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;
169 de = *ret ?: &proc_root;
170 while ((next = strchr(cp, '/')) != NULL) {
171 de = pde_subdir_find(de, cp, next - cp);
173 WARN(1, "name '%s'\n", name);
183 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
184 const char **residual)
188 read_lock(&proc_subdir_lock);
189 rv = __xlate_proc_name(name, ret, residual);
190 read_unlock(&proc_subdir_lock);
194 static DEFINE_IDA(proc_inum_ida);
196 #define PROC_DYNAMIC_FIRST 0xF0000000U
199 * Return an inode number between PROC_DYNAMIC_FIRST and
200 * 0xffffffff, or zero on failure.
202 int proc_alloc_inum(unsigned int *inum)
206 i = ida_simple_get(&proc_inum_ida, 0, UINT_MAX - PROC_DYNAMIC_FIRST + 1,
211 *inum = PROC_DYNAMIC_FIRST + (unsigned int)i;
215 void proc_free_inum(unsigned int inum)
217 ida_simple_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
220 static int proc_misc_d_revalidate(struct dentry *dentry, unsigned int flags)
222 if (flags & LOOKUP_RCU)
225 if (atomic_read(&PDE(d_inode(dentry))->in_use) < 0)
226 return 0; /* revalidate */
230 static int proc_misc_d_delete(const struct dentry *dentry)
232 return atomic_read(&PDE(d_inode(dentry))->in_use) < 0;
235 static const struct dentry_operations proc_misc_dentry_ops = {
236 .d_revalidate = proc_misc_d_revalidate,
237 .d_delete = proc_misc_d_delete,
241 * Don't create negative dentries here, return -ENOENT by hand
244 struct dentry *proc_lookup_de(struct inode *dir, struct dentry *dentry,
245 struct proc_dir_entry *de)
249 read_lock(&proc_subdir_lock);
250 de = pde_subdir_find(de, dentry->d_name.name, dentry->d_name.len);
253 read_unlock(&proc_subdir_lock);
254 inode = proc_get_inode(dir->i_sb, de);
256 return ERR_PTR(-ENOMEM);
257 d_set_d_op(dentry, de->proc_dops);
258 return d_splice_alias(inode, dentry);
260 read_unlock(&proc_subdir_lock);
261 return ERR_PTR(-ENOENT);
264 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
267 struct proc_fs_info *fs_info = proc_sb_info(dir->i_sb);
269 if (fs_info->pidonly == PROC_PIDONLY_ON)
270 return ERR_PTR(-ENOENT);
272 return proc_lookup_de(dir, dentry, PDE(dir));
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..
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
284 int proc_readdir_de(struct file *file, struct dir_context *ctx,
285 struct proc_dir_entry *de)
289 if (!dir_emit_dots(file, ctx))
293 read_lock(&proc_subdir_lock);
294 de = pde_subdir_first(de);
297 read_unlock(&proc_subdir_lock);
302 de = pde_subdir_next(de);
307 struct proc_dir_entry *next;
309 read_unlock(&proc_subdir_lock);
310 if (!dir_emit(ctx, de->name, de->namelen,
311 de->low_ino, de->mode >> 12)) {
316 read_lock(&proc_subdir_lock);
317 next = pde_subdir_next(de);
321 read_unlock(&proc_subdir_lock);
325 int proc_readdir(struct file *file, struct dir_context *ctx)
327 struct inode *inode = file_inode(file);
328 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
330 if (fs_info->pidonly == PROC_PIDONLY_ON)
333 return proc_readdir_de(file, ctx, PDE(inode));
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.
341 static const struct file_operations proc_dir_operations = {
342 .llseek = generic_file_llseek,
343 .read = generic_read_dir,
344 .iterate_shared = proc_readdir,
347 static int proc_net_d_revalidate(struct dentry *dentry, unsigned int flags)
352 const struct dentry_operations proc_net_dentry_ops = {
353 .d_revalidate = proc_net_d_revalidate,
354 .d_delete = always_delete_dentry,
358 * proc directories can do almost nothing..
360 static const struct inode_operations proc_dir_inode_operations = {
361 .lookup = proc_lookup,
362 .getattr = proc_getattr,
363 .setattr = proc_notify_change,
366 /* returns the registered entry, or frees dp and returns NULL on failure */
367 struct proc_dir_entry *proc_register(struct proc_dir_entry *dir,
368 struct proc_dir_entry *dp)
370 if (proc_alloc_inum(&dp->low_ino))
373 write_lock(&proc_subdir_lock);
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);
382 write_unlock(&proc_subdir_lock);
386 proc_free_inum(dp->low_ino);
392 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
397 struct proc_dir_entry *ent = NULL;
401 if (xlate_proc_name(name, parent, &fn) != 0)
404 qstr.len = strlen(fn);
405 if (qstr.len == 0 || qstr.len >= 256) {
406 WARN(1, "name len %u\n", qstr.len);
409 if (qstr.len == 1 && fn[0] == '.') {
410 WARN(1, "name '.'\n");
413 if (qstr.len == 2 && fn[0] == '.' && fn[1] == '.') {
414 WARN(1, "name '..'\n");
417 if (*parent == &proc_root && name_to_int(&qstr) != ~0U) {
418 WARN(1, "create '/proc/%s' by hand\n", qstr.name);
421 if (is_empty_pde(*parent)) {
422 WARN(1, "attempt to add to permanently empty directory");
426 ent = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL);
430 if (qstr.len + 1 <= SIZEOF_PDE_INLINE_NAME) {
431 ent->name = ent->inline_name;
433 ent->name = kmalloc(qstr.len + 1, GFP_KERNEL);
440 memcpy(ent->name, fn, qstr.len + 1);
441 ent->namelen = qstr.len;
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);
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);
459 struct proc_dir_entry *proc_symlink(const char *name,
460 struct proc_dir_entry *parent, const char *dest)
462 struct proc_dir_entry *ent;
464 ent = __proc_create(&parent, name,
465 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
468 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
470 strcpy((char*)ent->data,dest);
471 ent->proc_iops = &proc_link_inode_operations;
472 ent = proc_register(parent, ent);
480 EXPORT_SYMBOL(proc_symlink);
482 struct proc_dir_entry *_proc_mkdir(const char *name, umode_t mode,
483 struct proc_dir_entry *parent, void *data, bool force_lookup)
485 struct proc_dir_entry *ent;
488 mode = S_IRUGO | S_IXUGO;
490 ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
493 ent->proc_dir_ops = &proc_dir_operations;
494 ent->proc_iops = &proc_dir_inode_operations;
496 pde_force_lookup(ent);
498 ent = proc_register(parent, ent);
502 EXPORT_SYMBOL_GPL(_proc_mkdir);
504 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode,
505 struct proc_dir_entry *parent, void *data)
507 return _proc_mkdir(name, mode, parent, data, false);
509 EXPORT_SYMBOL_GPL(proc_mkdir_data);
511 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
512 struct proc_dir_entry *parent)
514 return proc_mkdir_data(name, mode, parent, NULL);
516 EXPORT_SYMBOL(proc_mkdir_mode);
518 struct proc_dir_entry *proc_mkdir(const char *name,
519 struct proc_dir_entry *parent)
521 return proc_mkdir_data(name, 0, parent, NULL);
523 EXPORT_SYMBOL(proc_mkdir);
525 struct proc_dir_entry *proc_create_mount_point(const char *name)
527 umode_t mode = S_IFDIR | S_IRUGO | S_IXUGO;
528 struct proc_dir_entry *ent, *parent = NULL;
530 ent = __proc_create(&parent, name, mode, 2);
533 ent->proc_dir_ops = NULL;
534 ent->proc_iops = NULL;
535 ent = proc_register(parent, ent);
539 EXPORT_SYMBOL(proc_create_mount_point);
541 struct proc_dir_entry *proc_create_reg(const char *name, umode_t mode,
542 struct proc_dir_entry **parent, void *data)
544 struct proc_dir_entry *p;
546 if ((mode & S_IFMT) == 0)
548 if ((mode & S_IALLUGO) == 0)
550 if (WARN_ON_ONCE(!S_ISREG(mode)))
553 p = __proc_create(parent, name, mode, 1);
555 p->proc_iops = &proc_file_inode_operations;
561 static inline void pde_set_flags(struct proc_dir_entry *pde)
563 if (pde->proc_ops->proc_flags & PROC_ENTRY_PERMANENT)
564 pde->flags |= PROC_ENTRY_PERMANENT;
567 struct 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)
571 struct proc_dir_entry *p;
573 p = proc_create_reg(name, mode, &parent, data);
576 p->proc_ops = proc_ops;
578 return proc_register(parent, p);
580 EXPORT_SYMBOL(proc_create_data);
582 struct proc_dir_entry *proc_create(const char *name, umode_t mode,
583 struct proc_dir_entry *parent,
584 const struct proc_ops *proc_ops)
586 return proc_create_data(name, mode, parent, proc_ops, NULL);
588 EXPORT_SYMBOL(proc_create);
590 static int proc_seq_open(struct inode *inode, struct file *file)
592 struct proc_dir_entry *de = PDE(inode);
595 return seq_open_private(file, de->seq_ops, de->state_size);
596 return seq_open(file, de->seq_ops);
599 static int proc_seq_release(struct inode *inode, struct file *file)
601 struct proc_dir_entry *de = PDE(inode);
604 return seq_release_private(inode, file);
605 return seq_release(inode, file);
608 static 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,
616 struct 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)
620 struct proc_dir_entry *p;
622 p = proc_create_reg(name, mode, &parent, data);
625 p->proc_ops = &proc_seq_ops;
627 p->state_size = state_size;
628 return proc_register(parent, p);
630 EXPORT_SYMBOL(proc_create_seq_private);
632 static int proc_single_open(struct inode *inode, struct file *file)
634 struct proc_dir_entry *de = PDE(inode);
636 return single_open(file, de->single_show, de->data);
639 static 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,
647 struct 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)
651 struct proc_dir_entry *p;
653 p = proc_create_reg(name, mode, &parent, data);
656 p->proc_ops = &proc_single_ops;
657 p->single_show = show;
658 return proc_register(parent, p);
660 EXPORT_SYMBOL(proc_create_single_data);
662 void proc_set_size(struct proc_dir_entry *de, loff_t size)
666 EXPORT_SYMBOL(proc_set_size);
668 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid)
673 EXPORT_SYMBOL(proc_set_user);
675 void pde_put(struct proc_dir_entry *pde)
677 if (refcount_dec_and_test(&pde->refcnt)) {
678 proc_free_inum(pde->low_ino);
684 * Remove a /proc entry and free it if it's not currently in use.
686 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
688 struct proc_dir_entry *de = NULL;
689 const char *fn = name;
692 write_lock(&proc_subdir_lock);
693 if (__xlate_proc_name(name, &parent, &fn) != 0) {
694 write_unlock(&proc_subdir_lock);
699 de = pde_subdir_find(parent, fn, len);
701 if (unlikely(pde_is_permanent(de))) {
702 WARN(1, "removing permanent /proc entry '%s'", de->name);
705 rb_erase(&de->subdir_node, &parent->subdir);
706 if (S_ISDIR(de->mode))
710 write_unlock(&proc_subdir_lock);
712 WARN(1, "name '%s'\n", name);
716 proc_entry_rundown(de);
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);
723 EXPORT_SYMBOL(remove_proc_entry);
725 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent)
727 struct proc_dir_entry *root = NULL, *de, *next;
728 const char *fn = name;
731 write_lock(&proc_subdir_lock);
732 if (__xlate_proc_name(name, &parent, &fn) != 0) {
733 write_unlock(&proc_subdir_lock);
738 root = pde_subdir_find(parent, fn, len);
740 write_unlock(&proc_subdir_lock);
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);
749 rb_erase(&root->subdir_node, &parent->subdir);
753 next = pde_subdir_first(de);
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);
761 rb_erase(&next->subdir_node, &de->subdir);
766 if (S_ISDIR(de->mode))
768 write_unlock(&proc_subdir_lock);
770 proc_entry_rundown(de);
775 write_lock(&proc_subdir_lock);
781 EXPORT_SYMBOL(remove_proc_subtree);
783 void *proc_get_parent_data(const struct inode *inode)
785 struct proc_dir_entry *de = PDE(inode);
786 return de->parent->data;
788 EXPORT_SYMBOL_GPL(proc_get_parent_data);
790 void proc_remove(struct proc_dir_entry *de)
793 remove_proc_subtree(de->name, de->parent);
795 EXPORT_SYMBOL(proc_remove);
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.
802 ssize_t proc_simple_write(struct file *f, const char __user *ubuf, size_t size,
805 struct proc_dir_entry *pde = PDE(file_inode(f));
811 if (size == 0 || size > PAGE_SIZE - 1)
813 buf = memdup_user_nul(ubuf, size);
816 ret = pde->write(f, buf, size);
818 return ret == 0 ? size : ret;