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>
25 static ssize_t proc_file_read(struct file *file, char __user *buf,
26 size_t nbytes, loff_t *ppos);
27 static ssize_t proc_file_write(struct file *file, const char __user *buffer,
28 size_t count, loff_t *ppos);
29 static loff_t proc_file_lseek(struct file *, loff_t, int);
31 int proc_match(int len, const char *name, struct proc_dir_entry *de)
33 if (de->namelen != len)
35 return !memcmp(name, de->name, len);
38 static struct file_operations proc_file_operations = {
39 .llseek = proc_file_lseek,
40 .read = proc_file_read,
41 .write = proc_file_write,
44 /* buffer size is one page but our output routines use some slack for overruns */
45 #define PROC_BLOCK_SIZE (PAGE_SIZE - 1024)
48 proc_file_read(struct file *file, char __user *buf, size_t nbytes,
51 struct inode * inode = file->f_dentry->d_inode;
57 struct proc_dir_entry * dp;
60 if (!(page = (char*) __get_free_page(GFP_KERNEL)))
63 while ((nbytes > 0) && !eof) {
64 count = min_t(size_t, PROC_BLOCK_SIZE, nbytes);
68 /* Handle old net routines */
69 n = dp->get_info(page, &start, *ppos, count);
72 } else if (dp->read_proc) {
74 * How to be a proc read function
75 * ------------------------------
77 * int f(char *buffer, char **start, off_t offset,
78 * int count, int *peof, void *dat)
80 * Assume that the buffer is "count" bytes in size.
82 * If you know you have supplied all the data you
85 * You have three ways to return data:
86 * 0) Leave *start = NULL. (This is the default.)
87 * Put the data of the requested offset at that
88 * offset within the buffer. Return the number (n)
89 * of bytes there are from the beginning of the
90 * buffer up to the last byte of data. If the
91 * number of supplied bytes (= n - offset) is
92 * greater than zero and you didn't signal eof
93 * and the reader is prepared to take more data
94 * you will be called again with the requested
95 * offset advanced by the number of bytes
96 * absorbed. This interface is useful for files
97 * no larger than the buffer.
98 * 1) Set *start = an unsigned long value less than
99 * the buffer address but greater than zero.
100 * Put the data of the requested offset at the
101 * beginning of the buffer. Return the number of
102 * bytes of data placed there. If this number 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 *start. This interface is
107 * useful when you have a large file consisting
108 * of a series of blocks which you want to count
109 * and return as wholes.
110 * (Hack by Paul.Russell@rustcorp.com.au)
111 * 2) Set *start = an address within the buffer.
112 * Put the data of the requested offset at *start.
113 * Return the number of bytes of data placed there.
114 * If this number is greater than zero and you
115 * didn't signal eof and the reader is prepared to
116 * take more data you will be called again with the
117 * requested offset advanced by the number of bytes
120 n = dp->read_proc(page, &start, *ppos,
121 count, &eof, dp->data);
125 if (n == 0) /* end of file */
127 if (n < 0) { /* error */
136 "proc_file_read: Apparent buffer overflow!\n");
144 start = page + *ppos;
145 } else if (start < page) {
148 "proc_file_read: Apparent buffer overflow!\n");
153 * Don't reduce n because doing so might
154 * cut off part of a data block.
157 "proc_file_read: Read count exceeded\n");
159 } else /* start >= page */ {
160 unsigned long startoff = (unsigned long)(start - page);
161 if (n > (PAGE_SIZE - startoff)) {
163 "proc_file_read: Apparent buffer overflow!\n");
164 n = PAGE_SIZE - startoff;
170 n -= copy_to_user(buf, start < page ? page : start, n);
177 *ppos += start < page ? (unsigned long)start : n;
182 free_page((unsigned long) page);
187 proc_file_write(struct file *file, const char __user *buffer,
188 size_t count, loff_t *ppos)
190 struct inode *inode = file->f_dentry->d_inode;
191 struct proc_dir_entry * dp;
198 /* FIXME: does this routine need ppos? probably... */
199 return dp->write_proc(file, buffer, count, dp->data);
204 proc_file_lseek(struct file *file, loff_t offset, int orig)
212 file->f_pos = offset;
216 if (offset + file->f_pos < 0)
218 file->f_pos += offset;
232 static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
234 struct inode *inode = dentry->d_inode;
235 struct proc_dir_entry *de = PDE(inode);
238 error = inode_change_ok(inode, iattr);
242 error = inode_setattr(inode, iattr);
246 de->uid = inode->i_uid;
247 de->gid = inode->i_gid;
248 de->mode = inode->i_mode;
253 static struct inode_operations proc_file_inode_operations = {
254 .setattr = proc_notify_change,
258 * This function parses a name such as "tty/driver/serial", and
259 * returns the struct proc_dir_entry for "/proc/tty/driver", and
260 * returns "serial" in residual.
262 static int xlate_proc_name(const char *name,
263 struct proc_dir_entry **ret, const char **residual)
265 const char *cp = name, *next;
266 struct proc_dir_entry *de;
271 next = strchr(cp, '/');
276 for (de = de->subdir; de ; de = de->next) {
277 if (proc_match(len, cp, de))
289 static DEFINE_IDR(proc_inum_idr);
290 static DEFINE_SPINLOCK(proc_inum_lock); /* protects the above */
292 #define PROC_DYNAMIC_FIRST 0xF0000000UL
295 * Return an inode number between PROC_DYNAMIC_FIRST and
296 * 0xffffffff, or zero on failure.
298 static unsigned int get_inode_number(void)
304 if (idr_pre_get(&proc_inum_idr, GFP_KERNEL) == 0)
307 spin_lock(&proc_inum_lock);
308 error = idr_get_new(&proc_inum_idr, NULL, &i);
309 spin_unlock(&proc_inum_lock);
310 if (error == -EAGAIN)
315 inum = (i & MAX_ID_MASK) + PROC_DYNAMIC_FIRST;
317 /* inum will never be more than 0xf0ffffff, so no check
324 static void release_inode_number(unsigned int inum)
326 int id = (inum - PROC_DYNAMIC_FIRST) | ~MAX_ID_MASK;
328 spin_lock(&proc_inum_lock);
329 idr_remove(&proc_inum_idr, id);
330 spin_unlock(&proc_inum_lock);
333 static int proc_follow_link(struct dentry *dentry, struct nameidata *nd)
335 nd_set_link(nd, PDE(dentry->d_inode)->data);
339 static struct inode_operations proc_link_inode_operations = {
340 .readlink = generic_readlink,
341 .follow_link = proc_follow_link,
345 * As some entries in /proc are volatile, we want to
346 * get rid of unused dentries. This could be made
347 * smarter: we could keep a "volatile" flag in the
348 * inode to indicate which ones to keep.
350 static int proc_delete_dentry(struct dentry * dentry)
355 static struct dentry_operations proc_dentry_operations =
357 .d_delete = proc_delete_dentry,
361 * Don't create negative dentries here, return -ENOENT by hand
364 struct dentry *proc_lookup(struct inode * dir, struct dentry *dentry, struct nameidata *nd)
366 struct inode *inode = NULL;
367 struct proc_dir_entry * de;
373 for (de = de->subdir; de ; de = de->next) {
374 if (de->namelen != dentry->d_name.len)
376 if (!vx_hide_check(0, de->vx_flags))
378 if (!memcmp(dentry->d_name.name, de->name, de->namelen)) {
379 unsigned int ino = de->low_ino;
382 inode = proc_get_inode(dir->i_sb, ino, de);
383 /* generic proc entries belong to the host */
392 dentry->d_op = &proc_dentry_operations;
393 d_add(dentry, inode);
396 return ERR_PTR(error);
400 * This returns non-zero if at EOF, so that the /proc
401 * root directory can use this and check if it should
402 * continue with the <pid> entries..
404 * Note that the VFS-layer doesn't care about the return
405 * value of the readdir() call, as long as it's non-negative
408 int proc_readdir(struct file * filp,
409 void * dirent, filldir_t filldir)
411 struct proc_dir_entry * de;
414 struct inode *inode = filp->f_dentry->d_inode;
428 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
434 if (filldir(dirent, "..", 2, i,
435 parent_ino(filp->f_dentry),
456 if (!vx_hide_check(0, de->vx_flags))
458 if (filldir(dirent, de->name, de->namelen, filp->f_pos,
459 de->low_ino, de->mode >> 12) < 0)
467 out: unlock_kernel();
472 * These are the generic /proc directory operations. They
473 * use the in-memory "struct proc_dir_entry" tree to parse
474 * the /proc directory.
476 static struct file_operations proc_dir_operations = {
477 .read = generic_read_dir,
478 .readdir = proc_readdir,
482 * proc directories can do almost nothing..
484 static struct inode_operations proc_dir_inode_operations = {
485 .lookup = proc_lookup,
486 .setattr = proc_notify_change,
489 static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp)
493 i = get_inode_number();
497 dp->next = dir->subdir;
500 if (S_ISDIR(dp->mode)) {
501 if (dp->proc_iops == NULL) {
502 dp->proc_fops = &proc_dir_operations;
503 dp->proc_iops = &proc_dir_inode_operations;
506 } else if (S_ISLNK(dp->mode)) {
507 if (dp->proc_iops == NULL)
508 dp->proc_iops = &proc_link_inode_operations;
509 } else if (S_ISREG(dp->mode)) {
510 if (dp->proc_fops == NULL)
511 dp->proc_fops = &proc_file_operations;
512 if (dp->proc_iops == NULL)
513 dp->proc_iops = &proc_file_inode_operations;
519 * Kill an inode that got unregistered..
521 static void proc_kill_inodes(struct proc_dir_entry *de)
524 struct super_block *sb = proc_mnt->mnt_sb;
527 * Actually it's a partial revoke().
530 list_for_each(p, &sb->s_files) {
531 struct file * filp = list_entry(p, struct file, f_list);
532 struct dentry * dentry = filp->f_dentry;
533 struct inode * inode;
534 struct file_operations *fops;
536 if (dentry->d_op != &proc_dentry_operations)
538 inode = dentry->d_inode;
539 if (PDE(inode) != de)
548 static struct proc_dir_entry *proc_create(struct proc_dir_entry **parent,
553 struct proc_dir_entry *ent = NULL;
554 const char *fn = name;
557 /* make sure name is valid */
558 if (!name || !strlen(name)) goto out;
560 if (!(*parent) && xlate_proc_name(name, parent, &fn) != 0)
563 /* At this point there must not be any '/' characters beyond *fn */
569 ent = kmalloc(sizeof(struct proc_dir_entry) + len + 1, GFP_KERNEL);
572 memset(ent, 0, sizeof(struct proc_dir_entry));
573 memcpy(((char *) ent) + sizeof(struct proc_dir_entry), fn, len + 1);
574 ent->name = ((char *) ent) + sizeof(*ent);
578 ent->vx_flags = IATTR_PROC_DEFAULT;
583 struct proc_dir_entry *proc_symlink(const char *name,
584 struct proc_dir_entry *parent, const char *dest)
586 struct proc_dir_entry *ent;
588 ent = proc_create(&parent,name,
589 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
592 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
594 strcpy((char*)ent->data,dest);
595 if (proc_register(parent, ent) < 0) {
600 ent->vx_flags = IATTR_PROC_SYMLINK;
609 struct proc_dir_entry *proc_mkdir_mode(const char *name, mode_t mode,
610 struct proc_dir_entry *parent)
612 struct proc_dir_entry *ent;
614 ent = proc_create(&parent, name, S_IFDIR | mode, 2);
616 ent->proc_fops = &proc_dir_operations;
617 ent->proc_iops = &proc_dir_inode_operations;
619 if (proc_register(parent, ent) < 0) {
627 struct proc_dir_entry *proc_mkdir(const char *name,
628 struct proc_dir_entry *parent)
630 return proc_mkdir_mode(name, S_IRUGO | S_IXUGO, parent);
633 struct proc_dir_entry *create_proc_entry(const char *name, mode_t mode,
634 struct proc_dir_entry *parent)
636 struct proc_dir_entry *ent;
640 if ((mode & S_IALLUGO) == 0)
641 mode |= S_IRUGO | S_IXUGO;
644 if ((mode & S_IFMT) == 0)
646 if ((mode & S_IALLUGO) == 0)
651 ent = proc_create(&parent,name,mode,nlink);
654 ent->proc_fops = &proc_dir_operations;
655 ent->proc_iops = &proc_dir_inode_operations;
657 if (proc_register(parent, ent) < 0) {
665 void free_proc_entry(struct proc_dir_entry *de)
667 unsigned int ino = de->low_ino;
669 if (ino < PROC_DYNAMIC_FIRST)
672 release_inode_number(ino);
674 if (S_ISLNK(de->mode) && de->data)
680 * Remove a /proc entry and free it if it's not currently in use.
681 * If it is in use, we set the 'deleted' flag.
683 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
685 struct proc_dir_entry **p;
686 struct proc_dir_entry *de;
687 const char *fn = name;
690 if (!parent && xlate_proc_name(name, &parent, &fn) != 0)
693 for (p = &parent->subdir; *p; p=&(*p)->next ) {
694 if (!proc_match(len, fn, *p))
699 if (S_ISDIR(de->mode))
701 proc_kill_inodes(de);
704 if (!atomic_read(&de->count))
708 printk("remove_proc_entry: %s/%s busy, count=%d\n",
709 parent->name, de->name, atomic_read(&de->count));