2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/buffer_head.h>
21 #include <linux/pagemap.h>
22 #include <linux/highmem.h>
23 #include <linux/time.h>
24 #include <linux/init.h>
25 #include <linux/string.h>
26 #include <linux/smp_lock.h>
27 #include <linux/backing-dev.h>
28 #include <linux/mpage.h>
29 #include <linux/swap.h>
30 #include <linux/writeback.h>
31 #include <linux/statfs.h>
32 #include <linux/compat.h>
33 #include <linux/version.h>
36 #include "transaction.h"
37 #include "btrfs_inode.h"
39 #include "print-tree.h"
42 static int btrfs_copy_from_user(loff_t pos
, int num_pages
, int write_bytes
,
43 struct page
**prepared_pages
,
44 const char __user
* buf
)
48 int offset
= pos
& (PAGE_CACHE_SIZE
- 1);
50 for (i
= 0; i
< num_pages
&& write_bytes
> 0; i
++, offset
= 0) {
51 size_t count
= min_t(size_t,
52 PAGE_CACHE_SIZE
- offset
, write_bytes
);
53 struct page
*page
= prepared_pages
[i
];
54 fault_in_pages_readable(buf
, count
);
56 /* Copy data from userspace to the current page */
58 page_fault
= __copy_from_user(page_address(page
) + offset
,
60 /* Flush processor's dcache for this page */
61 flush_dcache_page(page
);
69 return page_fault
? -EFAULT
: 0;
72 static void btrfs_drop_pages(struct page
**pages
, size_t num_pages
)
75 for (i
= 0; i
< num_pages
; i
++) {
78 unlock_page(pages
[i
]);
79 mark_page_accessed(pages
[i
]);
80 page_cache_release(pages
[i
]);
84 static int insert_inline_extent(struct btrfs_trans_handle
*trans
,
85 struct btrfs_root
*root
, struct inode
*inode
,
86 u64 offset
, ssize_t size
,
87 struct page
*page
, size_t page_offset
)
90 struct btrfs_path
*path
;
92 struct btrfs_file_extent_item
*ei
;
97 path
= btrfs_alloc_path();
101 btrfs_set_trans_block_group(trans
, inode
);
103 key
.objectid
= inode
->i_ino
;
106 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
107 BUG_ON(size
>= PAGE_CACHE_SIZE
);
108 datasize
= btrfs_file_extent_calc_inline_size(size
);
110 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
116 ei
= btrfs_item_ptr(btrfs_buffer_leaf(path
->nodes
[0]),
117 path
->slots
[0], struct btrfs_file_extent_item
);
118 btrfs_set_file_extent_generation(ei
, trans
->transid
);
119 btrfs_set_file_extent_type(ei
,
120 BTRFS_FILE_EXTENT_INLINE
);
121 ptr
= btrfs_file_extent_inline_start(ei
);
123 kaddr
= kmap_atomic(page
, KM_USER0
);
124 btrfs_memcpy(root
, path
->nodes
[0]->b_data
,
125 ptr
, kaddr
+ page_offset
, size
);
126 kunmap_atomic(kaddr
, KM_USER0
);
127 btrfs_mark_buffer_dirty(path
->nodes
[0]);
129 btrfs_free_path(path
);
133 static int dirty_and_release_pages(struct btrfs_trans_handle
*trans
,
134 struct btrfs_root
*root
,
143 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
144 struct extent_map
*em
;
145 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
149 u64 end_of_last_block
;
150 u64 end_pos
= pos
+ write_bytes
;
151 loff_t isize
= i_size_read(inode
);
153 em
= alloc_extent_map(GFP_NOFS
);
157 em
->bdev
= inode
->i_sb
->s_bdev
;
159 start_pos
= pos
& ~((u64
)root
->blocksize
- 1);
160 num_blocks
= (write_bytes
+ pos
- start_pos
+ root
->blocksize
- 1) >>
163 down_read(&BTRFS_I(inode
)->root
->snap_sem
);
164 end_of_last_block
= start_pos
+ (num_blocks
<< inode
->i_blkbits
) - 1;
165 lock_extent(em_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
166 mutex_lock(&root
->fs_info
->fs_mutex
);
167 trans
= btrfs_start_transaction(root
, 1);
172 btrfs_set_trans_block_group(trans
, inode
);
173 inode
->i_blocks
+= num_blocks
<< 3;
176 if ((end_of_last_block
& 4095) == 0) {
177 printk("strange end of last %Lu %zu %Lu\n", start_pos
, write_bytes
, end_of_last_block
);
179 set_extent_uptodate(em_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
181 /* FIXME...EIEIO, ENOSPC and more */
183 /* insert any holes we need to create */
184 if (inode
->i_size
< start_pos
) {
185 u64 last_pos_in_file
;
187 u64 mask
= root
->blocksize
- 1;
188 last_pos_in_file
= (isize
+ mask
) & ~mask
;
189 hole_size
= (start_pos
- last_pos_in_file
+ mask
) & ~mask
;
191 if (last_pos_in_file
< start_pos
) {
192 err
= btrfs_drop_extents(trans
, root
, inode
,
194 last_pos_in_file
+ hole_size
,
199 hole_size
>>= inode
->i_blkbits
;
200 err
= btrfs_insert_file_extent(trans
, root
,
210 * either allocate an extent for the new bytes or setup the key
211 * to show we are doing inline data in the extent
213 if (isize
>= PAGE_CACHE_SIZE
|| pos
+ write_bytes
< inode
->i_size
||
214 pos
+ write_bytes
- start_pos
> BTRFS_MAX_INLINE_DATA_SIZE(root
)) {
216 for (i
= 0; i
< num_pages
; i
++) {
217 struct page
*p
= pages
[i
];
221 last_end
= pages
[num_pages
-1]->index
<< PAGE_CACHE_SHIFT
;
222 last_end
+= PAGE_CACHE_SIZE
- 1;
223 set_extent_delalloc(em_tree
, start_pos
, end_of_last_block
,
226 struct page
*p
= pages
[0];
227 /* step one, delete the existing extents in this range */
228 /* FIXME blocksize != pagesize */
229 err
= btrfs_drop_extents(trans
, root
, inode
, start_pos
,
230 (pos
+ write_bytes
+ root
->blocksize
-1) &
231 ~((u64
)root
->blocksize
- 1), &hint_block
);
235 err
= insert_inline_extent(trans
, root
, inode
, start_pos
,
236 end_pos
- start_pos
, p
, 0);
238 em
->start
= start_pos
;
240 em
->block_start
= EXTENT_MAP_INLINE
;
241 em
->block_end
= EXTENT_MAP_INLINE
;
242 add_extent_mapping(em_tree
, em
);
244 if (end_pos
> isize
) {
245 i_size_write(inode
, end_pos
);
246 btrfs_update_inode(trans
, root
, inode
);
249 err
= btrfs_end_transaction(trans
, root
);
251 mutex_unlock(&root
->fs_info
->fs_mutex
);
252 unlock_extent(em_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
254 up_read(&BTRFS_I(inode
)->root
->snap_sem
);
258 int btrfs_drop_extent_cache(struct inode
*inode
, u64 start
, u64 end
)
260 struct extent_map
*em
;
261 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
264 em
= lookup_extent_mapping(em_tree
, start
, end
);
267 remove_extent_mapping(em_tree
, em
);
270 /* once for the tree*/
277 * this is very complex, but the basic idea is to drop all extents
278 * in the range start - end. hint_block is filled in with a block number
279 * that would be a good hint to the block allocator for this file.
281 * If an extent intersects the range but is not entirely inside the range
282 * it is either truncated or split. Anything entirely inside the range
283 * is deleted from the tree.
285 int btrfs_drop_extents(struct btrfs_trans_handle
*trans
,
286 struct btrfs_root
*root
, struct inode
*inode
,
287 u64 start
, u64 end
, u64
*hint_block
)
290 struct btrfs_key key
;
291 struct btrfs_leaf
*leaf
;
293 struct btrfs_file_extent_item
*extent
;
296 struct btrfs_file_extent_item old
;
297 struct btrfs_path
*path
;
298 u64 search_start
= start
;
305 btrfs_drop_extent_cache(inode
, start
, end
- 1);
307 path
= btrfs_alloc_path();
312 btrfs_release_path(root
, path
);
313 ret
= btrfs_lookup_file_extent(trans
, root
, path
, inode
->i_ino
,
318 if (path
->slots
[0] == 0) {
330 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
331 slot
= path
->slots
[0];
333 btrfs_disk_key_to_cpu(&key
, &leaf
->items
[slot
].key
);
334 if (key
.offset
>= end
|| key
.objectid
!= inode
->i_ino
) {
337 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
) {
341 search_start
= key
.offset
;
344 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
345 extent
= btrfs_item_ptr(leaf
, slot
,
346 struct btrfs_file_extent_item
);
347 found_type
= btrfs_file_extent_type(extent
);
348 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
349 extent_end
= key
.offset
+
350 (btrfs_file_extent_num_blocks(extent
) <<
353 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
355 extent_end
= key
.offset
+
356 btrfs_file_extent_inline_len(leaf
->items
+
360 extent_end
= search_start
;
363 /* we found nothing we can drop */
364 if ((!found_extent
&& !found_inline
) ||
365 search_start
>= extent_end
) {
368 nritems
= btrfs_header_nritems(
369 btrfs_buffer_header(path
->nodes
[0]));
370 if (slot
>= nritems
- 1) {
371 nextret
= btrfs_next_leaf(root
, path
);
381 /* FIXME, there's only one inline extent allowed right now */
383 u64 mask
= root
->blocksize
- 1;
384 search_start
= (extent_end
+ mask
) & ~mask
;
386 search_start
= extent_end
;
388 if (end
< extent_end
&& end
>= key
.offset
) {
391 btrfs_file_extent_disk_blocknr(extent
);
392 u64 disk_num_blocks
=
393 btrfs_file_extent_disk_num_blocks(extent
);
394 memcpy(&old
, extent
, sizeof(old
));
395 if (disk_blocknr
!= 0) {
396 ret
= btrfs_inc_extent_ref(trans
, root
,
397 disk_blocknr
, disk_num_blocks
);
401 WARN_ON(found_inline
);
404 /* truncate existing extent */
405 if (start
> key
.offset
) {
409 WARN_ON(start
& (root
->blocksize
- 1));
411 new_num
= (start
- key
.offset
) >>
413 old_num
= btrfs_file_extent_num_blocks(extent
);
415 btrfs_file_extent_disk_blocknr(extent
);
416 if (btrfs_file_extent_disk_blocknr(extent
)) {
418 (old_num
- new_num
) << 3;
420 btrfs_set_file_extent_num_blocks(extent
,
422 btrfs_mark_buffer_dirty(path
->nodes
[0]);
427 /* delete the entire extent */
429 u64 disk_blocknr
= 0;
430 u64 disk_num_blocks
= 0;
431 u64 extent_num_blocks
= 0;
434 btrfs_file_extent_disk_blocknr(extent
);
436 btrfs_file_extent_disk_num_blocks(extent
);
438 btrfs_file_extent_num_blocks(extent
);
440 btrfs_file_extent_disk_blocknr(extent
);
442 ret
= btrfs_del_item(trans
, root
, path
);
443 /* TODO update progress marker and return */
445 btrfs_release_path(root
, path
);
447 if (found_extent
&& disk_blocknr
!= 0) {
448 inode
->i_blocks
-= extent_num_blocks
<< 3;
449 ret
= btrfs_free_extent(trans
, root
,
455 if (!bookend
&& search_start
>= end
) {
462 /* create bookend, splitting the extent in two */
463 if (bookend
&& found_extent
) {
464 struct btrfs_key ins
;
465 ins
.objectid
= inode
->i_ino
;
468 btrfs_set_key_type(&ins
, BTRFS_EXTENT_DATA_KEY
);
469 btrfs_release_path(root
, path
);
470 ret
= btrfs_insert_empty_item(trans
, root
, path
, &ins
,
474 btrfs_print_leaf(root
, btrfs_buffer_leaf(path
->nodes
[0]));
475 printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret
, ins
.objectid
, ins
.flags
, ins
.offset
, start
, end
, key
.offset
, extent_end
, keep
);
478 extent
= btrfs_item_ptr(
479 btrfs_buffer_leaf(path
->nodes
[0]),
481 struct btrfs_file_extent_item
);
482 btrfs_set_file_extent_disk_blocknr(extent
,
483 btrfs_file_extent_disk_blocknr(&old
));
484 btrfs_set_file_extent_disk_num_blocks(extent
,
485 btrfs_file_extent_disk_num_blocks(&old
));
487 btrfs_set_file_extent_offset(extent
,
488 btrfs_file_extent_offset(&old
) +
489 ((end
- key
.offset
) >> inode
->i_blkbits
));
490 WARN_ON(btrfs_file_extent_num_blocks(&old
) <
491 (extent_end
- end
) >> inode
->i_blkbits
);
492 btrfs_set_file_extent_num_blocks(extent
,
493 (extent_end
- end
) >> inode
->i_blkbits
);
495 btrfs_set_file_extent_type(extent
,
496 BTRFS_FILE_EXTENT_REG
);
497 btrfs_set_file_extent_generation(extent
,
498 btrfs_file_extent_generation(&old
));
499 btrfs_mark_buffer_dirty(path
->nodes
[0]);
500 if (btrfs_file_extent_disk_blocknr(&old
) != 0) {
502 btrfs_file_extent_num_blocks(extent
) << 3;
509 btrfs_free_path(path
);
514 * this gets pages into the page cache and locks them down
516 static int prepare_pages(struct btrfs_root
*root
,
521 unsigned long first_index
,
522 unsigned long last_index
,
526 unsigned long index
= pos
>> PAGE_CACHE_SHIFT
;
527 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
532 start_pos
= pos
& ~((u64
)root
->blocksize
- 1);
533 num_blocks
= (write_bytes
+ pos
- start_pos
+ root
->blocksize
- 1) >>
536 memset(pages
, 0, num_pages
* sizeof(struct page
*));
538 for (i
= 0; i
< num_pages
; i
++) {
539 pages
[i
] = grab_cache_page(inode
->i_mapping
, index
+ i
);
544 cancel_dirty_page(pages
[i
], PAGE_CACHE_SIZE
);
545 wait_on_page_writeback(pages
[i
]);
546 if (!PagePrivate(pages
[i
])) {
547 SetPagePrivate(pages
[i
]);
548 set_page_private(pages
[i
], 1);
549 WARN_ON(!pages
[i
]->mapping
->a_ops
->invalidatepage
);
550 page_cache_get(pages
[i
]);
552 WARN_ON(!PageLocked(pages
[i
]));
557 static ssize_t
btrfs_file_write(struct file
*file
, const char __user
*buf
,
558 size_t count
, loff_t
*ppos
)
561 size_t num_written
= 0;
564 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
565 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
566 struct page
**pages
= NULL
;
568 struct page
*pinned
[2];
569 unsigned long first_index
;
570 unsigned long last_index
;
572 nrptrs
= min((count
+ PAGE_CACHE_SIZE
- 1) / PAGE_CACHE_SIZE
,
573 PAGE_CACHE_SIZE
/ (sizeof(struct page
*)));
576 if (file
->f_flags
& O_DIRECT
)
579 vfs_check_frozen(inode
->i_sb
, SB_FREEZE_WRITE
);
580 current
->backing_dev_info
= inode
->i_mapping
->backing_dev_info
;
581 err
= generic_write_checks(file
, &pos
, &count
, S_ISBLK(inode
->i_mode
));
586 err
= remove_suid(file
->f_path
.dentry
);
589 file_update_time(file
);
591 pages
= kmalloc(nrptrs
* sizeof(struct page
*), GFP_KERNEL
);
593 mutex_lock(&inode
->i_mutex
);
594 first_index
= pos
>> PAGE_CACHE_SHIFT
;
595 last_index
= (pos
+ count
) >> PAGE_CACHE_SHIFT
;
598 * there are lots of better ways to do this, but this code
599 * makes sure the first and last page in the file range are
600 * up to date and ready for cow
602 if ((pos
& (PAGE_CACHE_SIZE
- 1))) {
603 pinned
[0] = grab_cache_page(inode
->i_mapping
, first_index
);
604 if (!PageUptodate(pinned
[0])) {
605 ret
= btrfs_readpage(NULL
, pinned
[0]);
607 wait_on_page_locked(pinned
[0]);
609 unlock_page(pinned
[0]);
612 if ((pos
+ count
) & (PAGE_CACHE_SIZE
- 1)) {
613 pinned
[1] = grab_cache_page(inode
->i_mapping
, last_index
);
614 if (!PageUptodate(pinned
[1])) {
615 ret
= btrfs_readpage(NULL
, pinned
[1]);
617 wait_on_page_locked(pinned
[1]);
619 unlock_page(pinned
[1]);
624 size_t offset
= pos
& (PAGE_CACHE_SIZE
- 1);
625 size_t write_bytes
= min(count
, nrptrs
*
626 (size_t)PAGE_CACHE_SIZE
-
628 size_t num_pages
= (write_bytes
+ PAGE_CACHE_SIZE
- 1) >>
631 WARN_ON(num_pages
> nrptrs
);
632 memset(pages
, 0, sizeof(pages
));
633 ret
= prepare_pages(root
, file
, pages
, num_pages
,
634 pos
, first_index
, last_index
,
639 ret
= btrfs_copy_from_user(pos
, num_pages
,
640 write_bytes
, pages
, buf
);
642 btrfs_drop_pages(pages
, num_pages
);
646 ret
= dirty_and_release_pages(NULL
, root
, file
, pages
,
647 num_pages
, pos
, write_bytes
);
648 btrfs_drop_pages(pages
, num_pages
);
653 count
-= write_bytes
;
655 num_written
+= write_bytes
;
657 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, num_pages
);
658 btrfs_btree_balance_dirty(root
);
661 mutex_unlock(&inode
->i_mutex
);
665 page_cache_release(pinned
[0]);
667 page_cache_release(pinned
[1]);
669 current
->backing_dev_info
= NULL
;
670 return num_written
? num_written
: err
;
673 static int btrfs_sync_file(struct file
*file
,
674 struct dentry
*dentry
, int datasync
)
676 struct inode
*inode
= dentry
->d_inode
;
677 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
679 struct btrfs_trans_handle
*trans
;
682 * check the transaction that last modified this inode
683 * and see if its already been committed
685 mutex_lock(&root
->fs_info
->fs_mutex
);
686 if (!BTRFS_I(inode
)->last_trans
)
688 mutex_lock(&root
->fs_info
->trans_mutex
);
689 if (BTRFS_I(inode
)->last_trans
<=
690 root
->fs_info
->last_trans_committed
) {
691 BTRFS_I(inode
)->last_trans
= 0;
692 mutex_unlock(&root
->fs_info
->trans_mutex
);
695 mutex_unlock(&root
->fs_info
->trans_mutex
);
698 * ok we haven't committed the transaction yet, lets do a commit
700 trans
= btrfs_start_transaction(root
, 1);
705 ret
= btrfs_commit_transaction(trans
, root
);
707 mutex_unlock(&root
->fs_info
->fs_mutex
);
708 return ret
> 0 ? EIO
: ret
;
711 static struct vm_operations_struct btrfs_file_vm_ops
= {
712 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
713 .nopage
= filemap_nopage
,
714 .populate
= filemap_populate
,
716 .fault
= filemap_fault
,
718 .page_mkwrite
= btrfs_page_mkwrite
,
721 static int btrfs_file_mmap(struct file
*filp
, struct vm_area_struct
*vma
)
723 vma
->vm_ops
= &btrfs_file_vm_ops
;
728 struct file_operations btrfs_file_operations
= {
729 .llseek
= generic_file_llseek
,
730 .read
= do_sync_read
,
731 .aio_read
= generic_file_aio_read
,
732 .write
= btrfs_file_write
,
733 .mmap
= btrfs_file_mmap
,
734 .open
= generic_file_open
,
735 .ioctl
= btrfs_ioctl
,
736 .fsync
= btrfs_sync_file
,
738 .compat_ioctl
= btrfs_compat_ioctl
,