xref: /linux/fs/proc/generic.c (revision eeecea75ece2ba894bee4d3c53ed2e40abda8f0e)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * proc/fs/generic.c --- generic routines for the proc-fs
4  *
5  * This file contains generic proc-fs routines for handling
6  * directories and files.
7  *
8  * Copyright (C) 1991, 1992 Linus Torvalds.
9  * Copyright (C) 1997 Theodore Ts'o
10  */
11 
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>
17 #include <linux/mm.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>
30 
31 #include "internal.h"
32 
33 static DEFINE_RWLOCK(proc_subdir_lock);
34 
35 struct kmem_cache *proc_dir_entry_cache __ro_after_init;
36 
37 void pde_free(struct proc_dir_entry *pde)
38 {
39 	if (S_ISLNK(pde->mode))
40 		kfree(pde->data);
41 	if (pde->name != pde->inline_name)
42 		kfree(pde->name);
43 	kmem_cache_free(proc_dir_entry_cache, pde);
44 }
45 
46 static int proc_match(const char *name, struct proc_dir_entry *de, unsigned int len)
47 {
48 	if (len < de->namelen)
49 		return -1;
50 	if (len > de->namelen)
51 		return 1;
52 
53 	return memcmp(name, de->name, len);
54 }
55 
56 static struct proc_dir_entry *pde_subdir_first(struct proc_dir_entry *dir)
57 {
58 	return rb_entry_safe(rb_first(&dir->subdir), struct proc_dir_entry,
59 			     subdir_node);
60 }
61 
62 static struct proc_dir_entry *pde_subdir_next(struct proc_dir_entry *dir)
63 {
64 	return rb_entry_safe(rb_next(&dir->subdir_node), struct proc_dir_entry,
65 			     subdir_node);
66 }
67 
68 static struct proc_dir_entry *pde_subdir_find(struct proc_dir_entry *dir,
69 					      const char *name,
70 					      unsigned int len)
71 {
72 	struct rb_node *node = dir->subdir.rb_node;
73 
74 	while (node) {
75 		struct proc_dir_entry *de = rb_entry(node,
76 						     struct proc_dir_entry,
77 						     subdir_node);
78 		int result = proc_match(name, de, len);
79 
80 		if (result < 0)
81 			node = node->rb_left;
82 		else if (result > 0)
83 			node = node->rb_right;
84 		else
85 			return de;
86 	}
87 	return NULL;
88 }
89 
90 static bool pde_subdir_insert(struct proc_dir_entry *dir,
91 			      struct proc_dir_entry *de)
92 {
93 	struct rb_root *root = &dir->subdir;
94 	struct rb_node **new = &root->rb_node, *parent = NULL;
95 
96 	/* Figure out where to put new node */
97 	while (*new) {
98 		struct proc_dir_entry *this = rb_entry(*new,
99 						       struct proc_dir_entry,
100 						       subdir_node);
101 		int result = proc_match(de->name, this, de->namelen);
102 
103 		parent = *new;
104 		if (result < 0)
105 			new = &(*new)->rb_left;
106 		else if (result > 0)
107 			new = &(*new)->rb_right;
108 		else
109 			return false;
110 	}
111 
112 	/* Add new node and rebalance tree. */
113 	rb_link_node(&de->subdir_node, parent, new);
114 	rb_insert_color(&de->subdir_node, root);
115 	if (S_ISDIR(de->mode))
116 		dir->nlink++;
117 	return true;
118 }
119 
120 static int proc_setattr(const struct mnt_idmap *idmap, struct dentry *dentry,
121 		struct iattr *iattr)
122 {
123 	struct inode *inode = d_inode(dentry);
124 	struct proc_dir_entry *de = PDE(inode);
125 	int error;
126 
127 	error = setattr_prepare(&nop_mnt_idmap, dentry, iattr);
128 	if (error)
129 		return error;
130 
131 	setattr_copy(&nop_mnt_idmap, inode, iattr);
132 
133 	proc_set_user(de, inode->i_uid, inode->i_gid);
134 	de->mode = inode->i_mode;
135 	return 0;
136 }
137 
138 static int proc_getattr(const struct mnt_idmap *idmap,
139 			const struct path *path, struct kstat *stat,
140 			u32 request_mask, unsigned int query_flags)
141 {
142 	struct inode *inode = d_inode(path->dentry);
143 	struct proc_dir_entry *de = PDE(inode);
144 	if (de) {
145 		nlink_t nlink = READ_ONCE(de->nlink);
146 		if (nlink > 0) {
147 			set_nlink(inode, nlink);
148 		}
149 	}
150 
151 	generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
152 	return 0;
153 }
154 
155 static const struct inode_operations proc_file_inode_operations = {
156 	.setattr	= proc_setattr,
157 };
158 
159 /*
160  * This function parses a name such as "tty/driver/serial", and
161  * returns the struct proc_dir_entry for "/proc/tty/driver", and
162  * returns "serial" in residual.
163  */
164 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
165 			     const char **residual)
166 {
167 	const char     		*cp = name, *next;
168 	struct proc_dir_entry	*de;
169 
170 	de = *ret ?: &proc_root;
171 	while ((next = strchr(cp, '/')) != NULL) {
172 		de = pde_subdir_find(de, cp, next - cp);
173 		if (!de) {
174 			WARN(1, "name '%s'\n", name);
175 			return -ENOENT;
176 		}
177 		cp = next + 1;
178 	}
179 	*residual = cp;
180 	*ret = de;
181 	return 0;
182 }
183 
184 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
185 			   const char **residual)
186 {
187 	int rv;
188 
189 	read_lock(&proc_subdir_lock);
190 	rv = __xlate_proc_name(name, ret, residual);
191 	read_unlock(&proc_subdir_lock);
192 	return rv;
193 }
194 
195 static DEFINE_IDA(proc_inum_ida);
196 
197 #define PROC_DYNAMIC_FIRST 0xF0000000U
198 
199 /*
200  * Return an inode number between PROC_DYNAMIC_FIRST and
201  * 0xffffffff, or zero on failure.
202  */
203 int proc_alloc_inum(unsigned int *inum)
204 {
205 	int i;
206 
207 	i = ida_alloc_max(&proc_inum_ida, UINT_MAX - PROC_DYNAMIC_FIRST,
208 			  GFP_KERNEL);
209 	if (i < 0)
210 		return i;
211 
212 	*inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
213 	return 0;
214 }
215 
216 void proc_free_inum(unsigned int inum)
217 {
218 	ida_free(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
219 }
220 
221 static int proc_misc_d_revalidate(struct inode *dir, const struct qstr *name,
222 				  struct dentry *dentry, unsigned int flags)
223 {
224 	if (flags & LOOKUP_RCU)
225 		return -ECHILD;
226 
227 	if (atomic_read(&PDE(d_inode(dentry))->in_use) < 0)
228 		return 0; /* revalidate */
229 	return 1;
230 }
231 
232 static int proc_misc_d_delete(const struct dentry *dentry)
233 {
234 	return atomic_read(&PDE(d_inode(dentry))->in_use) < 0;
235 }
236 
237 static const struct dentry_operations proc_misc_dentry_ops = {
238 	.d_revalidate	= proc_misc_d_revalidate,
239 	.d_delete	= proc_misc_d_delete,
240 };
241 
242 /*
243  * Don't create negative dentries here, return -ENOENT by hand
244  * instead.
245  */
246 struct dentry *proc_lookup_de(struct inode *dir, struct dentry *dentry,
247 			      struct proc_dir_entry *de)
248 {
249 	struct inode *inode;
250 
251 	read_lock(&proc_subdir_lock);
252 	de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
253 	if (de) {
254 		pde_get(de);
255 		read_unlock(&proc_subdir_lock);
256 		inode = proc_get_inode(dir->i_sb, de);
257 		if (!inode)
258 			return ERR_PTR(-ENOMEM);
259 		/* never reused and so never a mountpoint for long */
260 		if (de->flags & PROC_ENTRY_FORCE_LOOKUP)
261 			return proc_splice_unmountable(inode, dentry,
262 						       &proc_net_dentry_ops);
263 		return d_splice_alias_ops(inode, dentry,
264 					  &proc_misc_dentry_ops);
265 	}
266 	read_unlock(&proc_subdir_lock);
267 	return ERR_PTR(-ENOENT);
268 }
269 
270 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
271 		unsigned int flags)
272 {
273 	struct proc_fs_info *fs_info = proc_sb_info(dir->i_sb);
274 
275 	if (fs_info->pidonly == PROC_PIDONLY_ON)
276 		return ERR_PTR(-ENOENT);
277 
278 	return proc_lookup_de(dir, dentry, PDE(dir));
279 }
280 
281 /*
282  * This returns non-zero if at EOF, so that the /proc
283  * root directory can use this and check if it should
284  * continue with the <pid> entries..
285  *
286  * Note that the VFS-layer doesn't care about the return
287  * value of the readdir() call, as long as it's non-negative
288  * for success..
289  */
290 int proc_readdir_de(struct file *file, struct dir_context *ctx,
291 		    struct proc_dir_entry *de)
292 {
293 	int i;
294 
295 	if (!dir_emit_dots(file, ctx))
296 		return 0;
297 
298 	i = ctx->pos - 2;
299 	read_lock(&proc_subdir_lock);
300 	de = pde_subdir_first(de);
301 	for (;;) {
302 		if (!de) {
303 			read_unlock(&proc_subdir_lock);
304 			return 0;
305 		}
306 		if (!i)
307 			break;
308 		de = pde_subdir_next(de);
309 		i--;
310 	}
311 
312 	do {
313 		struct proc_dir_entry *next;
314 		pde_get(de);
315 		read_unlock(&proc_subdir_lock);
316 		if (!dir_emit(ctx, de->name, de->namelen,
317 			    de->low_ino, de->mode >> 12)) {
318 			pde_put(de);
319 			return 0;
320 		}
321 		ctx->pos++;
322 		read_lock(&proc_subdir_lock);
323 		next = pde_subdir_next(de);
324 		pde_put(de);
325 		de = next;
326 	} while (de);
327 	read_unlock(&proc_subdir_lock);
328 	return 1;
329 }
330 
331 int proc_readdir(struct file *file, struct dir_context *ctx)
332 {
333 	struct inode *inode = file_inode(file);
334 	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
335 
336 	if (fs_info->pidonly == PROC_PIDONLY_ON)
337 		return 1;
338 
339 	return proc_readdir_de(file, ctx, PDE(inode));
340 }
341 
342 /*
343  * These are the generic /proc directory operations. They
344  * use the in-memory "struct proc_dir_entry" tree to parse
345  * the /proc directory.
346  */
347 static const struct file_operations proc_dir_operations = {
348 	.llseek			= generic_file_llseek,
349 	.read			= generic_read_dir,
350 	.iterate_shared		= proc_readdir,
351 };
352 
353 static int proc_net_d_revalidate(struct inode *dir, const struct qstr *name,
354 				 struct dentry *dentry, unsigned int flags)
355 {
356 	return 0;
357 }
358 
359 const struct dentry_operations proc_net_dentry_ops = {
360 	.d_revalidate	= proc_net_d_revalidate,
361 	.d_delete	= always_delete_dentry,
362 };
363 
364 /*
365  * proc directories can do almost nothing..
366  */
367 static const struct inode_operations proc_dir_inode_operations = {
368 	.lookup		= proc_lookup,
369 	.getattr	= proc_getattr,
370 	.setattr	= proc_setattr,
371 };
372 
373 static void pde_set_flags(struct proc_dir_entry *pde)
374 {
375 	const struct proc_ops *proc_ops = pde->proc_ops;
376 
377 	if (!proc_ops)
378 		return;
379 
380 	if (proc_ops->proc_flags & PROC_ENTRY_PERMANENT)
381 		pde->flags |= PROC_ENTRY_PERMANENT;
382 	if (proc_ops->proc_read_iter)
383 		pde->flags |= PROC_ENTRY_proc_read_iter;
384 #ifdef CONFIG_COMPAT
385 	if (proc_ops->proc_compat_ioctl)
386 		pde->flags |= PROC_ENTRY_proc_compat_ioctl;
387 #endif
388 	if (proc_ops->proc_lseek)
389 		pde->flags |= PROC_ENTRY_proc_lseek;
390 }
391 
392 /* returns the registered entry, or frees dp and returns NULL on failure */
393 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
394 		struct proc_dir_entry *dp)
395 {
396 	if (proc_alloc_inum(&dp->low_ino))
397 		goto out_free_entry;
398 
399 	if (!S_ISDIR(dp->mode))
400 		pde_set_flags(dp);
401 
402 	write_lock(&proc_subdir_lock);
403 	dp->parent = dir;
404 	if (pde_subdir_insert(dir, dp) == false) {
405 		WARN(1, "proc_dir_entry '%s/%s' already registered\n",
406 		     dir->name, dp->name);
407 		write_unlock(&proc_subdir_lock);
408 		goto out_free_inum;
409 	}
410 	write_unlock(&proc_subdir_lock);
411 
412 	return dp;
413 out_free_inum:
414 	proc_free_inum(dp->low_ino);
415 out_free_entry:
416 	pde_free(dp);
417 	return NULL;
418 }
419 
420 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
421 					  const char *name,
422 					  umode_t mode,
423 					  nlink_t nlink)
424 {
425 	struct proc_dir_entry *ent = NULL;
426 	const char *fn;
427 	struct qstr qstr;
428 
429 	if (xlate_proc_name(name, parent, &fn) != 0)
430 		goto out;
431 	qstr.name = fn;
432 	qstr.len = strnlen(fn, NAME_MAX + 1);
433 	if (qstr.len == 0) {
434 		WARN(1, "empty name\n");
435 		return NULL;
436 	}
437 	if (qstr.len > NAME_MAX) {
438 		WARN(1, "name too long\n");
439 		return NULL;
440 	}
441 	if (qstr.len == 1 && fn[0] == '.') {
442 		WARN(1, "name '.'\n");
443 		return NULL;
444 	}
445 	if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
446 		WARN(1, "name '..'\n");
447 		return NULL;
448 	}
449 	if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
450 		WARN(1, "create '/proc/%s' by hand\n", qstr.name);
451 		return NULL;
452 	}
453 	if (is_empty_pde(*parent)) {
454 		WARN(1, "attempt to add to permanently empty directory");
455 		return NULL;
456 	}
457 
458 	ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
459 	if (!ent)
460 		goto out;
461 
462 	if (qstr.len + 1 <= SIZEOF_PDE_INLINE_NAME) {
463 		ent->name = ent->inline_name;
464 	} else {
465 		ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
466 		if (!ent->name) {
467 			pde_free(ent);
468 			return NULL;
469 		}
470 	}
471 
472 	memcpy(ent->name, fn, qstr.len + 1);
473 	ent->namelen = qstr.len;
474 	ent->mode = mode;
475 	ent->nlink = nlink;
476 	ent->subdir = RB_ROOT;
477 	refcount_set(&ent->refcnt, 1);
478 	spin_lock_init(&ent->pde_unload_lock);
479 	INIT_LIST_HEAD(&ent->pde_openers);
480 	proc_set_user(ent, (*parent)->uid, (*parent)->gid);
481 
482 	/* Revalidate everything under /proc/${pid}/net */
483 	if ((*parent)->flags & PROC_ENTRY_FORCE_LOOKUP)
484 		pde_force_lookup(ent);
485 
486 out:
487 	return ent;
488 }
489 
490 struct proc_dir_entry *proc_symlink(const char *name,
491 		struct proc_dir_entry *parent, const char *dest)
492 {
493 	struct proc_dir_entry *ent;
494 
495 	ent = __proc_create(&parent, name,
496 			  (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
497 
498 	if (ent) {
499 		ent->size = strlen(dest);
500 		ent->data = kmemdup(dest, ent->size + 1, GFP_KERNEL);
501 		if (ent->data) {
502 			ent->proc_iops = &proc_link_inode_operations;
503 			ent = proc_register(parent, ent);
504 		} else {
505 			pde_free(ent);
506 			ent = NULL;
507 		}
508 	}
509 	return ent;
510 }
511 EXPORT_SYMBOL(proc_symlink);
512 
513 struct proc_dir_entry *_proc_mkdir(const char *name, umode_t mode,
514 		struct proc_dir_entry *parent, void *data, bool force_lookup)
515 {
516 	struct proc_dir_entry *ent;
517 
518 	if (mode == 0)
519 		mode = S_IRUGO | S_IXUGO;
520 
521 	ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
522 	if (ent) {
523 		ent->data = data;
524 		ent->proc_dir_ops = &proc_dir_operations;
525 		ent->proc_iops = &proc_dir_inode_operations;
526 		if (force_lookup) {
527 			pde_force_lookup(ent);
528 		}
529 		ent = proc_register(parent, ent);
530 	}
531 	return ent;
532 }
533 EXPORT_SYMBOL_GPL(_proc_mkdir);
534 
535 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
536 		struct proc_dir_entry *parent, void *data)
537 {
538 	return _proc_mkdir(name, mode, parent, data, false);
539 }
540 EXPORT_SYMBOL_GPL(proc_mkdir_data);
541 
542 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
543 				       struct proc_dir_entry *parent)
544 {
545 	return proc_mkdir_data(name, mode, parent, NULL);
546 }
547 EXPORT_SYMBOL(proc_mkdir_mode);
548 
549 struct proc_dir_entry *proc_mkdir(const char *name,
550 		struct proc_dir_entry *parent)
551 {
552 	return proc_mkdir_data(name, 0, parent, NULL);
553 }
554 EXPORT_SYMBOL(proc_mkdir);
555 
556 struct proc_dir_entry *proc_create_mount_point(const char *name)
557 {
558 	umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
559 	struct proc_dir_entry *ent, *parent = NULL;
560 
561 	ent = __proc_create(&parent, name, mode, 2);
562 	if (ent) {
563 		ent->data = NULL;
564 		ent->proc_dir_ops = NULL;
565 		ent->proc_iops = NULL;
566 		ent = proc_register(parent, ent);
567 	}
568 	return ent;
569 }
570 EXPORT_SYMBOL(proc_create_mount_point);
571 
572 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
573 		struct proc_dir_entry **parent, void *data)
574 {
575 	struct proc_dir_entry *p;
576 
577 	if ((mode & S_IFMT) == 0)
578 		mode |= S_IFREG;
579 	if ((mode & S_IALLUGO) == 0)
580 		mode |= S_IRUGO;
581 	if (WARN_ON_ONCE(!S_ISREG(mode)))
582 		return NULL;
583 
584 	p = __proc_create(parent, name, mode, 1);
585 	if (p) {
586 		p->proc_iops = &proc_file_inode_operations;
587 		p->data = data;
588 	}
589 	return p;
590 }
591 
592 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
593 		struct proc_dir_entry *parent,
594 		const struct proc_ops *proc_ops, void *data)
595 {
596 	struct proc_dir_entry *p;
597 
598 	p = proc_create_reg(name, mode, &parent, data);
599 	if (!p)
600 		return NULL;
601 	p->proc_ops = proc_ops;
602 	return proc_register(parent, p);
603 }
604 EXPORT_SYMBOL(proc_create_data);
605 
606 struct proc_dir_entry *proc_create(const char *name, umode_t mode,
607 				   struct proc_dir_entry *parent,
608 				   const struct proc_ops *proc_ops)
609 {
610 	return proc_create_data(name, mode, parent, proc_ops, NULL);
611 }
612 EXPORT_SYMBOL(proc_create);
613 
614 static int proc_seq_open(struct inode *inode, struct file *file)
615 {
616 	struct proc_dir_entry *de = PDE(inode);
617 
618 	if (de->state_size)
619 		return seq_open_private(file, de->seq_ops, de->state_size);
620 	return seq_open(file, de->seq_ops);
621 }
622 
623 static int proc_seq_release(struct inode *inode, struct file *file)
624 {
625 	struct proc_dir_entry *de = PDE(inode);
626 
627 	if (de->state_size)
628 		return seq_release_private(inode, file);
629 	return seq_release(inode, file);
630 }
631 
632 static const struct proc_ops proc_seq_ops = {
633 	/* not permanent -- can call into arbitrary seq_operations */
634 	.proc_open	= proc_seq_open,
635 	.proc_read_iter	= seq_read_iter,
636 	.proc_lseek	= seq_lseek,
637 	.proc_release	= proc_seq_release,
638 };
639 
640 struct proc_dir_entry *proc_create_seq_private(const char *name, umode_t mode,
641 		struct proc_dir_entry *parent, const struct seq_operations *ops,
642 		unsigned int state_size, void *data)
643 {
644 	struct proc_dir_entry *p;
645 
646 	p = proc_create_reg(name, mode, &parent, data);
647 	if (!p)
648 		return NULL;
649 	p->proc_ops = &proc_seq_ops;
650 	p->seq_ops = ops;
651 	p->state_size = state_size;
652 	return proc_register(parent, p);
653 }
654 EXPORT_SYMBOL(proc_create_seq_private);
655 
656 static int proc_single_open(struct inode *inode, struct file *file)
657 {
658 	struct proc_dir_entry *de = PDE(inode);
659 
660 	return single_open(file, de->single_show, de->data);
661 }
662 
663 static const struct proc_ops proc_single_ops = {
664 	/* not permanent -- can call into arbitrary ->single_show */
665 	.proc_open	= proc_single_open,
666 	.proc_read_iter = seq_read_iter,
667 	.proc_lseek	= seq_lseek,
668 	.proc_release	= single_release,
669 };
670 
671 struct proc_dir_entry *proc_create_single_data(const char *name, umode_t mode,
672 		struct proc_dir_entry *parent,
673 		int (*show)(struct seq_file *, void *), void *data)
674 {
675 	struct proc_dir_entry *p;
676 
677 	p = proc_create_reg(name, mode, &parent, data);
678 	if (!p)
679 		return NULL;
680 	p->proc_ops = &proc_single_ops;
681 	p->single_show = show;
682 	return proc_register(parent, p);
683 }
684 EXPORT_SYMBOL(proc_create_single_data);
685 
686 void proc_set_size(struct proc_dir_entry *de, loff_t size)
687 {
688 	de->size = size;
689 }
690 EXPORT_SYMBOL(proc_set_size);
691 
692 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
693 {
694 	de->uid = uid;
695 	de->gid = gid;
696 }
697 EXPORT_SYMBOL(proc_set_user);
698 
699 void pde_put(struct proc_dir_entry *pde)
700 {
701 	if (refcount_dec_and_test(&pde->refcnt)) {
702 		proc_free_inum(pde->low_ino);
703 		pde_free(pde);
704 	}
705 }
706 
707 static void pde_erase(struct proc_dir_entry *pde, struct proc_dir_entry *parent)
708 {
709 	rb_erase(&pde->subdir_node, &parent->subdir);
710 	RB_CLEAR_NODE(&pde->subdir_node);
711 	if (S_ISDIR(pde->mode))
712 		parent->nlink--;
713 }
714 
715 /*
716  * Remove a /proc entry and free it if it's not currently in use.
717  */
718 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
719 {
720 	struct proc_dir_entry *de = NULL;
721 	const char *fn = name;
722 	unsigned int len;
723 
724 	write_lock(&proc_subdir_lock);
725 	if (__xlate_proc_name(name, &parent, &fn) != 0) {
726 		write_unlock(&proc_subdir_lock);
727 		return;
728 	}
729 	len = strlen(fn);
730 
731 	de = pde_subdir_find(parent, fn, len);
732 	if (de) {
733 		if (unlikely(pde_is_permanent(de))) {
734 			WARN(1, "removing permanent /proc entry '%s'", de->name);
735 			de = NULL;
736 		} else {
737 			pde_erase(de, parent);
738 		}
739 	}
740 	write_unlock(&proc_subdir_lock);
741 	if (!de) {
742 		WARN(1, "name '%s'\n", name);
743 		return;
744 	}
745 
746 	proc_entry_rundown(de);
747 
748 	WARN(pde_subdir_first(de),
749 	     "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
750 	     __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
751 	pde_put(de);
752 }
753 EXPORT_SYMBOL(remove_proc_entry);
754 
755 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
756 {
757 	struct proc_dir_entry *root = NULL, *de, *next;
758 	const char *fn = name;
759 	unsigned int len;
760 
761 	write_lock(&proc_subdir_lock);
762 	if (__xlate_proc_name(name, &parent, &fn) != 0) {
763 		write_unlock(&proc_subdir_lock);
764 		return -ENOENT;
765 	}
766 	len = strlen(fn);
767 
768 	root = pde_subdir_find(parent, fn, len);
769 	if (!root) {
770 		write_unlock(&proc_subdir_lock);
771 		return -ENOENT;
772 	}
773 	if (unlikely(pde_is_permanent(root))) {
774 		write_unlock(&proc_subdir_lock);
775 		WARN(1, "removing permanent /proc entry '%s/%s'",
776 			root->parent->name, root->name);
777 		return -EINVAL;
778 	}
779 	pde_erase(root, parent);
780 
781 	de = root;
782 	while (1) {
783 		next = pde_subdir_first(de);
784 		if (next) {
785 			if (unlikely(pde_is_permanent(next))) {
786 				write_unlock(&proc_subdir_lock);
787 				WARN(1, "removing permanent /proc entry '%s/%s'",
788 					next->parent->name, next->name);
789 				return -EINVAL;
790 			}
791 			pde_erase(next, de);
792 			de = next;
793 			continue;
794 		}
795 		next = de->parent;
796 		write_unlock(&proc_subdir_lock);
797 
798 		proc_entry_rundown(de);
799 		if (de == root)
800 			break;
801 		pde_put(de);
802 
803 		write_lock(&proc_subdir_lock);
804 		de = next;
805 	}
806 	pde_put(root);
807 	return 0;
808 }
809 EXPORT_SYMBOL(remove_proc_subtree);
810 
811 void *proc_get_parent_data(const struct inode *inode)
812 {
813 	struct proc_dir_entry *de = PDE(inode);
814 	return de->parent->data;
815 }
816 EXPORT_SYMBOL_GPL(proc_get_parent_data);
817 
818 void proc_remove(struct proc_dir_entry *de)
819 {
820 	if (de)
821 		remove_proc_subtree(de->name, de->parent);
822 }
823 EXPORT_SYMBOL(proc_remove);
824 
825 /*
826  * Pull a user buffer into memory and pass it to the file's write handler if
827  * one is supplied.  The ->write() method is permitted to modify the
828  * kernel-side buffer.
829  */
830 ssize_t proc_simple_write(struct file *f, const char __user *ubuf, size_t size,
831 			  loff_t *_pos)
832 {
833 	struct proc_dir_entry *pde = PDE(file_inode(f));
834 	char *buf;
835 	int ret;
836 
837 	if (!pde->write)
838 		return -EACCES;
839 	if (size == 0 || size > PAGE_SIZE - 1)
840 		return -EINVAL;
841 	buf = memdup_user_nul(ubuf, size);
842 	if (IS_ERR(buf))
843 		return PTR_ERR(buf);
844 	ret = pde->write(f, buf, size);
845 	kfree(buf);
846 	return ret == 0 ? size : ret;
847 }
848 
849 /*
850  * Not exported to modules:
851  * modules' /proc files aren't permanent because modules aren't permanent.
852  */
853 void impl_proc_make_permanent(struct proc_dir_entry *pde)
854 {
855 	if (pde)
856 		pde_make_permanent(pde);
857 }
858