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/errno.h>
12 #include <linux/time.h>
13 #include <linux/proc_fs.h>
14 #include <linux/stat.h>
16 #include <linux/module.h>
17 #include <linux/slab.h>
18 #include <linux/printk.h>
19 #include <linux/mount.h>
20 #include <linux/init.h>
21 #include <linux/idr.h>
22 #include <linux/namei.h>
23 #include <linux/bitops.h>
24 #include <linux/spinlock.h>
25 #include <linux/completion.h>
26 #include <asm/uaccess.h>
30 static DEFINE_SPINLOCK(proc_subdir_lock);
32 static int proc_match(unsigned int len, const char *name, struct proc_dir_entry *de)
34 if (len < de->namelen)
36 if (len > de->namelen)
39 return memcmp(name, de->name, len);
42 static struct proc_dir_entry *pde_subdir_first(struct proc_dir_entry *dir)
44 return rb_entry_safe(rb_first(&dir->subdir), struct proc_dir_entry,
48 static struct proc_dir_entry *pde_subdir_next(struct proc_dir_entry *dir)
50 return rb_entry_safe(rb_next(&dir->subdir_node), struct proc_dir_entry,
54 static struct proc_dir_entry *pde_subdir_find(struct proc_dir_entry *dir,
58 struct rb_node *node = dir->subdir.rb_node;
61 struct proc_dir_entry *de = container_of(node,
62 struct proc_dir_entry,
64 int result = proc_match(len, name, de);
69 node = node->rb_right;
76 static bool pde_subdir_insert(struct proc_dir_entry *dir,
77 struct proc_dir_entry *de)
79 struct rb_root *root = &dir->subdir;
80 struct rb_node **new = &root->rb_node, *parent = NULL;
82 /* Figure out where to put new node */
84 struct proc_dir_entry *this =
85 container_of(*new, struct proc_dir_entry, subdir_node);
86 int result = proc_match(de->namelen, de->name, this);
90 new = &(*new)->rb_left;
92 new = &(*new)->rb_right;
97 /* Add new node and rebalance tree. */
98 rb_link_node(&de->subdir_node, parent, new);
99 rb_insert_color(&de->subdir_node, root);
103 static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
105 struct inode *inode = dentry->d_inode;
106 struct proc_dir_entry *de = PDE(inode);
109 error = inode_change_ok(inode, iattr);
113 setattr_copy(inode, iattr);
114 mark_inode_dirty(inode);
116 proc_set_user(de, inode->i_uid, inode->i_gid);
117 de->mode = inode->i_mode;
121 static int proc_getattr(struct vfsmount *mnt, struct dentry *dentry,
124 struct inode *inode = dentry->d_inode;
125 struct proc_dir_entry *de = PDE(inode);
127 set_nlink(inode, de->nlink);
129 generic_fillattr(inode, stat);
133 static const struct inode_operations proc_file_inode_operations = {
134 .setattr = proc_notify_change,
138 * This function parses a name such as "tty/driver/serial", and
139 * returns the struct proc_dir_entry for "/proc/tty/driver", and
140 * returns "serial" in residual.
142 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
143 const char **residual)
145 const char *cp = name, *next;
146 struct proc_dir_entry *de;
154 next = strchr(cp, '/');
159 de = pde_subdir_find(de, cp, len);
161 WARN(1, "name '%s'\n", name);
171 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
172 const char **residual)
176 spin_lock(&proc_subdir_lock);
177 rv = __xlate_proc_name(name, ret, residual);
178 spin_unlock(&proc_subdir_lock);
182 static DEFINE_IDA(proc_inum_ida);
183 static DEFINE_SPINLOCK(proc_inum_lock); /* protects the above */
185 #define PROC_DYNAMIC_FIRST 0xF0000000U
188 * Return an inode number between PROC_DYNAMIC_FIRST and
189 * 0xffffffff, or zero on failure.
191 int proc_alloc_inum(unsigned int *inum)
197 if (!ida_pre_get(&proc_inum_ida, GFP_KERNEL))
200 spin_lock_irq(&proc_inum_lock);
201 error = ida_get_new(&proc_inum_ida, &i);
202 spin_unlock_irq(&proc_inum_lock);
203 if (error == -EAGAIN)
208 if (i > UINT_MAX - PROC_DYNAMIC_FIRST) {
209 spin_lock_irq(&proc_inum_lock);
210 ida_remove(&proc_inum_ida, i);
211 spin_unlock_irq(&proc_inum_lock);
214 *inum = PROC_DYNAMIC_FIRST + i;
218 void proc_free_inum(unsigned int inum)
221 spin_lock_irqsave(&proc_inum_lock, flags);
222 ida_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
223 spin_unlock_irqrestore(&proc_inum_lock, flags);
226 static void *proc_follow_link(struct dentry *dentry, struct nameidata *nd)
228 nd_set_link(nd, __PDE_DATA(dentry->d_inode));
232 static const struct inode_operations proc_link_inode_operations = {
233 .readlink = generic_readlink,
234 .follow_link = proc_follow_link,
238 * Don't create negative dentries here, return -ENOENT by hand
241 struct dentry *proc_lookup_de(struct proc_dir_entry *de, struct inode *dir,
242 struct dentry *dentry)
246 spin_lock(&proc_subdir_lock);
247 de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
250 spin_unlock(&proc_subdir_lock);
251 inode = proc_get_inode(dir->i_sb, de);
253 return ERR_PTR(-ENOMEM);
254 d_set_d_op(dentry, &simple_dentry_operations);
255 d_add(dentry, inode);
258 spin_unlock(&proc_subdir_lock);
259 return ERR_PTR(-ENOENT);
262 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
265 return proc_lookup_de(PDE(dir), dir, dentry);
269 * This returns non-zero if at EOF, so that the /proc
270 * root directory can use this and check if it should
271 * continue with the <pid> entries..
273 * Note that the VFS-layer doesn't care about the return
274 * value of the readdir() call, as long as it's non-negative
277 int proc_readdir_de(struct proc_dir_entry *de, struct file *file,
278 struct dir_context *ctx)
282 if (!dir_emit_dots(file, ctx))
285 spin_lock(&proc_subdir_lock);
286 de = pde_subdir_first(de);
290 spin_unlock(&proc_subdir_lock);
295 de = pde_subdir_next(de);
300 struct proc_dir_entry *next;
302 spin_unlock(&proc_subdir_lock);
303 if (!dir_emit(ctx, de->name, de->namelen,
304 de->low_ino, de->mode >> 12)) {
308 spin_lock(&proc_subdir_lock);
310 next = pde_subdir_next(de);
314 spin_unlock(&proc_subdir_lock);
318 int proc_readdir(struct file *file, struct dir_context *ctx)
320 struct inode *inode = file_inode(file);
322 return proc_readdir_de(PDE(inode), file, ctx);
326 * These are the generic /proc directory operations. They
327 * use the in-memory "struct proc_dir_entry" tree to parse
328 * the /proc directory.
330 static const struct file_operations proc_dir_operations = {
331 .llseek = generic_file_llseek,
332 .read = generic_read_dir,
333 .iterate = proc_readdir,
337 * proc directories can do almost nothing..
339 static const struct inode_operations proc_dir_inode_operations = {
340 .lookup = proc_lookup,
341 .getattr = proc_getattr,
342 .setattr = proc_notify_change,
345 static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp)
349 ret = proc_alloc_inum(&dp->low_ino);
353 if (S_ISDIR(dp->mode)) {
354 dp->proc_fops = &proc_dir_operations;
355 dp->proc_iops = &proc_dir_inode_operations;
357 } else if (S_ISLNK(dp->mode)) {
358 dp->proc_iops = &proc_link_inode_operations;
359 } else if (S_ISREG(dp->mode)) {
360 BUG_ON(dp->proc_fops == NULL);
361 dp->proc_iops = &proc_file_inode_operations;
364 proc_free_inum(dp->low_ino);
368 spin_lock(&proc_subdir_lock);
370 if (pde_subdir_insert(dir, dp) == false) {
371 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
372 dir->name, dp->name);
373 spin_unlock(&proc_subdir_lock);
374 if (S_ISDIR(dp->mode))
376 proc_free_inum(dp->low_ino);
379 spin_unlock(&proc_subdir_lock);
384 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
389 struct proc_dir_entry *ent = NULL;
393 if (xlate_proc_name(name, parent, &fn) != 0)
396 qstr.len = strlen(fn);
397 if (qstr.len == 0 || qstr.len >= 256) {
398 WARN(1, "name len %u\n", qstr.len);
401 if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
402 WARN(1, "create '/proc/%s' by hand\n", qstr.name);
406 ent = kzalloc(sizeof(struct proc_dir_entry) + qstr.len + 1, GFP_KERNEL);
410 memcpy(ent->name, fn, qstr.len + 1);
411 ent->namelen = qstr.len;
414 ent->subdir = RB_ROOT;
415 atomic_set(&ent->count, 1);
416 spin_lock_init(&ent->pde_unload_lock);
417 INIT_LIST_HEAD(&ent->pde_openers);
422 struct proc_dir_entry *proc_symlink(const char *name,
423 struct proc_dir_entry *parent, const char *dest)
425 struct proc_dir_entry *ent;
427 ent = __proc_create(&parent, name,
428 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
431 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
433 strcpy((char*)ent->data,dest);
434 if (proc_register(parent, ent) < 0) {
446 EXPORT_SYMBOL(proc_symlink);
448 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
449 struct proc_dir_entry *parent, void *data)
451 struct proc_dir_entry *ent;
454 mode = S_IRUGO | S_IXUGO;
456 ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
459 if (proc_register(parent, ent) < 0) {
466 EXPORT_SYMBOL_GPL(proc_mkdir_data);
468 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
469 struct proc_dir_entry *parent)
471 return proc_mkdir_data(name, mode, parent, NULL);
473 EXPORT_SYMBOL(proc_mkdir_mode);
475 struct proc_dir_entry *proc_mkdir(const char *name,
476 struct proc_dir_entry *parent)
478 return proc_mkdir_data(name, 0, parent, NULL);
480 EXPORT_SYMBOL(proc_mkdir);
482 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
483 struct proc_dir_entry *parent,
484 const struct file_operations *proc_fops,
487 struct proc_dir_entry *pde;
488 if ((mode & S_IFMT) == 0)
491 if (!S_ISREG(mode)) {
492 WARN_ON(1); /* use proc_mkdir() */
496 if ((mode & S_IALLUGO) == 0)
498 pde = __proc_create(&parent, name, mode, 1);
501 pde->proc_fops = proc_fops;
503 if (proc_register(parent, pde) < 0)
511 EXPORT_SYMBOL(proc_create_data);
513 void proc_set_size(struct proc_dir_entry *de, loff_t size)
517 EXPORT_SYMBOL(proc_set_size);
519 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
524 EXPORT_SYMBOL(proc_set_user);
526 static void free_proc_entry(struct proc_dir_entry *de)
528 proc_free_inum(de->low_ino);
530 if (S_ISLNK(de->mode))
535 void pde_put(struct proc_dir_entry *pde)
537 if (atomic_dec_and_test(&pde->count))
538 free_proc_entry(pde);
542 * Remove a /proc entry and free it if it's not currently in use.
544 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
546 struct proc_dir_entry *de = NULL;
547 const char *fn = name;
550 spin_lock(&proc_subdir_lock);
551 if (__xlate_proc_name(name, &parent, &fn) != 0) {
552 spin_unlock(&proc_subdir_lock);
557 de = pde_subdir_find(parent, fn, len);
559 rb_erase(&de->subdir_node, &parent->subdir);
560 spin_unlock(&proc_subdir_lock);
562 WARN(1, "name '%s'\n", name);
566 proc_entry_rundown(de);
568 if (S_ISDIR(de->mode))
571 WARN(pde_subdir_first(de),
572 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
573 __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
576 EXPORT_SYMBOL(remove_proc_entry);
578 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
580 struct proc_dir_entry *root = NULL, *de, *next;
581 const char *fn = name;
584 spin_lock(&proc_subdir_lock);
585 if (__xlate_proc_name(name, &parent, &fn) != 0) {
586 spin_unlock(&proc_subdir_lock);
591 root = pde_subdir_find(parent, fn, len);
593 spin_unlock(&proc_subdir_lock);
596 rb_erase(&root->subdir_node, &parent->subdir);
600 next = pde_subdir_first(de);
602 rb_erase(&next->subdir_node, &de->subdir);
606 spin_unlock(&proc_subdir_lock);
608 proc_entry_rundown(de);
610 if (S_ISDIR(de->mode))
617 spin_lock(&proc_subdir_lock);
623 EXPORT_SYMBOL(remove_proc_subtree);
625 void *proc_get_parent_data(const struct inode *inode)
627 struct proc_dir_entry *de = PDE(inode);
628 return de->parent->data;
630 EXPORT_SYMBOL_GPL(proc_get_parent_data);
632 void proc_remove(struct proc_dir_entry *de)
635 remove_proc_subtree(de->name, de->parent);
637 EXPORT_SYMBOL(proc_remove);
639 void *PDE_DATA(const struct inode *inode)
641 return __PDE_DATA(inode);
643 EXPORT_SYMBOL(PDE_DATA);