2 * proc/fs/generic.c --- generic routines for the proc-fs
4 * This file contains generic proc-fs routines for handling
5 * directories and files.
7 * Copyright (C) 1991, 1992 Linus Torvalds.
8 * Copyright (C) 1997 Theodore Ts'o
11 #include <linux/errno.h>
12 #include <linux/time.h>
13 #include <linux/proc_fs.h>
14 #include <linux/stat.h>
16 #include <linux/module.h>
17 #include <linux/slab.h>
18 #include <linux/mount.h>
19 #include <linux/init.h>
20 #include <linux/idr.h>
21 #include <linux/namei.h>
22 #include <linux/bitops.h>
23 #include <linux/spinlock.h>
24 #include <linux/completion.h>
25 #include <asm/uaccess.h>
29 DEFINE_SPINLOCK(proc_subdir_lock);
31 static int proc_match(unsigned int len, const char *name, struct proc_dir_entry *de)
33 if (de->namelen != len)
35 return !memcmp(name, de->name, len);
38 /* buffer size is one page but our output routines use some slack for overruns */
39 #define PROC_BLOCK_SIZE (PAGE_SIZE - 1024)
42 __proc_file_read(struct file *file, char __user *buf, size_t nbytes,
45 struct inode * inode = file->f_path.dentry->d_inode;
51 struct proc_dir_entry * dp;
52 unsigned long long pos;
55 * Gaah, please just use "seq_file" instead. The legacy /proc
56 * interfaces cut loff_t down to off_t for reads, and ignore
57 * the offset entirely for writes..
60 if (pos > MAX_NON_LFS)
62 if (nbytes > MAX_NON_LFS - pos)
63 nbytes = MAX_NON_LFS - pos;
66 if (!(page = (char*) __get_free_page(GFP_TEMPORARY)))
69 while ((nbytes > 0) && !eof) {
70 count = min_t(size_t, PROC_BLOCK_SIZE, nbytes);
75 * How to be a proc read function
76 * ------------------------------
78 * int f(char *buffer, char **start, off_t offset,
79 * int count, int *peof, void *dat)
81 * Assume that the buffer is "count" bytes in size.
83 * If you know you have supplied all the data you
86 * You have three ways to return data:
87 * 0) Leave *start = NULL. (This is the default.)
88 * Put the data of the requested offset at that
89 * offset within the buffer. Return the number (n)
90 * of bytes there are from the beginning of the
91 * buffer up to the last byte of data. If the
92 * number of supplied bytes (= n - offset) is
93 * greater than zero and you didn't signal eof
94 * and the reader is prepared to take more data
95 * you will be called again with the requested
96 * offset advanced by the number of bytes
97 * absorbed. This interface is useful for files
98 * no larger than the buffer.
99 * 1) Set *start = an unsigned long value less than
100 * the buffer address but greater than zero.
101 * Put the data of the requested offset at the
102 * beginning of the buffer. Return the number of
103 * bytes of data placed there. If this number is
104 * greater than zero and you didn't signal eof
105 * and the reader is prepared to take more data
106 * you will be called again with the requested
107 * offset advanced by *start. This interface is
108 * useful when you have a large file consisting
109 * of a series of blocks which you want to count
110 * and return as wholes.
111 * (Hack by Paul.Russell@rustcorp.com.au)
112 * 2) Set *start = an address within the buffer.
113 * Put the data of the requested offset at *start.
114 * Return the number of bytes of data placed there.
115 * If this number is greater than zero and you
116 * didn't signal eof and the reader is prepared to
117 * take more data you will be called again with the
118 * requested offset advanced by the number of bytes
121 n = dp->read_proc(page, &start, *ppos,
122 count, &eof, dp->data);
126 if (n == 0) /* end of file */
128 if (n < 0) { /* error */
137 "proc_file_read: Apparent buffer overflow!\n");
145 start = page + *ppos;
146 } else if (start < page) {
149 "proc_file_read: Apparent buffer overflow!\n");
154 * Don't reduce n because doing so might
155 * cut off part of a data block.
158 "proc_file_read: Read count exceeded\n");
160 } else /* start >= page */ {
161 unsigned long startoff = (unsigned long)(start - page);
162 if (n > (PAGE_SIZE - startoff)) {
164 "proc_file_read: Apparent buffer overflow!\n");
165 n = PAGE_SIZE - startoff;
171 n -= copy_to_user(buf, start < page ? page : start, n);
178 *ppos += start < page ? (unsigned long)start : n;
183 free_page((unsigned long) page);
188 proc_file_read(struct file *file, char __user *buf, size_t nbytes,
191 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
194 spin_lock(&pde->pde_unload_lock);
195 if (!pde->proc_fops) {
196 spin_unlock(&pde->pde_unload_lock);
200 spin_unlock(&pde->pde_unload_lock);
202 rv = __proc_file_read(file, buf, nbytes, ppos);
209 proc_file_write(struct file *file, const char __user *buffer,
210 size_t count, loff_t *ppos)
212 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
215 if (pde->write_proc) {
216 spin_lock(&pde->pde_unload_lock);
217 if (!pde->proc_fops) {
218 spin_unlock(&pde->pde_unload_lock);
222 spin_unlock(&pde->pde_unload_lock);
224 /* FIXME: does this routine need ppos? probably... */
225 rv = pde->write_proc(file, buffer, count, pde->data);
233 proc_file_lseek(struct file *file, loff_t offset, int orig)
235 loff_t retval = -EINVAL;
238 offset += file->f_pos;
241 if (offset < 0 || offset > MAX_NON_LFS)
243 file->f_pos = retval = offset;
248 static const struct file_operations proc_file_operations = {
249 .llseek = proc_file_lseek,
250 .read = proc_file_read,
251 .write = proc_file_write,
254 static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
256 struct inode *inode = dentry->d_inode;
257 struct proc_dir_entry *de = PDE(inode);
260 error = inode_change_ok(inode, iattr);
264 setattr_copy(inode, iattr);
265 mark_inode_dirty(inode);
267 de->uid = inode->i_uid;
268 de->gid = inode->i_gid;
269 de->mode = inode->i_mode;
273 static int proc_getattr(struct vfsmount *mnt, struct dentry *dentry,
276 struct inode *inode = dentry->d_inode;
277 struct proc_dir_entry *de = PROC_I(inode)->pde;
279 set_nlink(inode, de->nlink);
281 generic_fillattr(inode, stat);
285 static const struct inode_operations proc_file_inode_operations = {
286 .setattr = proc_notify_change,
290 * This function parses a name such as "tty/driver/serial", and
291 * returns the struct proc_dir_entry for "/proc/tty/driver", and
292 * returns "serial" in residual.
294 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret,
295 const char **residual)
297 const char *cp = name, *next;
298 struct proc_dir_entry *de;
306 next = strchr(cp, '/');
311 for (de = de->subdir; de ; de = de->next) {
312 if (proc_match(len, cp, de))
316 WARN(1, "name '%s'\n", name);
326 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret,
327 const char **residual)
331 spin_lock(&proc_subdir_lock);
332 rv = __xlate_proc_name(name, ret, residual);
333 spin_unlock(&proc_subdir_lock);
337 static DEFINE_IDA(proc_inum_ida);
338 static DEFINE_SPINLOCK(proc_inum_lock); /* protects the above */
340 #define PROC_DYNAMIC_FIRST 0xF0000000U
343 * Return an inode number between PROC_DYNAMIC_FIRST and
344 * 0xffffffff, or zero on failure.
346 int proc_alloc_inum(unsigned int *inum)
352 if (!ida_pre_get(&proc_inum_ida, GFP_KERNEL))
355 spin_lock_bh(&proc_inum_lock);
356 error = ida_get_new(&proc_inum_ida, &i);
357 spin_unlock_bh(&proc_inum_lock);
358 if (error == -EAGAIN)
363 if (i > UINT_MAX - PROC_DYNAMIC_FIRST) {
364 spin_lock_bh(&proc_inum_lock);
365 ida_remove(&proc_inum_ida, i);
366 spin_unlock_bh(&proc_inum_lock);
369 *inum = PROC_DYNAMIC_FIRST + i;
373 void proc_free_inum(unsigned int inum)
375 spin_lock_bh(&proc_inum_lock);
376 ida_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
377 spin_unlock_bh(&proc_inum_lock);
380 static void *proc_follow_link(struct dentry *dentry, struct nameidata *nd)
382 nd_set_link(nd, PDE(dentry->d_inode)->data);
386 static const struct inode_operations proc_link_inode_operations = {
387 .readlink = generic_readlink,
388 .follow_link = proc_follow_link,
392 * As some entries in /proc are volatile, we want to
393 * get rid of unused dentries. This could be made
394 * smarter: we could keep a "volatile" flag in the
395 * inode to indicate which ones to keep.
397 static int proc_delete_dentry(const struct dentry * dentry)
402 static const struct dentry_operations proc_dentry_operations =
404 .d_delete = proc_delete_dentry,
408 * Don't create negative dentries here, return -ENOENT by hand
411 struct dentry *proc_lookup_de(struct proc_dir_entry *de, struct inode *dir,
412 struct dentry *dentry)
414 struct inode *inode = NULL;
417 spin_lock(&proc_subdir_lock);
418 for (de = de->subdir; de ; de = de->next) {
419 if (de->namelen != dentry->d_name.len)
421 if (!memcmp(dentry->d_name.name, de->name, de->namelen)) {
423 spin_unlock(&proc_subdir_lock);
425 inode = proc_get_inode(dir->i_sb, de);
429 spin_unlock(&proc_subdir_lock);
433 d_set_d_op(dentry, &proc_dentry_operations);
434 d_add(dentry, inode);
439 return ERR_PTR(error);
442 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
445 return proc_lookup_de(PDE(dir), dir, dentry);
449 * This returns non-zero if at EOF, so that the /proc
450 * root directory can use this and check if it should
451 * continue with the <pid> entries..
453 * Note that the VFS-layer doesn't care about the return
454 * value of the readdir() call, as long as it's non-negative
457 int proc_readdir_de(struct proc_dir_entry *de, struct file *filp, void *dirent,
462 struct inode *inode = filp->f_path.dentry->d_inode;
469 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
475 if (filldir(dirent, "..", 2, i,
476 parent_ino(filp->f_path.dentry),
483 spin_lock(&proc_subdir_lock);
489 spin_unlock(&proc_subdir_lock);
499 struct proc_dir_entry *next;
501 /* filldir passes info to user space */
503 spin_unlock(&proc_subdir_lock);
504 if (filldir(dirent, de->name, de->namelen, filp->f_pos,
505 de->low_ino, de->mode >> 12) < 0) {
509 spin_lock(&proc_subdir_lock);
515 spin_unlock(&proc_subdir_lock);
522 int proc_readdir(struct file *filp, void *dirent, filldir_t filldir)
524 struct inode *inode = filp->f_path.dentry->d_inode;
526 return proc_readdir_de(PDE(inode), filp, dirent, filldir);
530 * These are the generic /proc directory operations. They
531 * use the in-memory "struct proc_dir_entry" tree to parse
532 * the /proc directory.
534 static const struct file_operations proc_dir_operations = {
535 .llseek = generic_file_llseek,
536 .read = generic_read_dir,
537 .readdir = proc_readdir,
541 * proc directories can do almost nothing..
543 static const struct inode_operations proc_dir_inode_operations = {
544 .lookup = proc_lookup,
545 .getattr = proc_getattr,
546 .setattr = proc_notify_change,
549 static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp)
551 struct proc_dir_entry *tmp;
554 ret = proc_alloc_inum(&dp->low_ino);
558 if (S_ISDIR(dp->mode)) {
559 if (dp->proc_iops == NULL) {
560 dp->proc_fops = &proc_dir_operations;
561 dp->proc_iops = &proc_dir_inode_operations;
564 } else if (S_ISLNK(dp->mode)) {
565 if (dp->proc_iops == NULL)
566 dp->proc_iops = &proc_link_inode_operations;
567 } else if (S_ISREG(dp->mode)) {
568 if (dp->proc_fops == NULL)
569 dp->proc_fops = &proc_file_operations;
570 if (dp->proc_iops == NULL)
571 dp->proc_iops = &proc_file_inode_operations;
574 spin_lock(&proc_subdir_lock);
576 for (tmp = dir->subdir; tmp; tmp = tmp->next)
577 if (strcmp(tmp->name, dp->name) == 0) {
578 WARN(1, KERN_WARNING "proc_dir_entry '%s/%s' already registered\n",
579 dir->name, dp->name);
583 dp->next = dir->subdir;
586 spin_unlock(&proc_subdir_lock);
591 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
596 struct proc_dir_entry *ent = NULL;
597 const char *fn = name;
600 /* make sure name is valid */
601 if (!name || !strlen(name))
604 if (xlate_proc_name(name, parent, &fn) != 0)
607 /* At this point there must not be any '/' characters beyond *fn */
613 ent = kzalloc(sizeof(struct proc_dir_entry) + len + 1, GFP_KERNEL);
617 memcpy(ent->name, fn, len + 1);
621 atomic_set(&ent->count, 1);
622 spin_lock_init(&ent->pde_unload_lock);
623 INIT_LIST_HEAD(&ent->pde_openers);
628 struct proc_dir_entry *proc_symlink(const char *name,
629 struct proc_dir_entry *parent, const char *dest)
631 struct proc_dir_entry *ent;
633 ent = __proc_create(&parent, name,
634 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
637 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
639 strcpy((char*)ent->data,dest);
640 if (proc_register(parent, ent) < 0) {
652 EXPORT_SYMBOL(proc_symlink);
654 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode,
655 struct proc_dir_entry *parent)
657 struct proc_dir_entry *ent;
659 ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
661 if (proc_register(parent, ent) < 0) {
668 EXPORT_SYMBOL(proc_mkdir_mode);
670 struct proc_dir_entry *proc_net_mkdir(struct net *net, const char *name,
671 struct proc_dir_entry *parent)
673 struct proc_dir_entry *ent;
675 ent = __proc_create(&parent, name, S_IFDIR | S_IRUGO | S_IXUGO, 2);
678 if (proc_register(parent, ent) < 0) {
685 EXPORT_SYMBOL_GPL(proc_net_mkdir);
687 struct proc_dir_entry *proc_mkdir(const char *name,
688 struct proc_dir_entry *parent)
690 return proc_mkdir_mode(name, S_IRUGO | S_IXUGO, parent);
692 EXPORT_SYMBOL(proc_mkdir);
694 struct proc_dir_entry *create_proc_entry(const char *name, umode_t mode,
695 struct proc_dir_entry *parent)
697 struct proc_dir_entry *ent;
701 if ((mode & S_IALLUGO) == 0)
702 mode |= S_IRUGO | S_IXUGO;
705 if ((mode & S_IFMT) == 0)
707 if ((mode & S_IALLUGO) == 0)
712 ent = __proc_create(&parent, name, mode, nlink);
714 if (proc_register(parent, ent) < 0) {
721 EXPORT_SYMBOL(create_proc_entry);
723 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode,
724 struct proc_dir_entry *parent,
725 const struct file_operations *proc_fops,
728 struct proc_dir_entry *pde;
732 if ((mode & S_IALLUGO) == 0)
733 mode |= S_IRUGO | S_IXUGO;
736 if ((mode & S_IFMT) == 0)
738 if ((mode & S_IALLUGO) == 0)
743 pde = __proc_create(&parent, name, mode, nlink);
746 pde->proc_fops = proc_fops;
748 if (proc_register(parent, pde) < 0)
756 EXPORT_SYMBOL(proc_create_data);
758 static void free_proc_entry(struct proc_dir_entry *de)
760 proc_free_inum(de->low_ino);
762 if (S_ISLNK(de->mode))
767 void pde_put(struct proc_dir_entry *pde)
769 if (atomic_dec_and_test(&pde->count))
770 free_proc_entry(pde);
774 * Remove a /proc entry and free it if it's not currently in use.
776 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
778 struct proc_dir_entry **p;
779 struct proc_dir_entry *de = NULL;
780 const char *fn = name;
783 spin_lock(&proc_subdir_lock);
784 if (__xlate_proc_name(name, &parent, &fn) != 0) {
785 spin_unlock(&proc_subdir_lock);
790 for (p = &parent->subdir; *p; p=&(*p)->next ) {
791 if (proc_match(len, fn, *p)) {
798 spin_unlock(&proc_subdir_lock);
800 WARN(1, "name '%s'\n", name);
804 spin_lock(&de->pde_unload_lock);
806 * Stop accepting new callers into module. If you're
807 * dynamically allocating ->proc_fops, save a pointer somewhere.
809 de->proc_fops = NULL;
810 /* Wait until all existing callers into module are done. */
811 if (de->pde_users > 0) {
812 DECLARE_COMPLETION_ONSTACK(c);
814 if (!de->pde_unload_completion)
815 de->pde_unload_completion = &c;
817 spin_unlock(&de->pde_unload_lock);
819 wait_for_completion(de->pde_unload_completion);
821 spin_lock(&de->pde_unload_lock);
824 while (!list_empty(&de->pde_openers)) {
825 struct pde_opener *pdeo;
827 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
829 spin_unlock(&de->pde_unload_lock);
830 pdeo->release(pdeo->inode, pdeo->file);
832 spin_lock(&de->pde_unload_lock);
834 spin_unlock(&de->pde_unload_lock);
836 if (S_ISDIR(de->mode))
839 WARN(de->subdir, KERN_WARNING "%s: removing non-empty directory "
840 "'%s/%s', leaking at least '%s'\n", __func__,
841 de->parent->name, de->name, de->subdir->name);
844 EXPORT_SYMBOL(remove_proc_entry);