir_lock);
191 rv = __xlate_proc_name(name, ret, residual);
192 read_unlock(&proc_subdir_lock);
196 static DEFINE_IDA(proc_inum_ida);
198 #define PROC_DYNAMIC_FIRST 0xF0000000U
201 * Return an inode number between PROC_DYNAMIC_FIRST and
202 * 0xffffffff, or zero on failure.
204 int proc_alloc_inum(unsigned int *inum)
208 i = ida_simple_get(&proc_inum_ida, 0, UINT_MAX - PROC_DYNAMIC_FIRST + 1,
213 *inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
217 void proc_free_inum(unsigned int inum)
219 ida_simple_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
222 static int proc_misc_d_revalidate(struct dentry *dentry, unsigned int flags)
224 if (flags & LOOKUP_RCU)
227 if (atomic_read(&PDE(d_inode(dentry))->in_use) < 0)
228 return 0; /* revalidate */
232 static int proc_misc_d_delete(const struct dentry *dentry)
234 return atomic_read(&PDE(d_inode(dentry))->in_use) < 0;
237 static const struct dentry_operations proc_misc_dentry_ops = {
238 .d_revalidate = proc_misc_d_revalidate,
239 .d_delete = proc_misc_d_delete,
243 * Don't create negative dentries here, return -ENOENT by hand
246 struct dentry *proc_lookup_de(struct inode *dir, struct dentry *dentry,
247 struct proc_dir_entry *de)
251 read_lock(&proc_subdir_lock);
252 de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
255 read_unlock(&proc_subdir_lock);
256 inode = proc_get_inode(dir->i_sb, de);
258 return ERR_PTR(-ENOMEM);
259 d_set_d_op(dentry, de->proc_dops);
260 return d_splice_alias(inode, dentry);
262 read_unlock(&proc_subdir_lock);
263 return ERR_PTR(-ENOENT);
266 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
269 return proc_lookup_de(dir, dentry, PDE(dir));
273 * This returns non-zero if at EOF, so that the /proc
274 * root directory can use this and check if it should
275 * continue with the <pid> entries..
277 * Note that the VFS-layer doesn't care about the return
278 * value of the readdir() call, as long as it's non-negative
281 int proc_readdir_de(struct file *file, struct dir_context *ctx,
282 struct proc_dir_entry *de)
286 if (!dir_emit_dots(file, ctx))
290 read_lock(&proc_subdir_lock);
291 de = pde_subdir_first(de);
294 read_unlock(&proc_subdir_lock);
299 de = pde_subdir_next(de);
304 struct proc_dir_entry *next;
306 read_unlock(&proc_subdir_lock);
307 if (!dir_emit(ctx, de->name, de->namelen,
308 de->low_ino, de->mode >> 12)) {
313 read_lock(&proc_subdir_lock);
314 next = pde_subdir_next(de);
318 read_unlock(&proc_subdir_lock);
322 int proc_readdir(struct file *file, struct dir_context *ctx)
324 struct inode *inode = file_inode(file);
326 return proc_readdir_de(file, ctx, PDE(inode));
330 * These are the generic /proc directory operations. They
331 * use the in-memory "struct proc_dir_entry" tree to parse
332 * the /proc directory.
334 static const struct file_operations proc_dir_operations = {
335 .llseek = generic_file_llseek,
336 .read = generic_read_dir,
337 .iterate_shared = proc_readdir,
341 * proc directories can do almost nothing..
343 static const struct inode_operations proc_dir_inode_operations = {
344 .lookup = proc_lookup,
345 .getattr = proc_getattr,
346 .setattr = proc_notify_change,
349 /* returns the registered entry, or frees dp and returns NULL on failure */
350 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
351 struct proc_dir_entry *dp)
353 if (proc_alloc_inum(&dp->low_ino))
356 write_lock(&proc_subdir_lock);
358 if (pde_subdir_insert(dir, dp) == false) {
359 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
360 dir->name, dp->name);
361 write_unlock(&proc_subdir_lock);
364 write_unlock(&proc_subdir_lock);
368 proc_free_inum(dp->low_ino);
374 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
379 struct proc_dir_entry *ent = NULL;
383 if (xlate_proc_name(name, parent, &fn) != 0)
386 qstr.len = strlen(fn);
387 if (qstr.len == 0 || qstr.len >= 256) {
388 WARN(1, "name len %u\n", qstr.len);
391 if (qstr.len == 1 && fn[0] == '.') {
392 WARN(1, "name '.'\n");
395 if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
396 WARN(1, "name '..'\n");
399 if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
400 WARN(1, "create '/proc/%s' by hand\n", qstr.name);
403 if (is_empty_pde(*parent)) {
404 WARN(1, "attempt to add to permanently empty directory");
408 ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
412 if (qstr.len + 1 <= SIZEOF_PDE_INLINE_NAME) {
413 ent->name = ent->inline_name;
415 ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
422 memcpy(ent->name, fn, qstr.len + 1);
423 ent->namelen = qstr.len;
426 ent->subdir = RB_ROOT;
427 refcount_set(&ent->refcnt, 1);
428 spin_lock_init(&ent->pde_unload_lock);
429 INIT_LIST_HEAD(&ent->pde_openers);
430 proc_set_user(ent, (*parent)->uid, (*parent)->gid);
432 ent->proc_dops = &proc_misc_dentry_ops;
438 struct proc_dir_entry *proc_symlink(const char *name,
439 struct proc_dir_entry *parent, const char *dest)
441 struct proc_dir_entry *ent;
443 ent = __proc_create(&parent, name,
444 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
447 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
449 strcpy((char*)ent->data,dest);
450 ent->proc_iops = &proc_link_inode_operations;
451 ent = proc_register(parent, ent);
459 EXPORT_SYMBOL(proc_symlink);
461 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
462 struct proc_dir_entry *parent, void *data)
464 struct proc_dir_entry *ent;
467 mode = S_IRUGO | S_IXUGO;
469 ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
472 ent->proc_fops = &proc_dir_operations;
473 ent->proc_iops = &proc_dir_inode_operations;
475 ent = proc_register(parent, ent);
481 EXPORT_SYMBOL_GPL(proc_mkdir_data);
483 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
484 struct proc_dir_entry *parent)
486 return proc_mkdir_data(name, mode, parent, NULL);
488 EXPORT_SYMBOL(proc_mkdir_mode);
490 struct proc_dir_entry *proc_mkdir(const char *name,
491 struct proc_dir_entry *parent)
493 return proc_mkdir_data(name, 0, parent, NULL);
495 EXPORT_SYMBOL(proc_mkdir);
497 struct proc_dir_entry *proc_create_mount_point(const char *name)
499 umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
500 struct proc_dir_entry *ent, *parent = NULL;
502 ent = __proc_create(&parent, name, mode, 2);
505 ent->proc_fops = NULL;
506 ent->proc_iops = NULL;
508 ent = proc_register(parent, ent);
514 EXPORT_SYMBOL(proc_create_mount_point);
516 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
517 struct proc_dir_entry **parent, void *data)
519 struct proc_dir_entry *p;
521 if ((mode & S_IFMT) == 0)
523 if ((mode & S_IALLUGO) == 0)
525 if (WARN_ON_ONCE(!S_ISREG(mode)))
528 p = __proc_create(parent, name, mode, 1);
530 p->proc_iops = &proc_file_inode_operations;
536 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
537 struct proc_dir_entry *parent,
538 const struct file_operations *proc_fops, void *data)
540 struct proc_dir_entry *p;
542 BUG_ON(proc_fops == NULL);
544 p = proc_create_reg(name, mode, &parent, data);
547 p->proc_fops = proc_fops;
548 return proc_register(parent, p);
550 EXPORT_SYMBOL(proc_create_data);
552 struct proc_dir_entry *proc_create(const char *name, umode_t mode,
553 struct proc_dir_entry *parent,
554 const struct file_operations *proc_fops)
556 return proc_create_data(name, mode, parent, proc_fops, NULL);
558 EXPORT_SYMBOL(proc_create);
560 static int proc_seq_open(struct inode *inode, struct file *file)
562 struct proc_dir_entry *de = PDE(inode);
565 return seq_open_private(file, de->seq_ops, de->state_size);
566 return seq_open(file, de->seq_ops);
569 static int proc_seq_release(struct inode *inode, struct file *file)
571 struct proc_dir_entry *de = PDE(inode);
574 return seq_release_private(inode, file);
575 return seq_release(inode, file);
578 static const struct file_operations proc_seq_fops = {
579 .open = proc_seq_open,
582 .release = proc_seq_release,
585 struct proc_dir_entry *proc_create_seq_private(const char *name, umode_t mode,
586 struct proc_dir_entry *parent, const struct seq_operations *ops,
587 unsigned int state_size, void *data)
589 struct proc_dir_entry *p;
591 p = proc_create_reg(name, mode, &parent, data);
594 p->proc_fops = &proc_seq_fops;
596 p->state_size = state_size;
597 return proc_register(parent, p);
599 EXPORT_SYMBOL(proc_create_seq_private);
601 static int proc_single_open(struct inode *inode, struct file *file)
603 struct proc_dir_entry *de = PDE(inode);
605 return single_open(file, de->single_show, de->data);
608 static const struct file_operations proc_single_fops = {
609 .open = proc_single_open,
612 .release = single_release,
615 struct proc_dir_entry *proc_create_single_data(const char *name, umode_t mode,
616 struct proc_dir_entry *parent,
617 int (*show)(struct seq_file *, void *), void *data)
619 struct proc_dir_entry *p;
621 p = proc_create_reg(name, mode, &parent, data);
624 p->proc_fops = &proc_single_fops;
625 p->single_show = show;
626 return proc_register(parent, p);
628 EXPORT_SYMBOL(proc_create_single_data);
630 void proc_set_size(struct proc_dir_entry *de, loff_t size)
634 EXPORT_SYMBOL(proc_set_size);
636 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
641 EXPORT_SYMBOL(proc_set_user);
643 void pde_put(struct proc_dir_entry *pde)
645 if (refcount_dec_and_test(&pde->refcnt)) {
646 proc_free_inum(pde->low_ino);
652 * Remove a /proc entry and free it if it's not currently in use.
654 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
656 struct proc_dir_entry *de = NULL;
657 const char *fn = name;
660 write_lock(&proc_subdir_lock);
661 if (__xlate_proc_name(name, &parent, &fn) != 0) {
662 write_unlock(&proc_subdir_lock);
667 de = pde_subdir_find(parent, fn, len);
669 rb_erase(&de->subdir_node, &parent->subdir);
670 write_unlock(&proc_subdir_lock);
672 WARN(1, "name '%s'\n", name);
676 proc_entry_rundown(de);
678 if (S_ISDIR(de->mode))
681 WARN(pde_subdir_first(de),
682 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
683 __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
686 EXPORT_SYMBOL(remove_proc_entry);
688 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
690 struct proc_dir_entry *root = NULL, *de, *next;
691 const char *fn = name;
694 write_lock(&proc_subdir_lock);
695 if (__xlate_proc_name(name, &parent, &fn) != 0) {
696 write_unlock(&proc_subdir_lock);
701 root = pde_subdir_find(parent, fn, len);
703 write_unlock(&proc_subdir_lock);
706 rb_erase(&root->subdir_node, &parent->subdir);
710 next = pde_subdir_first(de);
712 rb_erase(&next->subdir_node, &de->subdir);
716 write_unlock(&proc_subdir_lock);
718 proc_entry_rundown(de);
720 if (S_ISDIR(de->mode))
727 write_lock(&proc_subdir_lock);
733 EXPORT_SYMBOL(remove_proc_subtree);
735 void *proc_get_parent_data(const struct inode *inode)
737 struct proc_dir_entry *de = PDE(inode);
738 return de->parent->data;
740 EXPORT_SYMBOL_GPL(proc_get_parent_data);
742 void proc_remove(struct proc_dir_entry *de)
745 remove_proc_subtree(de->name, de->parent);
747 EXPORT_SYMBOL(proc_remove);
749 void *PDE_DATA(const struct inode *inode)
751 return __PDE_DATA(inode);
753 EXPORT_SYMBOL(PDE_DATA);
756 * Pull a user buffer into memory and pass it to the file's write handler if
757 * one is supplied. The ->write() method is permitted to modify the
758 * kernel-side buffer.
760 ssize_t proc_simple_write(struct file *f, const char __user *ubuf, size_t size,
763 struct proc_dir_entry *pde = PDE(file_inode(f));
769 if (size == 0 || size > PAGE_SIZE - 1)
771 buf = memdup_user_nul(ubuf, size);
774 ret = pde->write(f, buf, size);
776 return ret == 0 ? size : ret;