btrfs: 32-bit type problems
[linux-2.6-block.git] / fs / btrfs / file.c
CommitLineData
6cbd5570
CM
1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
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.
7 *
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.
12 *
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.
17 */
18
39279cc3
CM
19#include <linux/fs.h>
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>
92fee66d 32#include <linux/version.h>
39279cc3
CM
33#include "ctree.h"
34#include "disk-io.h"
35#include "transaction.h"
36#include "btrfs_inode.h"
37#include "ioctl.h"
38#include "print-tree.h"
39
40
41static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes,
42 struct page **prepared_pages,
43 const char __user * buf)
44{
45 long page_fault = 0;
46 int i;
47 int offset = pos & (PAGE_CACHE_SIZE - 1);
48
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);
54
55 /* Copy data from userspace to the current page */
56 kmap(page);
57 page_fault = __copy_from_user(page_address(page) + offset,
58 buf, count);
59 /* Flush processor's dcache for this page */
60 flush_dcache_page(page);
61 kunmap(page);
62 buf += count;
63 write_bytes -= count;
64
65 if (page_fault)
66 break;
67 }
68 return page_fault ? -EFAULT : 0;
69}
70
71static void btrfs_drop_pages(struct page **pages, size_t num_pages)
72{
73 size_t i;
74 for (i = 0; i < num_pages; i++) {
75 if (!pages[i])
76 break;
77 unlock_page(pages[i]);
78 mark_page_accessed(pages[i]);
79 page_cache_release(pages[i]);
80 }
81}
82
a52d9a80
CM
83static int insert_inline_extent(struct btrfs_trans_handle *trans,
84 struct btrfs_root *root, struct inode *inode,
3326d1b0
CM
85 u64 offset, size_t size,
86 struct page **pages, size_t page_offset,
87 int num_pages)
54aa1f4d
CM
88{
89 struct btrfs_key key;
90 struct btrfs_path *path;
5f39d397
CM
91 struct extent_buffer *leaf;
92 char *kaddr;
93 unsigned long ptr;
54aa1f4d 94 struct btrfs_file_extent_item *ei;
3326d1b0 95 struct page *page;
54aa1f4d
CM
96 u32 datasize;
97 int err = 0;
98 int ret;
3326d1b0
CM
99 int i;
100 ssize_t cur_size;
54aa1f4d
CM
101
102 path = btrfs_alloc_path();
103 if (!path)
104 return -ENOMEM;
105
54aa1f4d
CM
106 btrfs_set_trans_block_group(trans, inode);
107
108 key.objectid = inode->i_ino;
109 key.offset = offset;
54aa1f4d 110 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
3326d1b0 111 datasize = btrfs_file_extent_calc_inline_size(offset + size);
54aa1f4d 112
3326d1b0
CM
113 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
114 if (ret < 0) {
54aa1f4d
CM
115 err = ret;
116 goto fail;
117 }
3326d1b0
CM
118 if (ret == 1) {
119 path->slots[0]--;
120 leaf = path->nodes[0];
121 ei = btrfs_item_ptr(leaf, path->slots[0],
122 struct btrfs_file_extent_item);
123
124 if (btrfs_file_extent_type(leaf, ei) !=
125 BTRFS_FILE_EXTENT_INLINE) {
126 goto insert;
127 }
128 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
129 ret = 0;
130 }
131 if (ret == 0) {
132 u32 found_size;
133 u64 found_start;
134
135 leaf = path->nodes[0];
136 ei = btrfs_item_ptr(leaf, path->slots[0],
137 struct btrfs_file_extent_item);
138
139 if (btrfs_file_extent_type(leaf, ei) !=
140 BTRFS_FILE_EXTENT_INLINE) {
141 err = ret;
142 btrfs_print_leaf(root, leaf);
143 printk("found wasn't inline offset %Lu inode %lu\n",
144 offset, inode->i_ino);
145 goto fail;
146 }
147 found_start = key.offset;
148 found_size = btrfs_file_extent_inline_len(leaf,
149 btrfs_item_nr(leaf, path->slots[0]));
150
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 -
155 found_start, 1);
156 BUG_ON(ret != 0);
157 ret = btrfs_extend_item(trans, root, path,
158 offset + size - found_size -
159 found_start);
160 if (ret) {
161 err = ret;
162 goto fail;
163 }
164 leaf = path->nodes[0];
165 ei = btrfs_item_ptr(leaf, path->slots[0],
166 struct btrfs_file_extent_item);
167 }
168 } else {
169insert:
170 btrfs_release_path(root, path);
171 ret = btrfs_insert_empty_item(trans, root, path, &key,
172 datasize);
173 if (ret) {
174 err = ret;
175 printk("got bad ret %d\n", ret);
176 goto fail;
177 }
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);
183 }
184 ptr = btrfs_file_extent_inline_start(ei) + offset;
185
186 cur_size = size;
187 i = 0;
188 while (size > 0) {
189 page = pages[i];
190 kaddr = kmap_atomic(page, KM_USER0);
ae2f5411 191 cur_size = min_t(size_t, PAGE_CACHE_SIZE - page_offset, size);
3326d1b0
CM
192 write_extent_buffer(leaf, kaddr + page_offset, ptr, cur_size);
193 kunmap_atomic(kaddr, KM_USER0);
194 page_offset = 0;
195 ptr += cur_size;
196 size -= cur_size;
197 if (i >= num_pages) {
198 printk("i %d num_pages %d\n", i, num_pages);
199 }
200 i++;
201 }
5f39d397 202 btrfs_mark_buffer_dirty(leaf);
54aa1f4d
CM
203fail:
204 btrfs_free_path(path);
54aa1f4d
CM
205 return err;
206}
207
39279cc3
CM
208static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
209 struct btrfs_root *root,
210 struct file *file,
211 struct page **pages,
212 size_t num_pages,
213 loff_t pos,
214 size_t write_bytes)
215{
39279cc3 216 int err = 0;
a52d9a80 217 int i;
39279cc3 218 struct inode *inode = file->f_path.dentry->d_inode;
a52d9a80
CM
219 struct extent_map *em;
220 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
db94535d
CM
221 u64 hint_byte;
222 u64 num_bytes;
a52d9a80
CM
223 u64 start_pos;
224 u64 end_of_last_block;
225 u64 end_pos = pos + write_bytes;
db94535d 226 u32 inline_size;
a52d9a80 227 loff_t isize = i_size_read(inode);
39279cc3 228
a52d9a80
CM
229 em = alloc_extent_map(GFP_NOFS);
230 if (!em)
231 return -ENOMEM;
39279cc3 232
a52d9a80 233 em->bdev = inode->i_sb->s_bdev;
39279cc3 234
5f39d397 235 start_pos = pos & ~((u64)root->sectorsize - 1);
db94535d
CM
236 num_bytes = (write_bytes + pos - start_pos +
237 root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
39279cc3 238
011410bd 239 down_read(&BTRFS_I(inode)->root->snap_sem);
db94535d
CM
240 end_of_last_block = start_pos + num_bytes - 1;
241
b888db2b 242 lock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS);
a52d9a80
CM
243 mutex_lock(&root->fs_info->fs_mutex);
244 trans = btrfs_start_transaction(root, 1);
245 if (!trans) {
246 err = -ENOMEM;
247 goto out_unlock;
248 }
249 btrfs_set_trans_block_group(trans, inode);
db94535d
CM
250 inode->i_blocks += num_bytes >> 9;
251 hint_byte = 0;
a52d9a80
CM
252
253 if ((end_of_last_block & 4095) == 0) {
9433063b 254 printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block);
a52d9a80
CM
255 }
256 set_extent_uptodate(em_tree, start_pos, end_of_last_block, GFP_NOFS);
257
258 /* FIXME...EIEIO, ENOSPC and more */
259
a52d9a80
CM
260 /* insert any holes we need to create */
261 if (inode->i_size < start_pos) {
262 u64 last_pos_in_file;
263 u64 hole_size;
5f39d397 264 u64 mask = root->sectorsize - 1;
a52d9a80
CM
265 last_pos_in_file = (isize + mask) & ~mask;
266 hole_size = (start_pos - last_pos_in_file + mask) & ~mask;
2bf5a725 267
a52d9a80 268 if (last_pos_in_file < start_pos) {
2bf5a725
CM
269 err = btrfs_drop_extents(trans, root, inode,
270 last_pos_in_file,
271 last_pos_in_file + hole_size,
3326d1b0 272 last_pos_in_file,
db94535d 273 &hint_byte);
2bf5a725
CM
274 if (err)
275 goto failed;
276
a52d9a80
CM
277 err = btrfs_insert_file_extent(trans, root,
278 inode->i_ino,
279 last_pos_in_file,
280 0, 0, hole_size);
281 }
282 if (err)
39279cc3 283 goto failed;
a52d9a80
CM
284 }
285
286 /*
287 * either allocate an extent for the new bytes or setup the key
288 * to show we are doing inline data in the extent
289 */
3326d1b0
CM
290 inline_size = end_pos;
291 if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
7936ca38 292 inline_size > 8192 ||
3326d1b0 293 inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
b888db2b 294 u64 last_end;
3326d1b0 295
a52d9a80
CM
296 for (i = 0; i < num_pages; i++) {
297 struct page *p = pages[i];
298 SetPageUptodate(p);
b888db2b 299 set_page_dirty(p);
39279cc3 300 }
b888db2b
CM
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,
304 GFP_NOFS);
a52d9a80 305 } else {
3326d1b0 306 u64 aligned_end;
b888db2b 307 /* step one, delete the existing extents in this range */
3326d1b0
CM
308 aligned_end = (pos + write_bytes + root->sectorsize - 1) &
309 ~((u64)root->sectorsize - 1);
2bf5a725 310 err = btrfs_drop_extents(trans, root, inode, start_pos,
3326d1b0 311 aligned_end, end_pos, &hint_byte);
2bf5a725
CM
312 if (err)
313 goto failed;
a52d9a80 314 err = insert_inline_extent(trans, root, inode, start_pos,
3326d1b0
CM
315 end_pos - start_pos, pages, 0,
316 num_pages);
a52d9a80 317 BUG_ON(err);
a52d9a80
CM
318 }
319 if (end_pos > isize) {
320 i_size_write(inode, end_pos);
321 btrfs_update_inode(trans, root, inode);
39279cc3
CM
322 }
323failed:
a52d9a80
CM
324 err = btrfs_end_transaction(trans, root);
325out_unlock:
326 mutex_unlock(&root->fs_info->fs_mutex);
b888db2b 327 unlock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS);
a52d9a80 328 free_extent_map(em);
011410bd 329 up_read(&BTRFS_I(inode)->root->snap_sem);
39279cc3
CM
330 return err;
331}
332
a52d9a80
CM
333int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end)
334{
335 struct extent_map *em;
336 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
337
338 while(1) {
339 em = lookup_extent_mapping(em_tree, start, end);
340 if (!em)
341 break;
342 remove_extent_mapping(em_tree, em);
343 /* once for us */
344 free_extent_map(em);
345 /* once for the tree*/
346 free_extent_map(em);
347 }
348 return 0;
349}
350
39279cc3
CM
351/*
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.
355 *
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.
359 */
360int btrfs_drop_extents(struct btrfs_trans_handle *trans,
361 struct btrfs_root *root, struct inode *inode,
3326d1b0 362 u64 start, u64 end, u64 inline_end, u64 *hint_byte)
39279cc3
CM
363{
364 int ret;
365 struct btrfs_key key;
5f39d397 366 struct extent_buffer *leaf;
39279cc3
CM
367 int slot;
368 struct btrfs_file_extent_item *extent;
369 u64 extent_end = 0;
370 int keep;
371 struct btrfs_file_extent_item old;
372 struct btrfs_path *path;
373 u64 search_start = start;
374 int bookend;
375 int found_type;
376 int found_extent;
377 int found_inline;
ccd467d6 378 int recow;
39279cc3 379
a52d9a80
CM
380 btrfs_drop_extent_cache(inode, start, end - 1);
381
39279cc3
CM
382 path = btrfs_alloc_path();
383 if (!path)
384 return -ENOMEM;
385 while(1) {
ccd467d6 386 recow = 0;
39279cc3
CM
387 btrfs_release_path(root, path);
388 ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
389 search_start, -1);
390 if (ret < 0)
391 goto out;
392 if (ret > 0) {
393 if (path->slots[0] == 0) {
394 ret = 0;
395 goto out;
396 }
397 path->slots[0]--;
398 }
8c2383c3 399next_slot:
39279cc3
CM
400 keep = 0;
401 bookend = 0;
402 found_extent = 0;
403 found_inline = 0;
404 extent = NULL;
5f39d397 405 leaf = path->nodes[0];
39279cc3 406 slot = path->slots[0];
8c2383c3 407 ret = 0;
5f39d397 408 btrfs_item_key_to_cpu(leaf, &key, slot);
39279cc3 409 if (key.offset >= end || key.objectid != inode->i_ino) {
39279cc3
CM
410 goto out;
411 }
8c2383c3 412 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY) {
39279cc3
CM
413 goto out;
414 }
ccd467d6
CM
415 if (recow) {
416 search_start = key.offset;
417 continue;
418 }
8c2383c3
CM
419 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
420 extent = btrfs_item_ptr(leaf, slot,
421 struct btrfs_file_extent_item);
5f39d397 422 found_type = btrfs_file_extent_type(leaf, extent);
8c2383c3
CM
423 if (found_type == BTRFS_FILE_EXTENT_REG) {
424 extent_end = key.offset +
db94535d 425 btrfs_file_extent_num_bytes(leaf, extent);
8c2383c3
CM
426 found_extent = 1;
427 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
5f39d397
CM
428 struct btrfs_item *item;
429 item = btrfs_item_nr(leaf, slot);
8c2383c3
CM
430 found_inline = 1;
431 extent_end = key.offset +
5f39d397 432 btrfs_file_extent_inline_len(leaf, item);
8c2383c3
CM
433 }
434 } else {
435 extent_end = search_start;
39279cc3
CM
436 }
437
438 /* we found nothing we can drop */
8c2383c3
CM
439 if ((!found_extent && !found_inline) ||
440 search_start >= extent_end) {
441 int nextret;
442 u32 nritems;
5f39d397 443 nritems = btrfs_header_nritems(leaf);
8c2383c3
CM
444 if (slot >= nritems - 1) {
445 nextret = btrfs_next_leaf(root, path);
446 if (nextret)
447 goto out;
ccd467d6 448 recow = 1;
8c2383c3
CM
449 } else {
450 path->slots[0]++;
451 }
452 goto next_slot;
39279cc3
CM
453 }
454
455 /* FIXME, there's only one inline extent allowed right now */
456 if (found_inline) {
5f39d397 457 u64 mask = root->sectorsize - 1;
39279cc3
CM
458 search_start = (extent_end + mask) & ~mask;
459 } else
460 search_start = extent_end;
461
462 if (end < extent_end && end >= key.offset) {
463 if (found_extent) {
db94535d
CM
464 u64 disk_bytenr =
465 btrfs_file_extent_disk_bytenr(leaf, extent);
466 u64 disk_num_bytes =
467 btrfs_file_extent_disk_num_bytes(leaf,
5f39d397
CM
468 extent);
469 read_extent_buffer(leaf, &old,
470 (unsigned long)extent,
471 sizeof(old));
db94535d 472 if (disk_bytenr != 0) {
39279cc3 473 ret = btrfs_inc_extent_ref(trans, root,
db94535d 474 disk_bytenr, disk_num_bytes);
39279cc3
CM
475 BUG_ON(ret);
476 }
477 }
3326d1b0
CM
478 if (!found_inline)
479 bookend = 1;
39279cc3 480 }
39279cc3
CM
481 /* truncate existing extent */
482 if (start > key.offset) {
483 u64 new_num;
484 u64 old_num;
485 keep = 1;
5f39d397 486 WARN_ON(start & (root->sectorsize - 1));
39279cc3 487 if (found_extent) {
db94535d
CM
488 new_num = start - key.offset;
489 old_num = btrfs_file_extent_num_bytes(leaf,
490 extent);
491 *hint_byte =
492 btrfs_file_extent_disk_bytenr(leaf,
493 extent);
494 if (btrfs_file_extent_disk_bytenr(leaf,
495 extent)) {
39279cc3 496 inode->i_blocks -=
db94535d 497 (old_num - new_num) >> 9;
39279cc3 498 }
db94535d
CM
499 btrfs_set_file_extent_num_bytes(leaf, extent,
500 new_num);
5f39d397 501 btrfs_mark_buffer_dirty(leaf);
3326d1b0
CM
502 } else if (end > extent_end &&
503 key.offset < inline_end &&
504 inline_end < extent_end) {
505 u32 new_size;
506 new_size = btrfs_file_extent_calc_inline_size(
507 inline_end - key.offset);
508 btrfs_truncate_item(trans, root, path,
509 new_size);
39279cc3
CM
510 }
511 }
512 /* delete the entire extent */
513 if (!keep) {
db94535d
CM
514 u64 disk_bytenr = 0;
515 u64 disk_num_bytes = 0;
516 u64 extent_num_bytes = 0;
39279cc3 517 if (found_extent) {
db94535d
CM
518 disk_bytenr =
519 btrfs_file_extent_disk_bytenr(leaf,
5f39d397 520 extent);
db94535d
CM
521 disk_num_bytes =
522 btrfs_file_extent_disk_num_bytes(leaf,
5f39d397 523 extent);
db94535d
CM
524 extent_num_bytes =
525 btrfs_file_extent_num_bytes(leaf, extent);
526 *hint_byte =
527 btrfs_file_extent_disk_bytenr(leaf,
528 extent);
39279cc3
CM
529 }
530 ret = btrfs_del_item(trans, root, path);
54aa1f4d 531 /* TODO update progress marker and return */
39279cc3
CM
532 BUG_ON(ret);
533 btrfs_release_path(root, path);
534 extent = NULL;
db94535d
CM
535 if (found_extent && disk_bytenr != 0) {
536 inode->i_blocks -= extent_num_bytes >> 9;
39279cc3 537 ret = btrfs_free_extent(trans, root,
db94535d
CM
538 disk_bytenr,
539 disk_num_bytes, 0);
39279cc3
CM
540 }
541
542 BUG_ON(ret);
543 if (!bookend && search_start >= end) {
544 ret = 0;
545 goto out;
546 }
547 if (!bookend)
548 continue;
549 }
550 /* create bookend, splitting the extent in two */
551 if (bookend && found_extent) {
552 struct btrfs_key ins;
553 ins.objectid = inode->i_ino;
554 ins.offset = end;
39279cc3 555 btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY);
39279cc3
CM
556 btrfs_release_path(root, path);
557 ret = btrfs_insert_empty_item(trans, root, path, &ins,
558 sizeof(*extent));
8c2383c3 559
5f39d397 560 leaf = path->nodes[0];
8c2383c3 561 if (ret) {
5f39d397
CM
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);
8c2383c3 564 }
39279cc3 565 BUG_ON(ret);
5f39d397
CM
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));
570
571 btrfs_set_file_extent_offset(leaf, extent,
db94535d
CM
572 le64_to_cpu(old.offset) + end - key.offset);
573 WARN_ON(le64_to_cpu(old.num_bytes) <
574 (extent_end - end));
575 btrfs_set_file_extent_num_bytes(leaf, extent,
576 extent_end - end);
5f39d397 577 btrfs_set_file_extent_type(leaf, extent,
39279cc3 578 BTRFS_FILE_EXTENT_REG);
db94535d 579
39279cc3 580 btrfs_mark_buffer_dirty(path->nodes[0]);
db94535d 581 if (le64_to_cpu(old.disk_bytenr) != 0) {
39279cc3 582 inode->i_blocks +=
db94535d
CM
583 btrfs_file_extent_num_bytes(leaf,
584 extent) >> 9;
39279cc3
CM
585 }
586 ret = 0;
587 goto out;
588 }
589 }
590out:
591 btrfs_free_path(path);
592 return ret;
593}
594
595/*
596 * this gets pages into the page cache and locks them down
597 */
598static int prepare_pages(struct btrfs_root *root,
599 struct file *file,
600 struct page **pages,
601 size_t num_pages,
602 loff_t pos,
603 unsigned long first_index,
604 unsigned long last_index,
8c2383c3 605 size_t write_bytes)
39279cc3
CM
606{
607 int i;
608 unsigned long index = pos >> PAGE_CACHE_SHIFT;
609 struct inode *inode = file->f_path.dentry->d_inode;
39279cc3 610 int err = 0;
8c2383c3 611 u64 start_pos;
8c2383c3 612
5f39d397 613 start_pos = pos & ~((u64)root->sectorsize - 1);
39279cc3
CM
614
615 memset(pages, 0, num_pages * sizeof(struct page *));
616
617 for (i = 0; i < num_pages; i++) {
618 pages[i] = grab_cache_page(inode->i_mapping, index + i);
619 if (!pages[i]) {
620 err = -ENOMEM;
a52d9a80 621 BUG_ON(1);
39279cc3 622 }
ccd467d6
CM
623 cancel_dirty_page(pages[i], PAGE_CACHE_SIZE);
624 wait_on_page_writeback(pages[i]);
b3cfa35a 625 set_page_extent_mapped(pages[i]);
b888db2b 626 WARN_ON(!PageLocked(pages[i]));
39279cc3
CM
627 }
628 return 0;
39279cc3
CM
629}
630
631static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
632 size_t count, loff_t *ppos)
633{
634 loff_t pos;
635 size_t num_written = 0;
636 int err = 0;
637 int ret = 0;
638 struct inode *inode = file->f_path.dentry->d_inode;
639 struct btrfs_root *root = BTRFS_I(inode)->root;
8c2383c3
CM
640 struct page **pages = NULL;
641 int nrptrs;
39279cc3
CM
642 struct page *pinned[2];
643 unsigned long first_index;
644 unsigned long last_index;
8c2383c3
CM
645
646 nrptrs = min((count + PAGE_CACHE_SIZE - 1) / PAGE_CACHE_SIZE,
647 PAGE_CACHE_SIZE / (sizeof(struct page *)));
39279cc3
CM
648 pinned[0] = NULL;
649 pinned[1] = NULL;
650 if (file->f_flags & O_DIRECT)
651 return -EINVAL;
652 pos = *ppos;
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));
656 if (err)
657 goto out;
658 if (count == 0)
659 goto out;
660 err = remove_suid(file->f_path.dentry);
661 if (err)
662 goto out;
663 file_update_time(file);
664
8c2383c3 665 pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
39279cc3
CM
666
667 mutex_lock(&inode->i_mutex);
668 first_index = pos >> PAGE_CACHE_SHIFT;
669 last_index = (pos + count) >> PAGE_CACHE_SHIFT;
670
671 /*
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
675 */
676 if ((pos & (PAGE_CACHE_SIZE - 1))) {
677 pinned[0] = grab_cache_page(inode->i_mapping, first_index);
678 if (!PageUptodate(pinned[0])) {
9ebefb18 679 ret = btrfs_readpage(NULL, pinned[0]);
39279cc3
CM
680 BUG_ON(ret);
681 wait_on_page_locked(pinned[0]);
682 } else {
683 unlock_page(pinned[0]);
684 }
685 }
686 if ((pos + count) & (PAGE_CACHE_SIZE - 1)) {
687 pinned[1] = grab_cache_page(inode->i_mapping, last_index);
688 if (!PageUptodate(pinned[1])) {
9ebefb18 689 ret = btrfs_readpage(NULL, pinned[1]);
39279cc3
CM
690 BUG_ON(ret);
691 wait_on_page_locked(pinned[1]);
692 } else {
693 unlock_page(pinned[1]);
694 }
695 }
696
39279cc3
CM
697 while(count > 0) {
698 size_t offset = pos & (PAGE_CACHE_SIZE - 1);
11bd143f
CM
699 size_t write_bytes = min(count, nrptrs *
700 (size_t)PAGE_CACHE_SIZE -
8c2383c3 701 offset);
39279cc3
CM
702 size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >>
703 PAGE_CACHE_SHIFT;
704
8c2383c3 705 WARN_ON(num_pages > nrptrs);
39279cc3
CM
706 memset(pages, 0, sizeof(pages));
707 ret = prepare_pages(root, file, pages, num_pages,
708 pos, first_index, last_index,
8c2383c3 709 write_bytes);
54aa1f4d
CM
710 if (ret)
711 goto out;
39279cc3 712
39279cc3
CM
713 ret = btrfs_copy_from_user(pos, num_pages,
714 write_bytes, pages, buf);
54aa1f4d
CM
715 if (ret) {
716 btrfs_drop_pages(pages, num_pages);
717 goto out;
718 }
39279cc3
CM
719
720 ret = dirty_and_release_pages(NULL, root, file, pages,
721 num_pages, pos, write_bytes);
39279cc3 722 btrfs_drop_pages(pages, num_pages);
54aa1f4d
CM
723 if (ret)
724 goto out;
39279cc3
CM
725
726 buf += write_bytes;
727 count -= write_bytes;
728 pos += write_bytes;
729 num_written += write_bytes;
730
8c2383c3 731 balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages);
d3c2fdcf 732 btrfs_btree_balance_dirty(root, 1);
39279cc3
CM
733 cond_resched();
734 }
39279cc3
CM
735 mutex_unlock(&inode->i_mutex);
736out:
8c2383c3 737 kfree(pages);
39279cc3
CM
738 if (pinned[0])
739 page_cache_release(pinned[0]);
740 if (pinned[1])
741 page_cache_release(pinned[1]);
742 *ppos = pos;
743 current->backing_dev_info = NULL;
39279cc3
CM
744 return num_written ? num_written : err;
745}
746
39279cc3
CM
747static int btrfs_sync_file(struct file *file,
748 struct dentry *dentry, int datasync)
749{
750 struct inode *inode = dentry->d_inode;
751 struct btrfs_root *root = BTRFS_I(inode)->root;
15ee9bc7 752 int ret = 0;
39279cc3
CM
753 struct btrfs_trans_handle *trans;
754
755 /*
15ee9bc7
JB
756 * check the transaction that last modified this inode
757 * and see if its already been committed
39279cc3
CM
758 */
759 mutex_lock(&root->fs_info->fs_mutex);
15ee9bc7
JB
760 if (!BTRFS_I(inode)->last_trans)
761 goto out;
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);
767 goto out;
768 }
769 mutex_unlock(&root->fs_info->trans_mutex);
770
771 /*
a52d9a80
CM
772 * ok we haven't committed the transaction yet, lets do a commit
773 */
39279cc3
CM
774 trans = btrfs_start_transaction(root, 1);
775 if (!trans) {
776 ret = -ENOMEM;
777 goto out;
778 }
779 ret = btrfs_commit_transaction(trans, root);
39279cc3 780out:
15ee9bc7 781 mutex_unlock(&root->fs_info->fs_mutex);
39279cc3
CM
782 return ret > 0 ? EIO : ret;
783}
784
9ebefb18 785static struct vm_operations_struct btrfs_file_vm_ops = {
92fee66d
CM
786#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
787 .nopage = filemap_nopage,
788 .populate = filemap_populate,
789#else
790 .fault = filemap_fault,
791#endif
9ebefb18
CM
792 .page_mkwrite = btrfs_page_mkwrite,
793};
794
795static int btrfs_file_mmap(struct file *filp, struct vm_area_struct *vma)
796{
797 vma->vm_ops = &btrfs_file_vm_ops;
798 file_accessed(filp);
799 return 0;
800}
801
39279cc3
CM
802struct file_operations btrfs_file_operations = {
803 .llseek = generic_file_llseek,
804 .read = do_sync_read,
9ebefb18 805 .aio_read = generic_file_aio_read,
39279cc3 806 .write = btrfs_file_write,
9ebefb18 807 .mmap = btrfs_file_mmap,
39279cc3 808 .open = generic_file_open,
39279cc3 809 .fsync = btrfs_sync_file,
34287aa3 810 .unlocked_ioctl = btrfs_ioctl,
39279cc3 811#ifdef CONFIG_COMPAT
34287aa3 812 .compat_ioctl = btrfs_ioctl,
39279cc3
CM
813#endif
814};
815