1 // SPDX-License-Identifier: GPL-2.0-only
3 * proc/fs/generic.c --- generic routines for the proc-fs
5 * This file contains generic proc-fs routines for handling
6 * directories and files.
8 * Copyright (C) 1991, 1992 Linus Torvalds.
9 * Copyright (C) 1997 Theodore Ts'o
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>
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>
33 static DEFINE_RWLOCK(proc_subdir_lock);
35 struct kmem_cache *proc_dir_entry_cache __ro_after_init;
37 void pde_free(struct proc_dir_entry *pde)
39 if (S_ISLNK(pde->mode))
41 if (pde->name != pde->inline_name)
43 kmem_cache_free(proc_dir_entry_cache, pde);
46 static int proc_match(const char *name, struct proc_dir_entry *de, unsigned int len)
48 if (len < de->namelen)
50 if (len > de->namelen)
53 return memcmp(name, de->name, len);
56 static struct proc_dir_entry *pde_subdir_first(struct proc_dir_entry *dir)
58 return rb_entry_safe(rb_first(&dir->subdir), struct proc_dir_entry,
62 static struct proc_dir_entry *pde_subdir_next(struct proc_dir_entry *dir)
64 return rb_entry_safe(rb_next(&dir->subdir_node), struct proc_dir_entry,
68 static struct proc_dir_entry *pde_subdir_find(struct proc_dir_entry *dir,
72 struct rb_node *node = dir->subdir.rb_node;
75 struct proc_dir_entry *de = rb_entry(node,
76 struct proc_dir_entry,
78 int result = proc_match(name, de, len);
83 node = node->rb_right;
90 static bool pde_subdir_insert(struct proc_dir_entry *dir,
91 struct proc_dir_entry *de)
93 struct rb_root *root = &dir->subdir;
94 struct rb_node **new = &root->rb_node, *parent = NULL;
96 /* Figure out where to put new node */
98 struct proc_dir_entry *this = rb_entry(*new,
99 struct proc_dir_entry,
101 int result = proc_match(de->name, this, de->namelen);
105 new = &(*new)->rb_left;
107 new = &(*new)->rb_right;
112 /* Add new node and rebalance tree. */
113 rb_link_node(&de->subdir_node, parent, new);
114 rb_insert_color(&de->subdir_node, root);
118 static int proc_notify_change(struct user_namespace *mnt_userns,
119 struct dentry *dentry, struct iattr *iattr)
121 struct inode *inode = d_inode(dentry);
122 struct proc_dir_entry *de = PDE(inode);
125 error = setattr_prepare(&init_user_ns, dentry, iattr);
129 setattr_copy(&init_user_ns, inode, iattr);
130 mark_inode_dirty(inode);
132 proc_set_user(de, inode->i_uid, inode->i_gid);
133 de->mode = inode->i_mode;
137 static int proc_getattr(struct user_namespace *mnt_userns,
138 const struct path *path, struct kstat *stat,
139 u32 request_mask, unsigned int query_flags)
141 struct inode *inode = d_inode(path->dentry);
142 struct proc_dir_entry *de = PDE(inode);
144 nlink_t nlink = READ_ONCE(de->nlink);
146 set_nlink(inode, nlink);
150 generic_fillattr(&init_user_ns, inode, stat);
154 static const struct inode_operations proc_file_inode_operations = {
155 .setattr = proc_notify_change,
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.
163 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
164 const char **residual)
166 const char *cp = name, *next;
167 struct proc_dir_entry *de;
174 next = strchr(cp, '/');
178 de = pde_subdir_find(de, cp, next - cp);
180 WARN(1, "name '%s'\n", name);
190 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
191 const char **residual)
195 read_lock(&proc_subdir_lock);
196 rv = __xlate_proc_name(name, ret, residual);
197 read_unlock(&proc_subdir_lock);
201 static DEFINE_IDA(proc_inum_ida);
203 #define PROC_DYNAMIC_FIRST 0xF0000000U
206 * Return an inode number between PROC_DYNAMIC_FIRST and
207 * 0xffffffff, or zero on failure.
209 int proc_alloc_inum(unsigned int *inum)
213 i = ida_simple_get(&proc_inum_ida, 0, UINT_MAX - PROC_DYNAMIC_FIRST + 1,
218 *inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
222 void proc_free_inum(unsigned int inum)
224 ida_simple_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
227 static int proc_misc_d_revalidate(struct dentry *dentry, unsigned int flags)
229 if (flags & LOOKUP_RCU)
232 if (atomic_read(&PDE(d_inode(dentry))->in_use) < 0)
233 return 0; /* revalidate */
237 static int proc_misc_d_delete(const struct dentry *dentry)
239 return atomic_read(&PDE(d_inode(dentry))->in_use) < 0;
242 static const struct dentry_operations proc_misc_dentry_ops = {
243 .d_revalidate = proc_misc_d_revalidate,
244 .d_delete = proc_misc_d_delete,
248 * Don't create negative dentries here, return -ENOENT by hand
251 struct dentry *proc_lookup_de(struct inode *dir, struct dentry *dentry,
252 struct proc_dir_entry *de)
256 read_lock(&proc_subdir_lock);
257 de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
260 read_unlock(&proc_subdir_lock);
261 inode = proc_get_inode(dir->i_sb, de);
263 return ERR_PTR(-ENOMEM);
264 d_set_d_op(dentry, de->proc_dops);
265 return d_splice_alias(inode, dentry);
267 read_unlock(&proc_subdir_lock);
268 return ERR_PTR(-ENOENT);
271 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
274 struct proc_fs_info *fs_info = proc_sb_info(dir->i_sb);
276 if (fs_info->pidonly == PROC_PIDONLY_ON)
277 return ERR_PTR(-ENOENT);
279 return proc_lookup_de(dir, dentry, PDE(dir));
283 * This returns non-zero if at EOF, so that the /proc
284 * root directory can use this and check if it should
285 * continue with the <pid> entries..
287 * Note that the VFS-layer doesn't care about the return
288 * value of the readdir() call, as long as it's non-negative
291 int proc_readdir_de(struct file *file, struct dir_context *ctx,
292 struct proc_dir_entry *de)
296 if (!dir_emit_dots(file, ctx))
300 read_lock(&proc_subdir_lock);
301 de = pde_subdir_first(de);
304 read_unlock(&proc_subdir_lock);
309 de = pde_subdir_next(de);
314 struct proc_dir_entry *next;
316 read_unlock(&proc_subdir_lock);
317 if (!dir_emit(ctx, de->name, de->namelen,
318 de->low_ino, de->mode >> 12)) {
323 read_lock(&proc_subdir_lock);
324 next = pde_subdir_next(de);
328 read_unlock(&proc_subdir_lock);
332 int proc_readdir(struct file *file, struct dir_context *ctx)
334 struct inode *inode = file_inode(file);
335 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
337 if (fs_info->pidonly == PROC_PIDONLY_ON)
340 return proc_readdir_de(file, ctx, PDE(inode));
344 * These are the generic /proc directory operations. They
345 * use the in-memory "struct proc_dir_entry" tree to parse
346 * the /proc directory.
348 static const struct file_operations proc_dir_operations = {
349 .llseek = generic_file_llseek,
350 .read = generic_read_dir,
351 .iterate_shared = proc_readdir,
354 static int proc_net_d_revalidate(struct dentry *dentry, unsigned int flags)
359 const struct dentry_operations proc_net_dentry_ops = {
360 .d_revalidate = proc_net_d_revalidate,
361 .d_delete = always_delete_dentry,
365 * proc directories can do almost nothing..
367 static const struct inode_operations proc_dir_inode_operations = {
368 .lookup = proc_lookup,
369 .getattr = proc_getattr,
370 .setattr = proc_notify_change,
373 /* returns the registered entry, or frees dp and returns NULL on failure */
374 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
375 struct proc_dir_entry *dp)
377 if (proc_alloc_inum(&dp->low_ino))
380 write_lock(&proc_subdir_lock);
382 if (pde_subdir_insert(dir, dp) == false) {
383 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
384 dir->name, dp->name);
385 write_unlock(&proc_subdir_lock);
389 write_unlock(&proc_subdir_lock);
393 proc_free_inum(dp->low_ino);
399 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
404 struct proc_dir_entry *ent = NULL;
408 if (xlate_proc_name(name, parent, &fn) != 0)
411 qstr.len = strlen(fn);
412 if (qstr.len == 0 || qstr.len >= 256) {
413 WARN(1, "name len %u\n", qstr.len);
416 if (qstr.len == 1 && fn[0] == '.') {
417 WARN(1, "name '.'\n");
420 if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
421 WARN(1, "name '..'\n");
424 if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
425 WARN(1, "create '/proc/%s' by hand\n", qstr.name);
428 if (is_empty_pde(*parent)) {
429 WARN(1, "attempt to add to permanently empty directory");
433 ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
437 if (qstr.len + 1 <= SIZEOF_PDE_INLINE_NAME) {
438 ent->name = ent->inline_name;
440 ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
447 memcpy(ent->name, fn, qstr.len + 1);
448 ent->namelen = qstr.len;
451 ent->subdir = RB_ROOT;
452 refcount_set(&ent->refcnt, 1);
453 spin_lock_init(&ent->pde_unload_lock);
454 INIT_LIST_HEAD(&ent->pde_openers);
455 proc_set_user(ent, (*parent)->uid, (*parent)->gid);
457 ent->proc_dops = &proc_misc_dentry_ops;
463 struct proc_dir_entry *proc_symlink(const char *name,
464 struct proc_dir_entry *parent, const char *dest)
466 struct proc_dir_entry *ent;
468 ent = __proc_create(&parent, name,
469 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
472 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
474 strcpy((char*)ent->data,dest);
475 ent->proc_iops = &proc_link_inode_operations;
476 ent = proc_register(parent, ent);
484 EXPORT_SYMBOL(proc_symlink);
486 struct proc_dir_entry *_proc_mkdir(const char *name, umode_t mode,
487 struct proc_dir_entry *parent, void *data, bool force_lookup)
489 struct proc_dir_entry *ent;
492 mode = S_IRUGO | S_IXUGO;
494 ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
497 ent->proc_dir_ops = &proc_dir_operations;
498 ent->proc_iops = &proc_dir_inode_operations;
500 pde_force_lookup(ent);
502 ent = proc_register(parent, ent);
506 EXPORT_SYMBOL_GPL(_proc_mkdir);
508 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
509 struct proc_dir_entry *parent, void *data)
511 return _proc_mkdir(name, mode, parent, data, false);
513 EXPORT_SYMBOL_GPL(proc_mkdir_data);
515 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
516 struct proc_dir_entry *parent)
518 return proc_mkdir_data(name, mode, parent, NULL);
520 EXPORT_SYMBOL(proc_mkdir_mode);
522 struct proc_dir_entry *proc_mkdir(const char *name,
523 struct proc_dir_entry *parent)
525 return proc_mkdir_data(name, 0, parent, NULL);
527 EXPORT_SYMBOL(proc_mkdir);
529 struct proc_dir_entry *proc_create_mount_point(const char *name)
531 umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
532 struct proc_dir_entry *ent, *parent = NULL;
534 ent = __proc_create(&parent, name, mode, 2);
537 ent->proc_dir_ops = NULL;
538 ent->proc_iops = NULL;
539 ent = proc_register(parent, ent);
543 EXPORT_SYMBOL(proc_create_mount_point);
545 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
546 struct proc_dir_entry **parent, void *data)
548 struct proc_dir_entry *p;
550 if ((mode & S_IFMT) == 0)
552 if ((mode & S_IALLUGO) == 0)
554 if (WARN_ON_ONCE(!S_ISREG(mode)))
557 p = __proc_create(parent, name, mode, 1);
559 p->proc_iops = &proc_file_inode_operations;
565 static inline void pde_set_flags(struct proc_dir_entry *pde)
567 if (pde->proc_ops->proc_flags & PROC_ENTRY_PERMANENT)
568 pde->flags |= PROC_ENTRY_PERMANENT;
571 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
572 struct proc_dir_entry *parent,
573 const struct proc_ops *proc_ops, void *data)
575 struct proc_dir_entry *p;
577 p = proc_create_reg(name, mode, &parent, data);
580 p->proc_ops = proc_ops;
582 return proc_register(parent, p);
584 EXPORT_SYMBOL(proc_create_data);
586 struct proc_dir_entry *proc_create(const char *name, umode_t mode,
587 struct proc_dir_entry *parent,
588 const struct proc_ops *proc_ops)
590 return proc_create_data(name, mode, parent, proc_ops, NULL);
592 EXPORT_SYMBOL(proc_create);
594 static int proc_seq_open(struct inode *inode, struct file *file)
596 struct proc_dir_entry *de = PDE(inode);
599 return seq_open_private(file, de->seq_ops, de->state_size);
600 return seq_open(file, de->seq_ops);
603 static int proc_seq_release(struct inode *inode, struct file *file)
605 struct proc_dir_entry *de = PDE(inode);
608 return seq_release_private(inode, file);
609 return seq_release(inode, file);
612 static const struct proc_ops proc_seq_ops = {
613 /* not permanent -- can call into arbitrary seq_operations */
614 .proc_open = proc_seq_open,
615 .proc_read_iter = seq_read_iter,
616 .proc_lseek = seq_lseek,
617 .proc_release = proc_seq_release,
620 struct proc_dir_entry *proc_create_seq_private(const char *name, umode_t mode,
621 struct proc_dir_entry *parent, const struct seq_operations *ops,
622 unsigned int state_size, void *data)
624 struct proc_dir_entry *p;
626 p = proc_create_reg(name, mode, &parent, data);
629 p->proc_ops = &proc_seq_ops;
631 p->state_size = state_size;
632 return proc_register(parent, p);
634 EXPORT_SYMBOL(proc_create_seq_private);
636 static int proc_single_open(struct inode *inode, struct file *file)
638 struct proc_dir_entry *de = PDE(inode);
640 return single_open(file, de->single_show, de->data);
643 static const struct proc_ops proc_single_ops = {
644 /* not permanent -- can call into arbitrary ->single_show */
645 .proc_open = proc_single_open,
646 .proc_read_iter = seq_read_iter,
647 .proc_lseek = seq_lseek,
648 .proc_release = single_release,
651 struct proc_dir_entry *proc_create_single_data(const char *name, umode_t mode,
652 struct proc_dir_entry *parent,
653 int (*show)(struct seq_file *, void *), void *data)
655 struct proc_dir_entry *p;
657 p = proc_create_reg(name, mode, &parent, data);
660 p->proc_ops = &proc_single_ops;
661 p->single_show = show;
662 return proc_register(parent, p);
664 EXPORT_SYMBOL(proc_create_single_data);
666 void proc_set_size(struct proc_dir_entry *de, loff_t size)
670 EXPORT_SYMBOL(proc_set_size);
672 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
677 EXPORT_SYMBOL(proc_set_user);
679 void pde_put(struct proc_dir_entry *pde)
681 if (refcount_dec_and_test(&pde->refcnt)) {
682 proc_free_inum(pde->low_ino);
688 * Remove a /proc entry and free it if it's not currently in use.
690 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
692 struct proc_dir_entry *de = NULL;
693 const char *fn = name;
696 write_lock(&proc_subdir_lock);
697 if (__xlate_proc_name(name, &parent, &fn) != 0) {
698 write_unlock(&proc_subdir_lock);
703 de = pde_subdir_find(parent, fn, len);
705 if (unlikely(pde_is_permanent(de))) {
706 WARN(1, "removing permanent /proc entry '%s'", de->name);
709 rb_erase(&de->subdir_node, &parent->subdir);
710 if (S_ISDIR(de->mode))
714 write_unlock(&proc_subdir_lock);
716 WARN(1, "name '%s'\n", name);
720 proc_entry_rundown(de);
722 WARN(pde_subdir_first(de),
723 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
724 __func__, de->parent->name, de->name, pde_subdir_first(de)->name);
727 EXPORT_SYMBOL(remove_proc_entry);
729 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
731 struct proc_dir_entry *root = NULL, *de, *next;
732 const char *fn = name;
735 write_lock(&proc_subdir_lock);
736 if (__xlate_proc_name(name, &parent, &fn) != 0) {
737 write_unlock(&proc_subdir_lock);
742 root = pde_subdir_find(parent, fn, len);
744 write_unlock(&proc_subdir_lock);
747 if (unlikely(pde_is_permanent(root))) {
748 write_unlock(&proc_subdir_lock);
749 WARN(1, "removing permanent /proc entry '%s/%s'",
750 root->parent->name, root->name);
753 rb_erase(&root->subdir_node, &parent->subdir);
757 next = pde_subdir_first(de);
759 if (unlikely(pde_is_permanent(root))) {
760 write_unlock(&proc_subdir_lock);
761 WARN(1, "removing permanent /proc entry '%s/%s'",
762 next->parent->name, next->name);
765 rb_erase(&next->subdir_node, &de->subdir);
770 if (S_ISDIR(de->mode))
772 write_unlock(&proc_subdir_lock);
774 proc_entry_rundown(de);
779 write_lock(&proc_subdir_lock);
785 EXPORT_SYMBOL(remove_proc_subtree);
787 void *proc_get_parent_data(const struct inode *inode)
789 struct proc_dir_entry *de = PDE(inode);
790 return de->parent->data;
792 EXPORT_SYMBOL_GPL(proc_get_parent_data);
794 void proc_remove(struct proc_dir_entry *de)
797 remove_proc_subtree(de->name, de->parent);
799 EXPORT_SYMBOL(proc_remove);
801 void *PDE_DATA(const struct inode *inode)
803 return __PDE_DATA(inode);
805 EXPORT_SYMBOL(PDE_DATA);
808 * Pull a user buffer into memory and pass it to the file's write handler if
809 * one is supplied. The ->write() method is permitted to modify the
810 * kernel-side buffer.
812 ssize_t proc_simple_write(struct file *f, const char __user *ubuf, size_t size,
815 struct proc_dir_entry *pde = PDE(file_inode(f));
821 if (size == 0 || size > PAGE_SIZE - 1)
823 buf = memdup_user_nul(ubuf, size);
826 ret = pde->write(f, buf, size);
828 return ret == 0 ? size : ret;