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 <linux/vserver/inode.h>
23 #include <asm/uaccess.h>
27 static ssize_t proc_file_read(struct file *file, char __user *buf,
28 size_t nbytes, loff_t *ppos);
29 static ssize_t proc_file_write(struct file *file, const char __user *buffer,
30 size_t count, loff_t *ppos);
31 static loff_t proc_file_lseek(struct file *, loff_t, int);
33 int proc_match(int len, const char *name, struct proc_dir_entry *de)
35 if (de->namelen != len)
37 return !memcmp(name, de->name, len);
40 static struct file_operations proc_file_operations = {
41 .llseek = proc_file_lseek,
42 .read = proc_file_read,
43 .write = proc_file_write,
46 /* buffer size is one page but our output routines use some slack for overruns */
47 #define PROC_BLOCK_SIZE (PAGE_SIZE - 1024)
50 proc_file_read(struct file *file, char __user *buf, size_t nbytes,
53 struct inode * inode = file->f_dentry->d_inode;
59 struct proc_dir_entry * dp;
60 unsigned long long pos;
63 * Gaah, please just use "seq_file" instead. The legacy /proc
64 * interfaces cut loff_t down to off_t for reads, and ignore
65 * the offset entirely for writes..
68 if (pos > MAX_NON_LFS)
70 if (nbytes > MAX_NON_LFS - pos)
71 nbytes = MAX_NON_LFS - pos;
74 if (!(page = (char*) __get_free_page(GFP_KERNEL)))
77 while ((nbytes > 0) && !eof) {
78 count = min_t(size_t, PROC_BLOCK_SIZE, nbytes);
82 /* Handle old net routines */
83 n = dp->get_info(page, &start, *ppos, count);
86 } else if (dp->read_proc) {
88 * How to be a proc read function
89 * ------------------------------
91 * int f(char *buffer, char **start, off_t offset,
92 * int count, int *peof, void *dat)
94 * Assume that the buffer is "count" bytes in size.
96 * If you know you have supplied all the data you
99 * You have three ways to return data:
100 * 0) Leave *start = NULL. (This is the default.)
101 * Put the data of the requested offset at that
102 * offset within the buffer. Return the number (n)
103 * of bytes there are from the beginning of the
104 * buffer up to the last byte of data. If the
105 * number of supplied bytes (= n - offset) is
106 * greater than zero and you didn't signal eof
107 * and the reader is prepared to take more data
108 * you will be called again with the requested
109 * offset advanced by the number of bytes
110 * absorbed. This interface is useful for files
111 * no larger than the buffer.
112 * 1) Set *start = an unsigned long value less than
113 * the buffer address but greater than zero.
114 * Put the data of the requested offset at the
115 * beginning of the buffer. Return the number of
116 * bytes of data placed there. If this number is
117 * greater than zero and you didn't signal eof
118 * and the reader is prepared to take more data
119 * you will be called again with the requested
120 * offset advanced by *start. This interface is
121 * useful when you have a large file consisting
122 * of a series of blocks which you want to count
123 * and return as wholes.
124 * (Hack by Paul.Russell@rustcorp.com.au)
125 * 2) Set *start = an address within the buffer.
126 * Put the data of the requested offset at *start.
127 * Return the number of bytes of data placed there.
128 * If this number is greater than zero and you
129 * didn't signal eof and the reader is prepared to
130 * take more data you will be called again with the
131 * requested offset advanced by the number of bytes
134 n = dp->read_proc(page, &start, *ppos,
135 count, &eof, dp->data);
139 if (n == 0) /* end of file */
141 if (n < 0) { /* error */
150 "proc_file_read: Apparent buffer overflow!\n");
158 start = page + *ppos;
159 } else if (start < page) {
162 "proc_file_read: Apparent buffer overflow!\n");
167 * Don't reduce n because doing so might
168 * cut off part of a data block.
171 "proc_file_read: Read count exceeded\n");
173 } else /* start >= page */ {
174 unsigned long startoff = (unsigned long)(start - page);
175 if (n > (PAGE_SIZE - startoff)) {
177 "proc_file_read: Apparent buffer overflow!\n");
178 n = PAGE_SIZE - startoff;
184 n -= copy_to_user(buf, start < page ? page : start, n);
191 *ppos += start < page ? (unsigned long)start : n;
196 free_page((unsigned long) page);
201 proc_file_write(struct file *file, const char __user *buffer,
202 size_t count, loff_t *ppos)
204 struct inode *inode = file->f_dentry->d_inode;
205 struct proc_dir_entry * dp;
212 /* FIXME: does this routine need ppos? probably... */
213 return dp->write_proc(file, buffer, count, dp->data);
218 proc_file_lseek(struct file *file, loff_t offset, int orig)
220 loff_t retval = -EINVAL;
223 offset += file->f_pos;
226 if (offset < 0 || offset > MAX_NON_LFS)
228 file->f_pos = retval = offset;
233 static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
235 struct inode *inode = dentry->d_inode;
236 struct proc_dir_entry *de = PDE(inode);
239 error = inode_change_ok(inode, iattr);
243 error = inode_setattr(inode, iattr);
247 de->uid = inode->i_uid;
248 de->gid = inode->i_gid;
249 de->mode = inode->i_mode;
254 static int proc_getattr(struct vfsmount *mnt, struct dentry *dentry,
257 struct inode *inode = dentry->d_inode;
258 struct proc_dir_entry *de = PROC_I(inode)->pde;
260 inode->i_nlink = de->nlink;
262 generic_fillattr(inode, stat);
266 static struct inode_operations proc_file_inode_operations = {
267 .setattr = proc_notify_change,
271 * This function parses a name such as "tty/driver/serial", and
272 * returns the struct proc_dir_entry for "/proc/tty/driver", and
273 * returns "serial" in residual.
275 static int xlate_proc_name(const char *name,
276 struct proc_dir_entry **ret, const char **residual)
278 const char *cp = name, *next;
279 struct proc_dir_entry *de;
284 next = strchr(cp, '/');
289 for (de = de->subdir; de ; de = de->next) {
290 if (proc_match(len, cp, de))
302 static DEFINE_IDR(proc_inum_idr);
303 static DEFINE_SPINLOCK(proc_inum_lock); /* protects the above */
305 #define PROC_DYNAMIC_FIRST 0xF0000000UL
308 * Return an inode number between PROC_DYNAMIC_FIRST and
309 * 0xffffffff, or zero on failure.
311 static unsigned int get_inode_number(void)
317 if (idr_pre_get(&proc_inum_idr, GFP_KERNEL) == 0)
320 spin_lock(&proc_inum_lock);
321 error = idr_get_new(&proc_inum_idr, NULL, &i);
322 spin_unlock(&proc_inum_lock);
323 if (error == -EAGAIN)
328 inum = (i & MAX_ID_MASK) + PROC_DYNAMIC_FIRST;
330 /* inum will never be more than 0xf0ffffff, so no check
337 static void release_inode_number(unsigned int inum)
339 int id = (inum - PROC_DYNAMIC_FIRST) | ~MAX_ID_MASK;
341 spin_lock(&proc_inum_lock);
342 idr_remove(&proc_inum_idr, id);
343 spin_unlock(&proc_inum_lock);
346 static void *proc_follow_link(struct dentry *dentry, struct nameidata *nd)
348 nd_set_link(nd, PDE(dentry->d_inode)->data);
352 static struct inode_operations proc_link_inode_operations = {
353 .readlink = generic_readlink,
354 .follow_link = proc_follow_link,
358 * As some entries in /proc are volatile, we want to
359 * get rid of unused dentries. This could be made
360 * smarter: we could keep a "volatile" flag in the
361 * inode to indicate which ones to keep.
363 static int proc_delete_dentry(struct dentry * dentry)
368 static struct dentry_operations proc_dentry_operations =
370 .d_delete = proc_delete_dentry,
374 * Don't create negative dentries here, return -ENOENT by hand
377 struct dentry *proc_lookup(struct inode * dir, struct dentry *dentry, struct nameidata *nd)
379 struct inode *inode = NULL;
380 struct proc_dir_entry * de;
386 for (de = de->subdir; de ; de = de->next) {
387 if (de->namelen != dentry->d_name.len)
389 if (!vx_hide_check(0, de->vx_flags))
391 if (!memcmp(dentry->d_name.name, de->name, de->namelen)) {
392 unsigned int ino = de->low_ino;
395 inode = proc_get_inode(dir->i_sb, ino, de);
396 /* generic proc entries belong to the host */
405 dentry->d_op = &proc_dentry_operations;
406 d_add(dentry, inode);
409 return ERR_PTR(error);
413 * This returns non-zero if at EOF, so that the /proc
414 * root directory can use this and check if it should
415 * continue with the <pid> entries..
417 * Note that the VFS-layer doesn't care about the return
418 * value of the readdir() call, as long as it's non-negative
421 int proc_readdir(struct file * filp,
422 void * dirent, filldir_t filldir)
424 struct proc_dir_entry * de;
427 struct inode *inode = filp->f_dentry->d_inode;
441 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
447 if (filldir(dirent, "..", 2, i,
448 parent_ino(filp->f_dentry),
469 if (!vx_hide_check(0, de->vx_flags))
471 if (filldir(dirent, de->name, de->namelen, filp->f_pos,
472 de->low_ino, de->mode >> 12) < 0)
480 out: unlock_kernel();
485 * These are the generic /proc directory operations. They
486 * use the in-memory "struct proc_dir_entry" tree to parse
487 * the /proc directory.
489 static struct file_operations proc_dir_operations = {
490 .read = generic_read_dir,
491 .readdir = proc_readdir,
495 * proc directories can do almost nothing..
497 static struct inode_operations proc_dir_inode_operations = {
498 .lookup = proc_lookup,
499 .getattr = proc_getattr,
500 .setattr = proc_notify_change,
503 static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp)
507 i = get_inode_number();
511 dp->next = dir->subdir;
514 if (S_ISDIR(dp->mode)) {
515 if (dp->proc_iops == NULL) {
516 dp->proc_fops = &proc_dir_operations;
517 dp->proc_iops = &proc_dir_inode_operations;
520 } else if (S_ISLNK(dp->mode)) {
521 if (dp->proc_iops == NULL)
522 dp->proc_iops = &proc_link_inode_operations;
523 } else if (S_ISREG(dp->mode)) {
524 if (dp->proc_fops == NULL)
525 dp->proc_fops = &proc_file_operations;
526 if (dp->proc_iops == NULL)
527 dp->proc_iops = &proc_file_inode_operations;
533 * Kill an inode that got unregistered..
535 static void proc_kill_inodes(struct proc_dir_entry *de)
538 struct super_block *sb = proc_mnt->mnt_sb;
541 * Actually it's a partial revoke().
544 list_for_each(p, &sb->s_files) {
545 struct file * filp = list_entry(p, struct file, f_u.fu_list);
546 struct dentry * dentry = filp->f_dentry;
547 struct inode * inode;
548 struct file_operations *fops;
550 if (dentry->d_op != &proc_dentry_operations)
552 inode = dentry->d_inode;
553 if (PDE(inode) != de)
562 static struct proc_dir_entry *proc_create(struct proc_dir_entry **parent,
567 struct proc_dir_entry *ent = NULL;
568 const char *fn = name;
571 /* make sure name is valid */
572 if (!name || !strlen(name)) goto out;
574 if (!(*parent) && xlate_proc_name(name, parent, &fn) != 0)
577 /* At this point there must not be any '/' characters beyond *fn */
583 ent = kmalloc(sizeof(struct proc_dir_entry) + len + 1, GFP_KERNEL);
586 memset(ent, 0, sizeof(struct proc_dir_entry));
587 memcpy(((char *) ent) + sizeof(struct proc_dir_entry), fn, len + 1);
588 ent->name = ((char *) ent) + sizeof(*ent);
592 ent->vx_flags = IATTR_PROC_DEFAULT;
597 struct proc_dir_entry *proc_symlink(const char *name,
598 struct proc_dir_entry *parent, const char *dest)
600 struct proc_dir_entry *ent;
602 ent = proc_create(&parent,name,
603 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
606 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
608 strcpy((char*)ent->data,dest);
609 if (proc_register(parent, ent) < 0) {
614 ent->vx_flags = IATTR_PROC_SYMLINK;
623 struct proc_dir_entry *proc_mkdir_mode(const char *name, mode_t mode,
624 struct proc_dir_entry *parent)
626 struct proc_dir_entry *ent;
628 ent = proc_create(&parent, name, S_IFDIR | mode, 2);
630 ent->proc_fops = &proc_dir_operations;
631 ent->proc_iops = &proc_dir_inode_operations;
633 if (proc_register(parent, ent) < 0) {
641 struct proc_dir_entry *proc_mkdir(const char *name,
642 struct proc_dir_entry *parent)
644 return proc_mkdir_mode(name, S_IRUGO | S_IXUGO, parent);
647 struct proc_dir_entry *create_proc_entry(const char *name, mode_t mode,
648 struct proc_dir_entry *parent)
650 struct proc_dir_entry *ent;
654 if ((mode & S_IALLUGO) == 0)
655 mode |= S_IRUGO | S_IXUGO;
658 if ((mode & S_IFMT) == 0)
660 if ((mode & S_IALLUGO) == 0)
665 ent = proc_create(&parent,name,mode,nlink);
668 ent->proc_fops = &proc_dir_operations;
669 ent->proc_iops = &proc_dir_inode_operations;
671 if (proc_register(parent, ent) < 0) {
679 void free_proc_entry(struct proc_dir_entry *de)
681 unsigned int ino = de->low_ino;
683 if (ino < PROC_DYNAMIC_FIRST)
686 release_inode_number(ino);
688 if (S_ISLNK(de->mode) && de->data)
694 * Remove a /proc entry and free it if it's not currently in use.
695 * If it is in use, we set the 'deleted' flag.
697 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
699 struct proc_dir_entry **p;
700 struct proc_dir_entry *de;
701 const char *fn = name;
704 if (!parent && xlate_proc_name(name, &parent, &fn) != 0)
707 for (p = &parent->subdir; *p; p=&(*p)->next ) {
708 if (!proc_match(len, fn, *p))
713 if (S_ISDIR(de->mode))
715 proc_kill_inodes(de);
718 if (!atomic_read(&de->count))
722 printk("remove_proc_entry: %s/%s busy, count=%d\n",
723 parent->name, de->name, atomic_read(&de->count));