ia64/linux-2.6.18-xen.hg

view fs/proc/generic.c @ 524:7f8b544237bf

netfront: Allow netfront in domain 0.

This is useful if your physical network device is in a utility domain.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Tue Apr 15 15:18:58 2008 +0100 (2008-04-15)
parents 831230e53067
children
line source
1 /*
2 * proc/fs/generic.c --- generic routines for the proc-fs
3 *
4 * This file contains generic proc-fs routines for handling
5 * directories and files.
6 *
7 * Copyright (C) 1991, 1992 Linus Torvalds.
8 * Copyright (C) 1997 Theodore Ts'o
9 */
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/spinlock.h>
23 #include <asm/uaccess.h>
25 #include "internal.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 DEFINE_SPINLOCK(proc_subdir_lock);
35 int proc_match(int len, const char *name, struct proc_dir_entry *de)
36 {
37 if (de->namelen != len)
38 return 0;
39 return !memcmp(name, de->name, len);
40 }
42 static struct file_operations proc_file_operations = {
43 .llseek = proc_file_lseek,
44 .read = proc_file_read,
45 .write = proc_file_write,
46 };
48 /* buffer size is one page but our output routines use some slack for overruns */
49 #define PROC_BLOCK_SIZE (PAGE_SIZE - 1024)
51 static ssize_t
52 proc_file_read(struct file *file, char __user *buf, size_t nbytes,
53 loff_t *ppos)
54 {
55 struct inode * inode = file->f_dentry->d_inode;
56 char *page;
57 ssize_t retval=0;
58 int eof=0;
59 ssize_t n, count;
60 char *start;
61 struct proc_dir_entry * dp;
62 unsigned long long pos;
64 /*
65 * Gaah, please just use "seq_file" instead. The legacy /proc
66 * interfaces cut loff_t down to off_t for reads, and ignore
67 * the offset entirely for writes..
68 */
69 pos = *ppos;
70 if (pos > MAX_NON_LFS)
71 return 0;
72 if (nbytes > MAX_NON_LFS - pos)
73 nbytes = MAX_NON_LFS - pos;
75 dp = PDE(inode);
76 if (!(page = (char*) __get_free_page(GFP_KERNEL)))
77 return -ENOMEM;
79 while ((nbytes > 0) && !eof) {
80 count = min_t(size_t, PROC_BLOCK_SIZE, nbytes);
82 start = NULL;
83 if (dp->get_info) {
84 /* Handle old net routines */
85 n = dp->get_info(page, &start, *ppos, count);
86 if (n < count)
87 eof = 1;
88 } else if (dp->read_proc) {
89 /*
90 * How to be a proc read function
91 * ------------------------------
92 * Prototype:
93 * int f(char *buffer, char **start, off_t offset,
94 * int count, int *peof, void *dat)
95 *
96 * Assume that the buffer is "count" bytes in size.
97 *
98 * If you know you have supplied all the data you
99 * have, set *peof.
100 *
101 * You have three ways to return data:
102 * 0) Leave *start = NULL. (This is the default.)
103 * Put the data of the requested offset at that
104 * offset within the buffer. Return the number (n)
105 * of bytes there are from the beginning of the
106 * buffer up to the last byte of data. If the
107 * number of supplied bytes (= n - offset) is
108 * greater than zero and you didn't signal eof
109 * and the reader is prepared to take more data
110 * you will be called again with the requested
111 * offset advanced by the number of bytes
112 * absorbed. This interface is useful for files
113 * no larger than the buffer.
114 * 1) Set *start = an unsigned long value less than
115 * the buffer address but greater than zero.
116 * Put the data of the requested offset at the
117 * beginning of the buffer. Return the number of
118 * bytes of data placed there. If this number is
119 * greater than zero and you didn't signal eof
120 * and the reader is prepared to take more data
121 * you will be called again with the requested
122 * offset advanced by *start. This interface is
123 * useful when you have a large file consisting
124 * of a series of blocks which you want to count
125 * and return as wholes.
126 * (Hack by Paul.Russell@rustcorp.com.au)
127 * 2) Set *start = an address within the buffer.
128 * Put the data of the requested offset at *start.
129 * Return the number of bytes of data placed there.
130 * If this number is greater than zero and you
131 * didn't signal eof and the reader is prepared to
132 * take more data you will be called again with the
133 * requested offset advanced by the number of bytes
134 * absorbed.
135 */
136 n = dp->read_proc(page, &start, *ppos,
137 count, &eof, dp->data);
138 } else
139 break;
141 if (n == 0) /* end of file */
142 break;
143 if (n < 0) { /* error */
144 if (retval == 0)
145 retval = n;
146 break;
147 }
149 if (start == NULL) {
150 if (n > PAGE_SIZE) {
151 printk(KERN_ERR
152 "proc_file_read: Apparent buffer overflow!\n");
153 n = PAGE_SIZE;
154 }
155 n -= *ppos;
156 if (n <= 0)
157 break;
158 if (n > count)
159 n = count;
160 start = page + *ppos;
161 } else if (start < page) {
162 if (n > PAGE_SIZE) {
163 printk(KERN_ERR
164 "proc_file_read: Apparent buffer overflow!\n");
165 n = PAGE_SIZE;
166 }
167 if (n > count) {
168 /*
169 * Don't reduce n because doing so might
170 * cut off part of a data block.
171 */
172 printk(KERN_WARNING
173 "proc_file_read: Read count exceeded\n");
174 }
175 } else /* start >= page */ {
176 unsigned long startoff = (unsigned long)(start - page);
177 if (n > (PAGE_SIZE - startoff)) {
178 printk(KERN_ERR
179 "proc_file_read: Apparent buffer overflow!\n");
180 n = PAGE_SIZE - startoff;
181 }
182 if (n > count)
183 n = count;
184 }
186 n -= copy_to_user(buf, start < page ? page : start, n);
187 if (n == 0) {
188 if (retval == 0)
189 retval = -EFAULT;
190 break;
191 }
193 *ppos += start < page ? (unsigned long)start : n;
194 nbytes -= n;
195 buf += n;
196 retval += n;
197 }
198 free_page((unsigned long) page);
199 return retval;
200 }
202 static ssize_t
203 proc_file_write(struct file *file, const char __user *buffer,
204 size_t count, loff_t *ppos)
205 {
206 struct inode *inode = file->f_dentry->d_inode;
207 struct proc_dir_entry * dp;
209 dp = PDE(inode);
211 if (!dp->write_proc)
212 return -EIO;
214 /* FIXME: does this routine need ppos? probably... */
215 return dp->write_proc(file, buffer, count, dp->data);
216 }
219 static loff_t
220 proc_file_lseek(struct file *file, loff_t offset, int orig)
221 {
222 loff_t retval = -EINVAL;
223 switch (orig) {
224 case 1:
225 offset += file->f_pos;
226 /* fallthrough */
227 case 0:
228 if (offset < 0 || offset > MAX_NON_LFS)
229 break;
230 file->f_pos = retval = offset;
231 }
232 return retval;
233 }
235 static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
236 {
237 struct inode *inode = dentry->d_inode;
238 struct proc_dir_entry *de = PDE(inode);
239 int error;
241 error = inode_change_ok(inode, iattr);
242 if (error)
243 goto out;
245 error = inode_setattr(inode, iattr);
246 if (error)
247 goto out;
249 de->uid = inode->i_uid;
250 de->gid = inode->i_gid;
251 de->mode = inode->i_mode;
252 out:
253 return error;
254 }
256 static int proc_getattr(struct vfsmount *mnt, struct dentry *dentry,
257 struct kstat *stat)
258 {
259 struct inode *inode = dentry->d_inode;
260 struct proc_dir_entry *de = PROC_I(inode)->pde;
261 if (de && de->nlink)
262 inode->i_nlink = de->nlink;
264 generic_fillattr(inode, stat);
265 return 0;
266 }
268 static struct inode_operations proc_file_inode_operations = {
269 .setattr = proc_notify_change,
270 };
272 /*
273 * This function parses a name such as "tty/driver/serial", and
274 * returns the struct proc_dir_entry for "/proc/tty/driver", and
275 * returns "serial" in residual.
276 */
277 static int xlate_proc_name(const char *name,
278 struct proc_dir_entry **ret, const char **residual)
279 {
280 const char *cp = name, *next;
281 struct proc_dir_entry *de;
282 int len;
283 int rtn = 0;
285 spin_lock(&proc_subdir_lock);
286 de = &proc_root;
287 while (1) {
288 next = strchr(cp, '/');
289 if (!next)
290 break;
292 len = next - cp;
293 for (de = de->subdir; de ; de = de->next) {
294 if (proc_match(len, cp, de))
295 break;
296 }
297 if (!de) {
298 rtn = -ENOENT;
299 goto out;
300 }
301 cp += len + 1;
302 }
303 *residual = cp;
304 *ret = de;
305 out:
306 spin_unlock(&proc_subdir_lock);
307 return rtn;
308 }
310 static DEFINE_IDR(proc_inum_idr);
311 static DEFINE_SPINLOCK(proc_inum_lock); /* protects the above */
313 #define PROC_DYNAMIC_FIRST 0xF0000000UL
315 /*
316 * Return an inode number between PROC_DYNAMIC_FIRST and
317 * 0xffffffff, or zero on failure.
318 */
319 static unsigned int get_inode_number(void)
320 {
321 int i, inum = 0;
322 int error;
324 retry:
325 if (idr_pre_get(&proc_inum_idr, GFP_KERNEL) == 0)
326 return 0;
328 spin_lock(&proc_inum_lock);
329 error = idr_get_new(&proc_inum_idr, NULL, &i);
330 spin_unlock(&proc_inum_lock);
331 if (error == -EAGAIN)
332 goto retry;
333 else if (error)
334 return 0;
336 inum = (i & MAX_ID_MASK) + PROC_DYNAMIC_FIRST;
338 /* inum will never be more than 0xf0ffffff, so no check
339 * for overflow.
340 */
342 return inum;
343 }
345 static void release_inode_number(unsigned int inum)
346 {
347 int id = (inum - PROC_DYNAMIC_FIRST) | ~MAX_ID_MASK;
349 spin_lock(&proc_inum_lock);
350 idr_remove(&proc_inum_idr, id);
351 spin_unlock(&proc_inum_lock);
352 }
354 static void *proc_follow_link(struct dentry *dentry, struct nameidata *nd)
355 {
356 nd_set_link(nd, PDE(dentry->d_inode)->data);
357 return NULL;
358 }
360 static struct inode_operations proc_link_inode_operations = {
361 .readlink = generic_readlink,
362 .follow_link = proc_follow_link,
363 };
365 /*
366 * As some entries in /proc are volatile, we want to
367 * get rid of unused dentries. This could be made
368 * smarter: we could keep a "volatile" flag in the
369 * inode to indicate which ones to keep.
370 */
371 static int proc_delete_dentry(struct dentry * dentry)
372 {
373 return 1;
374 }
376 static struct dentry_operations proc_dentry_operations =
377 {
378 .d_delete = proc_delete_dentry,
379 };
381 /*
382 * Don't create negative dentries here, return -ENOENT by hand
383 * instead.
384 */
385 struct dentry *proc_lookup(struct inode * dir, struct dentry *dentry, struct nameidata *nd)
386 {
387 struct inode *inode = NULL;
388 struct proc_dir_entry * de;
389 int error = -ENOENT;
391 lock_kernel();
392 spin_lock(&proc_subdir_lock);
393 de = PDE(dir);
394 if (de) {
395 for (de = de->subdir; de ; de = de->next) {
396 if (de->namelen != dentry->d_name.len)
397 continue;
398 if (!memcmp(dentry->d_name.name, de->name, de->namelen)) {
399 unsigned int ino = de->low_ino;
401 spin_unlock(&proc_subdir_lock);
402 error = -EINVAL;
403 inode = proc_get_inode(dir->i_sb, ino, de);
404 spin_lock(&proc_subdir_lock);
405 break;
406 }
407 }
408 }
409 spin_unlock(&proc_subdir_lock);
410 unlock_kernel();
412 if (inode) {
413 dentry->d_op = &proc_dentry_operations;
414 d_add(dentry, inode);
415 return NULL;
416 }
417 return ERR_PTR(error);
418 }
420 /*
421 * This returns non-zero if at EOF, so that the /proc
422 * root directory can use this and check if it should
423 * continue with the <pid> entries..
424 *
425 * Note that the VFS-layer doesn't care about the return
426 * value of the readdir() call, as long as it's non-negative
427 * for success..
428 */
429 int proc_readdir(struct file * filp,
430 void * dirent, filldir_t filldir)
431 {
432 struct proc_dir_entry * de;
433 unsigned int ino;
434 int i;
435 struct inode *inode = filp->f_dentry->d_inode;
436 int ret = 0;
438 lock_kernel();
440 ino = inode->i_ino;
441 de = PDE(inode);
442 if (!de) {
443 ret = -EINVAL;
444 goto out;
445 }
446 i = filp->f_pos;
447 switch (i) {
448 case 0:
449 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
450 goto out;
451 i++;
452 filp->f_pos++;
453 /* fall through */
454 case 1:
455 if (filldir(dirent, "..", 2, i,
456 parent_ino(filp->f_dentry),
457 DT_DIR) < 0)
458 goto out;
459 i++;
460 filp->f_pos++;
461 /* fall through */
462 default:
463 spin_lock(&proc_subdir_lock);
464 de = de->subdir;
465 i -= 2;
466 for (;;) {
467 if (!de) {
468 ret = 1;
469 spin_unlock(&proc_subdir_lock);
470 goto out;
471 }
472 if (!i)
473 break;
474 de = de->next;
475 i--;
476 }
478 do {
479 /* filldir passes info to user space */
480 spin_unlock(&proc_subdir_lock);
481 if (filldir(dirent, de->name, de->namelen, filp->f_pos,
482 de->low_ino, de->mode >> 12) < 0)
483 goto out;
484 spin_lock(&proc_subdir_lock);
485 filp->f_pos++;
486 de = de->next;
487 } while (de);
488 spin_unlock(&proc_subdir_lock);
489 }
490 ret = 1;
491 out: unlock_kernel();
492 return ret;
493 }
495 /*
496 * These are the generic /proc directory operations. They
497 * use the in-memory "struct proc_dir_entry" tree to parse
498 * the /proc directory.
499 */
500 static struct file_operations proc_dir_operations = {
501 .read = generic_read_dir,
502 .readdir = proc_readdir,
503 };
505 /*
506 * proc directories can do almost nothing..
507 */
508 static struct inode_operations proc_dir_inode_operations = {
509 .lookup = proc_lookup,
510 .getattr = proc_getattr,
511 .setattr = proc_notify_change,
512 };
514 static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp)
515 {
516 unsigned int i;
518 i = get_inode_number();
519 if (i == 0)
520 return -EAGAIN;
521 dp->low_ino = i;
523 spin_lock(&proc_subdir_lock);
524 dp->next = dir->subdir;
525 dp->parent = dir;
526 dir->subdir = dp;
527 spin_unlock(&proc_subdir_lock);
529 if (S_ISDIR(dp->mode)) {
530 if (dp->proc_iops == NULL) {
531 dp->proc_fops = &proc_dir_operations;
532 dp->proc_iops = &proc_dir_inode_operations;
533 }
534 dir->nlink++;
535 } else if (S_ISLNK(dp->mode)) {
536 if (dp->proc_iops == NULL)
537 dp->proc_iops = &proc_link_inode_operations;
538 } else if (S_ISREG(dp->mode)) {
539 if (dp->proc_fops == NULL)
540 dp->proc_fops = &proc_file_operations;
541 if (dp->proc_iops == NULL)
542 dp->proc_iops = &proc_file_inode_operations;
543 }
544 return 0;
545 }
547 /*
548 * Kill an inode that got unregistered..
549 */
550 static void proc_kill_inodes(struct proc_dir_entry *de)
551 {
552 struct list_head *p;
553 struct super_block *sb = proc_mnt->mnt_sb;
555 /*
556 * Actually it's a partial revoke().
557 */
558 file_list_lock();
559 list_for_each(p, &sb->s_files) {
560 struct file * filp = list_entry(p, struct file, f_u.fu_list);
561 struct dentry * dentry = filp->f_dentry;
562 struct inode * inode;
563 const struct file_operations *fops;
565 if (dentry->d_op != &proc_dentry_operations)
566 continue;
567 inode = dentry->d_inode;
568 if (PDE(inode) != de)
569 continue;
570 fops = filp->f_op;
571 filp->f_op = NULL;
572 fops_put(fops);
573 }
574 file_list_unlock();
575 }
577 static struct proc_dir_entry *proc_create(struct proc_dir_entry **parent,
578 const char *name,
579 mode_t mode,
580 nlink_t nlink)
581 {
582 struct proc_dir_entry *ent = NULL;
583 const char *fn = name;
584 int len;
586 /* make sure name is valid */
587 if (!name || !strlen(name)) goto out;
589 if (!(*parent) && xlate_proc_name(name, parent, &fn) != 0)
590 goto out;
592 /* At this point there must not be any '/' characters beyond *fn */
593 if (strchr(fn, '/'))
594 goto out;
596 len = strlen(fn);
598 ent = kmalloc(sizeof(struct proc_dir_entry) + len + 1, GFP_KERNEL);
599 if (!ent) goto out;
601 memset(ent, 0, sizeof(struct proc_dir_entry));
602 memcpy(((char *) ent) + sizeof(struct proc_dir_entry), fn, len + 1);
603 ent->name = ((char *) ent) + sizeof(*ent);
604 ent->namelen = len;
605 ent->mode = mode;
606 ent->nlink = nlink;
607 out:
608 return ent;
609 }
611 struct proc_dir_entry *proc_symlink(const char *name,
612 struct proc_dir_entry *parent, const char *dest)
613 {
614 struct proc_dir_entry *ent;
616 ent = proc_create(&parent,name,
617 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
619 if (ent) {
620 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
621 if (ent->data) {
622 strcpy((char*)ent->data,dest);
623 if (proc_register(parent, ent) < 0) {
624 kfree(ent->data);
625 kfree(ent);
626 ent = NULL;
627 }
628 } else {
629 kfree(ent);
630 ent = NULL;
631 }
632 }
633 return ent;
634 }
636 struct proc_dir_entry *proc_mkdir_mode(const char *name, mode_t mode,
637 struct proc_dir_entry *parent)
638 {
639 struct proc_dir_entry *ent;
641 ent = proc_create(&parent, name, S_IFDIR | mode, 2);
642 if (ent) {
643 ent->proc_fops = &proc_dir_operations;
644 ent->proc_iops = &proc_dir_inode_operations;
646 if (proc_register(parent, ent) < 0) {
647 kfree(ent);
648 ent = NULL;
649 }
650 }
651 return ent;
652 }
654 struct proc_dir_entry *proc_mkdir(const char *name,
655 struct proc_dir_entry *parent)
656 {
657 return proc_mkdir_mode(name, S_IRUGO | S_IXUGO, parent);
658 }
660 struct proc_dir_entry *create_proc_entry(const char *name, mode_t mode,
661 struct proc_dir_entry *parent)
662 {
663 struct proc_dir_entry *ent;
664 nlink_t nlink;
666 if (S_ISDIR(mode)) {
667 if ((mode & S_IALLUGO) == 0)
668 mode |= S_IRUGO | S_IXUGO;
669 nlink = 2;
670 } else {
671 if ((mode & S_IFMT) == 0)
672 mode |= S_IFREG;
673 if ((mode & S_IALLUGO) == 0)
674 mode |= S_IRUGO;
675 nlink = 1;
676 }
678 ent = proc_create(&parent,name,mode,nlink);
679 if (ent) {
680 if (S_ISDIR(mode)) {
681 ent->proc_fops = &proc_dir_operations;
682 ent->proc_iops = &proc_dir_inode_operations;
683 }
684 if (proc_register(parent, ent) < 0) {
685 kfree(ent);
686 ent = NULL;
687 }
688 }
689 return ent;
690 }
692 void free_proc_entry(struct proc_dir_entry *de)
693 {
694 unsigned int ino = de->low_ino;
696 if (ino < PROC_DYNAMIC_FIRST)
697 return;
699 release_inode_number(ino);
701 if (S_ISLNK(de->mode) && de->data)
702 kfree(de->data);
703 kfree(de);
704 }
706 /*
707 * Remove a /proc entry and free it if it's not currently in use.
708 * If it is in use, we set the 'deleted' flag.
709 */
710 void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
711 {
712 struct proc_dir_entry **p;
713 struct proc_dir_entry *de;
714 const char *fn = name;
715 int len;
717 if (!parent && xlate_proc_name(name, &parent, &fn) != 0)
718 goto out;
719 len = strlen(fn);
721 spin_lock(&proc_subdir_lock);
722 for (p = &parent->subdir; *p; p=&(*p)->next ) {
723 if (!proc_match(len, fn, *p))
724 continue;
725 de = *p;
726 *p = de->next;
727 de->next = NULL;
728 if (S_ISDIR(de->mode))
729 parent->nlink--;
730 proc_kill_inodes(de);
731 de->nlink = 0;
732 WARN_ON(de->subdir);
733 if (!atomic_read(&de->count))
734 free_proc_entry(de);
735 else {
736 de->deleted = 1;
737 printk("remove_proc_entry: %s/%s busy, count=%d\n",
738 parent->name, de->name, atomic_read(&de->count));
739 }
740 break;
741 }
742 spin_unlock(&proc_subdir_lock);
743 out:
744 return;
745 }