Linux 3.11-rc3
[linux-2.6.git] / fs / ext4 / ioctl.c
blob9491ac0590f746b9abe56e484e7646022a0bd1c3
1 /*
2 * linux/fs/ext4/ioctl.c
4 * Copyright (C) 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 */
10 #include <linux/fs.h>
11 #include <linux/jbd2.h>
12 #include <linux/capability.h>
13 #include <linux/time.h>
14 #include <linux/compat.h>
15 #include <linux/mount.h>
16 #include <linux/file.h>
17 #include <asm/uaccess.h>
18 #include "ext4_jbd2.h"
19 #include "ext4.h"
20 #include "ext4_extents.h"
22 #define MAX_32_NUM ((((unsigned long long) 1) << 32) - 1)
24 /**
25 * Swap memory between @a and @b for @len bytes.
27 * @a: pointer to first memory area
28 * @b: pointer to second memory area
29 * @len: number of bytes to swap
32 static void memswap(void *a, void *b, size_t len)
34 unsigned char *ap, *bp;
35 unsigned char tmp;
37 ap = (unsigned char *)a;
38 bp = (unsigned char *)b;
39 while (len-- > 0) {
40 tmp = *ap;
41 *ap = *bp;
42 *bp = tmp;
43 ap++;
44 bp++;
48 /**
49 * Swap i_data and associated attributes between @inode1 and @inode2.
50 * This function is used for the primary swap between inode1 and inode2
51 * and also to revert this primary swap in case of errors.
53 * Therefore you have to make sure, that calling this method twice
54 * will revert all changes.
56 * @inode1: pointer to first inode
57 * @inode2: pointer to second inode
59 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
61 loff_t isize;
62 struct ext4_inode_info *ei1;
63 struct ext4_inode_info *ei2;
65 ei1 = EXT4_I(inode1);
66 ei2 = EXT4_I(inode2);
68 memswap(&inode1->i_flags, &inode2->i_flags, sizeof(inode1->i_flags));
69 memswap(&inode1->i_version, &inode2->i_version,
70 sizeof(inode1->i_version));
71 memswap(&inode1->i_blocks, &inode2->i_blocks,
72 sizeof(inode1->i_blocks));
73 memswap(&inode1->i_bytes, &inode2->i_bytes, sizeof(inode1->i_bytes));
74 memswap(&inode1->i_atime, &inode2->i_atime, sizeof(inode1->i_atime));
75 memswap(&inode1->i_mtime, &inode2->i_mtime, sizeof(inode1->i_mtime));
77 memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
78 memswap(&ei1->i_flags, &ei2->i_flags, sizeof(ei1->i_flags));
79 memswap(&ei1->i_disksize, &ei2->i_disksize, sizeof(ei1->i_disksize));
80 memswap(&ei1->i_es_tree, &ei2->i_es_tree, sizeof(ei1->i_es_tree));
81 memswap(&ei1->i_es_lru_nr, &ei2->i_es_lru_nr, sizeof(ei1->i_es_lru_nr));
83 isize = i_size_read(inode1);
84 i_size_write(inode1, i_size_read(inode2));
85 i_size_write(inode2, isize);
88 /**
89 * Swap the information from the given @inode and the inode
90 * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
91 * important fields of the inodes.
93 * @sb: the super block of the filesystem
94 * @inode: the inode to swap with EXT4_BOOT_LOADER_INO
97 static long swap_inode_boot_loader(struct super_block *sb,
98 struct inode *inode)
100 handle_t *handle;
101 int err;
102 struct inode *inode_bl;
103 struct ext4_inode_info *ei;
104 struct ext4_inode_info *ei_bl;
105 struct ext4_sb_info *sbi;
107 if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode)) {
108 err = -EINVAL;
109 goto swap_boot_out;
112 if (!inode_owner_or_capable(inode) || !capable(CAP_SYS_ADMIN)) {
113 err = -EPERM;
114 goto swap_boot_out;
117 sbi = EXT4_SB(sb);
118 ei = EXT4_I(inode);
120 inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO);
121 if (IS_ERR(inode_bl)) {
122 err = PTR_ERR(inode_bl);
123 goto swap_boot_out;
125 ei_bl = EXT4_I(inode_bl);
127 filemap_flush(inode->i_mapping);
128 filemap_flush(inode_bl->i_mapping);
130 /* Protect orig inodes against a truncate and make sure,
131 * that only 1 swap_inode_boot_loader is running. */
132 ext4_inode_double_lock(inode, inode_bl);
134 truncate_inode_pages(&inode->i_data, 0);
135 truncate_inode_pages(&inode_bl->i_data, 0);
137 /* Wait for all existing dio workers */
138 ext4_inode_block_unlocked_dio(inode);
139 ext4_inode_block_unlocked_dio(inode_bl);
140 inode_dio_wait(inode);
141 inode_dio_wait(inode_bl);
143 handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
144 if (IS_ERR(handle)) {
145 err = -EINVAL;
146 goto swap_boot_out;
149 /* Protect extent tree against block allocations via delalloc */
150 ext4_double_down_write_data_sem(inode, inode_bl);
152 if (inode_bl->i_nlink == 0) {
153 /* this inode has never been used as a BOOT_LOADER */
154 set_nlink(inode_bl, 1);
155 i_uid_write(inode_bl, 0);
156 i_gid_write(inode_bl, 0);
157 inode_bl->i_flags = 0;
158 ei_bl->i_flags = 0;
159 inode_bl->i_version = 1;
160 i_size_write(inode_bl, 0);
161 inode_bl->i_mode = S_IFREG;
162 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
163 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
164 ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
165 ext4_ext_tree_init(handle, inode_bl);
166 } else
167 memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
170 swap_inode_data(inode, inode_bl);
172 inode->i_ctime = inode_bl->i_ctime = ext4_current_time(inode);
174 spin_lock(&sbi->s_next_gen_lock);
175 inode->i_generation = sbi->s_next_generation++;
176 inode_bl->i_generation = sbi->s_next_generation++;
177 spin_unlock(&sbi->s_next_gen_lock);
179 ext4_discard_preallocations(inode);
181 err = ext4_mark_inode_dirty(handle, inode);
182 if (err < 0) {
183 ext4_warning(inode->i_sb,
184 "couldn't mark inode #%lu dirty (err %d)",
185 inode->i_ino, err);
186 /* Revert all changes: */
187 swap_inode_data(inode, inode_bl);
188 } else {
189 err = ext4_mark_inode_dirty(handle, inode_bl);
190 if (err < 0) {
191 ext4_warning(inode_bl->i_sb,
192 "couldn't mark inode #%lu dirty (err %d)",
193 inode_bl->i_ino, err);
194 /* Revert all changes: */
195 swap_inode_data(inode, inode_bl);
196 ext4_mark_inode_dirty(handle, inode);
200 ext4_journal_stop(handle);
202 ext4_double_up_write_data_sem(inode, inode_bl);
204 ext4_inode_resume_unlocked_dio(inode);
205 ext4_inode_resume_unlocked_dio(inode_bl);
207 ext4_inode_double_unlock(inode, inode_bl);
209 iput(inode_bl);
211 swap_boot_out:
212 return err;
215 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
217 struct inode *inode = file_inode(filp);
218 struct super_block *sb = inode->i_sb;
219 struct ext4_inode_info *ei = EXT4_I(inode);
220 unsigned int flags;
222 ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
224 switch (cmd) {
225 case EXT4_IOC_GETFLAGS:
226 ext4_get_inode_flags(ei);
227 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
228 return put_user(flags, (int __user *) arg);
229 case EXT4_IOC_SETFLAGS: {
230 handle_t *handle = NULL;
231 int err, migrate = 0;
232 struct ext4_iloc iloc;
233 unsigned int oldflags, mask, i;
234 unsigned int jflag;
236 if (!inode_owner_or_capable(inode))
237 return -EACCES;
239 if (get_user(flags, (int __user *) arg))
240 return -EFAULT;
242 err = mnt_want_write_file(filp);
243 if (err)
244 return err;
246 flags = ext4_mask_flags(inode->i_mode, flags);
248 err = -EPERM;
249 mutex_lock(&inode->i_mutex);
250 /* Is it quota file? Do not allow user to mess with it */
251 if (IS_NOQUOTA(inode))
252 goto flags_out;
254 oldflags = ei->i_flags;
256 /* The JOURNAL_DATA flag is modifiable only by root */
257 jflag = flags & EXT4_JOURNAL_DATA_FL;
260 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
261 * the relevant capability.
263 * This test looks nicer. Thanks to Pauline Middelink
265 if ((flags ^ oldflags) & (EXT4_APPEND_FL | EXT4_IMMUTABLE_FL)) {
266 if (!capable(CAP_LINUX_IMMUTABLE))
267 goto flags_out;
271 * The JOURNAL_DATA flag can only be changed by
272 * the relevant capability.
274 if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
275 if (!capable(CAP_SYS_RESOURCE))
276 goto flags_out;
278 if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
279 migrate = 1;
281 if (flags & EXT4_EOFBLOCKS_FL) {
282 /* we don't support adding EOFBLOCKS flag */
283 if (!(oldflags & EXT4_EOFBLOCKS_FL)) {
284 err = -EOPNOTSUPP;
285 goto flags_out;
287 } else if (oldflags & EXT4_EOFBLOCKS_FL)
288 ext4_truncate(inode);
290 handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
291 if (IS_ERR(handle)) {
292 err = PTR_ERR(handle);
293 goto flags_out;
295 if (IS_SYNC(inode))
296 ext4_handle_sync(handle);
297 err = ext4_reserve_inode_write(handle, inode, &iloc);
298 if (err)
299 goto flags_err;
301 for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
302 if (!(mask & EXT4_FL_USER_MODIFIABLE))
303 continue;
304 if (mask & flags)
305 ext4_set_inode_flag(inode, i);
306 else
307 ext4_clear_inode_flag(inode, i);
310 ext4_set_inode_flags(inode);
311 inode->i_ctime = ext4_current_time(inode);
313 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
314 flags_err:
315 ext4_journal_stop(handle);
316 if (err)
317 goto flags_out;
319 if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL))
320 err = ext4_change_inode_journal_flag(inode, jflag);
321 if (err)
322 goto flags_out;
323 if (migrate) {
324 if (flags & EXT4_EXTENTS_FL)
325 err = ext4_ext_migrate(inode);
326 else
327 err = ext4_ind_migrate(inode);
330 flags_out:
331 mutex_unlock(&inode->i_mutex);
332 mnt_drop_write_file(filp);
333 return err;
335 case EXT4_IOC_GETVERSION:
336 case EXT4_IOC_GETVERSION_OLD:
337 return put_user(inode->i_generation, (int __user *) arg);
338 case EXT4_IOC_SETVERSION:
339 case EXT4_IOC_SETVERSION_OLD: {
340 handle_t *handle;
341 struct ext4_iloc iloc;
342 __u32 generation;
343 int err;
345 if (!inode_owner_or_capable(inode))
346 return -EPERM;
348 if (EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
349 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
350 ext4_warning(sb, "Setting inode version is not "
351 "supported with metadata_csum enabled.");
352 return -ENOTTY;
355 err = mnt_want_write_file(filp);
356 if (err)
357 return err;
358 if (get_user(generation, (int __user *) arg)) {
359 err = -EFAULT;
360 goto setversion_out;
363 mutex_lock(&inode->i_mutex);
364 handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
365 if (IS_ERR(handle)) {
366 err = PTR_ERR(handle);
367 goto unlock_out;
369 err = ext4_reserve_inode_write(handle, inode, &iloc);
370 if (err == 0) {
371 inode->i_ctime = ext4_current_time(inode);
372 inode->i_generation = generation;
373 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
375 ext4_journal_stop(handle);
377 unlock_out:
378 mutex_unlock(&inode->i_mutex);
379 setversion_out:
380 mnt_drop_write_file(filp);
381 return err;
383 case EXT4_IOC_GROUP_EXTEND: {
384 ext4_fsblk_t n_blocks_count;
385 int err, err2=0;
387 err = ext4_resize_begin(sb);
388 if (err)
389 return err;
391 if (get_user(n_blocks_count, (__u32 __user *)arg)) {
392 err = -EFAULT;
393 goto group_extend_out;
396 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
397 EXT4_FEATURE_RO_COMPAT_BIGALLOC)) {
398 ext4_msg(sb, KERN_ERR,
399 "Online resizing not supported with bigalloc");
400 err = -EOPNOTSUPP;
401 goto group_extend_out;
404 err = mnt_want_write_file(filp);
405 if (err)
406 goto group_extend_out;
408 err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
409 if (EXT4_SB(sb)->s_journal) {
410 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
411 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
412 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
414 if (err == 0)
415 err = err2;
416 mnt_drop_write_file(filp);
417 group_extend_out:
418 ext4_resize_end(sb);
419 return err;
422 case EXT4_IOC_MOVE_EXT: {
423 struct move_extent me;
424 struct fd donor;
425 int err;
427 if (!(filp->f_mode & FMODE_READ) ||
428 !(filp->f_mode & FMODE_WRITE))
429 return -EBADF;
431 if (copy_from_user(&me,
432 (struct move_extent __user *)arg, sizeof(me)))
433 return -EFAULT;
434 me.moved_len = 0;
436 donor = fdget(me.donor_fd);
437 if (!donor.file)
438 return -EBADF;
440 if (!(donor.file->f_mode & FMODE_WRITE)) {
441 err = -EBADF;
442 goto mext_out;
445 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
446 EXT4_FEATURE_RO_COMPAT_BIGALLOC)) {
447 ext4_msg(sb, KERN_ERR,
448 "Online defrag not supported with bigalloc");
449 err = -EOPNOTSUPP;
450 goto mext_out;
453 err = mnt_want_write_file(filp);
454 if (err)
455 goto mext_out;
457 err = ext4_move_extents(filp, donor.file, me.orig_start,
458 me.donor_start, me.len, &me.moved_len);
459 mnt_drop_write_file(filp);
461 if (copy_to_user((struct move_extent __user *)arg,
462 &me, sizeof(me)))
463 err = -EFAULT;
464 mext_out:
465 fdput(donor);
466 return err;
469 case EXT4_IOC_GROUP_ADD: {
470 struct ext4_new_group_data input;
471 int err, err2=0;
473 err = ext4_resize_begin(sb);
474 if (err)
475 return err;
477 if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
478 sizeof(input))) {
479 err = -EFAULT;
480 goto group_add_out;
483 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
484 EXT4_FEATURE_RO_COMPAT_BIGALLOC)) {
485 ext4_msg(sb, KERN_ERR,
486 "Online resizing not supported with bigalloc");
487 err = -EOPNOTSUPP;
488 goto group_add_out;
491 err = mnt_want_write_file(filp);
492 if (err)
493 goto group_add_out;
495 err = ext4_group_add(sb, &input);
496 if (EXT4_SB(sb)->s_journal) {
497 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
498 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
499 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
501 if (err == 0)
502 err = err2;
503 mnt_drop_write_file(filp);
504 if (!err && ext4_has_group_desc_csum(sb) &&
505 test_opt(sb, INIT_INODE_TABLE))
506 err = ext4_register_li_request(sb, input.group);
507 group_add_out:
508 ext4_resize_end(sb);
509 return err;
512 case EXT4_IOC_MIGRATE:
514 int err;
515 if (!inode_owner_or_capable(inode))
516 return -EACCES;
518 err = mnt_want_write_file(filp);
519 if (err)
520 return err;
522 * inode_mutex prevent write and truncate on the file.
523 * Read still goes through. We take i_data_sem in
524 * ext4_ext_swap_inode_data before we switch the
525 * inode format to prevent read.
527 mutex_lock(&(inode->i_mutex));
528 err = ext4_ext_migrate(inode);
529 mutex_unlock(&(inode->i_mutex));
530 mnt_drop_write_file(filp);
531 return err;
534 case EXT4_IOC_ALLOC_DA_BLKS:
536 int err;
537 if (!inode_owner_or_capable(inode))
538 return -EACCES;
540 err = mnt_want_write_file(filp);
541 if (err)
542 return err;
543 err = ext4_alloc_da_blocks(inode);
544 mnt_drop_write_file(filp);
545 return err;
548 case EXT4_IOC_SWAP_BOOT:
549 if (!(filp->f_mode & FMODE_WRITE))
550 return -EBADF;
551 return swap_inode_boot_loader(sb, inode);
553 case EXT4_IOC_RESIZE_FS: {
554 ext4_fsblk_t n_blocks_count;
555 int err = 0, err2 = 0;
556 ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
558 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
559 EXT4_FEATURE_RO_COMPAT_BIGALLOC)) {
560 ext4_msg(sb, KERN_ERR,
561 "Online resizing not (yet) supported with bigalloc");
562 return -EOPNOTSUPP;
565 if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
566 sizeof(__u64))) {
567 return -EFAULT;
570 err = ext4_resize_begin(sb);
571 if (err)
572 return err;
574 err = mnt_want_write_file(filp);
575 if (err)
576 goto resizefs_out;
578 err = ext4_resize_fs(sb, n_blocks_count);
579 if (EXT4_SB(sb)->s_journal) {
580 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
581 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
582 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
584 if (err == 0)
585 err = err2;
586 mnt_drop_write_file(filp);
587 if (!err && (o_group > EXT4_SB(sb)->s_groups_count) &&
588 ext4_has_group_desc_csum(sb) &&
589 test_opt(sb, INIT_INODE_TABLE))
590 err = ext4_register_li_request(sb, o_group);
592 resizefs_out:
593 ext4_resize_end(sb);
594 return err;
597 case FITRIM:
599 struct request_queue *q = bdev_get_queue(sb->s_bdev);
600 struct fstrim_range range;
601 int ret = 0;
603 if (!capable(CAP_SYS_ADMIN))
604 return -EPERM;
606 if (!blk_queue_discard(q))
607 return -EOPNOTSUPP;
609 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
610 sizeof(range)))
611 return -EFAULT;
613 range.minlen = max((unsigned int)range.minlen,
614 q->limits.discard_granularity);
615 ret = ext4_trim_fs(sb, &range);
616 if (ret < 0)
617 return ret;
619 if (copy_to_user((struct fstrim_range __user *)arg, &range,
620 sizeof(range)))
621 return -EFAULT;
623 return 0;
626 default:
627 return -ENOTTY;
631 #ifdef CONFIG_COMPAT
632 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
634 /* These are just misnamed, they actually get/put from/to user an int */
635 switch (cmd) {
636 case EXT4_IOC32_GETFLAGS:
637 cmd = EXT4_IOC_GETFLAGS;
638 break;
639 case EXT4_IOC32_SETFLAGS:
640 cmd = EXT4_IOC_SETFLAGS;
641 break;
642 case EXT4_IOC32_GETVERSION:
643 cmd = EXT4_IOC_GETVERSION;
644 break;
645 case EXT4_IOC32_SETVERSION:
646 cmd = EXT4_IOC_SETVERSION;
647 break;
648 case EXT4_IOC32_GROUP_EXTEND:
649 cmd = EXT4_IOC_GROUP_EXTEND;
650 break;
651 case EXT4_IOC32_GETVERSION_OLD:
652 cmd = EXT4_IOC_GETVERSION_OLD;
653 break;
654 case EXT4_IOC32_SETVERSION_OLD:
655 cmd = EXT4_IOC_SETVERSION_OLD;
656 break;
657 case EXT4_IOC32_GETRSVSZ:
658 cmd = EXT4_IOC_GETRSVSZ;
659 break;
660 case EXT4_IOC32_SETRSVSZ:
661 cmd = EXT4_IOC_SETRSVSZ;
662 break;
663 case EXT4_IOC32_GROUP_ADD: {
664 struct compat_ext4_new_group_input __user *uinput;
665 struct ext4_new_group_input input;
666 mm_segment_t old_fs;
667 int err;
669 uinput = compat_ptr(arg);
670 err = get_user(input.group, &uinput->group);
671 err |= get_user(input.block_bitmap, &uinput->block_bitmap);
672 err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
673 err |= get_user(input.inode_table, &uinput->inode_table);
674 err |= get_user(input.blocks_count, &uinput->blocks_count);
675 err |= get_user(input.reserved_blocks,
676 &uinput->reserved_blocks);
677 if (err)
678 return -EFAULT;
679 old_fs = get_fs();
680 set_fs(KERNEL_DS);
681 err = ext4_ioctl(file, EXT4_IOC_GROUP_ADD,
682 (unsigned long) &input);
683 set_fs(old_fs);
684 return err;
686 case EXT4_IOC_MOVE_EXT:
687 case FITRIM:
688 case EXT4_IOC_RESIZE_FS:
689 break;
690 default:
691 return -ENOIOCTLCMD;
693 return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
695 #endif