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.
20 #include <linux/pagemap.h>
21 #include <linux/highmem.h>
22 #include <linux/time.h>
23 #include <linux/init.h>
24 #include <linux/string.h>
25 #include <linux/smp_lock.h>
26 #include <linux/backing-dev.h>
27 #include <linux/mpage.h>
28 #include <linux/swap.h>
29 #include <linux/writeback.h>
30 #include <linux/statfs.h>
31 #include <linux/compat.h>
32 #include <linux/version.h>
35 #include "transaction.h"
36 #include "btrfs_inode.h"
37 #include "ordered-data.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 noinline
insert_inline_extent(struct btrfs_trans_handle
*trans
,
85 struct btrfs_root
*root
, struct inode
*inode
,
86 u64 offset
, size_t size
,
87 struct page
**pages
, size_t page_offset
,
91 struct btrfs_path
*path
;
92 struct extent_buffer
*leaf
;
95 struct btrfs_file_extent_item
*ei
;
103 path
= btrfs_alloc_path();
107 btrfs_set_trans_block_group(trans
, inode
);
109 key
.objectid
= inode
->i_ino
;
111 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
113 ret
= btrfs_search_slot(trans
, root
, &key
, path
, 0, 1);
119 struct btrfs_key found_key
;
121 if (path
->slots
[0] == 0)
125 leaf
= path
->nodes
[0];
126 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
128 if (found_key
.objectid
!= inode
->i_ino
)
131 if (found_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
133 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
134 struct btrfs_file_extent_item
);
136 if (btrfs_file_extent_type(leaf
, ei
) !=
137 BTRFS_FILE_EXTENT_INLINE
) {
140 btrfs_item_key_to_cpu(leaf
, &key
, path
->slots
[0]);
147 leaf
= path
->nodes
[0];
148 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
149 struct btrfs_file_extent_item
);
151 if (btrfs_file_extent_type(leaf
, ei
) !=
152 BTRFS_FILE_EXTENT_INLINE
) {
154 btrfs_print_leaf(root
, leaf
);
155 printk("found wasn't inline offset %Lu inode %lu\n",
156 offset
, inode
->i_ino
);
159 found_size
= btrfs_file_extent_inline_len(leaf
,
160 btrfs_item_nr(leaf
, path
->slots
[0]));
161 found_end
= key
.offset
+ found_size
;
163 if (found_end
< offset
+ size
) {
164 btrfs_release_path(root
, path
);
165 ret
= btrfs_search_slot(trans
, root
, &key
, path
,
166 offset
+ size
- found_end
, 1);
169 ret
= btrfs_extend_item(trans
, root
, path
,
170 offset
+ size
- found_end
);
175 leaf
= path
->nodes
[0];
176 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
177 struct btrfs_file_extent_item
);
178 inode
->i_blocks
+= (offset
+ size
- found_end
) >> 9;
180 if (found_end
< offset
) {
181 ptr
= btrfs_file_extent_inline_start(ei
) + found_size
;
182 memset_extent_buffer(leaf
, 0, ptr
, offset
- found_end
);
186 btrfs_release_path(root
, path
);
187 datasize
= offset
+ size
- key
.offset
;
188 inode
->i_blocks
+= datasize
>> 9;
189 datasize
= btrfs_file_extent_calc_inline_size(datasize
);
190 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
194 printk("got bad ret %d\n", ret
);
197 leaf
= path
->nodes
[0];
198 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
199 struct btrfs_file_extent_item
);
200 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
201 btrfs_set_file_extent_type(leaf
, ei
, BTRFS_FILE_EXTENT_INLINE
);
203 ptr
= btrfs_file_extent_inline_start(ei
) + offset
- key
.offset
;
209 kaddr
= kmap_atomic(page
, KM_USER0
);
210 cur_size
= min_t(size_t, PAGE_CACHE_SIZE
- page_offset
, size
);
211 write_extent_buffer(leaf
, kaddr
+ page_offset
, ptr
, cur_size
);
212 kunmap_atomic(kaddr
, KM_USER0
);
216 if (i
>= num_pages
) {
217 printk("i %d num_pages %d\n", i
, num_pages
);
221 btrfs_mark_buffer_dirty(leaf
);
223 btrfs_free_path(path
);
227 static int noinline
dirty_and_release_pages(struct btrfs_trans_handle
*trans
,
228 struct btrfs_root
*root
,
237 struct inode
*inode
= fdentry(file
)->d_inode
;
238 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
242 u64 end_of_last_block
;
243 u64 end_pos
= pos
+ write_bytes
;
245 loff_t isize
= i_size_read(inode
);
247 start_pos
= pos
& ~((u64
)root
->sectorsize
- 1);
248 num_bytes
= (write_bytes
+ pos
- start_pos
+
249 root
->sectorsize
- 1) & ~((u64
)root
->sectorsize
- 1);
251 end_of_last_block
= start_pos
+ num_bytes
- 1;
253 lock_extent(io_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
254 mutex_lock(&root
->fs_info
->fs_mutex
);
255 trans
= btrfs_start_transaction(root
, 1);
260 btrfs_set_trans_block_group(trans
, inode
);
263 if ((end_of_last_block
& 4095) == 0) {
264 printk("strange end of last %Lu %zu %Lu\n", start_pos
, write_bytes
, end_of_last_block
);
266 set_extent_uptodate(io_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
268 /* FIXME...EIEIO, ENOSPC and more */
269 /* insert any holes we need to create */
270 if (isize
< end_pos
) {
271 u64 last_pos_in_file
;
273 u64 mask
= root
->sectorsize
- 1;
274 last_pos_in_file
= (isize
+ mask
) & ~mask
;
275 hole_size
= (end_pos
- last_pos_in_file
+ mask
) & ~mask
;
276 if (last_pos_in_file
< end_pos
) {
277 err
= btrfs_drop_extents(trans
, root
, inode
,
279 last_pos_in_file
+ hole_size
,
285 err
= btrfs_insert_file_extent(trans
, root
,
289 btrfs_drop_extent_cache(inode
, last_pos_in_file
,
290 last_pos_in_file
+ hole_size
-1);
291 btrfs_check_file(root
, inode
);
298 * either allocate an extent for the new bytes or setup the key
299 * to show we are doing inline data in the extent
301 inline_size
= end_pos
;
302 if (isize
>= BTRFS_MAX_INLINE_DATA_SIZE(root
) ||
303 inline_size
> root
->fs_info
->max_inline
||
304 (inline_size
& (root
->sectorsize
-1)) == 0 ||
305 inline_size
>= BTRFS_MAX_INLINE_DATA_SIZE(root
)) {
307 u64 existing_delalloc
= 0;
309 for (i
= 0; i
< num_pages
; i
++) {
310 struct page
*p
= pages
[i
];
314 last_end
= (u64
)(pages
[num_pages
-1]->index
) <<
316 last_end
+= PAGE_CACHE_SIZE
- 1;
317 if (start_pos
< isize
) {
318 u64 delalloc_start
= start_pos
;
319 existing_delalloc
= count_range_bits(io_tree
,
321 end_of_last_block
, (u64
)-1,
324 set_extent_delalloc(io_tree
, start_pos
, end_of_last_block
,
326 btrfs_add_ordered_inode(inode
);
329 /* step one, delete the existing extents in this range */
330 aligned_end
= (pos
+ write_bytes
+ root
->sectorsize
- 1) &
331 ~((u64
)root
->sectorsize
- 1);
332 err
= btrfs_drop_extents(trans
, root
, inode
, start_pos
,
333 aligned_end
, aligned_end
, &hint_byte
);
336 if (isize
> inline_size
)
337 inline_size
= min_t(u64
, isize
, aligned_end
);
338 inline_size
-= start_pos
;
339 err
= insert_inline_extent(trans
, root
, inode
, start_pos
,
340 inline_size
, pages
, 0, num_pages
);
341 btrfs_drop_extent_cache(inode
, start_pos
, aligned_end
- 1);
344 if (end_pos
> isize
) {
345 i_size_write(inode
, end_pos
);
346 btrfs_update_inode(trans
, root
, inode
);
349 err
= btrfs_end_transaction(trans
, root
);
351 mutex_unlock(&root
->fs_info
->fs_mutex
);
352 unlock_extent(io_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
356 int btrfs_drop_extent_cache(struct inode
*inode
, u64 start
, u64 end
)
358 struct extent_map
*em
;
359 struct extent_map
*split
= NULL
;
360 struct extent_map
*split2
= NULL
;
361 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
362 u64 len
= end
- start
+ 1;
366 if (end
== (u64
)-1) {
372 split
= alloc_extent_map(GFP_NOFS
);
374 split2
= alloc_extent_map(GFP_NOFS
);
376 spin_lock(&em_tree
->lock
);
377 em
= lookup_extent_mapping(em_tree
, start
, len
);
379 spin_unlock(&em_tree
->lock
);
382 remove_extent_mapping(em_tree
, em
);
384 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
&&
386 split
->start
= em
->start
;
387 split
->len
= start
- em
->start
;
388 split
->block_start
= em
->block_start
;
389 split
->bdev
= em
->bdev
;
390 split
->flags
= em
->flags
;
391 ret
= add_extent_mapping(em_tree
, split
);
393 free_extent_map(split
);
397 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
&&
398 testend
&& em
->start
+ em
->len
> start
+ len
) {
399 u64 diff
= start
+ len
- em
->start
;
401 split
->start
= start
+ len
;
402 split
->len
= em
->start
+ em
->len
- (start
+ len
);
403 split
->bdev
= em
->bdev
;
404 split
->flags
= em
->flags
;
406 split
->block_start
= em
->block_start
+ diff
;
408 ret
= add_extent_mapping(em_tree
, split
);
410 free_extent_map(split
);
413 spin_unlock(&em_tree
->lock
);
417 /* once for the tree*/
421 free_extent_map(split
);
423 free_extent_map(split2
);
427 int btrfs_check_file(struct btrfs_root
*root
, struct inode
*inode
)
431 struct btrfs_path
*path
;
432 struct btrfs_key found_key
;
433 struct extent_buffer
*leaf
;
434 struct btrfs_file_extent_item
*extent
;
443 path
= btrfs_alloc_path();
444 ret
= btrfs_lookup_file_extent(NULL
, root
, path
, inode
->i_ino
,
447 nritems
= btrfs_header_nritems(path
->nodes
[0]);
448 if (path
->slots
[0] >= nritems
) {
449 ret
= btrfs_next_leaf(root
, path
);
452 nritems
= btrfs_header_nritems(path
->nodes
[0]);
454 slot
= path
->slots
[0];
455 leaf
= path
->nodes
[0];
456 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
457 if (found_key
.objectid
!= inode
->i_ino
)
459 if (found_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
462 if (found_key
.offset
< last_offset
) {
464 btrfs_print_leaf(root
, leaf
);
465 printk("inode %lu found offset %Lu expected %Lu\n",
466 inode
->i_ino
, found_key
.offset
, last_offset
);
470 extent
= btrfs_item_ptr(leaf
, slot
,
471 struct btrfs_file_extent_item
);
472 found_type
= btrfs_file_extent_type(leaf
, extent
);
473 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
474 extent_end
= found_key
.offset
+
475 btrfs_file_extent_num_bytes(leaf
, extent
);
476 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
477 struct btrfs_item
*item
;
478 item
= btrfs_item_nr(leaf
, slot
);
479 extent_end
= found_key
.offset
+
480 btrfs_file_extent_inline_len(leaf
, item
);
481 extent_end
= (extent_end
+ root
->sectorsize
- 1) &
482 ~((u64
)root
->sectorsize
-1 );
484 last_offset
= extent_end
;
487 if (0 && last_offset
< inode
->i_size
) {
489 btrfs_print_leaf(root
, leaf
);
490 printk("inode %lu found offset %Lu size %Lu\n", inode
->i_ino
,
491 last_offset
, inode
->i_size
);
496 btrfs_free_path(path
);
502 * this is very complex, but the basic idea is to drop all extents
503 * in the range start - end. hint_block is filled in with a block number
504 * that would be a good hint to the block allocator for this file.
506 * If an extent intersects the range but is not entirely inside the range
507 * it is either truncated or split. Anything entirely inside the range
508 * is deleted from the tree.
510 int btrfs_drop_extents(struct btrfs_trans_handle
*trans
,
511 struct btrfs_root
*root
, struct inode
*inode
,
512 u64 start
, u64 end
, u64 inline_limit
, u64
*hint_byte
)
515 u64 search_start
= start
;
516 struct extent_buffer
*leaf
;
517 struct btrfs_file_extent_item
*extent
;
518 struct btrfs_path
*path
;
519 struct btrfs_key key
;
520 struct btrfs_file_extent_item old
;
530 btrfs_drop_extent_cache(inode
, start
, end
- 1);
532 path
= btrfs_alloc_path();
537 btrfs_release_path(root
, path
);
538 ret
= btrfs_lookup_file_extent(trans
, root
, path
, inode
->i_ino
,
543 if (path
->slots
[0] == 0) {
555 leaf
= path
->nodes
[0];
556 slot
= path
->slots
[0];
558 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
559 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
&&
563 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
564 key
.objectid
!= inode
->i_ino
) {
568 search_start
= key
.offset
;
571 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
572 extent
= btrfs_item_ptr(leaf
, slot
,
573 struct btrfs_file_extent_item
);
574 found_type
= btrfs_file_extent_type(leaf
, extent
);
575 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
577 btrfs_file_extent_disk_bytenr(leaf
,
580 *hint_byte
= extent_end
;
582 extent_end
= key
.offset
+
583 btrfs_file_extent_num_bytes(leaf
, extent
);
585 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
586 struct btrfs_item
*item
;
587 item
= btrfs_item_nr(leaf
, slot
);
589 extent_end
= key
.offset
+
590 btrfs_file_extent_inline_len(leaf
, item
);
593 extent_end
= search_start
;
596 /* we found nothing we can drop */
597 if ((!found_extent
&& !found_inline
) ||
598 search_start
>= extent_end
) {
601 nritems
= btrfs_header_nritems(leaf
);
602 if (slot
>= nritems
- 1) {
603 nextret
= btrfs_next_leaf(root
, path
);
614 u64 mask
= root
->sectorsize
- 1;
615 search_start
= (extent_end
+ mask
) & ~mask
;
617 search_start
= extent_end
;
618 if (end
<= extent_end
&& start
>= key
.offset
&& found_inline
) {
619 *hint_byte
= EXTENT_MAP_INLINE
;
622 if (end
< extent_end
&& end
>= key
.offset
) {
625 btrfs_file_extent_disk_bytenr(leaf
, extent
);
627 btrfs_file_extent_disk_num_bytes(leaf
,
629 read_extent_buffer(leaf
, &old
,
630 (unsigned long)extent
,
632 if (disk_bytenr
!= 0) {
633 ret
= btrfs_inc_extent_ref(trans
, root
,
634 disk_bytenr
, disk_num_bytes
,
635 root
->root_key
.objectid
,
642 if (found_inline
&& start
<= key
.offset
)
645 /* truncate existing extent */
646 if (start
> key
.offset
) {
650 WARN_ON(start
& (root
->sectorsize
- 1));
652 new_num
= start
- key
.offset
;
653 old_num
= btrfs_file_extent_num_bytes(leaf
,
656 btrfs_file_extent_disk_bytenr(leaf
,
658 if (btrfs_file_extent_disk_bytenr(leaf
,
660 dec_i_blocks(inode
, old_num
- new_num
);
662 btrfs_set_file_extent_num_bytes(leaf
, extent
,
664 btrfs_mark_buffer_dirty(leaf
);
665 } else if (key
.offset
< inline_limit
&&
666 (end
> extent_end
) &&
667 (inline_limit
< extent_end
)) {
669 new_size
= btrfs_file_extent_calc_inline_size(
670 inline_limit
- key
.offset
);
671 dec_i_blocks(inode
, (extent_end
- key
.offset
) -
672 (inline_limit
- key
.offset
));
673 btrfs_truncate_item(trans
, root
, path
,
677 /* delete the entire extent */
680 u64 disk_num_bytes
= 0;
681 u64 extent_num_bytes
= 0;
685 root_gen
= btrfs_header_generation(leaf
);
686 root_owner
= btrfs_header_owner(leaf
);
689 btrfs_file_extent_disk_bytenr(leaf
,
692 btrfs_file_extent_disk_num_bytes(leaf
,
695 btrfs_file_extent_num_bytes(leaf
, extent
);
697 btrfs_file_extent_disk_bytenr(leaf
,
700 ret
= btrfs_del_item(trans
, root
, path
);
701 /* TODO update progress marker and return */
703 btrfs_release_path(root
, path
);
705 if (found_extent
&& disk_bytenr
!= 0) {
706 dec_i_blocks(inode
, extent_num_bytes
);
707 ret
= btrfs_free_extent(trans
, root
,
711 root_gen
, inode
->i_ino
,
716 if (!bookend
&& search_start
>= end
) {
723 if (bookend
&& found_inline
&& start
<= key
.offset
) {
725 new_size
= btrfs_file_extent_calc_inline_size(
727 dec_i_blocks(inode
, (extent_end
- key
.offset
) -
729 btrfs_truncate_item(trans
, root
, path
, new_size
, 0);
731 /* create bookend, splitting the extent in two */
732 if (bookend
&& found_extent
) {
733 struct btrfs_key ins
;
734 ins
.objectid
= inode
->i_ino
;
736 btrfs_set_key_type(&ins
, BTRFS_EXTENT_DATA_KEY
);
737 btrfs_release_path(root
, path
);
738 ret
= btrfs_insert_empty_item(trans
, root
, path
, &ins
,
741 leaf
= path
->nodes
[0];
743 btrfs_print_leaf(root
, leaf
);
744 printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret
, ins
.objectid
, ins
.type
, ins
.offset
, start
, end
, key
.offset
, extent_end
, keep
);
747 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
748 struct btrfs_file_extent_item
);
749 write_extent_buffer(leaf
, &old
,
750 (unsigned long)extent
, sizeof(old
));
752 btrfs_set_file_extent_offset(leaf
, extent
,
753 le64_to_cpu(old
.offset
) + end
- key
.offset
);
754 WARN_ON(le64_to_cpu(old
.num_bytes
) <
756 btrfs_set_file_extent_num_bytes(leaf
, extent
,
758 btrfs_set_file_extent_type(leaf
, extent
,
759 BTRFS_FILE_EXTENT_REG
);
761 btrfs_mark_buffer_dirty(path
->nodes
[0]);
762 if (le64_to_cpu(old
.disk_bytenr
) != 0) {
764 btrfs_file_extent_num_bytes(leaf
,
772 btrfs_free_path(path
);
773 btrfs_check_file(root
, inode
);
778 * this gets pages into the page cache and locks them down
780 static int prepare_pages(struct btrfs_root
*root
, struct file
*file
,
781 struct page
**pages
, size_t num_pages
,
782 loff_t pos
, unsigned long first_index
,
783 unsigned long last_index
, size_t write_bytes
)
786 unsigned long index
= pos
>> PAGE_CACHE_SHIFT
;
787 struct inode
*inode
= fdentry(file
)->d_inode
;
791 start_pos
= pos
& ~((u64
)root
->sectorsize
- 1);
793 memset(pages
, 0, num_pages
* sizeof(struct page
*));
795 for (i
= 0; i
< num_pages
; i
++) {
796 pages
[i
] = grab_cache_page(inode
->i_mapping
, index
+ i
);
801 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
802 ClearPageDirty(pages
[i
]);
804 cancel_dirty_page(pages
[i
], PAGE_CACHE_SIZE
);
806 wait_on_page_writeback(pages
[i
]);
807 set_page_extent_mapped(pages
[i
]);
808 WARN_ON(!PageLocked(pages
[i
]));
810 if (start_pos
< inode
->i_size
) {
812 last_pos
= ((u64
)index
+ num_pages
) << PAGE_CACHE_SHIFT
;
813 lock_extent(&BTRFS_I(inode
)->io_tree
,
814 start_pos
, last_pos
- 1, GFP_NOFS
);
815 clear_extent_bits(&BTRFS_I(inode
)->io_tree
, start_pos
,
816 last_pos
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
,
818 unlock_extent(&BTRFS_I(inode
)->io_tree
,
819 start_pos
, last_pos
- 1, GFP_NOFS
);
824 static ssize_t
btrfs_file_write(struct file
*file
, const char __user
*buf
,
825 size_t count
, loff_t
*ppos
)
829 ssize_t num_written
= 0;
832 struct inode
*inode
= fdentry(file
)->d_inode
;
833 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
834 struct page
**pages
= NULL
;
836 struct page
*pinned
[2];
837 unsigned long first_index
;
838 unsigned long last_index
;
840 nrptrs
= min((count
+ PAGE_CACHE_SIZE
- 1) / PAGE_CACHE_SIZE
,
841 PAGE_CACHE_SIZE
/ (sizeof(struct page
*)));
848 vfs_check_frozen(inode
->i_sb
, SB_FREEZE_WRITE
);
849 current
->backing_dev_info
= inode
->i_mapping
->backing_dev_info
;
850 err
= generic_write_checks(file
, &pos
, &count
, S_ISBLK(inode
->i_mode
));
855 err
= remove_suid(fdentry(file
));
858 file_update_time(file
);
860 pages
= kmalloc(nrptrs
* sizeof(struct page
*), GFP_KERNEL
);
862 mutex_lock(&inode
->i_mutex
);
863 first_index
= pos
>> PAGE_CACHE_SHIFT
;
864 last_index
= (pos
+ count
) >> PAGE_CACHE_SHIFT
;
867 * if this is a nodatasum mount, force summing off for the inode
868 * all the time. That way a later mount with summing on won't
871 if (btrfs_test_opt(root
, NODATASUM
))
872 btrfs_set_flag(inode
, NODATASUM
);
875 * there are lots of better ways to do this, but this code
876 * makes sure the first and last page in the file range are
877 * up to date and ready for cow
879 if ((pos
& (PAGE_CACHE_SIZE
- 1))) {
880 pinned
[0] = grab_cache_page(inode
->i_mapping
, first_index
);
881 if (!PageUptodate(pinned
[0])) {
882 ret
= btrfs_readpage(NULL
, pinned
[0]);
884 wait_on_page_locked(pinned
[0]);
886 unlock_page(pinned
[0]);
889 if ((pos
+ count
) & (PAGE_CACHE_SIZE
- 1)) {
890 pinned
[1] = grab_cache_page(inode
->i_mapping
, last_index
);
891 if (!PageUptodate(pinned
[1])) {
892 ret
= btrfs_readpage(NULL
, pinned
[1]);
894 wait_on_page_locked(pinned
[1]);
896 unlock_page(pinned
[1]);
901 size_t offset
= pos
& (PAGE_CACHE_SIZE
- 1);
902 size_t write_bytes
= min(count
, nrptrs
*
903 (size_t)PAGE_CACHE_SIZE
-
905 size_t num_pages
= (write_bytes
+ PAGE_CACHE_SIZE
- 1) >>
908 WARN_ON(num_pages
> nrptrs
);
909 memset(pages
, 0, sizeof(pages
));
911 mutex_lock(&root
->fs_info
->fs_mutex
);
912 ret
= btrfs_check_free_space(root
, write_bytes
, 0);
913 mutex_unlock(&root
->fs_info
->fs_mutex
);
917 ret
= prepare_pages(root
, file
, pages
, num_pages
,
918 pos
, first_index
, last_index
,
923 ret
= btrfs_copy_from_user(pos
, num_pages
,
924 write_bytes
, pages
, buf
);
926 btrfs_drop_pages(pages
, num_pages
);
930 ret
= dirty_and_release_pages(NULL
, root
, file
, pages
,
931 num_pages
, pos
, write_bytes
);
932 btrfs_drop_pages(pages
, num_pages
);
937 count
-= write_bytes
;
939 num_written
+= write_bytes
;
941 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, num_pages
);
942 if (num_pages
< (root
->leafsize
>> PAGE_CACHE_SHIFT
) + 1)
943 btrfs_btree_balance_dirty(root
, 1);
944 btrfs_throttle(root
);
948 mutex_unlock(&inode
->i_mutex
);
953 page_cache_release(pinned
[0]);
955 page_cache_release(pinned
[1]);
958 if (num_written
> 0 && ((file
->f_flags
& O_SYNC
) || IS_SYNC(inode
))) {
959 err
= sync_page_range(inode
, inode
->i_mapping
,
960 start_pos
, num_written
);
963 } else if (num_written
> 0 && (file
->f_flags
& O_DIRECT
)) {
964 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22)
965 do_sync_file_range(file
, start_pos
,
966 start_pos
+ num_written
- 1,
967 SYNC_FILE_RANGE_WRITE
|
968 SYNC_FILE_RANGE_WAIT_AFTER
);
970 do_sync_mapping_range(inode
->i_mapping
, start_pos
,
971 start_pos
+ num_written
- 1,
972 SYNC_FILE_RANGE_WRITE
|
973 SYNC_FILE_RANGE_WAIT_AFTER
);
975 invalidate_mapping_pages(inode
->i_mapping
,
976 start_pos
>> PAGE_CACHE_SHIFT
,
977 (start_pos
+ num_written
- 1) >> PAGE_CACHE_SHIFT
);
979 current
->backing_dev_info
= NULL
;
980 btrfs_ordered_throttle(root
, inode
);
981 return num_written
? num_written
: err
;
984 static int btrfs_sync_file(struct file
*file
,
985 struct dentry
*dentry
, int datasync
)
987 struct inode
*inode
= dentry
->d_inode
;
988 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
990 struct btrfs_trans_handle
*trans
;
993 * check the transaction that last modified this inode
994 * and see if its already been committed
996 mutex_lock(&root
->fs_info
->fs_mutex
);
997 if (!BTRFS_I(inode
)->last_trans
)
999 mutex_lock(&root
->fs_info
->trans_mutex
);
1000 if (BTRFS_I(inode
)->last_trans
<=
1001 root
->fs_info
->last_trans_committed
) {
1002 BTRFS_I(inode
)->last_trans
= 0;
1003 mutex_unlock(&root
->fs_info
->trans_mutex
);
1006 mutex_unlock(&root
->fs_info
->trans_mutex
);
1009 * ok we haven't committed the transaction yet, lets do a commit
1011 trans
= btrfs_start_transaction(root
, 1);
1016 ret
= btrfs_commit_transaction(trans
, root
);
1018 mutex_unlock(&root
->fs_info
->fs_mutex
);
1019 return ret
> 0 ? EIO
: ret
;
1022 static struct vm_operations_struct btrfs_file_vm_ops
= {
1023 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
1024 .nopage
= filemap_nopage
,
1025 .populate
= filemap_populate
,
1027 .fault
= filemap_fault
,
1029 .page_mkwrite
= btrfs_page_mkwrite
,
1032 static int btrfs_file_mmap(struct file
*filp
, struct vm_area_struct
*vma
)
1034 vma
->vm_ops
= &btrfs_file_vm_ops
;
1035 file_accessed(filp
);
1039 struct file_operations btrfs_file_operations
= {
1040 .llseek
= generic_file_llseek
,
1041 .read
= do_sync_read
,
1042 .aio_read
= generic_file_aio_read
,
1043 .splice_read
= generic_file_splice_read
,
1044 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1045 .sendfile
= generic_file_sendfile
,
1047 .write
= btrfs_file_write
,
1048 .mmap
= btrfs_file_mmap
,
1049 .open
= generic_file_open
,
1050 .fsync
= btrfs_sync_file
,
1051 .unlocked_ioctl
= btrfs_ioctl
,
1052 #ifdef CONFIG_COMPAT
1053 .compat_ioctl
= btrfs_ioctl
,