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>
15 #include <linux/module.h>
16 #include <linux/mount.h>
17 #include <linux/smp_lock.h>
18 #include <linux/init.h>
19 #include <linux/idr.h>
20 #include <linux/namei.h>
21 #include <linux/bitops.h>
22 #include <asm/uaccess.h>
24 static ssize_t proc_file_read(struct file *file, char __user *buf,
25 size_t nbytes, loff_t *ppos);
26 static ssize_t proc_file_write(struct file *file, const char __user *buffer,
27 size_t count, loff_t *ppos);
28 static loff_t proc_file_lseek(struct file *, loff_t, int);
30 int proc_match(int len, const char *name, struct proc_dir_entry *de)
32 if (de->namelen != len)
34 return !memcmp(name, de->name, len);
37 static struct file_operations proc_file_operations = {
38 .llseek = proc_file_lseek,
39 .read = proc_file_read,
40 .write = proc_file_write,
43 /* buffer size is one page but our output routines use some slack for overruns */
44 #define PROC_BLOCK_SIZE (PAGE_SIZE - 1024)
47 proc_file_read(struct file *file, char __user *buf, size_t nbytes,
50 struct inode * inode = file->f_dentry->d_inode;
56 struct proc_dir_entry * dp;
59 if (!(page = (char*) __get_free_page(GFP_KERNEL)))
62 while ((nbytes > 0) && !eof) {
63 count = min_t(size_t, PROC_BLOCK_SIZE, nbytes);
67 /* Handle old net routines */
68 n = dp->get_info(page, &start, *ppos, count);
71 } else if (dp->read_proc) {
73 * How to be a proc read function
74 * ------------------------------
76 * int f(char *buffer, char **start, off_t offset,
77 * int count, int *peof, void *dat)
79 * Assume that the buffer is "count" bytes in size.
81 * If you know you have supplied all the data you
84 * You have three ways to return data:
85 * 0) Leave *start = NULL. (This is the default.)
86 * Put the data of the requested offset at that
87 * offset within the buffer. Return the number (n)
88 * of bytes there are from the beginning of the
89 * buffer up to the last byte of data. If the
90 * number of supplied bytes (= n - offset) is
91 * greater than zero and you didn't signal eof
92 * and the reader is prepared to take more data
93 * you will be called again with the requested
94 * offset advanced by the number of bytes
95 * absorbed. This interface is useful for files
96 * no larger than the buffer.
97 * 1) Set *start = an unsigned long value less than
98 * the buffer address but greater than zero.
99 * Put the data of the requested offset at the
100 * beginning of the buffer. Return the number of
101 * bytes of data placed there. If this number is
102 * greater than zero and you didn't signal eof
103 * and the reader is prepared to take more data
104 * you will be called again with the requested
105 * offset advanced by *start. This interface is
106 * useful when you have a large file consisting
107 * of a series of blocks which you want to count
108 * and return as wholes.
109 * (Hack by Paul.Russell@rustcorp.com.au)
110 * 2) Set *start = an address within the buffer.
111 * Put the data of the requested offset at *start.
112 * Return the number of bytes of data placed there.
113 * If this number is greater than zero and you
114 * didn't signal eof and the reader is prepared to
115 * take more data you will be called again with the
116 * requested offset advanced by the number of bytes
119 n = dp->read_proc(page, &start, *ppos,
120 count, &eof, dp->data);
124 if (n == 0) /* end of file */
126 if (n < 0) { /* error */
135 "proc_file_read: Apparent buffer overflow!\n");
143 start = page + *ppos;
144 } else if (start < page) {
147 "proc_file_read: Apparent buffer overflow!\n");
152 * Don't reduce n because doing so might
153 * cut off part of a data block.
156 "proc_file_read: Read count exceeded\n");
158 } else /* start >= page */ {
159 unsigned long startoff = (unsigned long)(start - page);
160 if (n > (PAGE_SIZE - startoff)) {
162 "proc_file_read: Apparent buffer overflow!\n");
163 n = PAGE_SIZE - startoff;
169 n -= copy_to_user(buf, start < page ? page : start, n);
176 *ppos += start < page ? (unsigned long)start : n;
181 free_page((unsigned long) page);
186 proc_file_write(struct file *file, const char __user *buffer,
187 size_t count, loff_t *ppos)
189 struct inode *inode = file->f_dentry->d_inode;
190 struct proc_dir_entry * dp;
197 /* FIXME: does this routine need ppos? probably... */
198 return dp->write_proc(file, buffer, count, dp->data);
203 proc_file_lseek(struct file *file, loff_t offset, int orig)
211 file->f_pos = offset;
215 if (offset + file->f_pos < 0)
217 file->f_pos += offset;
231 static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
233 struct inode *inode = dentry->d_inode;
234 struct proc_dir_entry *de = PDE(inode);
237 error = inode_change_ok(inode, iattr);
241 error = inode_setattr(inode, iattr);
245 de->uid = inode->i_uid;
246 de->gid = inode->i_gid;
247 de->mode = inode->i_mode;
252 static struct inode_operations proc_file_inode_operations = {
253 .setattr = proc_notify_change,
257 * This function parses a name such as "tty/driver/serial", and
258 * returns the struct proc_dir_entry for "/proc/tty/driver", and
259 * returns "serial" in residual.
261 static int xlate_proc_name(const char *name,
262 struct proc_dir_entry **ret, const char **residual)
264 const char *cp = name, *next;
265 struct proc_dir_entry *de;
270 next = strchr(cp, '/');
275 for (de = de->subdir; de ; de = de->next) {
276 if (proc_match(len, cp, de))
288 static DEFINE_IDR(proc_inum_idr);
289 static DEFINE_SPINLOCK(proc_inum_lock); /* protects the above */
291 #define PROC_DYNAMIC_FIRST 0xF0000000UL
294 * Return an inode number between PROC_DYNAMIC_FIRST and
295 * 0xffffffff, or zero on failure.
297 static unsigned int get_inode_number(void)
303 if (idr_pre_get(&proc_inum_idr, GFP_KERNEL) == 0)
306 spin_lock(&proc_inum_lock);
307 error = idr_get_new(&proc_inum_idr, NULL, &i);
308 spin_unlock(&proc_inum_lock);
309 if (error == -EAGAIN)
314 inum = (i & MAX_ID_MASK) + PROC_DYNAMIC_FIRST;
316 /* inum will never be more than 0xf0ffffff, so no check
323 static void release_inode_number(unsigned int inum)
325 int id = (inum - PROC_DYNAMIC_FIRST) | ~MAX_ID_MASK;
327 spin_lock(&proc_inum_lock);
328 idr_remove(&proc_inum_idr, id);
329 spin_unlock(&proc_inum_lock);
332 static int proc_follow_link(struct dentry *dentry, struct nameidata *nd)
334 nd_set_link(nd, PDE(dentry->d_inode)->data);
338 static struct inode_operations proc_link_inode_operations = {
339 .readlink = generic_readlink,
340 .follow_link = proc_follow_link,
344 * As some entries in /proc are volatile, we want to
345 * get rid of unused dentries. This could be made
346 * smarter: we could keep a "volatile" flag in the
347 * inode to indicate which ones to keep.
349 static int proc_delete_dentry(struct dentry * dentry)
354 static struct dentry_operations proc_dentry_operations =
356 .d_delete = proc_delete_dentry,
360 * Don't create negative dentries here, return -ENOENT by hand
363 struct dentry *proc_lookup(struct inode * dir, struct dentry *dentry, struct nameidata *nd)
365 struct inode *inode = NULL;
366 struct proc_dir_entry * de;
372 for (de = de->subdir; de ; de = de->next) {
373 if (de->namelen != dentry->d_name.len)
375 if (!memcmp(dentry->d_name.name, de->name, de->namelen)) {
376 unsigned int ino = de->low_ino;
379 inode = proc_get_inode(dir->i_sb, ino, de);
387 dentry->d_op = &proc_dentry_operations;
388 d_add(dentry, inode);
391 return ERR_PTR(error);
395 * This returns non-zero if at EOF, so that the /proc
396 * root directory can use this and check if it should
397 * continue with the <pid> entries..
399 * Note that the VFS-layer doesn't care about the return
400 * value of the readdir() call, as long as it's non-negative
403 int proc_readdir(struct file * filp,
404 void * dirent, filldir_t filldir)
406 struct proc_dir_entry * de;
409 struct inode *inode = filp->f_dentry->d_inode;
423 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
429 if (filldir(dirent, "..", 2, i,
430 parent_ino(filp->f_dentry),
451 if (filldir(dirent, de->name, de->namelen, filp->f_pos,
452 de->low_ino, de->mode >> 12) < 0)
459 out: unlock_kernel();
464 * These are the generic /proc directory operations. They
465 * use the in-memory "struct proc_dir_entry" tree to parse
466 * the /proc directory.
468 static struct file_operations proc_dir_operations = {
469 .read = generic_read_dir,
470 .readdir = proc_readdir,
474 * proc directories can do almost nothing..
476 static struct inode_operations proc_dir_inode_operations = {
477 .lookup = proc_lookup,
478 .setattr = proc_notify_change,
481 static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp)
485 i = get_inode_number();
489 dp->next = dir->subdir;
492 if (S_ISDIR(dp->mode)) {
493 if (dp->proc_iops == NULL) {
494 dp->proc_fops = &proc_dir_operations;
495 dp->proc_iops = &proc_dir_inode_operations;
498 } else if (S_ISLNK(dp->mode)) {
499 if (dp->proc_iops == NULL)
500 dp->proc_iops = &proc_link_inode_operations;
501 } else if (S_ISREG(dp->mode)) {
502 if (dp->proc_fops == NULL)
503 dp->proc_fops = &proc_file_operations;
504 if (dp->proc_iops == NULL)
505 dp->proc_iops = &proc_file_inode_operations;
511 * Kill an inode that got unregistered..
513 static void proc_kill_inodes(struct proc_dir_entry *de)
516 struct super_block *sb = proc_mnt->mnt_sb;
519 * Actually it's a partial revoke().
522 list_for_each(p, &sb->s_files) {
523 struct file * filp = list_entry(p, struct file, f_list);
524 struct dentry * dentry = filp->f_dentry;
525 struct inode * inode;
526 struct file_operations *fops;
528 if (dentry->d_op != &proc_dentry_operations)
530 inode = dentry->d_inode;
531 if (PDE(inode) != de)
540 static struct proc_dir_entry *proc_create(struct proc_dir_entry **parent,
545 struct proc_dir_entry *ent = NULL;
546 const char *fn = name;
549 /* make sure name is valid */
550 if (!name || !strlen(name)) goto out;
552 if (!(*parent) && xlate_proc_name(name, parent, &fn) != 0)
555 /* At this point there must not be any '/' characters beyond *fn */
561 ent = kmalloc(sizeof(struct proc_dir_entry) + len + 1, GFP_KERNEL);
564 memset(ent, 0, sizeof(struct proc_dir_entry));
565 memcpy(((char *) ent) + sizeof(struct proc_dir_entry), fn, len + 1);
566 ent->name = ((char *) ent) + sizeof(*ent);
574 struct proc_dir_entry *proc_symlink(const char *name,
575 struct proc_dir_entry *parent, const char *dest)
577 struct proc_dir_entry *ent;
579 ent = proc_create(&parent,name,
580 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
583 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
585 strcpy((char*)ent->data,dest);
586 if (proc_register(parent, ent) < 0) {
599 struct proc_dir_entry *proc_mkdir_mode(const char *name, mode_t mode,
600 struct proc_dir_entry *parent)
602 struct proc_dir_entry *ent;
604 ent = proc_create(&parent, name, S_IFDIR | mode, 2);
606 ent->proc_fops = &proc_dir_operations;
607 ent->proc_iops = &proc_dir_inode_operations;
609 if (proc_register(parent, ent) < 0) {
617 struct proc_dir_entry *proc_mkdir(const char *name,
618 struct proc_dir_entry *parent)
620 return proc_mkdir_mode(name, S_IRUGO | S_IXUGO, parent);
623 struct proc_dir_entry *create_proc_entry(const char *name, mode_t mode,
624 struct proc_dir_entry *parent)
626 struct proc_dir_entry *ent;
630 if ((mode & S_IALLUGO) == 0)
631 mode |= S_IRUGO | S_IXUGO;
634 if ((mode & S_IFMT) == 0)
636 if ((mode & S_IALLUGO) == 0)
641 ent = proc_create(&parent,name,mode,nlink);
644 ent->proc_fops = &proc_dir_operations;
645 ent->proc_iops = &proc_dir_inode_operations;
647 if (proc_register(parent, ent) < 0) {
655 void free_proc_entry(struct proc_dir_entry *de)
657 unsigned int ino = de->low_ino;
659 if (ino < PROC_DYNAMIC_FIRST)
662 release_inode_number(ino);
664 if (S_ISLNK(de->mode) && de->data)
670 * Remove a /proc entry and free it if it's not currently in use.
671 * If it is in use, we set the 'deleted' flag.
673 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
675 struct proc_dir_entry **p;
676 struct proc_dir_entry *de;
677 const char *fn = name;
680 if (!parent && xlate_proc_name(name, &parent, &fn) != 0)
683 for (p = &parent->subdir; *p; p=&(*p)->next ) {
684 if (!proc_match(len, fn, *p))
689 if (S_ISDIR(de->mode))
691 proc_kill_inodes(de);
694 if (!atomic_read(&de->count))
698 printk("remove_proc_entry: %s/%s busy, count=%d\n",
699 parent->name, de->name, atomic_read(&de->count));