Linux Audio

Check our new training course

Loading...
v6.2
  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
 33static DEFINE_RWLOCK(proc_subdir_lock);
 34
 35struct kmem_cache *proc_dir_entry_cache __ro_after_init;
 36
 37void 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
 46static 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
 56static 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
 62static 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
 68static 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
 90static 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
118static int proc_notify_change(struct user_namespace *mnt_userns,
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(&init_user_ns, dentry, iattr);
126	if (error)
127		return error;
128
129	setattr_copy(&init_user_ns, inode, iattr);
130	mark_inode_dirty(inode);
131
132	proc_set_user(de, inode->i_uid, inode->i_gid);
133	de->mode = inode->i_mode;
134	return 0;
135}
136
137static int proc_getattr(struct user_namespace *mnt_userns,
138			const struct path *path, struct kstat *stat,
139			u32 request_mask, unsigned int query_flags)
140{
141	struct inode *inode = d_inode(path->dentry);
142	struct proc_dir_entry *de = PDE(inode);
143	if (de) {
144		nlink_t nlink = READ_ONCE(de->nlink);
145		if (nlink > 0) {
146			set_nlink(inode, nlink);
147		}
148	}
149
150	generic_fillattr(&init_user_ns, inode, stat);
151	return 0;
152}
153
154static const struct inode_operations proc_file_inode_operations = {
155	.setattr	= proc_notify_change,
156};
157
158/*
159 * This function parses a name such as "tty/driver/serial", and
160 * returns the struct proc_dir_entry for "/proc/tty/driver", and
161 * returns "serial" in residual.
162 */
163static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
164			     const char **residual)
165{
166	const char     		*cp = name, *next;
167	struct proc_dir_entry	*de;
168
169	de = *ret ?: &proc_root;
170	while ((next = strchr(cp, '/')) != NULL) {
 
 
 
 
 
 
 
171		de = pde_subdir_find(de, cp, next - cp);
172		if (!de) {
173			WARN(1, "name '%s'\n", name);
174			return -ENOENT;
175		}
176		cp = next + 1;
177	}
178	*residual = cp;
179	*ret = de;
180	return 0;
181}
182
183static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
184			   const char **residual)
185{
186	int rv;
187
188	read_lock(&proc_subdir_lock);
189	rv = __xlate_proc_name(name, ret, residual);
190	read_unlock(&proc_subdir_lock);
191	return rv;
192}
193
194static DEFINE_IDA(proc_inum_ida);
195
196#define PROC_DYNAMIC_FIRST 0xF0000000U
197
198/*
199 * Return an inode number between PROC_DYNAMIC_FIRST and
200 * 0xffffffff, or zero on failure.
201 */
202int proc_alloc_inum(unsigned int *inum)
203{
204	int i;
205
206	i = ida_simple_get(&proc_inum_ida, 0, UINT_MAX - PROC_DYNAMIC_FIRST + 1,
207			   GFP_KERNEL);
208	if (i < 0)
209		return i;
210
211	*inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
212	return 0;
213}
214
215void proc_free_inum(unsigned int inum)
216{
217	ida_simple_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
218}
219
220static int proc_misc_d_revalidate(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
230static int proc_misc_d_delete(const struct dentry *dentry)
231{
232	return atomic_read(&PDE(d_inode(dentry))->in_use) < 0;
233}
234
235static 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 */
244struct 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		d_set_d_op(dentry, de->proc_dops);
258		return d_splice_alias(inode, dentry);
259	}
260	read_unlock(&proc_subdir_lock);
261	return ERR_PTR(-ENOENT);
262}
263
264struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
265		unsigned int flags)
266{
267	struct proc_fs_info *fs_info = proc_sb_info(dir->i_sb);
268
269	if (fs_info->pidonly == PROC_PIDONLY_ON)
270		return ERR_PTR(-ENOENT);
271
272	return proc_lookup_de(dir, dentry, PDE(dir));
273}
274
275/*
276 * This returns non-zero if at EOF, so that the /proc
277 * root directory can use this and check if it should
278 * continue with the <pid> entries..
279 *
280 * Note that the VFS-layer doesn't care about the return
281 * value of the readdir() call, as long as it's non-negative
282 * for success..
283 */
284int proc_readdir_de(struct file *file, struct dir_context *ctx,
285		    struct proc_dir_entry *de)
286{
287	int i;
288
289	if (!dir_emit_dots(file, ctx))
290		return 0;
291
292	i = ctx->pos - 2;
293	read_lock(&proc_subdir_lock);
294	de = pde_subdir_first(de);
295	for (;;) {
296		if (!de) {
297			read_unlock(&proc_subdir_lock);
298			return 0;
299		}
300		if (!i)
301			break;
302		de = pde_subdir_next(de);
303		i--;
304	}
305
306	do {
307		struct proc_dir_entry *next;
308		pde_get(de);
309		read_unlock(&proc_subdir_lock);
310		if (!dir_emit(ctx, de->name, de->namelen,
311			    de->low_ino, de->mode >> 12)) {
312			pde_put(de);
313			return 0;
314		}
315		ctx->pos++;
316		read_lock(&proc_subdir_lock);
317		next = pde_subdir_next(de);
318		pde_put(de);
319		de = next;
320	} while (de);
321	read_unlock(&proc_subdir_lock);
322	return 1;
323}
324
325int proc_readdir(struct file *file, struct dir_context *ctx)
326{
327	struct inode *inode = file_inode(file);
328	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
329
330	if (fs_info->pidonly == PROC_PIDONLY_ON)
331		return 1;
332
333	return proc_readdir_de(file, ctx, PDE(inode));
334}
335
336/*
337 * These are the generic /proc directory operations. They
338 * use the in-memory "struct proc_dir_entry" tree to parse
339 * the /proc directory.
340 */
341static const struct file_operations proc_dir_operations = {
342	.llseek			= generic_file_llseek,
343	.read			= generic_read_dir,
344	.iterate_shared		= proc_readdir,
345};
346
347static int proc_net_d_revalidate(struct dentry *dentry, unsigned int flags)
348{
349	return 0;
350}
351
352const struct dentry_operations proc_net_dentry_ops = {
353	.d_revalidate	= proc_net_d_revalidate,
354	.d_delete	= always_delete_dentry,
355};
356
357/*
358 * proc directories can do almost nothing..
359 */
360static const struct inode_operations proc_dir_inode_operations = {
361	.lookup		= proc_lookup,
362	.getattr	= proc_getattr,
363	.setattr	= proc_notify_change,
364};
365
366/* returns the registered entry, or frees dp and returns NULL on failure */
367struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
368		struct proc_dir_entry *dp)
369{
370	if (proc_alloc_inum(&dp->low_ino))
371		goto out_free_entry;
372
373	write_lock(&proc_subdir_lock);
374	dp->parent = dir;
375	if (pde_subdir_insert(dir, dp) == false) {
376		WARN(1, "proc_dir_entry '%s/%s' already registered\n",
377		     dir->name, dp->name);
378		write_unlock(&proc_subdir_lock);
379		goto out_free_inum;
380	}
381	dir->nlink++;
382	write_unlock(&proc_subdir_lock);
383
384	return dp;
385out_free_inum:
386	proc_free_inum(dp->low_ino);
387out_free_entry:
388	pde_free(dp);
389	return NULL;
390}
391
392static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
393					  const char *name,
394					  umode_t mode,
395					  nlink_t nlink)
396{
397	struct proc_dir_entry *ent = NULL;
398	const char *fn;
399	struct qstr qstr;
400
401	if (xlate_proc_name(name, parent, &fn) != 0)
402		goto out;
403	qstr.name = fn;
404	qstr.len = strlen(fn);
405	if (qstr.len == 0 || qstr.len >= 256) {
406		WARN(1, "name len %u\n", qstr.len);
407		return NULL;
408	}
409	if (qstr.len == 1 && fn[0] == '.') {
410		WARN(1, "name '.'\n");
411		return NULL;
412	}
413	if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
414		WARN(1, "name '..'\n");
415		return NULL;
416	}
417	if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
418		WARN(1, "create '/proc/%s' by hand\n", qstr.name);
419		return NULL;
420	}
421	if (is_empty_pde(*parent)) {
422		WARN(1, "attempt to add to permanently empty directory");
423		return NULL;
424	}
425
426	ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
427	if (!ent)
428		goto out;
429
430	if (qstr.len + 1 <= SIZEOF_PDE_INLINE_NAME) {
431		ent->name = ent->inline_name;
432	} else {
433		ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
434		if (!ent->name) {
435			pde_free(ent);
436			return NULL;
437		}
438	}
439
440	memcpy(ent->name, fn, qstr.len + 1);
441	ent->namelen = qstr.len;
442	ent->mode = mode;
443	ent->nlink = nlink;
444	ent->subdir = RB_ROOT;
445	refcount_set(&ent->refcnt, 1);
446	spin_lock_init(&ent->pde_unload_lock);
447	INIT_LIST_HEAD(&ent->pde_openers);
448	proc_set_user(ent, (*parent)->uid, (*parent)->gid);
449
450	ent->proc_dops = &proc_misc_dentry_ops;
451	/* Revalidate everything under /proc/${pid}/net */
452	if ((*parent)->proc_dops == &proc_net_dentry_ops)
453		pde_force_lookup(ent);
454
455out:
456	return ent;
457}
458
459struct proc_dir_entry *proc_symlink(const char *name,
460		struct proc_dir_entry *parent, const char *dest)
461{
462	struct proc_dir_entry *ent;
463
464	ent = __proc_create(&parent, name,
465			  (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
466
467	if (ent) {
468		ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
469		if (ent->data) {
470			strcpy((char*)ent->data,dest);
471			ent->proc_iops = &proc_link_inode_operations;
472			ent = proc_register(parent, ent);
473		} else {
474			pde_free(ent);
475			ent = NULL;
476		}
477	}
478	return ent;
479}
480EXPORT_SYMBOL(proc_symlink);
481
482struct proc_dir_entry *_proc_mkdir(const char *name, umode_t mode,
483		struct proc_dir_entry *parent, void *data, bool force_lookup)
484{
485	struct proc_dir_entry *ent;
486
487	if (mode == 0)
488		mode = S_IRUGO | S_IXUGO;
489
490	ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
491	if (ent) {
492		ent->data = data;
493		ent->proc_dir_ops = &proc_dir_operations;
494		ent->proc_iops = &proc_dir_inode_operations;
495		if (force_lookup) {
496			pde_force_lookup(ent);
497		}
498		ent = proc_register(parent, ent);
499	}
500	return ent;
501}
502EXPORT_SYMBOL_GPL(_proc_mkdir);
503
504struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
505		struct proc_dir_entry *parent, void *data)
506{
507	return _proc_mkdir(name, mode, parent, data, false);
508}
509EXPORT_SYMBOL_GPL(proc_mkdir_data);
510
511struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
512				       struct proc_dir_entry *parent)
513{
514	return proc_mkdir_data(name, mode, parent, NULL);
515}
516EXPORT_SYMBOL(proc_mkdir_mode);
517
518struct proc_dir_entry *proc_mkdir(const char *name,
519		struct proc_dir_entry *parent)
520{
521	return proc_mkdir_data(name, 0, parent, NULL);
522}
523EXPORT_SYMBOL(proc_mkdir);
524
525struct proc_dir_entry *proc_create_mount_point(const char *name)
526{
527	umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
528	struct proc_dir_entry *ent, *parent = NULL;
529
530	ent = __proc_create(&parent, name, mode, 2);
531	if (ent) {
532		ent->data = NULL;
533		ent->proc_dir_ops = NULL;
534		ent->proc_iops = NULL;
535		ent = proc_register(parent, ent);
536	}
537	return ent;
538}
539EXPORT_SYMBOL(proc_create_mount_point);
540
541struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
542		struct proc_dir_entry **parent, void *data)
543{
544	struct proc_dir_entry *p;
545
546	if ((mode & S_IFMT) == 0)
547		mode |= S_IFREG;
548	if ((mode & S_IALLUGO) == 0)
549		mode |= S_IRUGO;
550	if (WARN_ON_ONCE(!S_ISREG(mode)))
551		return NULL;
552
553	p = __proc_create(parent, name, mode, 1);
554	if (p) {
555		p->proc_iops = &proc_file_inode_operations;
556		p->data = data;
557	}
558	return p;
559}
560
561static inline void pde_set_flags(struct proc_dir_entry *pde)
562{
563	if (pde->proc_ops->proc_flags & PROC_ENTRY_PERMANENT)
564		pde->flags |= PROC_ENTRY_PERMANENT;
565}
566
567struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
568		struct proc_dir_entry *parent,
569		const struct proc_ops *proc_ops, void *data)
570{
571	struct proc_dir_entry *p;
572
573	p = proc_create_reg(name, mode, &parent, data);
574	if (!p)
575		return NULL;
576	p->proc_ops = proc_ops;
577	pde_set_flags(p);
578	return proc_register(parent, p);
579}
580EXPORT_SYMBOL(proc_create_data);
581 
582struct proc_dir_entry *proc_create(const char *name, umode_t mode,
583				   struct proc_dir_entry *parent,
584				   const struct proc_ops *proc_ops)
585{
586	return proc_create_data(name, mode, parent, proc_ops, NULL);
587}
588EXPORT_SYMBOL(proc_create);
589
590static int proc_seq_open(struct inode *inode, struct file *file)
591{
592	struct proc_dir_entry *de = PDE(inode);
593
594	if (de->state_size)
595		return seq_open_private(file, de->seq_ops, de->state_size);
596	return seq_open(file, de->seq_ops);
597}
598
599static int proc_seq_release(struct inode *inode, struct file *file)
600{
601	struct proc_dir_entry *de = PDE(inode);
602
603	if (de->state_size)
604		return seq_release_private(inode, file);
605	return seq_release(inode, file);
606}
607
608static const struct proc_ops proc_seq_ops = {
609	/* not permanent -- can call into arbitrary seq_operations */
610	.proc_open	= proc_seq_open,
611	.proc_read_iter	= seq_read_iter,
612	.proc_lseek	= seq_lseek,
613	.proc_release	= proc_seq_release,
614};
615
616struct proc_dir_entry *proc_create_seq_private(const char *name, umode_t mode,
617		struct proc_dir_entry *parent, const struct seq_operations *ops,
618		unsigned int state_size, void *data)
619{
620	struct proc_dir_entry *p;
621
622	p = proc_create_reg(name, mode, &parent, data);
623	if (!p)
624		return NULL;
625	p->proc_ops = &proc_seq_ops;
626	p->seq_ops = ops;
627	p->state_size = state_size;
628	return proc_register(parent, p);
629}
630EXPORT_SYMBOL(proc_create_seq_private);
631
632static int proc_single_open(struct inode *inode, struct file *file)
633{
634	struct proc_dir_entry *de = PDE(inode);
635
636	return single_open(file, de->single_show, de->data);
637}
638
639static const struct proc_ops proc_single_ops = {
640	/* not permanent -- can call into arbitrary ->single_show */
641	.proc_open	= proc_single_open,
642	.proc_read_iter = seq_read_iter,
643	.proc_lseek	= seq_lseek,
644	.proc_release	= single_release,
645};
646
647struct proc_dir_entry *proc_create_single_data(const char *name, umode_t mode,
648		struct proc_dir_entry *parent,
649		int (*show)(struct seq_file *, void *), void *data)
650{
651	struct proc_dir_entry *p;
652
653	p = proc_create_reg(name, mode, &parent, data);
654	if (!p)
655		return NULL;
656	p->proc_ops = &proc_single_ops;
657	p->single_show = show;
658	return proc_register(parent, p);
659}
660EXPORT_SYMBOL(proc_create_single_data);
661
662void proc_set_size(struct proc_dir_entry *de, loff_t size)
663{
664	de->size = size;
665}
666EXPORT_SYMBOL(proc_set_size);
667
668void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
669{
670	de->uid = uid;
671	de->gid = gid;
672}
673EXPORT_SYMBOL(proc_set_user);
674
675void pde_put(struct proc_dir_entry *pde)
676{
677	if (refcount_dec_and_test(&pde->refcnt)) {
678		proc_free_inum(pde->low_ino);
679		pde_free(pde);
680	}
681}
682
683/*
684 * Remove a /proc entry and free it if it's not currently in use.
685 */
686void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
687{
688	struct proc_dir_entry *de = NULL;
689	const char *fn = name;
690	unsigned int len;
691
692	write_lock(&proc_subdir_lock);
693	if (__xlate_proc_name(name, &parent, &fn) != 0) {
694		write_unlock(&proc_subdir_lock);
695		return;
696	}
697	len = strlen(fn);
698
699	de = pde_subdir_find(parent, fn, len);
700	if (de) {
701		if (unlikely(pde_is_permanent(de))) {
702			WARN(1, "removing permanent /proc entry '%s'", de->name);
703			de = NULL;
704		} else {
705			rb_erase(&de->subdir_node, &parent->subdir);
706			if (S_ISDIR(de->mode))
707				parent->nlink--;
708		}
709	}
710	write_unlock(&proc_subdir_lock);
711	if (!de) {
712		WARN(1, "name '%s'\n", name);
713		return;
714	}
715
716	proc_entry_rundown(de);
717
718	WARN(pde_subdir_first(de),
719	     "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
720	     __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
721	pde_put(de);
722}
723EXPORT_SYMBOL(remove_proc_entry);
724
725int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
726{
727	struct proc_dir_entry *root = NULL, *de, *next;
728	const char *fn = name;
729	unsigned int len;
730
731	write_lock(&proc_subdir_lock);
732	if (__xlate_proc_name(name, &parent, &fn) != 0) {
733		write_unlock(&proc_subdir_lock);
734		return -ENOENT;
735	}
736	len = strlen(fn);
737
738	root = pde_subdir_find(parent, fn, len);
739	if (!root) {
740		write_unlock(&proc_subdir_lock);
741		return -ENOENT;
742	}
743	if (unlikely(pde_is_permanent(root))) {
744		write_unlock(&proc_subdir_lock);
745		WARN(1, "removing permanent /proc entry '%s/%s'",
746			root->parent->name, root->name);
747		return -EINVAL;
748	}
749	rb_erase(&root->subdir_node, &parent->subdir);
750
751	de = root;
752	while (1) {
753		next = pde_subdir_first(de);
754		if (next) {
755			if (unlikely(pde_is_permanent(next))) {
756				write_unlock(&proc_subdir_lock);
757				WARN(1, "removing permanent /proc entry '%s/%s'",
758					next->parent->name, next->name);
759				return -EINVAL;
760			}
761			rb_erase(&next->subdir_node, &de->subdir);
762			de = next;
763			continue;
764		}
765		next = de->parent;
766		if (S_ISDIR(de->mode))
767			next->nlink--;
768		write_unlock(&proc_subdir_lock);
769
770		proc_entry_rundown(de);
771		if (de == root)
772			break;
773		pde_put(de);
774
775		write_lock(&proc_subdir_lock);
776		de = next;
777	}
778	pde_put(root);
779	return 0;
780}
781EXPORT_SYMBOL(remove_proc_subtree);
782
783void *proc_get_parent_data(const struct inode *inode)
784{
785	struct proc_dir_entry *de = PDE(inode);
786	return de->parent->data;
787}
788EXPORT_SYMBOL_GPL(proc_get_parent_data);
789
790void proc_remove(struct proc_dir_entry *de)
791{
792	if (de)
793		remove_proc_subtree(de->name, de->parent);
794}
795EXPORT_SYMBOL(proc_remove);
 
 
 
 
 
 
796
797/*
798 * Pull a user buffer into memory and pass it to the file's write handler if
799 * one is supplied.  The ->write() method is permitted to modify the
800 * kernel-side buffer.
801 */
802ssize_t proc_simple_write(struct file *f, const char __user *ubuf, size_t size,
803			  loff_t *_pos)
804{
805	struct proc_dir_entry *pde = PDE(file_inode(f));
806	char *buf;
807	int ret;
808
809	if (!pde->write)
810		return -EACCES;
811	if (size == 0 || size > PAGE_SIZE - 1)
812		return -EINVAL;
813	buf = memdup_user_nul(ubuf, size);
814	if (IS_ERR(buf))
815		return PTR_ERR(buf);
816	ret = pde->write(f, buf, size);
817	kfree(buf);
818	return ret == 0 ? size : ret;
819}
v5.9
  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
 33static DEFINE_RWLOCK(proc_subdir_lock);
 34
 35struct kmem_cache *proc_dir_entry_cache __ro_after_init;
 36
 37void 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
 46static 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
 56static 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
 62static 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
 68static 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
 90static 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
118static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
 
119{
120	struct inode *inode = d_inode(dentry);
121	struct proc_dir_entry *de = PDE(inode);
122	int error;
123
124	error = setattr_prepare(dentry, iattr);
125	if (error)
126		return error;
127
128	setattr_copy(inode, iattr);
129	mark_inode_dirty(inode);
130
131	proc_set_user(de, inode->i_uid, inode->i_gid);
132	de->mode = inode->i_mode;
133	return 0;
134}
135
136static int proc_getattr(const struct path *path, struct kstat *stat,
 
137			u32 request_mask, unsigned int query_flags)
138{
139	struct inode *inode = d_inode(path->dentry);
140	struct proc_dir_entry *de = PDE(inode);
141	if (de) {
142		nlink_t nlink = READ_ONCE(de->nlink);
143		if (nlink > 0) {
144			set_nlink(inode, nlink);
145		}
146	}
147
148	generic_fillattr(inode, stat);
149	return 0;
150}
151
152static const struct inode_operations proc_file_inode_operations = {
153	.setattr	= proc_notify_change,
154};
155
156/*
157 * This function parses a name such as "tty/driver/serial", and
158 * returns the struct proc_dir_entry for "/proc/tty/driver", and
159 * returns "serial" in residual.
160 */
161static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
162			     const char **residual)
163{
164	const char     		*cp = name, *next;
165	struct proc_dir_entry	*de;
166
167	de = *ret;
168	if (!de)
169		de = &proc_root;
170
171	while (1) {
172		next = strchr(cp, '/');
173		if (!next)
174			break;
175
176		de = pde_subdir_find(de, cp, next - cp);
177		if (!de) {
178			WARN(1, "name '%s'\n", name);
179			return -ENOENT;
180		}
181		cp = next + 1;
182	}
183	*residual = cp;
184	*ret = de;
185	return 0;
186}
187
188static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
189			   const char **residual)
190{
191	int rv;
192
193	read_lock(&proc_subdir_lock);
194	rv = __xlate_proc_name(name, ret, residual);
195	read_unlock(&proc_subdir_lock);
196	return rv;
197}
198
199static DEFINE_IDA(proc_inum_ida);
200
201#define PROC_DYNAMIC_FIRST 0xF0000000U
202
203/*
204 * Return an inode number between PROC_DYNAMIC_FIRST and
205 * 0xffffffff, or zero on failure.
206 */
207int proc_alloc_inum(unsigned int *inum)
208{
209	int i;
210
211	i = ida_simple_get(&proc_inum_ida, 0, UINT_MAX - PROC_DYNAMIC_FIRST + 1,
212			   GFP_KERNEL);
213	if (i < 0)
214		return i;
215
216	*inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
217	return 0;
218}
219
220void proc_free_inum(unsigned int inum)
221{
222	ida_simple_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
223}
224
225static int proc_misc_d_revalidate(struct dentry *dentry, unsigned int flags)
226{
227	if (flags & LOOKUP_RCU)
228		return -ECHILD;
229
230	if (atomic_read(&PDE(d_inode(dentry))->in_use) < 0)
231		return 0; /* revalidate */
232	return 1;
233}
234
235static int proc_misc_d_delete(const struct dentry *dentry)
236{
237	return atomic_read(&PDE(d_inode(dentry))->in_use) < 0;
238}
239
240static const struct dentry_operations proc_misc_dentry_ops = {
241	.d_revalidate	= proc_misc_d_revalidate,
242	.d_delete	= proc_misc_d_delete,
243};
244
245/*
246 * Don't create negative dentries here, return -ENOENT by hand
247 * instead.
248 */
249struct dentry *proc_lookup_de(struct inode *dir, struct dentry *dentry,
250			      struct proc_dir_entry *de)
251{
252	struct inode *inode;
253
254	read_lock(&proc_subdir_lock);
255	de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
256	if (de) {
257		pde_get(de);
258		read_unlock(&proc_subdir_lock);
259		inode = proc_get_inode(dir->i_sb, de);
260		if (!inode)
261			return ERR_PTR(-ENOMEM);
262		d_set_d_op(dentry, de->proc_dops);
263		return d_splice_alias(inode, dentry);
264	}
265	read_unlock(&proc_subdir_lock);
266	return ERR_PTR(-ENOENT);
267}
268
269struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
270		unsigned int flags)
271{
272	struct proc_fs_info *fs_info = proc_sb_info(dir->i_sb);
273
274	if (fs_info->pidonly == PROC_PIDONLY_ON)
275		return ERR_PTR(-ENOENT);
276
277	return proc_lookup_de(dir, dentry, PDE(dir));
278}
279
280/*
281 * This returns non-zero if at EOF, so that the /proc
282 * root directory can use this and check if it should
283 * continue with the <pid> entries..
284 *
285 * Note that the VFS-layer doesn't care about the return
286 * value of the readdir() call, as long as it's non-negative
287 * for success..
288 */
289int proc_readdir_de(struct file *file, struct dir_context *ctx,
290		    struct proc_dir_entry *de)
291{
292	int i;
293
294	if (!dir_emit_dots(file, ctx))
295		return 0;
296
297	i = ctx->pos - 2;
298	read_lock(&proc_subdir_lock);
299	de = pde_subdir_first(de);
300	for (;;) {
301		if (!de) {
302			read_unlock(&proc_subdir_lock);
303			return 0;
304		}
305		if (!i)
306			break;
307		de = pde_subdir_next(de);
308		i--;
309	}
310
311	do {
312		struct proc_dir_entry *next;
313		pde_get(de);
314		read_unlock(&proc_subdir_lock);
315		if (!dir_emit(ctx, de->name, de->namelen,
316			    de->low_ino, de->mode >> 12)) {
317			pde_put(de);
318			return 0;
319		}
320		ctx->pos++;
321		read_lock(&proc_subdir_lock);
322		next = pde_subdir_next(de);
323		pde_put(de);
324		de = next;
325	} while (de);
326	read_unlock(&proc_subdir_lock);
327	return 1;
328}
329
330int proc_readdir(struct file *file, struct dir_context *ctx)
331{
332	struct inode *inode = file_inode(file);
333	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
334
335	if (fs_info->pidonly == PROC_PIDONLY_ON)
336		return 1;
337
338	return proc_readdir_de(file, ctx, PDE(inode));
339}
340
341/*
342 * These are the generic /proc directory operations. They
343 * use the in-memory "struct proc_dir_entry" tree to parse
344 * the /proc directory.
345 */
346static const struct file_operations proc_dir_operations = {
347	.llseek			= generic_file_llseek,
348	.read			= generic_read_dir,
349	.iterate_shared		= proc_readdir,
350};
351
 
 
 
 
 
 
 
 
 
 
352/*
353 * proc directories can do almost nothing..
354 */
355static const struct inode_operations proc_dir_inode_operations = {
356	.lookup		= proc_lookup,
357	.getattr	= proc_getattr,
358	.setattr	= proc_notify_change,
359};
360
361/* returns the registered entry, or frees dp and returns NULL on failure */
362struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
363		struct proc_dir_entry *dp)
364{
365	if (proc_alloc_inum(&dp->low_ino))
366		goto out_free_entry;
367
368	write_lock(&proc_subdir_lock);
369	dp->parent = dir;
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		write_unlock(&proc_subdir_lock);
374		goto out_free_inum;
375	}
376	dir->nlink++;
377	write_unlock(&proc_subdir_lock);
378
379	return dp;
380out_free_inum:
381	proc_free_inum(dp->low_ino);
382out_free_entry:
383	pde_free(dp);
384	return NULL;
385}
386
387static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
388					  const char *name,
389					  umode_t mode,
390					  nlink_t nlink)
391{
392	struct proc_dir_entry *ent = NULL;
393	const char *fn;
394	struct qstr qstr;
395
396	if (xlate_proc_name(name, parent, &fn) != 0)
397		goto out;
398	qstr.name = fn;
399	qstr.len = strlen(fn);
400	if (qstr.len == 0 || qstr.len >= 256) {
401		WARN(1, "name len %u\n", qstr.len);
402		return NULL;
403	}
404	if (qstr.len == 1 && fn[0] == '.') {
405		WARN(1, "name '.'\n");
406		return NULL;
407	}
408	if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
409		WARN(1, "name '..'\n");
410		return NULL;
411	}
412	if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
413		WARN(1, "create '/proc/%s' by hand\n", qstr.name);
414		return NULL;
415	}
416	if (is_empty_pde(*parent)) {
417		WARN(1, "attempt to add to permanently empty directory");
418		return NULL;
419	}
420
421	ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
422	if (!ent)
423		goto out;
424
425	if (qstr.len + 1 <= SIZEOF_PDE_INLINE_NAME) {
426		ent->name = ent->inline_name;
427	} else {
428		ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
429		if (!ent->name) {
430			pde_free(ent);
431			return NULL;
432		}
433	}
434
435	memcpy(ent->name, fn, qstr.len + 1);
436	ent->namelen = qstr.len;
437	ent->mode = mode;
438	ent->nlink = nlink;
439	ent->subdir = RB_ROOT;
440	refcount_set(&ent->refcnt, 1);
441	spin_lock_init(&ent->pde_unload_lock);
442	INIT_LIST_HEAD(&ent->pde_openers);
443	proc_set_user(ent, (*parent)->uid, (*parent)->gid);
444
445	ent->proc_dops = &proc_misc_dentry_ops;
 
 
 
446
447out:
448	return ent;
449}
450
451struct proc_dir_entry *proc_symlink(const char *name,
452		struct proc_dir_entry *parent, const char *dest)
453{
454	struct proc_dir_entry *ent;
455
456	ent = __proc_create(&parent, name,
457			  (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
458
459	if (ent) {
460		ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
461		if (ent->data) {
462			strcpy((char*)ent->data,dest);
463			ent->proc_iops = &proc_link_inode_operations;
464			ent = proc_register(parent, ent);
465		} else {
466			pde_free(ent);
467			ent = NULL;
468		}
469	}
470	return ent;
471}
472EXPORT_SYMBOL(proc_symlink);
473
474struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
475		struct proc_dir_entry *parent, void *data)
476{
477	struct proc_dir_entry *ent;
478
479	if (mode == 0)
480		mode = S_IRUGO | S_IXUGO;
481
482	ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
483	if (ent) {
484		ent->data = data;
485		ent->proc_dir_ops = &proc_dir_operations;
486		ent->proc_iops = &proc_dir_inode_operations;
 
 
 
487		ent = proc_register(parent, ent);
488	}
489	return ent;
490}
 
 
 
 
 
 
 
491EXPORT_SYMBOL_GPL(proc_mkdir_data);
492
493struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
494				       struct proc_dir_entry *parent)
495{
496	return proc_mkdir_data(name, mode, parent, NULL);
497}
498EXPORT_SYMBOL(proc_mkdir_mode);
499
500struct proc_dir_entry *proc_mkdir(const char *name,
501		struct proc_dir_entry *parent)
502{
503	return proc_mkdir_data(name, 0, parent, NULL);
504}
505EXPORT_SYMBOL(proc_mkdir);
506
507struct proc_dir_entry *proc_create_mount_point(const char *name)
508{
509	umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
510	struct proc_dir_entry *ent, *parent = NULL;
511
512	ent = __proc_create(&parent, name, mode, 2);
513	if (ent) {
514		ent->data = NULL;
515		ent->proc_dir_ops = NULL;
516		ent->proc_iops = NULL;
517		ent = proc_register(parent, ent);
518	}
519	return ent;
520}
521EXPORT_SYMBOL(proc_create_mount_point);
522
523struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
524		struct proc_dir_entry **parent, void *data)
525{
526	struct proc_dir_entry *p;
527
528	if ((mode & S_IFMT) == 0)
529		mode |= S_IFREG;
530	if ((mode & S_IALLUGO) == 0)
531		mode |= S_IRUGO;
532	if (WARN_ON_ONCE(!S_ISREG(mode)))
533		return NULL;
534
535	p = __proc_create(parent, name, mode, 1);
536	if (p) {
537		p->proc_iops = &proc_file_inode_operations;
538		p->data = data;
539	}
540	return p;
541}
542
543static inline void pde_set_flags(struct proc_dir_entry *pde)
544{
545	if (pde->proc_ops->proc_flags & PROC_ENTRY_PERMANENT)
546		pde->flags |= PROC_ENTRY_PERMANENT;
547}
548
549struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
550		struct proc_dir_entry *parent,
551		const struct proc_ops *proc_ops, void *data)
552{
553	struct proc_dir_entry *p;
554
555	p = proc_create_reg(name, mode, &parent, data);
556	if (!p)
557		return NULL;
558	p->proc_ops = proc_ops;
559	pde_set_flags(p);
560	return proc_register(parent, p);
561}
562EXPORT_SYMBOL(proc_create_data);
563 
564struct proc_dir_entry *proc_create(const char *name, umode_t mode,
565				   struct proc_dir_entry *parent,
566				   const struct proc_ops *proc_ops)
567{
568	return proc_create_data(name, mode, parent, proc_ops, NULL);
569}
570EXPORT_SYMBOL(proc_create);
571
572static int proc_seq_open(struct inode *inode, struct file *file)
573{
574	struct proc_dir_entry *de = PDE(inode);
575
576	if (de->state_size)
577		return seq_open_private(file, de->seq_ops, de->state_size);
578	return seq_open(file, de->seq_ops);
579}
580
581static int proc_seq_release(struct inode *inode, struct file *file)
582{
583	struct proc_dir_entry *de = PDE(inode);
584
585	if (de->state_size)
586		return seq_release_private(inode, file);
587	return seq_release(inode, file);
588}
589
590static const struct proc_ops proc_seq_ops = {
591	/* not permanent -- can call into arbitrary seq_operations */
592	.proc_open	= proc_seq_open,
593	.proc_read	= seq_read,
594	.proc_lseek	= seq_lseek,
595	.proc_release	= proc_seq_release,
596};
597
598struct proc_dir_entry *proc_create_seq_private(const char *name, umode_t mode,
599		struct proc_dir_entry *parent, const struct seq_operations *ops,
600		unsigned int state_size, void *data)
601{
602	struct proc_dir_entry *p;
603
604	p = proc_create_reg(name, mode, &parent, data);
605	if (!p)
606		return NULL;
607	p->proc_ops = &proc_seq_ops;
608	p->seq_ops = ops;
609	p->state_size = state_size;
610	return proc_register(parent, p);
611}
612EXPORT_SYMBOL(proc_create_seq_private);
613
614static int proc_single_open(struct inode *inode, struct file *file)
615{
616	struct proc_dir_entry *de = PDE(inode);
617
618	return single_open(file, de->single_show, de->data);
619}
620
621static const struct proc_ops proc_single_ops = {
622	/* not permanent -- can call into arbitrary ->single_show */
623	.proc_open	= proc_single_open,
624	.proc_read	= seq_read,
625	.proc_lseek	= seq_lseek,
626	.proc_release	= single_release,
627};
628
629struct proc_dir_entry *proc_create_single_data(const char *name, umode_t mode,
630		struct proc_dir_entry *parent,
631		int (*show)(struct seq_file *, void *), void *data)
632{
633	struct proc_dir_entry *p;
634
635	p = proc_create_reg(name, mode, &parent, data);
636	if (!p)
637		return NULL;
638	p->proc_ops = &proc_single_ops;
639	p->single_show = show;
640	return proc_register(parent, p);
641}
642EXPORT_SYMBOL(proc_create_single_data);
643
644void proc_set_size(struct proc_dir_entry *de, loff_t size)
645{
646	de->size = size;
647}
648EXPORT_SYMBOL(proc_set_size);
649
650void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
651{
652	de->uid = uid;
653	de->gid = gid;
654}
655EXPORT_SYMBOL(proc_set_user);
656
657void pde_put(struct proc_dir_entry *pde)
658{
659	if (refcount_dec_and_test(&pde->refcnt)) {
660		proc_free_inum(pde->low_ino);
661		pde_free(pde);
662	}
663}
664
665/*
666 * Remove a /proc entry and free it if it's not currently in use.
667 */
668void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
669{
670	struct proc_dir_entry *de = NULL;
671	const char *fn = name;
672	unsigned int len;
673
674	write_lock(&proc_subdir_lock);
675	if (__xlate_proc_name(name, &parent, &fn) != 0) {
676		write_unlock(&proc_subdir_lock);
677		return;
678	}
679	len = strlen(fn);
680
681	de = pde_subdir_find(parent, fn, len);
682	if (de) {
683		if (unlikely(pde_is_permanent(de))) {
684			WARN(1, "removing permanent /proc entry '%s'", de->name);
685			de = NULL;
686		} else {
687			rb_erase(&de->subdir_node, &parent->subdir);
688			if (S_ISDIR(de->mode))
689				parent->nlink--;
690		}
691	}
692	write_unlock(&proc_subdir_lock);
693	if (!de) {
694		WARN(1, "name '%s'\n", name);
695		return;
696	}
697
698	proc_entry_rundown(de);
699
700	WARN(pde_subdir_first(de),
701	     "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
702	     __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
703	pde_put(de);
704}
705EXPORT_SYMBOL(remove_proc_entry);
706
707int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
708{
709	struct proc_dir_entry *root = NULL, *de, *next;
710	const char *fn = name;
711	unsigned int len;
712
713	write_lock(&proc_subdir_lock);
714	if (__xlate_proc_name(name, &parent, &fn) != 0) {
715		write_unlock(&proc_subdir_lock);
716		return -ENOENT;
717	}
718	len = strlen(fn);
719
720	root = pde_subdir_find(parent, fn, len);
721	if (!root) {
722		write_unlock(&proc_subdir_lock);
723		return -ENOENT;
724	}
725	if (unlikely(pde_is_permanent(root))) {
726		write_unlock(&proc_subdir_lock);
727		WARN(1, "removing permanent /proc entry '%s/%s'",
728			root->parent->name, root->name);
729		return -EINVAL;
730	}
731	rb_erase(&root->subdir_node, &parent->subdir);
732
733	de = root;
734	while (1) {
735		next = pde_subdir_first(de);
736		if (next) {
737			if (unlikely(pde_is_permanent(root))) {
738				write_unlock(&proc_subdir_lock);
739				WARN(1, "removing permanent /proc entry '%s/%s'",
740					next->parent->name, next->name);
741				return -EINVAL;
742			}
743			rb_erase(&next->subdir_node, &de->subdir);
744			de = next;
745			continue;
746		}
747		next = de->parent;
748		if (S_ISDIR(de->mode))
749			next->nlink--;
750		write_unlock(&proc_subdir_lock);
751
752		proc_entry_rundown(de);
753		if (de == root)
754			break;
755		pde_put(de);
756
757		write_lock(&proc_subdir_lock);
758		de = next;
759	}
760	pde_put(root);
761	return 0;
762}
763EXPORT_SYMBOL(remove_proc_subtree);
764
765void *proc_get_parent_data(const struct inode *inode)
766{
767	struct proc_dir_entry *de = PDE(inode);
768	return de->parent->data;
769}
770EXPORT_SYMBOL_GPL(proc_get_parent_data);
771
772void proc_remove(struct proc_dir_entry *de)
773{
774	if (de)
775		remove_proc_subtree(de->name, de->parent);
776}
777EXPORT_SYMBOL(proc_remove);
778
779void *PDE_DATA(const struct inode *inode)
780{
781	return __PDE_DATA(inode);
782}
783EXPORT_SYMBOL(PDE_DATA);
784
785/*
786 * Pull a user buffer into memory and pass it to the file's write handler if
787 * one is supplied.  The ->write() method is permitted to modify the
788 * kernel-side buffer.
789 */
790ssize_t proc_simple_write(struct file *f, const char __user *ubuf, size_t size,
791			  loff_t *_pos)
792{
793	struct proc_dir_entry *pde = PDE(file_inode(f));
794	char *buf;
795	int ret;
796
797	if (!pde->write)
798		return -EACCES;
799	if (size == 0 || size > PAGE_SIZE - 1)
800		return -EINVAL;
801	buf = memdup_user_nul(ubuf, size);
802	if (IS_ERR(buf))
803		return PTR_ERR(buf);
804	ret = pde->write(f, buf, size);
805	kfree(buf);
806	return ret == 0 ? size : ret;
807}