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/slab.h>
19 #include <linux/printk.h>
20 #include <linux/mount.h>
21 #include <linux/init.h>
22 #include <linux/idr.h>
23 #include <linux/bitops.h>
24 #include <linux/spinlock.h>
25 #include <linux/completion.h>
26 #include <linux/uaccess.h>
30 static DEFINE_RWLOCK(proc_subdir_lock);
32 struct kmem_cache *proc_dir_entry_cache __ro_after_init;
34 void pde_free(struct proc_dir_entry *pde)
36 if (S_ISLNK(pde->mode))
38 if (pde->name != pde->inline_name)
40 kmem_cache_free(proc_dir_entry_cache, pde);
43 static int proc_match(const char *name, struct proc_dir_entry *de, unsigned int len)
45 if (len < de->namelen)
47 if (len > de->namelen)
50 return memcmp(name, de->name, len);
53 static struct proc_dir_entry *pde_subdir_first(struct proc_dir_entry *dir)
55 return rb_entry_safe(rb_first(&dir->subdir), struct proc_dir_entry,
59 static struct proc_dir_entry *pde_subdir_next(struct proc_dir_entry *dir)
61 return rb_entry_safe(rb_next(&dir->subdir_node), struct proc_dir_entry,
65 static struct proc_dir_entry *pde_subdir_find(struct proc_dir_entry *dir,
69 struct rb_node *node = dir->subdir.rb_node;
72 struct proc_dir_entry *de = rb_entry(node,
73 struct proc_dir_entry,
75 int result = proc_match(name, de, len);
80 node = node->rb_right;
87 static bool pde_subdir_insert(struct proc_dir_entry *dir,
88 struct proc_dir_entry *de)
90 struct rb_root *root = &dir->subdir;
91 struct rb_node **new = &root->rb_node, *parent = NULL;
93 /* Figure out where to put new node */
95 struct proc_dir_entry *this = rb_entry(*new,
96 struct proc_dir_entry,
98 int result = proc_match(de->name, this, de->namelen);
102 new = &(*new)->rb_left;
104 new = &(*new)->rb_right;
109 /* Add new node and rebalance tree. */
110 rb_link_node(&de->subdir_node, parent, new);
111 rb_insert_color(&de->subdir_node, root);
115 static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
117 struct inode *inode = d_inode(dentry);
118 struct proc_dir_entry *de = PDE(inode);
121 error = setattr_prepare(dentry, iattr);
125 setattr_copy(inode, iattr);
126 mark_inode_dirty(inode);
128 proc_set_user(de, inode->i_uid, inode->i_gid);
129 de->mode = inode->i_mode;
133 static int proc_getattr(const struct path *path, struct kstat *stat,
134 u32 request_mask, unsigned int query_flags)
136 struct inode *inode = d_inode(path->dentry);
137 struct proc_dir_entry *de = PDE(inode);
139 set_nlink(inode, de->nlink);
141 generic_fillattr(inode, stat);
145 static const struct inode_operations proc_file_inode_operations = {
146 .setattr = proc_notify_change,
150 * This function parses a name such as "tty/driver/serial", and
151 * returns the struct proc_dir_entry for "/proc/tty/driver", and
152 * returns "serial" in residual.
154 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
155 const char **residual)
157 const char *cp = name, *next;
158 struct proc_dir_entry *de;
166 next = strchr(cp, '/');
171 de = pde_subdir_find(de, cp, len);
173 WARN(1, "name '%s'\n", name);
183 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
184 const char **residual)
188 read_lock(&proc_subdir_lock);
189 rv = __xlate_proc_name(name, ret, residual);
190 read_unlock(&proc_subdir_lock);
194 static DEFINE_IDA(proc_inum_ida);
196 #define PROC_DYNAMIC_FIRST 0xF0000000U
199 * Return an inode number between PROC_DYNAMIC_FIRST and
200 * 0xffffffff, or zero on failure.
202 int proc_alloc_inum(unsigned int *inum)
206 i = ida_simple_get(&proc_inum_ida, 0, UINT_MAX - PROC_DYNAMIC_FIRST + 1,
211 *inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
215 void proc_free_inum(unsigned int inum)
217 ida_simple_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
221 * Don't create negative dentries here, return -ENOENT by hand
224 struct dentry *proc_lookup_de(struct inode *dir, struct dentry *dentry,
225 struct proc_dir_entry *de)
229 read_lock(&proc_subdir_lock);
230 de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
233 read_unlock(&proc_subdir_lock);
234 inode = proc_get_inode(dir->i_sb, de);
236 return ERR_PTR(-ENOMEM);
237 d_set_d_op(dentry, &simple_dentry_operations);
238 d_add(dentry, inode);
241 read_unlock(&proc_subdir_lock);
242 return ERR_PTR(-ENOENT);
245 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
248 return proc_lookup_de(dir, dentry, PDE(dir));
252 * This returns non-zero if at EOF, so that the /proc
253 * root directory can use this and check if it should
254 * continue with the <pid> entries..
256 * Note that the VFS-layer doesn't care about the return
257 * value of the readdir() call, as long as it's non-negative
260 int proc_readdir_de(struct file *file, struct dir_context *ctx,
261 struct proc_dir_entry *de)
265 if (!dir_emit_dots(file, ctx))
268 read_lock(&proc_subdir_lock);
269 de = pde_subdir_first(de);
273 read_unlock(&proc_subdir_lock);
278 de = pde_subdir_next(de);
283 struct proc_dir_entry *next;
285 read_unlock(&proc_subdir_lock);
286 if (!dir_emit(ctx, de->name, de->namelen,
287 de->low_ino, de->mode >> 12)) {
291 read_lock(&proc_subdir_lock);
293 next = pde_subdir_next(de);
297 read_unlock(&proc_subdir_lock);
301 int proc_readdir(struct file *file, struct dir_context *ctx)
303 struct inode *inode = file_inode(file);
305 return proc_readdir_de(file, ctx, PDE(inode));
309 * These are the generic /proc directory operations. They
310 * use the in-memory "struct proc_dir_entry" tree to parse
311 * the /proc directory.
313 static const struct file_operations proc_dir_operations = {
314 .llseek = generic_file_llseek,
315 .read = generic_read_dir,
316 .iterate_shared = proc_readdir,
320 * proc directories can do almost nothing..
322 static const struct inode_operations proc_dir_inode_operations = {
323 .lookup = proc_lookup,
324 .getattr = proc_getattr,
325 .setattr = proc_notify_change,
328 static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp)
332 ret = proc_alloc_inum(&dp->low_ino);
336 write_lock(&proc_subdir_lock);
338 if (pde_subdir_insert(dir, dp) == false) {
339 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
340 dir->name, dp->name);
341 write_unlock(&proc_subdir_lock);
342 proc_free_inum(dp->low_ino);
345 write_unlock(&proc_subdir_lock);
350 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
355 struct proc_dir_entry *ent = NULL;
359 if (xlate_proc_name(name, parent, &fn) != 0)
362 qstr.len = strlen(fn);
363 if (qstr.len == 0 || qstr.len >= 256) {
364 WARN(1, "name len %u\n", qstr.len);
367 if (qstr.len == 1 && fn[0] == '.') {
368 WARN(1, "name '.'\n");
371 if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
372 WARN(1, "name '..'\n");
375 if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
376 WARN(1, "create '/proc/%s' by hand\n", qstr.name);
379 if (is_empty_pde(*parent)) {
380 WARN(1, "attempt to add to permanently empty directory");
384 ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
388 if (qstr.len + 1 <= sizeof(ent->inline_name)) {
389 ent->name = ent->inline_name;
391 ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
398 memcpy(ent->name, fn, qstr.len + 1);
399 ent->namelen = qstr.len;
402 ent->subdir = RB_ROOT;
403 refcount_set(&ent->refcnt, 1);
404 spin_lock_init(&ent->pde_unload_lock);
405 INIT_LIST_HEAD(&ent->pde_openers);
406 proc_set_user(ent, (*parent)->uid, (*parent)->gid);
412 struct proc_dir_entry *proc_symlink(const char *name,
413 struct proc_dir_entry *parent, const char *dest)
415 struct proc_dir_entry *ent;
417 ent = __proc_create(&parent, name,
418 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
421 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
423 strcpy((char*)ent->data,dest);
424 ent->proc_iops = &proc_link_inode_operations;
425 if (proc_register(parent, ent) < 0) {
436 EXPORT_SYMBOL(proc_symlink);
438 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
439 struct proc_dir_entry *parent, void *data)
441 struct proc_dir_entry *ent;
444 mode = S_IRUGO | S_IXUGO;
446 ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
449 ent->proc_fops = &proc_dir_operations;
450 ent->proc_iops = &proc_dir_inode_operations;
452 if (proc_register(parent, ent) < 0) {
460 EXPORT_SYMBOL_GPL(proc_mkdir_data);
462 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
463 struct proc_dir_entry *parent)
465 return proc_mkdir_data(name, mode, parent, NULL);
467 EXPORT_SYMBOL(proc_mkdir_mode);
469 struct proc_dir_entry *proc_mkdir(const char *name,
470 struct proc_dir_entry *parent)
472 return proc_mkdir_data(name, 0, parent, NULL);
474 EXPORT_SYMBOL(proc_mkdir);
476 struct proc_dir_entry *proc_create_mount_point(const char *name)
478 umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
479 struct proc_dir_entry *ent, *parent = NULL;
481 ent = __proc_create(&parent, name, mode, 2);
484 ent->proc_fops = NULL;
485 ent->proc_iops = NULL;
487 if (proc_register(parent, ent) < 0) {
495 EXPORT_SYMBOL(proc_create_mount_point);
497 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
498 struct proc_dir_entry *parent,
499 const struct file_operations *proc_fops,
502 struct proc_dir_entry *pde;
503 if ((mode & S_IFMT) == 0)
506 if (!S_ISREG(mode)) {
507 WARN_ON(1); /* use proc_mkdir() */
511 BUG_ON(proc_fops == NULL);
513 if ((mode & S_IALLUGO) == 0)
515 pde = __proc_create(&parent, name, mode, 1);
518 pde->proc_fops = proc_fops;
520 pde->proc_iops = &proc_file_inode_operations;
521 if (proc_register(parent, pde) < 0)
529 EXPORT_SYMBOL(proc_create_data);
531 struct proc_dir_entry *proc_create(const char *name, umode_t mode,
532 struct proc_dir_entry *parent,
533 const struct file_operations *proc_fops)
535 return proc_create_data(name, mode, parent, proc_fops, NULL);
537 EXPORT_SYMBOL(proc_create);
539 void proc_set_size(struct proc_dir_entry *de, loff_t size)
543 EXPORT_SYMBOL(proc_set_size);
545 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
550 EXPORT_SYMBOL(proc_set_user);
552 void pde_put(struct proc_dir_entry *pde)
554 if (refcount_dec_and_test(&pde->refcnt)) {
555 proc_free_inum(pde->low_ino);
561 * Remove a /proc entry and free it if it's not currently in use.
563 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
565 struct proc_dir_entry *de = NULL;
566 const char *fn = name;
569 write_lock(&proc_subdir_lock);
570 if (__xlate_proc_name(name, &parent, &fn) != 0) {
571 write_unlock(&proc_subdir_lock);
576 de = pde_subdir_find(parent, fn, len);
578 rb_erase(&de->subdir_node, &parent->subdir);
579 write_unlock(&proc_subdir_lock);
581 WARN(1, "name '%s'\n", name);
585 proc_entry_rundown(de);
587 if (S_ISDIR(de->mode))
590 WARN(pde_subdir_first(de),
591 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
592 __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
595 EXPORT_SYMBOL(remove_proc_entry);
597 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
599 struct proc_dir_entry *root = NULL, *de, *next;
600 const char *fn = name;
603 write_lock(&proc_subdir_lock);
604 if (__xlate_proc_name(name, &parent, &fn) != 0) {
605 write_unlock(&proc_subdir_lock);
610 root = pde_subdir_find(parent, fn, len);
612 write_unlock(&proc_subdir_lock);
615 rb_erase(&root->subdir_node, &parent->subdir);
619 next = pde_subdir_first(de);
621 rb_erase(&next->subdir_node, &de->subdir);
625 write_unlock(&proc_subdir_lock);
627 proc_entry_rundown(de);
629 if (S_ISDIR(de->mode))
636 write_lock(&proc_subdir_lock);
642 EXPORT_SYMBOL(remove_proc_subtree);
644 void *proc_get_parent_data(const struct inode *inode)
646 struct proc_dir_entry *de = PDE(inode);
647 return de->parent->data;
649 EXPORT_SYMBOL_GPL(proc_get_parent_data);
651 void proc_remove(struct proc_dir_entry *de)
654 remove_proc_subtree(de->name, de->parent);
656 EXPORT_SYMBOL(proc_remove);
658 void *PDE_DATA(const struct inode *inode)
660 return __PDE_DATA(inode);
662 EXPORT_SYMBOL(PDE_DATA);