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"
38 #include "print-tree.h"
41 static int btrfs_copy_from_user(loff_t pos
, int num_pages
, int write_bytes
,
42 struct page
**prepared_pages
,
43 const char __user
* buf
)
47 int offset
= pos
& (PAGE_CACHE_SIZE
- 1);
49 for (i
= 0; i
< num_pages
&& write_bytes
> 0; i
++, offset
= 0) {
50 size_t count
= min_t(size_t,
51 PAGE_CACHE_SIZE
- offset
, write_bytes
);
52 struct page
*page
= prepared_pages
[i
];
53 fault_in_pages_readable(buf
, count
);
55 /* Copy data from userspace to the current page */
57 page_fault
= __copy_from_user(page_address(page
) + offset
,
59 /* Flush processor's dcache for this page */
60 flush_dcache_page(page
);
68 return page_fault
? -EFAULT
: 0;
71 static void btrfs_drop_pages(struct page
**pages
, size_t num_pages
)
74 for (i
= 0; i
< num_pages
; i
++) {
77 unlock_page(pages
[i
]);
78 mark_page_accessed(pages
[i
]);
79 page_cache_release(pages
[i
]);
83 static int insert_inline_extent(struct btrfs_trans_handle
*trans
,
84 struct btrfs_root
*root
, struct inode
*inode
,
85 u64 offset
, size_t size
,
86 struct page
**pages
, size_t page_offset
,
90 struct btrfs_path
*path
;
91 struct extent_buffer
*leaf
;
94 struct btrfs_file_extent_item
*ei
;
102 path
= btrfs_alloc_path();
106 btrfs_set_trans_block_group(trans
, inode
);
108 key
.objectid
= inode
->i_ino
;
110 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
111 datasize
= btrfs_file_extent_calc_inline_size(offset
+ size
);
113 ret
= btrfs_search_slot(trans
, root
, &key
, path
, 0, 1);
120 leaf
= path
->nodes
[0];
121 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
122 struct btrfs_file_extent_item
);
124 if (btrfs_file_extent_type(leaf
, ei
) !=
125 BTRFS_FILE_EXTENT_INLINE
) {
128 btrfs_item_key_to_cpu(leaf
, &key
, path
->slots
[0]);
135 leaf
= path
->nodes
[0];
136 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
137 struct btrfs_file_extent_item
);
139 if (btrfs_file_extent_type(leaf
, ei
) !=
140 BTRFS_FILE_EXTENT_INLINE
) {
142 btrfs_print_leaf(root
, leaf
);
143 printk("found wasn't inline offset %Lu inode %lu\n",
144 offset
, inode
->i_ino
);
147 found_start
= key
.offset
;
148 found_size
= btrfs_file_extent_inline_len(leaf
,
149 btrfs_item_nr(leaf
, path
->slots
[0]));
151 if (found_size
< offset
+ size
) {
152 btrfs_release_path(root
, path
);
153 ret
= btrfs_search_slot(trans
, root
, &key
, path
,
154 offset
+ size
- found_size
-
157 ret
= btrfs_extend_item(trans
, root
, path
,
158 offset
+ size
- found_size
-
164 leaf
= path
->nodes
[0];
165 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
166 struct btrfs_file_extent_item
);
170 btrfs_release_path(root
, path
);
171 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
175 printk("got bad ret %d\n", ret
);
178 leaf
= path
->nodes
[0];
179 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
180 struct btrfs_file_extent_item
);
181 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
182 btrfs_set_file_extent_type(leaf
, ei
, BTRFS_FILE_EXTENT_INLINE
);
184 ptr
= btrfs_file_extent_inline_start(ei
) + offset
;
190 kaddr
= kmap_atomic(page
, KM_USER0
);
191 cur_size
= min_t(size_t, PAGE_CACHE_SIZE
- page_offset
, size
);
192 write_extent_buffer(leaf
, kaddr
+ page_offset
, ptr
, cur_size
);
193 kunmap_atomic(kaddr
, KM_USER0
);
197 if (i
>= num_pages
) {
198 printk("i %d num_pages %d\n", i
, num_pages
);
202 btrfs_mark_buffer_dirty(leaf
);
204 btrfs_free_path(path
);
208 static int dirty_and_release_pages(struct btrfs_trans_handle
*trans
,
209 struct btrfs_root
*root
,
218 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
219 struct extent_map
*em
;
220 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
224 u64 end_of_last_block
;
225 u64 end_pos
= pos
+ write_bytes
;
227 loff_t isize
= i_size_read(inode
);
229 em
= alloc_extent_map(GFP_NOFS
);
233 em
->bdev
= inode
->i_sb
->s_bdev
;
235 start_pos
= pos
& ~((u64
)root
->sectorsize
- 1);
236 num_bytes
= (write_bytes
+ pos
- start_pos
+
237 root
->sectorsize
- 1) & ~((u64
)root
->sectorsize
- 1);
239 down_read(&BTRFS_I(inode
)->root
->snap_sem
);
240 end_of_last_block
= start_pos
+ num_bytes
- 1;
242 lock_extent(em_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
243 mutex_lock(&root
->fs_info
->fs_mutex
);
244 trans
= btrfs_start_transaction(root
, 1);
249 btrfs_set_trans_block_group(trans
, inode
);
250 inode
->i_blocks
+= num_bytes
>> 9;
253 if ((end_of_last_block
& 4095) == 0) {
254 printk("strange end of last %Lu %zu %Lu\n", start_pos
, write_bytes
, end_of_last_block
);
256 set_extent_uptodate(em_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
258 /* FIXME...EIEIO, ENOSPC and more */
260 /* insert any holes we need to create */
261 if (inode
->i_size
< start_pos
) {
262 u64 last_pos_in_file
;
264 u64 mask
= root
->sectorsize
- 1;
265 last_pos_in_file
= (isize
+ mask
) & ~mask
;
266 hole_size
= (start_pos
- last_pos_in_file
+ mask
) & ~mask
;
268 if (last_pos_in_file
< start_pos
) {
269 err
= btrfs_drop_extents(trans
, root
, inode
,
271 last_pos_in_file
+ hole_size
,
277 err
= btrfs_insert_file_extent(trans
, root
,
287 * either allocate an extent for the new bytes or setup the key
288 * to show we are doing inline data in the extent
290 inline_size
= end_pos
;
291 if (isize
>= BTRFS_MAX_INLINE_DATA_SIZE(root
) ||
292 inline_size
> 8192 ||
293 inline_size
>= BTRFS_MAX_INLINE_DATA_SIZE(root
)) {
296 for (i
= 0; i
< num_pages
; i
++) {
297 struct page
*p
= pages
[i
];
301 last_end
= pages
[num_pages
-1]->index
<< PAGE_CACHE_SHIFT
;
302 last_end
+= PAGE_CACHE_SIZE
- 1;
303 set_extent_delalloc(em_tree
, start_pos
, end_of_last_block
,
307 /* step one, delete the existing extents in this range */
308 aligned_end
= (pos
+ write_bytes
+ root
->sectorsize
- 1) &
309 ~((u64
)root
->sectorsize
- 1);
310 err
= btrfs_drop_extents(trans
, root
, inode
, start_pos
,
311 aligned_end
, end_pos
, &hint_byte
);
314 err
= insert_inline_extent(trans
, root
, inode
, start_pos
,
315 end_pos
- start_pos
, pages
, 0,
319 if (end_pos
> isize
) {
320 i_size_write(inode
, end_pos
);
321 btrfs_update_inode(trans
, root
, inode
);
324 err
= btrfs_end_transaction(trans
, root
);
326 mutex_unlock(&root
->fs_info
->fs_mutex
);
327 unlock_extent(em_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
329 up_read(&BTRFS_I(inode
)->root
->snap_sem
);
333 int btrfs_drop_extent_cache(struct inode
*inode
, u64 start
, u64 end
)
335 struct extent_map
*em
;
336 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
339 em
= lookup_extent_mapping(em_tree
, start
, end
);
342 remove_extent_mapping(em_tree
, em
);
345 /* once for the tree*/
352 * this is very complex, but the basic idea is to drop all extents
353 * in the range start - end. hint_block is filled in with a block number
354 * that would be a good hint to the block allocator for this file.
356 * If an extent intersects the range but is not entirely inside the range
357 * it is either truncated or split. Anything entirely inside the range
358 * is deleted from the tree.
360 int btrfs_drop_extents(struct btrfs_trans_handle
*trans
,
361 struct btrfs_root
*root
, struct inode
*inode
,
362 u64 start
, u64 end
, u64 inline_end
, u64
*hint_byte
)
365 struct btrfs_key key
;
366 struct extent_buffer
*leaf
;
368 struct btrfs_file_extent_item
*extent
;
371 struct btrfs_file_extent_item old
;
372 struct btrfs_path
*path
;
373 u64 search_start
= start
;
380 btrfs_drop_extent_cache(inode
, start
, end
- 1);
382 path
= btrfs_alloc_path();
387 btrfs_release_path(root
, path
);
388 ret
= btrfs_lookup_file_extent(trans
, root
, path
, inode
->i_ino
,
393 if (path
->slots
[0] == 0) {
405 leaf
= path
->nodes
[0];
406 slot
= path
->slots
[0];
408 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
409 if (key
.offset
>= end
|| key
.objectid
!= inode
->i_ino
) {
412 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
) {
416 search_start
= key
.offset
;
419 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
420 extent
= btrfs_item_ptr(leaf
, slot
,
421 struct btrfs_file_extent_item
);
422 found_type
= btrfs_file_extent_type(leaf
, extent
);
423 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
424 extent_end
= key
.offset
+
425 btrfs_file_extent_num_bytes(leaf
, extent
);
427 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
428 struct btrfs_item
*item
;
429 item
= btrfs_item_nr(leaf
, slot
);
431 extent_end
= key
.offset
+
432 btrfs_file_extent_inline_len(leaf
, item
);
435 extent_end
= search_start
;
438 /* we found nothing we can drop */
439 if ((!found_extent
&& !found_inline
) ||
440 search_start
>= extent_end
) {
443 nritems
= btrfs_header_nritems(leaf
);
444 if (slot
>= nritems
- 1) {
445 nextret
= btrfs_next_leaf(root
, path
);
455 /* FIXME, there's only one inline extent allowed right now */
457 u64 mask
= root
->sectorsize
- 1;
458 search_start
= (extent_end
+ mask
) & ~mask
;
460 search_start
= extent_end
;
462 if (end
< extent_end
&& end
>= key
.offset
) {
465 btrfs_file_extent_disk_bytenr(leaf
, extent
);
467 btrfs_file_extent_disk_num_bytes(leaf
,
469 read_extent_buffer(leaf
, &old
,
470 (unsigned long)extent
,
472 if (disk_bytenr
!= 0) {
473 ret
= btrfs_inc_extent_ref(trans
, root
,
474 disk_bytenr
, disk_num_bytes
);
481 /* truncate existing extent */
482 if (start
> key
.offset
) {
486 WARN_ON(start
& (root
->sectorsize
- 1));
488 new_num
= start
- key
.offset
;
489 old_num
= btrfs_file_extent_num_bytes(leaf
,
492 btrfs_file_extent_disk_bytenr(leaf
,
494 if (btrfs_file_extent_disk_bytenr(leaf
,
497 (old_num
- new_num
) >> 9;
499 btrfs_set_file_extent_num_bytes(leaf
, extent
,
501 btrfs_mark_buffer_dirty(leaf
);
502 } else if (end
> extent_end
&&
503 key
.offset
< inline_end
&&
504 inline_end
< extent_end
) {
506 new_size
= btrfs_file_extent_calc_inline_size(
507 inline_end
- key
.offset
);
508 btrfs_truncate_item(trans
, root
, path
,
512 /* delete the entire extent */
515 u64 disk_num_bytes
= 0;
516 u64 extent_num_bytes
= 0;
519 btrfs_file_extent_disk_bytenr(leaf
,
522 btrfs_file_extent_disk_num_bytes(leaf
,
525 btrfs_file_extent_num_bytes(leaf
, extent
);
527 btrfs_file_extent_disk_bytenr(leaf
,
530 ret
= btrfs_del_item(trans
, root
, path
);
531 /* TODO update progress marker and return */
533 btrfs_release_path(root
, path
);
535 if (found_extent
&& disk_bytenr
!= 0) {
536 inode
->i_blocks
-= extent_num_bytes
>> 9;
537 ret
= btrfs_free_extent(trans
, root
,
543 if (!bookend
&& search_start
>= end
) {
550 /* create bookend, splitting the extent in two */
551 if (bookend
&& found_extent
) {
552 struct btrfs_key ins
;
553 ins
.objectid
= inode
->i_ino
;
555 btrfs_set_key_type(&ins
, BTRFS_EXTENT_DATA_KEY
);
556 btrfs_release_path(root
, path
);
557 ret
= btrfs_insert_empty_item(trans
, root
, path
, &ins
,
560 leaf
= path
->nodes
[0];
562 btrfs_print_leaf(root
, leaf
);
563 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
);
566 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
567 struct btrfs_file_extent_item
);
568 write_extent_buffer(leaf
, &old
,
569 (unsigned long)extent
, sizeof(old
));
571 btrfs_set_file_extent_offset(leaf
, extent
,
572 le64_to_cpu(old
.offset
) + end
- key
.offset
);
573 WARN_ON(le64_to_cpu(old
.num_bytes
) <
575 btrfs_set_file_extent_num_bytes(leaf
, extent
,
577 btrfs_set_file_extent_type(leaf
, extent
,
578 BTRFS_FILE_EXTENT_REG
);
580 btrfs_mark_buffer_dirty(path
->nodes
[0]);
581 if (le64_to_cpu(old
.disk_bytenr
) != 0) {
583 btrfs_file_extent_num_bytes(leaf
,
591 btrfs_free_path(path
);
596 * this gets pages into the page cache and locks them down
598 static int prepare_pages(struct btrfs_root
*root
,
603 unsigned long first_index
,
604 unsigned long last_index
,
608 unsigned long index
= pos
>> PAGE_CACHE_SHIFT
;
609 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
613 start_pos
= pos
& ~((u64
)root
->sectorsize
- 1);
615 memset(pages
, 0, num_pages
* sizeof(struct page
*));
617 for (i
= 0; i
< num_pages
; i
++) {
618 pages
[i
] = grab_cache_page(inode
->i_mapping
, index
+ i
);
623 cancel_dirty_page(pages
[i
], PAGE_CACHE_SIZE
);
624 wait_on_page_writeback(pages
[i
]);
625 set_page_extent_mapped(pages
[i
]);
626 WARN_ON(!PageLocked(pages
[i
]));
631 static ssize_t
btrfs_file_write(struct file
*file
, const char __user
*buf
,
632 size_t count
, loff_t
*ppos
)
635 size_t num_written
= 0;
638 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
639 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
640 struct page
**pages
= NULL
;
642 struct page
*pinned
[2];
643 unsigned long first_index
;
644 unsigned long last_index
;
646 nrptrs
= min((count
+ PAGE_CACHE_SIZE
- 1) / PAGE_CACHE_SIZE
,
647 PAGE_CACHE_SIZE
/ (sizeof(struct page
*)));
650 if (file
->f_flags
& O_DIRECT
)
653 vfs_check_frozen(inode
->i_sb
, SB_FREEZE_WRITE
);
654 current
->backing_dev_info
= inode
->i_mapping
->backing_dev_info
;
655 err
= generic_write_checks(file
, &pos
, &count
, S_ISBLK(inode
->i_mode
));
660 err
= remove_suid(file
->f_path
.dentry
);
663 file_update_time(file
);
665 pages
= kmalloc(nrptrs
* sizeof(struct page
*), GFP_KERNEL
);
667 mutex_lock(&inode
->i_mutex
);
668 first_index
= pos
>> PAGE_CACHE_SHIFT
;
669 last_index
= (pos
+ count
) >> PAGE_CACHE_SHIFT
;
672 * there are lots of better ways to do this, but this code
673 * makes sure the first and last page in the file range are
674 * up to date and ready for cow
676 if ((pos
& (PAGE_CACHE_SIZE
- 1))) {
677 pinned
[0] = grab_cache_page(inode
->i_mapping
, first_index
);
678 if (!PageUptodate(pinned
[0])) {
679 ret
= btrfs_readpage(NULL
, pinned
[0]);
681 wait_on_page_locked(pinned
[0]);
683 unlock_page(pinned
[0]);
686 if ((pos
+ count
) & (PAGE_CACHE_SIZE
- 1)) {
687 pinned
[1] = grab_cache_page(inode
->i_mapping
, last_index
);
688 if (!PageUptodate(pinned
[1])) {
689 ret
= btrfs_readpage(NULL
, pinned
[1]);
691 wait_on_page_locked(pinned
[1]);
693 unlock_page(pinned
[1]);
698 size_t offset
= pos
& (PAGE_CACHE_SIZE
- 1);
699 size_t write_bytes
= min(count
, nrptrs
*
700 (size_t)PAGE_CACHE_SIZE
-
702 size_t num_pages
= (write_bytes
+ PAGE_CACHE_SIZE
- 1) >>
705 WARN_ON(num_pages
> nrptrs
);
706 memset(pages
, 0, sizeof(pages
));
707 ret
= prepare_pages(root
, file
, pages
, num_pages
,
708 pos
, first_index
, last_index
,
713 ret
= btrfs_copy_from_user(pos
, num_pages
,
714 write_bytes
, pages
, buf
);
716 btrfs_drop_pages(pages
, num_pages
);
720 ret
= dirty_and_release_pages(NULL
, root
, file
, pages
,
721 num_pages
, pos
, write_bytes
);
722 btrfs_drop_pages(pages
, num_pages
);
727 count
-= write_bytes
;
729 num_written
+= write_bytes
;
731 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, num_pages
);
732 btrfs_btree_balance_dirty(root
, 1);
735 mutex_unlock(&inode
->i_mutex
);
739 page_cache_release(pinned
[0]);
741 page_cache_release(pinned
[1]);
743 current
->backing_dev_info
= NULL
;
744 return num_written
? num_written
: err
;
747 static int btrfs_sync_file(struct file
*file
,
748 struct dentry
*dentry
, int datasync
)
750 struct inode
*inode
= dentry
->d_inode
;
751 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
753 struct btrfs_trans_handle
*trans
;
756 * check the transaction that last modified this inode
757 * and see if its already been committed
759 mutex_lock(&root
->fs_info
->fs_mutex
);
760 if (!BTRFS_I(inode
)->last_trans
)
762 mutex_lock(&root
->fs_info
->trans_mutex
);
763 if (BTRFS_I(inode
)->last_trans
<=
764 root
->fs_info
->last_trans_committed
) {
765 BTRFS_I(inode
)->last_trans
= 0;
766 mutex_unlock(&root
->fs_info
->trans_mutex
);
769 mutex_unlock(&root
->fs_info
->trans_mutex
);
772 * ok we haven't committed the transaction yet, lets do a commit
774 trans
= btrfs_start_transaction(root
, 1);
779 ret
= btrfs_commit_transaction(trans
, root
);
781 mutex_unlock(&root
->fs_info
->fs_mutex
);
782 return ret
> 0 ? EIO
: ret
;
785 static struct vm_operations_struct btrfs_file_vm_ops
= {
786 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
787 .nopage
= filemap_nopage
,
788 .populate
= filemap_populate
,
790 .fault
= filemap_fault
,
792 .page_mkwrite
= btrfs_page_mkwrite
,
795 static int btrfs_file_mmap(struct file
*filp
, struct vm_area_struct
*vma
)
797 vma
->vm_ops
= &btrfs_file_vm_ops
;
802 struct file_operations btrfs_file_operations
= {
803 .llseek
= generic_file_llseek
,
804 .read
= do_sync_read
,
805 .aio_read
= generic_file_aio_read
,
806 .write
= btrfs_file_write
,
807 .mmap
= btrfs_file_mmap
,
808 .open
= generic_file_open
,
809 .fsync
= btrfs_sync_file
,
810 .unlocked_ioctl
= btrfs_ioctl
,
812 .compat_ioctl
= btrfs_ioctl
,