xref: /linux/fs/proc/generic.c (revision 989fe6771266bdb82a815d78802c5aa7c918fdfd) 
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 	return true;
116 }
117 
118 static int proc_notify_change(struct mnt_idmap *idmap,
119 			      struct dentry *dentry, struct iattr *iattr)
120 {
121 	struct inode *inode = d_inode(dentry);
122 	struct proc_dir_entry *de = PDE(inode);
123 	int error;
124 
125 	error = setattr_prepare(&nop_mnt_idmap, dentry, iattr);
126 	if (error)
127 		return error;
128 
129 	setattr_copy(&nop_mnt_idmap, inode, iattr);
130 
131 	proc_set_user(de, inode->i_uid, inode->i_gid);
132 	de->mode = inode->i_mode;
133 	return 0;
134 }
135 
136 static int proc_getattr(struct mnt_idmap *idmap,
137 			const struct path *path, struct kstat *stat,
138 			u32 request_mask, unsigned int query_flags)
139 {
140 	struct inode *inode = d_inode(path->dentry);
141 	struct proc_dir_entry *de = PDE(inode);
142 	if (de) {
143 		nlink_t nlink = READ_ONCE(de->nlink);
144 		if (nlink > 0) {
145 			set_nlink(inode, nlink);
146 		}
147 	}
148 
149 	generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
150 	return 0;
151 }
152 
153 static const struct inode_operations proc_file_inode_operations = {
154 	.setattr	= proc_notify_change,
155 };
156 
157 /*
158  * This function parses a name such as "tty/driver/serial", and
159  * returns the struct proc_dir_entry for "/proc/tty/driver", and
160  * returns "serial" in residual.
161  */
162 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
163 			     const char **residual)
164 {
165 	const char     		*cp = name, *next;
166 	struct proc_dir_entry	*de;
167 
168 	de = *ret ?: &proc_root;
169 	while ((next = strchr(cp, '/')) != NULL) {
170 		de = pde_subdir_find(de, cp, next - cp);
171 		if (!de) {
172 			WARN(1, "name '%s'\n", name);
173 			return -ENOENT;
174 		}
175 		cp = next + 1;
176 	}
177 	*residual = cp;
178 	*ret = de;
179 	return 0;
180 }
181 
182 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
183 			   const char **residual)
184 {
185 	int rv;
186 
187 	read_lock(&proc_subdir_lock);
188 	rv = __xlate_proc_name(name, ret, residual);
189 	read_unlock(&proc_subdir_lock);
190 	return rv;
191 }
192 
193 static DEFINE_IDA(proc_inum_ida);
194 
195 #define PROC_DYNAMIC_FIRST 0xF0000000U
196 
197 /*
198  * Return an inode number between PROC_DYNAMIC_FIRST and
199  * 0xffffffff, or zero on failure.
200  */
201 int proc_alloc_inum(unsigned int *inum)
202 {
203 	int i;
204 
205 	i = ida_alloc_max(&proc_inum_ida, UINT_MAX - PROC_DYNAMIC_FIRST,
206 			  GFP_KERNEL);
207 	if (i < 0)
208 		return i;
209 
210 	*inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
211 	return 0;
212 }
213 
214 void proc_free_inum(unsigned int inum)
215 {
216 	ida_free(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
217 }
218 
219 static int proc_misc_d_revalidate(struct inode *dir, const struct qstr *name,
220 				  struct dentry *dentry, unsigned int flags)
221 {
222 	if (flags & LOOKUP_RCU)
223 		return -ECHILD;
224 
225 	if (atomic_read(&PDE(d_inode(dentry))->in_use) < 0)
226 		return 0; /* revalidate */
227 	return 1;
228 }
229 
230 static int proc_misc_d_delete(const struct dentry *dentry)
231 {
232 	return atomic_read(&PDE(d_inode(dentry))->in_use) < 0;
233 }
234 
235 static const struct dentry_operations proc_misc_dentry_ops = {
236 	.d_revalidate	= proc_misc_d_revalidate,
237 	.d_delete	= proc_misc_d_delete,
238 };
239 
240 /*
241  * Don't create negative dentries here, return -ENOENT by hand
242  * instead.
243  */
244 struct dentry *proc_lookup_de(struct inode *dir, struct dentry *dentry,
245 			      struct proc_dir_entry *de)
246 {
247 	struct inode *inode;
248 
249 	read_lock(&proc_subdir_lock);
250 	de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
251 	if (de) {
252 		pde_get(de);
253 		read_unlock(&proc_subdir_lock);
254 		inode = proc_get_inode(dir->i_sb, de);
255 		if (!inode)
256 			return ERR_PTR(-ENOMEM);
257 		if (de->flags & PROC_ENTRY_FORCE_LOOKUP)
258 			return d_splice_alias_ops(inode, dentry,
259 						  &proc_net_dentry_ops);
260 		return d_splice_alias_ops(inode, dentry,
261 					  &proc_misc_dentry_ops);
262 	}
263 	read_unlock(&proc_subdir_lock);
264 	return ERR_PTR(-ENOENT);
265 }
266 
267 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
268 		unsigned int flags)
269 {
270 	struct proc_fs_info *fs_info = proc_sb_info(dir->i_sb);
271 
272 	if (fs_info->pidonly == PROC_PIDONLY_ON)
273 		return ERR_PTR(-ENOENT);
274 
275 	return proc_lookup_de(dir, dentry, PDE(dir));
276 }
277 
278 /*
279  * This returns non-zero if at EOF, so that the /proc
280  * root directory can use this and check if it should
281  * continue with the <pid> entries..
282  *
283  * Note that the VFS-layer doesn't care about the return
284  * value of the readdir() call, as long as it's non-negative
285  * for success..
286  */
287 int proc_readdir_de(struct file *file, struct dir_context *ctx,
288 		    struct proc_dir_entry *de)
289 {
290 	int i;
291 
292 	if (!dir_emit_dots(file, ctx))
293 		return 0;
294 
295 	i = ctx->pos - 2;
296 	read_lock(&proc_subdir_lock);
297 	de = pde_subdir_first(de);
298 	for (;;) {
299 		if (!de) {
300 			read_unlock(&proc_subdir_lock);
301 			return 0;
302 		}
303 		if (!i)
304 			break;
305 		de = pde_subdir_next(de);
306 		i--;
307 	}
308 
309 	do {
310 		struct proc_dir_entry *next;
311 		pde_get(de);
312 		read_unlock(&proc_subdir_lock);
313 		if (!dir_emit(ctx, de->name, de->namelen,
314 			    de->low_ino, de->mode >> 12)) {
315 			pde_put(de);
316 			return 0;
317 		}
318 		ctx->pos++;
319 		read_lock(&proc_subdir_lock);
320 		next = pde_subdir_next(de);
321 		pde_put(de);
322 		de = next;
323 	} while (de);
324 	read_unlock(&proc_subdir_lock);
325 	return 1;
326 }
327 
328 int proc_readdir(struct file *file, struct dir_context *ctx)
329 {
330 	struct inode *inode = file_inode(file);
331 	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
332 
333 	if (fs_info->pidonly == PROC_PIDONLY_ON)
334 		return 1;
335 
336 	return proc_readdir_de(file, ctx, PDE(inode));
337 }
338 
339 /*
340  * These are the generic /proc directory operations. They
341  * use the in-memory "struct proc_dir_entry" tree to parse
342  * the /proc directory.
343  */
344 static const struct file_operations proc_dir_operations = {
345 	.llseek			= generic_file_llseek,
346 	.read			= generic_read_dir,
347 	.iterate_shared		= proc_readdir,
348 };
349 
350 static int proc_net_d_revalidate(struct inode *dir, const struct qstr *name,
351 				 struct dentry *dentry, unsigned int flags)
352 {
353 	return 0;
354 }
355 
356 const struct dentry_operations proc_net_dentry_ops = {
357 	.d_revalidate	= proc_net_d_revalidate,
358 	.d_delete	= always_delete_dentry,
359 };
360 
361 /*
362  * proc directories can do almost nothing..
363  */
364 static const struct inode_operations proc_dir_inode_operations = {
365 	.lookup		= proc_lookup,
366 	.getattr	= proc_getattr,
367 	.setattr	= proc_notify_change,
368 };
369 
370 /* returns the registered entry, or frees dp and returns NULL on failure */
371 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
372 		struct proc_dir_entry *dp)
373 {
374 	if (proc_alloc_inum(&dp->low_ino))
375 		goto out_free_entry;
376 
377 	write_lock(&proc_subdir_lock);
378 	dp->parent = dir;
379 	if (pde_subdir_insert(dir, dp) == false) {
380 		WARN(1, "proc_dir_entry '%s/%s' already registered\n",
381 		     dir->name, dp->name);
382 		write_unlock(&proc_subdir_lock);
383 		goto out_free_inum;
384 	}
385 	dir->nlink++;
386 	write_unlock(&proc_subdir_lock);
387 
388 	return dp;
389 out_free_inum:
390 	proc_free_inum(dp->low_ino);
391 out_free_entry:
392 	pde_free(dp);
393 	return NULL;
394 }
395 
396 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
397 					  const char *name,
398 					  umode_t mode,
399 					  nlink_t nlink)
400 {
401 	struct proc_dir_entry *ent = NULL;
402 	const char *fn;
403 	struct qstr qstr;
404 
405 	if (xlate_proc_name(name, parent, &fn) != 0)
406 		goto out;
407 	qstr.name = fn;
408 	qstr.len = strlen(fn);
409 	if (qstr.len == 0 || qstr.len >= 256) {
410 		WARN(1, "name len %u\n", qstr.len);
411 		return NULL;
412 	}
413 	if (qstr.len == 1 && fn[0] == '.') {
414 		WARN(1, "name '.'\n");
415 		return NULL;
416 	}
417 	if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
418 		WARN(1, "name '..'\n");
419 		return NULL;
420 	}
421 	if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
422 		WARN(1, "create '/proc/%s' by hand\n", qstr.name);
423 		return NULL;
424 	}
425 	if (is_empty_pde(*parent)) {
426 		WARN(1, "attempt to add to permanently empty directory");
427 		return NULL;
428 	}
429 
430 	ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
431 	if (!ent)
432 		goto out;
433 
434 	if (qstr.len + 1 <= SIZEOF_PDE_INLINE_NAME) {
435 		ent->name = ent->inline_name;
436 	} else {
437 		ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
438 		if (!ent->name) {
439 			pde_free(ent);
440 			return NULL;
441 		}
442 	}
443 
444 	memcpy(ent->name, fn, qstr.len + 1);
445 	ent->namelen = qstr.len;
446 	ent->mode = mode;
447 	ent->nlink = nlink;
448 	ent->subdir = RB_ROOT;
449 	refcount_set(&ent->refcnt, 1);
450 	spin_lock_init(&ent->pde_unload_lock);
451 	INIT_LIST_HEAD(&ent->pde_openers);
452 	proc_set_user(ent, (*parent)->uid, (*parent)->gid);
453 
454 	/* Revalidate everything under /proc/${pid}/net */
455 	if ((*parent)->flags & PROC_ENTRY_FORCE_LOOKUP)
456 		pde_force_lookup(ent);
457 
458 out:
459 	return ent;
460 }
461 
462 struct proc_dir_entry *proc_symlink(const char *name,
463 		struct proc_dir_entry *parent, const char *dest)
464 {
465 	struct proc_dir_entry *ent;
466 
467 	ent = __proc_create(&parent, name,
468 			  (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
469 
470 	if (ent) {
471 		ent->size = strlen(dest);
472 		ent->data = kmemdup(dest, ent->size + 1, GFP_KERNEL);
473 		if (ent->data) {
474 			ent->proc_iops = &proc_link_inode_operations;
475 			ent = proc_register(parent, ent);
476 		} else {
477 			pde_free(ent);
478 			ent = NULL;
479 		}
480 	}
481 	return ent;
482 }
483 EXPORT_SYMBOL(proc_symlink);
484 
485 struct proc_dir_entry *_proc_mkdir(const char *name, umode_t mode,
486 		struct proc_dir_entry *parent, void *data, bool force_lookup)
487 {
488 	struct proc_dir_entry *ent;
489 
490 	if (mode == 0)
491 		mode = S_IRUGO | S_IXUGO;
492 
493 	ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
494 	if (ent) {
495 		ent->data = data;
496 		ent->proc_dir_ops = &proc_dir_operations;
497 		ent->proc_iops = &proc_dir_inode_operations;
498 		if (force_lookup) {
499 			pde_force_lookup(ent);
500 		}
501 		ent = proc_register(parent, ent);
502 	}
503 	return ent;
504 }
505 EXPORT_SYMBOL_GPL(_proc_mkdir);
506 
507 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
508 		struct proc_dir_entry *parent, void *data)
509 {
510 	return _proc_mkdir(name, mode, parent, data, false);
511 }
512 EXPORT_SYMBOL_GPL(proc_mkdir_data);
513 
514 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
515 				       struct proc_dir_entry *parent)
516 {
517 	return proc_mkdir_data(name, mode, parent, NULL);
518 }
519 EXPORT_SYMBOL(proc_mkdir_mode);
520 
521 struct proc_dir_entry *proc_mkdir(const char *name,
522 		struct proc_dir_entry *parent)
523 {
524 	return proc_mkdir_data(name, 0, parent, NULL);
525 }
526 EXPORT_SYMBOL(proc_mkdir);
527 
528 struct proc_dir_entry *proc_create_mount_point(const char *name)
529 {
530 	umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
531 	struct proc_dir_entry *ent, *parent = NULL;
532 
533 	ent = __proc_create(&parent, name, mode, 2);
534 	if (ent) {
535 		ent->data = NULL;
536 		ent->proc_dir_ops = NULL;
537 		ent->proc_iops = NULL;
538 		ent = proc_register(parent, ent);
539 	}
540 	return ent;
541 }
542 EXPORT_SYMBOL(proc_create_mount_point);
543 
544 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
545 		struct proc_dir_entry **parent, void *data)
546 {
547 	struct proc_dir_entry *p;
548 
549 	if ((mode & S_IFMT) == 0)
550 		mode |= S_IFREG;
551 	if ((mode & S_IALLUGO) == 0)
552 		mode |= S_IRUGO;
553 	if (WARN_ON_ONCE(!S_ISREG(mode)))
554 		return NULL;
555 
556 	p = __proc_create(parent, name, mode, 1);
557 	if (p) {
558 		p->proc_iops = &proc_file_inode_operations;
559 		p->data = data;
560 	}
561 	return p;
562 }
563 
564 static void pde_set_flags(struct proc_dir_entry *pde)
565 {
566 	if (pde->proc_ops->proc_flags & PROC_ENTRY_PERMANENT)
567 		pde->flags |= PROC_ENTRY_PERMANENT;
568 	if (pde->proc_ops->proc_read_iter)
569 		pde->flags |= PROC_ENTRY_proc_read_iter;
570 #ifdef CONFIG_COMPAT
571 	if (pde->proc_ops->proc_compat_ioctl)
572 		pde->flags |= PROC_ENTRY_proc_compat_ioctl;
573 #endif
574 	if (pde->proc_ops->proc_lseek)
575 		pde->flags |= PROC_ENTRY_proc_lseek;
576 }
577 
578 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
579 		struct proc_dir_entry *parent,
580 		const struct proc_ops *proc_ops, void *data)
581 {
582 	struct proc_dir_entry *p;
583 
584 	p = proc_create_reg(name, mode, &parent, data);
585 	if (!p)
586 		return NULL;
587 	p->proc_ops = proc_ops;
588 	pde_set_flags(p);
589 	return proc_register(parent, p);
590 }
591 EXPORT_SYMBOL(proc_create_data);
592 
593 struct proc_dir_entry *proc_create(const char *name, umode_t mode,
594 				   struct proc_dir_entry *parent,
595 				   const struct proc_ops *proc_ops)
596 {
597 	return proc_create_data(name, mode, parent, proc_ops, NULL);
598 }
599 EXPORT_SYMBOL(proc_create);
600 
601 static int proc_seq_open(struct inode *inode, struct file *file)
602 {
603 	struct proc_dir_entry *de = PDE(inode);
604 
605 	if (de->state_size)
606 		return seq_open_private(file, de->seq_ops, de->state_size);
607 	return seq_open(file, de->seq_ops);
608 }
609 
610 static int proc_seq_release(struct inode *inode, struct file *file)
611 {
612 	struct proc_dir_entry *de = PDE(inode);
613 
614 	if (de->state_size)
615 		return seq_release_private(inode, file);
616 	return seq_release(inode, file);
617 }
618 
619 static const struct proc_ops proc_seq_ops = {
620 	/* not permanent -- can call into arbitrary seq_operations */
621 	.proc_open	= proc_seq_open,
622 	.proc_read_iter	= seq_read_iter,
623 	.proc_lseek	= seq_lseek,
624 	.proc_release	= proc_seq_release,
625 };
626 
627 struct proc_dir_entry *proc_create_seq_private(const char *name, umode_t mode,
628 		struct proc_dir_entry *parent, const struct seq_operations *ops,
629 		unsigned int state_size, void *data)
630 {
631 	struct proc_dir_entry *p;
632 
633 	p = proc_create_reg(name, mode, &parent, data);
634 	if (!p)
635 		return NULL;
636 	p->proc_ops = &proc_seq_ops;
637 	p->seq_ops = ops;
638 	p->state_size = state_size;
639 	pde_set_flags(p);
640 	return proc_register(parent, p);
641 }
642 EXPORT_SYMBOL(proc_create_seq_private);
643 
644 static int proc_single_open(struct inode *inode, struct file *file)
645 {
646 	struct proc_dir_entry *de = PDE(inode);
647 
648 	return single_open(file, de->single_show, de->data);
649 }
650 
651 static const struct proc_ops proc_single_ops = {
652 	/* not permanent -- can call into arbitrary ->single_show */
653 	.proc_open	= proc_single_open,
654 	.proc_read_iter = seq_read_iter,
655 	.proc_lseek	= seq_lseek,
656 	.proc_release	= single_release,
657 };
658 
659 struct proc_dir_entry *proc_create_single_data(const char *name, umode_t mode,
660 		struct proc_dir_entry *parent,
661 		int (*show)(struct seq_file *, void *), void *data)
662 {
663 	struct proc_dir_entry *p;
664 
665 	p = proc_create_reg(name, mode, &parent, data);
666 	if (!p)
667 		return NULL;
668 	p->proc_ops = &proc_single_ops;
669 	p->single_show = show;
670 	pde_set_flags(p);
671 	return proc_register(parent, p);
672 }
673 EXPORT_SYMBOL(proc_create_single_data);
674 
675 void proc_set_size(struct proc_dir_entry *de, loff_t size)
676 {
677 	de->size = size;
678 }
679 EXPORT_SYMBOL(proc_set_size);
680 
681 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
682 {
683 	de->uid = uid;
684 	de->gid = gid;
685 }
686 EXPORT_SYMBOL(proc_set_user);
687 
688 void pde_put(struct proc_dir_entry *pde)
689 {
690 	if (refcount_dec_and_test(&pde->refcnt)) {
691 		proc_free_inum(pde->low_ino);
692 		pde_free(pde);
693 	}
694 }
695 
696 /*
697  * Remove a /proc entry and free it if it's not currently in use.
698  */
699 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
700 {
701 	struct proc_dir_entry *de = NULL;
702 	const char *fn = name;
703 	unsigned int len;
704 
705 	write_lock(&proc_subdir_lock);
706 	if (__xlate_proc_name(name, &parent, &fn) != 0) {
707 		write_unlock(&proc_subdir_lock);
708 		return;
709 	}
710 	len = strlen(fn);
711 
712 	de = pde_subdir_find(parent, fn, len);
713 	if (de) {
714 		if (unlikely(pde_is_permanent(de))) {
715 			WARN(1, "removing permanent /proc entry '%s'", de->name);
716 			de = NULL;
717 		} else {
718 			rb_erase(&de->subdir_node, &parent->subdir);
719 			if (S_ISDIR(de->mode))
720 				parent->nlink--;
721 		}
722 	}
723 	write_unlock(&proc_subdir_lock);
724 	if (!de) {
725 		WARN(1, "name '%s'\n", name);
726 		return;
727 	}
728 
729 	proc_entry_rundown(de);
730 
731 	WARN(pde_subdir_first(de),
732 	     "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
733 	     __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
734 	pde_put(de);
735 }
736 EXPORT_SYMBOL(remove_proc_entry);
737 
738 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
739 {
740 	struct proc_dir_entry *root = NULL, *de, *next;
741 	const char *fn = name;
742 	unsigned int len;
743 
744 	write_lock(&proc_subdir_lock);
745 	if (__xlate_proc_name(name, &parent, &fn) != 0) {
746 		write_unlock(&proc_subdir_lock);
747 		return -ENOENT;
748 	}
749 	len = strlen(fn);
750 
751 	root = pde_subdir_find(parent, fn, len);
752 	if (!root) {
753 		write_unlock(&proc_subdir_lock);
754 		return -ENOENT;
755 	}
756 	if (unlikely(pde_is_permanent(root))) {
757 		write_unlock(&proc_subdir_lock);
758 		WARN(1, "removing permanent /proc entry '%s/%s'",
759 			root->parent->name, root->name);
760 		return -EINVAL;
761 	}
762 	rb_erase(&root->subdir_node, &parent->subdir);
763 
764 	de = root;
765 	while (1) {
766 		next = pde_subdir_first(de);
767 		if (next) {
768 			if (unlikely(pde_is_permanent(next))) {
769 				write_unlock(&proc_subdir_lock);
770 				WARN(1, "removing permanent /proc entry '%s/%s'",
771 					next->parent->name, next->name);
772 				return -EINVAL;
773 			}
774 			rb_erase(&next->subdir_node, &de->subdir);
775 			de = next;
776 			continue;
777 		}
778 		next = de->parent;
779 		if (S_ISDIR(de->mode))
780 			next->nlink--;
781 		write_unlock(&proc_subdir_lock);
782 
783 		proc_entry_rundown(de);
784 		if (de == root)
785 			break;
786 		pde_put(de);
787 
788 		write_lock(&proc_subdir_lock);
789 		de = next;
790 	}
791 	pde_put(root);
792 	return 0;
793 }
794 EXPORT_SYMBOL(remove_proc_subtree);
795 
796 void *proc_get_parent_data(const struct inode *inode)
797 {
798 	struct proc_dir_entry *de = PDE(inode);
799 	return de->parent->data;
800 }
801 EXPORT_SYMBOL_GPL(proc_get_parent_data);
802 
803 void proc_remove(struct proc_dir_entry *de)
804 {
805 	if (de)
806 		remove_proc_subtree(de->name, de->parent);
807 }
808 EXPORT_SYMBOL(proc_remove);
809 
810 /*
811  * Pull a user buffer into memory and pass it to the file's write handler if
812  * one is supplied.  The ->write() method is permitted to modify the
813  * kernel-side buffer.
814  */
815 ssize_t proc_simple_write(struct file *f, const char __user *ubuf, size_t size,
816 			  loff_t *_pos)
817 {
818 	struct proc_dir_entry *pde = PDE(file_inode(f));
819 	char *buf;
820 	int ret;
821 
822 	if (!pde->write)
823 		return -EACCES;
824 	if (size == 0 || size > PAGE_SIZE - 1)
825 		return -EINVAL;
826 	buf = memdup_user_nul(ubuf, size);
827 	if (IS_ERR(buf))
828 		return PTR_ERR(buf);
829 	ret = pde->write(f, buf, size);
830 	kfree(buf);
831 	return ret == 0 ? size : ret;
832 }
833