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;
57 unsigned long long pos;
60 * Gaah, please just use "seq_file" instead. The legacy /proc
61 * interfaces cut loff_t down to off_t for reads, and ignore
62 * the offset entirely for writes..
65 if (pos > MAX_NON_LFS)
67 if (nbytes > MAX_NON_LFS - pos)
68 nbytes = MAX_NON_LFS - pos;
71 if (!(page = (char*) __get_free_page(GFP_KERNEL)))
74 while ((nbytes > 0) && !eof) {
75 count = min_t(size_t, PROC_BLOCK_SIZE, nbytes);
79 /* Handle old net routines */
80 n = dp->get_info(page, &start, *ppos, count);
83 } else if (dp->read_proc) {
85 * How to be a proc read function
86 * ------------------------------
88 * int f(char *buffer, char **start, off_t offset,
89 * int count, int *peof, void *dat)
91 * Assume that the buffer is "count" bytes in size.
93 * If you know you have supplied all the data you
96 * You have three ways to return data:
97 * 0) Leave *start = NULL. (This is the default.)
98 * Put the data of the requested offset at that
99 * offset within the buffer. Return the number (n)
100 * of bytes there are from the beginning of the
101 * buffer up to the last byte of data. If the
102 * number of supplied bytes (= n - offset) is
103 * greater than zero and you didn't signal eof
104 * and the reader is prepared to take more data
105 * you will be called again with the requested
106 * offset advanced by the number of bytes
107 * absorbed. This interface is useful for files
108 * no larger than the buffer.
109 * 1) Set *start = an unsigned long value less than
110 * the buffer address but greater than zero.
111 * Put the data of the requested offset at the
112 * beginning of the buffer. Return the number of
113 * bytes of data placed there. If this number is
114 * greater than zero and you didn't signal eof
115 * and the reader is prepared to take more data
116 * you will be called again with the requested
117 * offset advanced by *start. This interface is
118 * useful when you have a large file consisting
119 * of a series of blocks which you want to count
120 * and return as wholes.
121 * (Hack by Paul.Russell@rustcorp.com.au)
122 * 2) Set *start = an address within the buffer.
123 * Put the data of the requested offset at *start.
124 * Return the number of bytes of data placed there.
125 * If this number is greater than zero and you
126 * didn't signal eof and the reader is prepared to
127 * take more data you will be called again with the
128 * requested offset advanced by the number of bytes
131 n = dp->read_proc(page, &start, *ppos,
132 count, &eof, dp->data);
136 if (n == 0) /* end of file */
138 if (n < 0) { /* error */
147 "proc_file_read: Apparent buffer overflow!\n");
155 start = page + *ppos;
156 } else if (start < page) {
159 "proc_file_read: Apparent buffer overflow!\n");
164 * Don't reduce n because doing so might
165 * cut off part of a data block.
168 "proc_file_read: Read count exceeded\n");
170 } else /* start >= page */ {
171 unsigned long startoff = (unsigned long)(start - page);
172 if (n > (PAGE_SIZE - startoff)) {
174 "proc_file_read: Apparent buffer overflow!\n");
175 n = PAGE_SIZE - startoff;
181 n -= copy_to_user(buf, start < page ? page : start, n);
188 *ppos += start < page ? (unsigned long)start : n;
193 free_page((unsigned long) page);
198 proc_file_write(struct file *file, const char __user *buffer,
199 size_t count, loff_t *ppos)
201 struct inode *inode = file->f_dentry->d_inode;
202 struct proc_dir_entry * dp;
209 /* FIXME: does this routine need ppos? probably... */
210 return dp->write_proc(file, buffer, count, dp->data);
215 proc_file_lseek(struct file *file, loff_t offset, int orig)
217 loff_t retval = -EINVAL;
220 offset += file->f_pos;
223 if (offset < 0 || offset > MAX_NON_LFS)
225 file->f_pos = retval = offset;
230 static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
232 struct inode *inode = dentry->d_inode;
233 struct proc_dir_entry *de = PDE(inode);
236 error = inode_change_ok(inode, iattr);
240 error = inode_setattr(inode, iattr);
244 de->uid = inode->i_uid;
245 de->gid = inode->i_gid;
246 de->mode = inode->i_mode;
251 static int proc_getattr(struct vfsmount *mnt, struct dentry *dentry,
254 struct inode *inode = dentry->d_inode;
255 struct proc_dir_entry *de = PROC_I(inode)->pde;
257 inode->i_nlink = de->nlink;
259 generic_fillattr(inode, stat);
263 static struct inode_operations proc_file_inode_operations = {
264 .setattr = proc_notify_change,
268 * This function parses a name such as "tty/driver/serial", and
269 * returns the struct proc_dir_entry for "/proc/tty/driver", and
270 * returns "serial" in residual.
272 static int xlate_proc_name(const char *name,
273 struct proc_dir_entry **ret, const char **residual)
275 const char *cp = name, *next;
276 struct proc_dir_entry *de;
281 next = strchr(cp, '/');
286 for (de = de->subdir; de ; de = de->next) {
287 if (proc_match(len, cp, de))
299 static DEFINE_IDR(proc_inum_idr);
300 static DEFINE_SPINLOCK(proc_inum_lock); /* protects the above */
302 #define PROC_DYNAMIC_FIRST 0xF0000000UL
305 * Return an inode number between PROC_DYNAMIC_FIRST and
306 * 0xffffffff, or zero on failure.
308 static unsigned int get_inode_number(void)
314 if (idr_pre_get(&proc_inum_idr, GFP_KERNEL) == 0)
317 spin_lock(&proc_inum_lock);
318 error = idr_get_new(&proc_inum_idr, NULL, &i);
319 spin_unlock(&proc_inum_lock);
320 if (error == -EAGAIN)
325 inum = (i & MAX_ID_MASK) + PROC_DYNAMIC_FIRST;
327 /* inum will never be more than 0xf0ffffff, so no check
334 static void release_inode_number(unsigned int inum)
336 int id = (inum - PROC_DYNAMIC_FIRST) | ~MAX_ID_MASK;
338 spin_lock(&proc_inum_lock);
339 idr_remove(&proc_inum_idr, id);
340 spin_unlock(&proc_inum_lock);
343 static void *proc_follow_link(struct dentry *dentry, struct nameidata *nd)
345 nd_set_link(nd, PDE(dentry->d_inode)->data);
349 static struct inode_operations proc_link_inode_operations = {
350 .readlink = generic_readlink,
351 .follow_link = proc_follow_link,
355 * As some entries in /proc are volatile, we want to
356 * get rid of unused dentries. This could be made
357 * smarter: we could keep a "volatile" flag in the
358 * inode to indicate which ones to keep.
360 static int proc_delete_dentry(struct dentry * dentry)
365 static struct dentry_operations proc_dentry_operations =
367 .d_delete = proc_delete_dentry,
371 * Don't create negative dentries here, return -ENOENT by hand
374 struct dentry *proc_lookup(struct inode * dir, struct dentry *dentry, struct nameidata *nd)
376 struct inode *inode = NULL;
377 struct proc_dir_entry * de;
383 for (de = de->subdir; de ; de = de->next) {
384 if (de->namelen != dentry->d_name.len)
386 if (!memcmp(dentry->d_name.name, de->name, de->namelen)) {
387 unsigned int ino = de->low_ino;
390 inode = proc_get_inode(dir->i_sb, ino, de);
398 dentry->d_op = &proc_dentry_operations;
399 d_add(dentry, inode);
402 return ERR_PTR(error);
406 * This returns non-zero if at EOF, so that the /proc
407 * root directory can use this and check if it should
408 * continue with the <pid> entries..
410 * Note that the VFS-layer doesn't care about the return
411 * value of the readdir() call, as long as it's non-negative
414 int proc_readdir(struct file * filp,
415 void * dirent, filldir_t filldir)
417 struct proc_dir_entry * de;
420 struct inode *inode = filp->f_dentry->d_inode;
434 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
440 if (filldir(dirent, "..", 2, i,
441 parent_ino(filp->f_dentry),
462 if (filldir(dirent, de->name, de->namelen, filp->f_pos,
463 de->low_ino, de->mode >> 12) < 0)
470 out: unlock_kernel();
475 * These are the generic /proc directory operations. They
476 * use the in-memory "struct proc_dir_entry" tree to parse
477 * the /proc directory.
479 static struct file_operations proc_dir_operations = {
480 .read = generic_read_dir,
481 .readdir = proc_readdir,
485 * proc directories can do almost nothing..
487 static struct inode_operations proc_dir_inode_operations = {
488 .lookup = proc_lookup,
489 .getattr = proc_getattr,
490 .setattr = proc_notify_change,
493 static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp)
497 i = get_inode_number();
501 dp->next = dir->subdir;
504 if (S_ISDIR(dp->mode)) {
505 if (dp->proc_iops == NULL) {
506 dp->proc_fops = &proc_dir_operations;
507 dp->proc_iops = &proc_dir_inode_operations;
510 } else if (S_ISLNK(dp->mode)) {
511 if (dp->proc_iops == NULL)
512 dp->proc_iops = &proc_link_inode_operations;
513 } else if (S_ISREG(dp->mode)) {
514 if (dp->proc_fops == NULL)
515 dp->proc_fops = &proc_file_operations;
516 if (dp->proc_iops == NULL)
517 dp->proc_iops = &proc_file_inode_operations;
523 * Kill an inode that got unregistered..
525 static void proc_kill_inodes(struct proc_dir_entry *de)
528 struct super_block *sb = proc_mnt->mnt_sb;
531 * Actually it's a partial revoke().
534 list_for_each(p, &sb->s_files) {
535 struct file * filp = list_entry(p, struct file, f_u.fu_list);
536 struct dentry * dentry = filp->f_dentry;
537 struct inode * inode;
538 struct file_operations *fops;
540 if (dentry->d_op != &proc_dentry_operations)
542 inode = dentry->d_inode;
543 if (PDE(inode) != de)
552 static struct proc_dir_entry *proc_create(struct proc_dir_entry **parent,
557 struct proc_dir_entry *ent = NULL;
558 const char *fn = name;
561 /* make sure name is valid */
562 if (!name || !strlen(name)) goto out;
564 if (!(*parent) && xlate_proc_name(name, parent, &fn) != 0)
567 /* At this point there must not be any '/' characters beyond *fn */
573 ent = kmalloc(sizeof(struct proc_dir_entry) + len + 1, GFP_KERNEL);
576 memset(ent, 0, sizeof(struct proc_dir_entry));
577 memcpy(((char *) ent) + sizeof(struct proc_dir_entry), fn, len + 1);
578 ent->name = ((char *) ent) + sizeof(*ent);
586 struct proc_dir_entry *proc_symlink(const char *name,
587 struct proc_dir_entry *parent, const char *dest)
589 struct proc_dir_entry *ent;
591 ent = proc_create(&parent,name,
592 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
595 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
597 strcpy((char*)ent->data,dest);
598 if (proc_register(parent, ent) < 0) {
611 struct proc_dir_entry *proc_mkdir_mode(const char *name, mode_t mode,
612 struct proc_dir_entry *parent)
614 struct proc_dir_entry *ent;
616 ent = proc_create(&parent, name, S_IFDIR | mode, 2);
618 ent->proc_fops = &proc_dir_operations;
619 ent->proc_iops = &proc_dir_inode_operations;
621 if (proc_register(parent, ent) < 0) {
629 struct proc_dir_entry *proc_mkdir(const char *name,
630 struct proc_dir_entry *parent)
632 return proc_mkdir_mode(name, S_IRUGO | S_IXUGO, parent);
635 struct proc_dir_entry *create_proc_entry(const char *name, mode_t mode,
636 struct proc_dir_entry *parent)
638 struct proc_dir_entry *ent;
642 if ((mode & S_IALLUGO) == 0)
643 mode |= S_IRUGO | S_IXUGO;
646 if ((mode & S_IFMT) == 0)
648 if ((mode & S_IALLUGO) == 0)
653 ent = proc_create(&parent,name,mode,nlink);
656 ent->proc_fops = &proc_dir_operations;
657 ent->proc_iops = &proc_dir_inode_operations;
659 if (proc_register(parent, ent) < 0) {
667 void free_proc_entry(struct proc_dir_entry *de)
669 unsigned int ino = de->low_ino;
671 if (ino < PROC_DYNAMIC_FIRST)
674 release_inode_number(ino);
676 if (S_ISLNK(de->mode) && de->data)
682 * Remove a /proc entry and free it if it's not currently in use.
683 * If it is in use, we set the 'deleted' flag.
685 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
687 struct proc_dir_entry **p;
688 struct proc_dir_entry *de;
689 const char *fn = name;
692 if (!parent && xlate_proc_name(name, &parent, &fn) != 0)
695 for (p = &parent->subdir; *p; p=&(*p)->next ) {
696 if (!proc_match(len, fn, *p))
701 if (S_ISDIR(de->mode))
703 proc_kill_inodes(de);
706 if (!atomic_read(&de->count))
710 printk("remove_proc_entry: %s/%s busy, count=%d\n",
711 parent->name, de->name, atomic_read(&de->count));