]>
Commit | Line | Data |
---|---|---|
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" | |
e02119d5 CM |
39 | #include "tree-log.h" |
40 | #include "locking.h" | |
12fa8ec6 | 41 | #include "compat.h" |
39279cc3 CM |
42 | |
43 | ||
d352ac68 CM |
44 | /* simple helper to fault in pages and copy. This should go away |
45 | * and be replaced with calls into generic code. | |
46 | */ | |
a1b32a59 CM |
47 | static int noinline btrfs_copy_from_user(loff_t pos, int num_pages, |
48 | int write_bytes, | |
49 | struct page **prepared_pages, | |
50 | const char __user * buf) | |
39279cc3 CM |
51 | { |
52 | long page_fault = 0; | |
53 | int i; | |
54 | int offset = pos & (PAGE_CACHE_SIZE - 1); | |
55 | ||
56 | for (i = 0; i < num_pages && write_bytes > 0; i++, offset = 0) { | |
57 | size_t count = min_t(size_t, | |
58 | PAGE_CACHE_SIZE - offset, write_bytes); | |
59 | struct page *page = prepared_pages[i]; | |
60 | fault_in_pages_readable(buf, count); | |
61 | ||
62 | /* Copy data from userspace to the current page */ | |
63 | kmap(page); | |
64 | page_fault = __copy_from_user(page_address(page) + offset, | |
65 | buf, count); | |
66 | /* Flush processor's dcache for this page */ | |
67 | flush_dcache_page(page); | |
68 | kunmap(page); | |
69 | buf += count; | |
70 | write_bytes -= count; | |
71 | ||
72 | if (page_fault) | |
73 | break; | |
74 | } | |
75 | return page_fault ? -EFAULT : 0; | |
76 | } | |
77 | ||
d352ac68 CM |
78 | /* |
79 | * unlocks pages after btrfs_file_write is done with them | |
80 | */ | |
a1b32a59 | 81 | static void noinline btrfs_drop_pages(struct page **pages, size_t num_pages) |
39279cc3 CM |
82 | { |
83 | size_t i; | |
84 | for (i = 0; i < num_pages; i++) { | |
85 | if (!pages[i]) | |
86 | break; | |
d352ac68 CM |
87 | /* page checked is some magic around finding pages that |
88 | * have been modified without going through btrfs_set_page_dirty | |
89 | * clear it here | |
90 | */ | |
4a096752 | 91 | ClearPageChecked(pages[i]); |
39279cc3 CM |
92 | unlock_page(pages[i]); |
93 | mark_page_accessed(pages[i]); | |
94 | page_cache_release(pages[i]); | |
95 | } | |
96 | } | |
97 | ||
d352ac68 CM |
98 | /* this does all the hard work for inserting an inline extent into |
99 | * the btree. Any existing inline extent is extended as required to make room, | |
100 | * otherwise things are inserted as required into the btree | |
101 | */ | |
98ed5174 | 102 | static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, |
a52d9a80 | 103 | struct btrfs_root *root, struct inode *inode, |
3326d1b0 CM |
104 | u64 offset, size_t size, |
105 | struct page **pages, size_t page_offset, | |
106 | int num_pages) | |
54aa1f4d CM |
107 | { |
108 | struct btrfs_key key; | |
109 | struct btrfs_path *path; | |
5f39d397 CM |
110 | struct extent_buffer *leaf; |
111 | char *kaddr; | |
112 | unsigned long ptr; | |
54aa1f4d | 113 | struct btrfs_file_extent_item *ei; |
3326d1b0 | 114 | struct page *page; |
54aa1f4d CM |
115 | u32 datasize; |
116 | int err = 0; | |
117 | int ret; | |
3326d1b0 CM |
118 | int i; |
119 | ssize_t cur_size; | |
54aa1f4d CM |
120 | |
121 | path = btrfs_alloc_path(); | |
122 | if (!path) | |
123 | return -ENOMEM; | |
124 | ||
54aa1f4d CM |
125 | btrfs_set_trans_block_group(trans, inode); |
126 | ||
127 | key.objectid = inode->i_ino; | |
128 | key.offset = offset; | |
54aa1f4d | 129 | btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); |
54aa1f4d | 130 | |
3326d1b0 CM |
131 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); |
132 | if (ret < 0) { | |
54aa1f4d CM |
133 | err = ret; |
134 | goto fail; | |
135 | } | |
3326d1b0 | 136 | if (ret == 1) { |
179e29e4 CM |
137 | struct btrfs_key found_key; |
138 | ||
139 | if (path->slots[0] == 0) | |
140 | goto insert; | |
141 | ||
3326d1b0 CM |
142 | path->slots[0]--; |
143 | leaf = path->nodes[0]; | |
179e29e4 CM |
144 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); |
145 | ||
146 | if (found_key.objectid != inode->i_ino) | |
147 | goto insert; | |
148 | ||
149 | if (found_key.type != BTRFS_EXTENT_DATA_KEY) | |
150 | goto insert; | |
3326d1b0 CM |
151 | ei = btrfs_item_ptr(leaf, path->slots[0], |
152 | struct btrfs_file_extent_item); | |
153 | ||
154 | if (btrfs_file_extent_type(leaf, ei) != | |
155 | BTRFS_FILE_EXTENT_INLINE) { | |
156 | goto insert; | |
157 | } | |
158 | btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); | |
159 | ret = 0; | |
160 | } | |
161 | if (ret == 0) { | |
162 | u32 found_size; | |
18f16f7b | 163 | u64 found_end; |
3326d1b0 CM |
164 | |
165 | leaf = path->nodes[0]; | |
166 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
167 | struct btrfs_file_extent_item); | |
168 | ||
169 | if (btrfs_file_extent_type(leaf, ei) != | |
170 | BTRFS_FILE_EXTENT_INLINE) { | |
171 | err = ret; | |
172 | btrfs_print_leaf(root, leaf); | |
173 | printk("found wasn't inline offset %Lu inode %lu\n", | |
174 | offset, inode->i_ino); | |
175 | goto fail; | |
176 | } | |
3326d1b0 CM |
177 | found_size = btrfs_file_extent_inline_len(leaf, |
178 | btrfs_item_nr(leaf, path->slots[0])); | |
18f16f7b | 179 | found_end = key.offset + found_size; |
3326d1b0 | 180 | |
18f16f7b | 181 | if (found_end < offset + size) { |
3326d1b0 CM |
182 | btrfs_release_path(root, path); |
183 | ret = btrfs_search_slot(trans, root, &key, path, | |
18f16f7b | 184 | offset + size - found_end, 1); |
3326d1b0 | 185 | BUG_ON(ret != 0); |
179e29e4 | 186 | |
3326d1b0 | 187 | ret = btrfs_extend_item(trans, root, path, |
18f16f7b | 188 | offset + size - found_end); |
3326d1b0 CM |
189 | if (ret) { |
190 | err = ret; | |
191 | goto fail; | |
192 | } | |
193 | leaf = path->nodes[0]; | |
194 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
195 | struct btrfs_file_extent_item); | |
a76a3cd4 | 196 | inode_add_bytes(inode, offset + size - found_end); |
3326d1b0 | 197 | } |
18f16f7b Y |
198 | if (found_end < offset) { |
199 | ptr = btrfs_file_extent_inline_start(ei) + found_size; | |
200 | memset_extent_buffer(leaf, 0, ptr, offset - found_end); | |
201 | } | |
3326d1b0 CM |
202 | } else { |
203 | insert: | |
204 | btrfs_release_path(root, path); | |
18f16f7b | 205 | datasize = offset + size - key.offset; |
a76a3cd4 | 206 | inode_add_bytes(inode, datasize); |
18f16f7b | 207 | datasize = btrfs_file_extent_calc_inline_size(datasize); |
3326d1b0 CM |
208 | ret = btrfs_insert_empty_item(trans, root, path, &key, |
209 | datasize); | |
210 | if (ret) { | |
211 | err = ret; | |
212 | printk("got bad ret %d\n", ret); | |
213 | goto fail; | |
214 | } | |
215 | leaf = path->nodes[0]; | |
216 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
217 | struct btrfs_file_extent_item); | |
218 | btrfs_set_file_extent_generation(leaf, ei, trans->transid); | |
219 | btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); | |
220 | } | |
18f16f7b | 221 | ptr = btrfs_file_extent_inline_start(ei) + offset - key.offset; |
3326d1b0 CM |
222 | |
223 | cur_size = size; | |
224 | i = 0; | |
225 | while (size > 0) { | |
226 | page = pages[i]; | |
227 | kaddr = kmap_atomic(page, KM_USER0); | |
ae2f5411 | 228 | cur_size = min_t(size_t, PAGE_CACHE_SIZE - page_offset, size); |
3326d1b0 CM |
229 | write_extent_buffer(leaf, kaddr + page_offset, ptr, cur_size); |
230 | kunmap_atomic(kaddr, KM_USER0); | |
231 | page_offset = 0; | |
232 | ptr += cur_size; | |
233 | size -= cur_size; | |
234 | if (i >= num_pages) { | |
235 | printk("i %d num_pages %d\n", i, num_pages); | |
236 | } | |
237 | i++; | |
238 | } | |
5f39d397 | 239 | btrfs_mark_buffer_dirty(leaf); |
54aa1f4d CM |
240 | fail: |
241 | btrfs_free_path(path); | |
54aa1f4d CM |
242 | return err; |
243 | } | |
244 | ||
d352ac68 CM |
245 | /* |
246 | * after copy_from_user, pages need to be dirtied and we need to make | |
247 | * sure holes are created between the current EOF and the start of | |
248 | * any next extents (if required). | |
249 | * | |
250 | * this also makes the decision about creating an inline extent vs | |
251 | * doing real data extents, marking pages dirty and delalloc as required. | |
252 | */ | |
98ed5174 | 253 | static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, |
39279cc3 CM |
254 | struct btrfs_root *root, |
255 | struct file *file, | |
256 | struct page **pages, | |
257 | size_t num_pages, | |
258 | loff_t pos, | |
259 | size_t write_bytes) | |
260 | { | |
39279cc3 | 261 | int err = 0; |
a52d9a80 | 262 | int i; |
6da6abae | 263 | struct inode *inode = fdentry(file)->d_inode; |
d1310b2e | 264 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
db94535d CM |
265 | u64 hint_byte; |
266 | u64 num_bytes; | |
a52d9a80 CM |
267 | u64 start_pos; |
268 | u64 end_of_last_block; | |
269 | u64 end_pos = pos + write_bytes; | |
dcfec0dc | 270 | u64 inline_size; |
ee6e6504 | 271 | int did_inline = 0; |
a52d9a80 | 272 | loff_t isize = i_size_read(inode); |
39279cc3 | 273 | |
5f39d397 | 274 | start_pos = pos & ~((u64)root->sectorsize - 1); |
db94535d CM |
275 | num_bytes = (write_bytes + pos - start_pos + |
276 | root->sectorsize - 1) & ~((u64)root->sectorsize - 1); | |
39279cc3 | 277 | |
db94535d CM |
278 | end_of_last_block = start_pos + num_bytes - 1; |
279 | ||
d1310b2e | 280 | lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); |
37d1aeee | 281 | trans = btrfs_join_transaction(root, 1); |
a52d9a80 CM |
282 | if (!trans) { |
283 | err = -ENOMEM; | |
284 | goto out_unlock; | |
285 | } | |
286 | btrfs_set_trans_block_group(trans, inode); | |
db94535d | 287 | hint_byte = 0; |
a52d9a80 CM |
288 | |
289 | if ((end_of_last_block & 4095) == 0) { | |
9433063b | 290 | printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block); |
a52d9a80 | 291 | } |
d1310b2e | 292 | set_extent_uptodate(io_tree, start_pos, end_of_last_block, GFP_NOFS); |
a52d9a80 CM |
293 | |
294 | /* FIXME...EIEIO, ENOSPC and more */ | |
a52d9a80 | 295 | /* insert any holes we need to create */ |
1b1e2135 | 296 | if (isize < start_pos) { |
a52d9a80 CM |
297 | u64 last_pos_in_file; |
298 | u64 hole_size; | |
5f39d397 | 299 | u64 mask = root->sectorsize - 1; |
a52d9a80 | 300 | last_pos_in_file = (isize + mask) & ~mask; |
1b1e2135 | 301 | hole_size = (start_pos - last_pos_in_file + mask) & ~mask; |
e6dcd2dc CM |
302 | if (hole_size > 0) { |
303 | btrfs_wait_ordered_range(inode, last_pos_in_file, | |
304 | last_pos_in_file + hole_size); | |
ee6e6504 | 305 | mutex_lock(&BTRFS_I(inode)->extent_mutex); |
2bf5a725 CM |
306 | err = btrfs_drop_extents(trans, root, inode, |
307 | last_pos_in_file, | |
308 | last_pos_in_file + hole_size, | |
3326d1b0 | 309 | last_pos_in_file, |
db94535d | 310 | &hint_byte); |
2bf5a725 CM |
311 | if (err) |
312 | goto failed; | |
313 | ||
a52d9a80 CM |
314 | err = btrfs_insert_file_extent(trans, root, |
315 | inode->i_ino, | |
316 | last_pos_in_file, | |
f2eb0a24 | 317 | 0, 0, hole_size, 0); |
d1310b2e | 318 | btrfs_drop_extent_cache(inode, last_pos_in_file, |
5b21f2ed | 319 | last_pos_in_file + hole_size - 1, 0); |
ee6e6504 | 320 | mutex_unlock(&BTRFS_I(inode)->extent_mutex); |
5f56406a | 321 | btrfs_check_file(root, inode); |
a52d9a80 CM |
322 | } |
323 | if (err) | |
39279cc3 | 324 | goto failed; |
a52d9a80 CM |
325 | } |
326 | ||
327 | /* | |
328 | * either allocate an extent for the new bytes or setup the key | |
329 | * to show we are doing inline data in the extent | |
330 | */ | |
3326d1b0 CM |
331 | inline_size = end_pos; |
332 | if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) || | |
6f568d35 CM |
333 | inline_size > root->fs_info->max_inline || |
334 | (inline_size & (root->sectorsize -1)) == 0 || | |
3326d1b0 | 335 | inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { |
e6dcd2dc CM |
336 | /* check for reserved extents on each page, we don't want |
337 | * to reset the delalloc bit on things that already have | |
338 | * extents reserved. | |
339 | */ | |
ea8c2819 | 340 | btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block); |
a52d9a80 CM |
341 | for (i = 0; i < num_pages; i++) { |
342 | struct page *p = pages[i]; | |
343 | SetPageUptodate(p); | |
247e743c | 344 | ClearPageChecked(p); |
b888db2b | 345 | set_page_dirty(p); |
39279cc3 | 346 | } |
a52d9a80 | 347 | } else { |
3326d1b0 | 348 | u64 aligned_end; |
b888db2b | 349 | /* step one, delete the existing extents in this range */ |
3326d1b0 CM |
350 | aligned_end = (pos + write_bytes + root->sectorsize - 1) & |
351 | ~((u64)root->sectorsize - 1); | |
ee6e6504 | 352 | mutex_lock(&BTRFS_I(inode)->extent_mutex); |
2bf5a725 | 353 | err = btrfs_drop_extents(trans, root, inode, start_pos, |
179e29e4 | 354 | aligned_end, aligned_end, &hint_byte); |
2bf5a725 CM |
355 | if (err) |
356 | goto failed; | |
dcfec0dc Y |
357 | if (isize > inline_size) |
358 | inline_size = min_t(u64, isize, aligned_end); | |
359 | inline_size -= start_pos; | |
a52d9a80 | 360 | err = insert_inline_extent(trans, root, inode, start_pos, |
dcfec0dc | 361 | inline_size, pages, 0, num_pages); |
5b21f2ed | 362 | btrfs_drop_extent_cache(inode, start_pos, aligned_end - 1, 0); |
a52d9a80 | 363 | BUG_ON(err); |
ee6e6504 | 364 | mutex_unlock(&BTRFS_I(inode)->extent_mutex); |
f87f057b CM |
365 | |
366 | /* | |
367 | * an ugly way to do all the prop accounting around | |
368 | * the page bits and mapping tags | |
369 | */ | |
370 | set_page_writeback(pages[0]); | |
371 | end_page_writeback(pages[0]); | |
ee6e6504 | 372 | did_inline = 1; |
a52d9a80 CM |
373 | } |
374 | if (end_pos > isize) { | |
375 | i_size_write(inode, end_pos); | |
ee6e6504 CM |
376 | if (did_inline) |
377 | BTRFS_I(inode)->disk_i_size = end_pos; | |
a52d9a80 | 378 | btrfs_update_inode(trans, root, inode); |
39279cc3 CM |
379 | } |
380 | failed: | |
017e5369 | 381 | err = btrfs_end_transaction(trans, root); |
a52d9a80 | 382 | out_unlock: |
d1310b2e | 383 | unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); |
39279cc3 CM |
384 | return err; |
385 | } | |
386 | ||
d352ac68 CM |
387 | /* |
388 | * this drops all the extents in the cache that intersect the range | |
389 | * [start, end]. Existing extents are split as required. | |
390 | */ | |
5b21f2ed ZY |
391 | int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, |
392 | int skip_pinned) | |
a52d9a80 CM |
393 | { |
394 | struct extent_map *em; | |
3b951516 CM |
395 | struct extent_map *split = NULL; |
396 | struct extent_map *split2 = NULL; | |
a52d9a80 | 397 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
39b5637f | 398 | u64 len = end - start + 1; |
3b951516 CM |
399 | int ret; |
400 | int testend = 1; | |
5b21f2ed | 401 | unsigned long flags; |
a52d9a80 | 402 | |
e6dcd2dc | 403 | WARN_ON(end < start); |
3b951516 | 404 | if (end == (u64)-1) { |
39b5637f | 405 | len = (u64)-1; |
3b951516 CM |
406 | testend = 0; |
407 | } | |
a52d9a80 | 408 | while(1) { |
3b951516 CM |
409 | if (!split) |
410 | split = alloc_extent_map(GFP_NOFS); | |
411 | if (!split2) | |
412 | split2 = alloc_extent_map(GFP_NOFS); | |
413 | ||
d1310b2e | 414 | spin_lock(&em_tree->lock); |
39b5637f | 415 | em = lookup_extent_mapping(em_tree, start, len); |
d1310b2e CM |
416 | if (!em) { |
417 | spin_unlock(&em_tree->lock); | |
a52d9a80 | 418 | break; |
d1310b2e | 419 | } |
5b21f2ed ZY |
420 | flags = em->flags; |
421 | if (skip_pinned && test_bit(EXTENT_FLAG_PINNED, &em->flags)) { | |
422 | spin_unlock(&em_tree->lock); | |
423 | if (em->start <= start && | |
424 | (!testend || em->start + em->len >= start + len)) { | |
425 | free_extent_map(em); | |
426 | break; | |
427 | } | |
428 | if (start < em->start) { | |
429 | len = em->start - start; | |
430 | } else { | |
431 | len = start + len - (em->start + em->len); | |
432 | start = em->start + em->len; | |
433 | } | |
434 | free_extent_map(em); | |
435 | continue; | |
436 | } | |
3ce7e67a | 437 | clear_bit(EXTENT_FLAG_PINNED, &em->flags); |
a52d9a80 | 438 | remove_extent_mapping(em_tree, em); |
3b951516 CM |
439 | |
440 | if (em->block_start < EXTENT_MAP_LAST_BYTE && | |
441 | em->start < start) { | |
442 | split->start = em->start; | |
443 | split->len = start - em->start; | |
444 | split->block_start = em->block_start; | |
445 | split->bdev = em->bdev; | |
5b21f2ed | 446 | split->flags = flags; |
3b951516 CM |
447 | ret = add_extent_mapping(em_tree, split); |
448 | BUG_ON(ret); | |
449 | free_extent_map(split); | |
450 | split = split2; | |
451 | split2 = NULL; | |
452 | } | |
453 | if (em->block_start < EXTENT_MAP_LAST_BYTE && | |
454 | testend && em->start + em->len > start + len) { | |
455 | u64 diff = start + len - em->start; | |
456 | ||
457 | split->start = start + len; | |
458 | split->len = em->start + em->len - (start + len); | |
459 | split->bdev = em->bdev; | |
5b21f2ed | 460 | split->flags = flags; |
3b951516 CM |
461 | |
462 | split->block_start = em->block_start + diff; | |
463 | ||
464 | ret = add_extent_mapping(em_tree, split); | |
465 | BUG_ON(ret); | |
466 | free_extent_map(split); | |
467 | split = NULL; | |
468 | } | |
d1310b2e CM |
469 | spin_unlock(&em_tree->lock); |
470 | ||
a52d9a80 CM |
471 | /* once for us */ |
472 | free_extent_map(em); | |
473 | /* once for the tree*/ | |
474 | free_extent_map(em); | |
475 | } | |
3b951516 CM |
476 | if (split) |
477 | free_extent_map(split); | |
478 | if (split2) | |
479 | free_extent_map(split2); | |
a52d9a80 CM |
480 | return 0; |
481 | } | |
482 | ||
5f56406a CM |
483 | int btrfs_check_file(struct btrfs_root *root, struct inode *inode) |
484 | { | |
485 | return 0; | |
486 | #if 0 | |
487 | struct btrfs_path *path; | |
488 | struct btrfs_key found_key; | |
489 | struct extent_buffer *leaf; | |
490 | struct btrfs_file_extent_item *extent; | |
491 | u64 last_offset = 0; | |
492 | int nritems; | |
493 | int slot; | |
494 | int found_type; | |
495 | int ret; | |
496 | int err = 0; | |
497 | u64 extent_end = 0; | |
498 | ||
499 | path = btrfs_alloc_path(); | |
500 | ret = btrfs_lookup_file_extent(NULL, root, path, inode->i_ino, | |
501 | last_offset, 0); | |
502 | while(1) { | |
503 | nritems = btrfs_header_nritems(path->nodes[0]); | |
504 | if (path->slots[0] >= nritems) { | |
505 | ret = btrfs_next_leaf(root, path); | |
506 | if (ret) | |
507 | goto out; | |
508 | nritems = btrfs_header_nritems(path->nodes[0]); | |
509 | } | |
510 | slot = path->slots[0]; | |
511 | leaf = path->nodes[0]; | |
512 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
513 | if (found_key.objectid != inode->i_ino) | |
514 | break; | |
515 | if (found_key.type != BTRFS_EXTENT_DATA_KEY) | |
516 | goto out; | |
517 | ||
9069218d | 518 | if (found_key.offset < last_offset) { |
5f56406a CM |
519 | WARN_ON(1); |
520 | btrfs_print_leaf(root, leaf); | |
521 | printk("inode %lu found offset %Lu expected %Lu\n", | |
522 | inode->i_ino, found_key.offset, last_offset); | |
523 | err = 1; | |
524 | goto out; | |
525 | } | |
526 | extent = btrfs_item_ptr(leaf, slot, | |
527 | struct btrfs_file_extent_item); | |
528 | found_type = btrfs_file_extent_type(leaf, extent); | |
529 | if (found_type == BTRFS_FILE_EXTENT_REG) { | |
530 | extent_end = found_key.offset + | |
531 | btrfs_file_extent_num_bytes(leaf, extent); | |
532 | } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { | |
533 | struct btrfs_item *item; | |
534 | item = btrfs_item_nr(leaf, slot); | |
535 | extent_end = found_key.offset + | |
536 | btrfs_file_extent_inline_len(leaf, item); | |
537 | extent_end = (extent_end + root->sectorsize - 1) & | |
538 | ~((u64)root->sectorsize -1 ); | |
539 | } | |
540 | last_offset = extent_end; | |
541 | path->slots[0]++; | |
542 | } | |
9069218d | 543 | if (0 && last_offset < inode->i_size) { |
5f56406a CM |
544 | WARN_ON(1); |
545 | btrfs_print_leaf(root, leaf); | |
546 | printk("inode %lu found offset %Lu size %Lu\n", inode->i_ino, | |
547 | last_offset, inode->i_size); | |
548 | err = 1; | |
549 | ||
550 | } | |
551 | out: | |
552 | btrfs_free_path(path); | |
553 | return err; | |
554 | #endif | |
555 | } | |
556 | ||
39279cc3 CM |
557 | /* |
558 | * this is very complex, but the basic idea is to drop all extents | |
559 | * in the range start - end. hint_block is filled in with a block number | |
560 | * that would be a good hint to the block allocator for this file. | |
561 | * | |
562 | * If an extent intersects the range but is not entirely inside the range | |
563 | * it is either truncated or split. Anything entirely inside the range | |
564 | * is deleted from the tree. | |
d352ac68 CM |
565 | * |
566 | * inline_limit is used to tell this code which offsets in the file to keep | |
567 | * if they contain inline extents. | |
39279cc3 | 568 | */ |
a1b32a59 | 569 | int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, |
39279cc3 | 570 | struct btrfs_root *root, struct inode *inode, |
00f5c795 | 571 | u64 start, u64 end, u64 inline_limit, u64 *hint_byte) |
39279cc3 | 572 | { |
00f5c795 CM |
573 | u64 extent_end = 0; |
574 | u64 search_start = start; | |
31840ae1 ZY |
575 | u64 leaf_start; |
576 | u64 root_gen; | |
577 | u64 root_owner; | |
5f39d397 | 578 | struct extent_buffer *leaf; |
39279cc3 | 579 | struct btrfs_file_extent_item *extent; |
39279cc3 | 580 | struct btrfs_path *path; |
00f5c795 CM |
581 | struct btrfs_key key; |
582 | struct btrfs_file_extent_item old; | |
583 | int keep; | |
584 | int slot; | |
39279cc3 CM |
585 | int bookend; |
586 | int found_type; | |
587 | int found_extent; | |
588 | int found_inline; | |
ccd467d6 | 589 | int recow; |
00f5c795 | 590 | int ret; |
39279cc3 | 591 | |
5b21f2ed | 592 | btrfs_drop_extent_cache(inode, start, end - 1, 0); |
a52d9a80 | 593 | |
39279cc3 CM |
594 | path = btrfs_alloc_path(); |
595 | if (!path) | |
596 | return -ENOMEM; | |
597 | while(1) { | |
ccd467d6 | 598 | recow = 0; |
39279cc3 CM |
599 | btrfs_release_path(root, path); |
600 | ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino, | |
601 | search_start, -1); | |
602 | if (ret < 0) | |
603 | goto out; | |
604 | if (ret > 0) { | |
605 | if (path->slots[0] == 0) { | |
606 | ret = 0; | |
607 | goto out; | |
608 | } | |
609 | path->slots[0]--; | |
610 | } | |
8c2383c3 | 611 | next_slot: |
39279cc3 CM |
612 | keep = 0; |
613 | bookend = 0; | |
614 | found_extent = 0; | |
615 | found_inline = 0; | |
31840ae1 ZY |
616 | leaf_start = 0; |
617 | root_gen = 0; | |
618 | root_owner = 0; | |
39279cc3 | 619 | extent = NULL; |
5f39d397 | 620 | leaf = path->nodes[0]; |
39279cc3 | 621 | slot = path->slots[0]; |
8c2383c3 | 622 | ret = 0; |
5f39d397 | 623 | btrfs_item_key_to_cpu(leaf, &key, slot); |
7261009c Y |
624 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY && |
625 | key.offset >= end) { | |
39279cc3 CM |
626 | goto out; |
627 | } | |
7261009c Y |
628 | if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || |
629 | key.objectid != inode->i_ino) { | |
39279cc3 CM |
630 | goto out; |
631 | } | |
ccd467d6 CM |
632 | if (recow) { |
633 | search_start = key.offset; | |
634 | continue; | |
635 | } | |
8c2383c3 CM |
636 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { |
637 | extent = btrfs_item_ptr(leaf, slot, | |
638 | struct btrfs_file_extent_item); | |
5f39d397 | 639 | found_type = btrfs_file_extent_type(leaf, extent); |
8c2383c3 | 640 | if (found_type == BTRFS_FILE_EXTENT_REG) { |
257d0ce3 CM |
641 | extent_end = |
642 | btrfs_file_extent_disk_bytenr(leaf, | |
643 | extent); | |
644 | if (extent_end) | |
645 | *hint_byte = extent_end; | |
646 | ||
8c2383c3 | 647 | extent_end = key.offset + |
db94535d | 648 | btrfs_file_extent_num_bytes(leaf, extent); |
8c2383c3 CM |
649 | found_extent = 1; |
650 | } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { | |
5f39d397 CM |
651 | struct btrfs_item *item; |
652 | item = btrfs_item_nr(leaf, slot); | |
8c2383c3 CM |
653 | found_inline = 1; |
654 | extent_end = key.offset + | |
5f39d397 | 655 | btrfs_file_extent_inline_len(leaf, item); |
8c2383c3 CM |
656 | } |
657 | } else { | |
658 | extent_end = search_start; | |
39279cc3 CM |
659 | } |
660 | ||
661 | /* we found nothing we can drop */ | |
8c2383c3 CM |
662 | if ((!found_extent && !found_inline) || |
663 | search_start >= extent_end) { | |
664 | int nextret; | |
665 | u32 nritems; | |
5f39d397 | 666 | nritems = btrfs_header_nritems(leaf); |
8c2383c3 CM |
667 | if (slot >= nritems - 1) { |
668 | nextret = btrfs_next_leaf(root, path); | |
669 | if (nextret) | |
670 | goto out; | |
ccd467d6 | 671 | recow = 1; |
8c2383c3 CM |
672 | } else { |
673 | path->slots[0]++; | |
674 | } | |
675 | goto next_slot; | |
39279cc3 CM |
676 | } |
677 | ||
39279cc3 | 678 | if (found_inline) { |
5f39d397 | 679 | u64 mask = root->sectorsize - 1; |
39279cc3 CM |
680 | search_start = (extent_end + mask) & ~mask; |
681 | } else | |
682 | search_start = extent_end; | |
6e3b9666 | 683 | if (end <= extent_end && start >= key.offset && found_inline) { |
179e29e4 | 684 | *hint_byte = EXTENT_MAP_INLINE; |
31840ae1 ZY |
685 | goto out; |
686 | } | |
687 | ||
688 | if (found_extent) { | |
689 | read_extent_buffer(leaf, &old, (unsigned long)extent, | |
690 | sizeof(old)); | |
691 | root_gen = btrfs_header_generation(leaf); | |
692 | root_owner = btrfs_header_owner(leaf); | |
693 | leaf_start = leaf->start; | |
179e29e4 | 694 | } |
31840ae1 | 695 | |
39279cc3 | 696 | if (end < extent_end && end >= key.offset) { |
179e29e4 | 697 | bookend = 1; |
0181e58f | 698 | if (found_inline && start <= key.offset) |
179e29e4 | 699 | keep = 1; |
39279cc3 | 700 | } |
39279cc3 CM |
701 | /* truncate existing extent */ |
702 | if (start > key.offset) { | |
703 | u64 new_num; | |
704 | u64 old_num; | |
705 | keep = 1; | |
5f39d397 | 706 | WARN_ON(start & (root->sectorsize - 1)); |
39279cc3 | 707 | if (found_extent) { |
db94535d CM |
708 | new_num = start - key.offset; |
709 | old_num = btrfs_file_extent_num_bytes(leaf, | |
710 | extent); | |
711 | *hint_byte = | |
712 | btrfs_file_extent_disk_bytenr(leaf, | |
713 | extent); | |
714 | if (btrfs_file_extent_disk_bytenr(leaf, | |
715 | extent)) { | |
a76a3cd4 YZ |
716 | inode_sub_bytes(inode, old_num - |
717 | new_num); | |
39279cc3 | 718 | } |
db94535d CM |
719 | btrfs_set_file_extent_num_bytes(leaf, extent, |
720 | new_num); | |
5f39d397 | 721 | btrfs_mark_buffer_dirty(leaf); |
00f5c795 CM |
722 | } else if (key.offset < inline_limit && |
723 | (end > extent_end) && | |
724 | (inline_limit < extent_end)) { | |
3326d1b0 CM |
725 | u32 new_size; |
726 | new_size = btrfs_file_extent_calc_inline_size( | |
00f5c795 | 727 | inline_limit - key.offset); |
a76a3cd4 YZ |
728 | inode_sub_bytes(inode, extent_end - |
729 | inline_limit); | |
3326d1b0 | 730 | btrfs_truncate_item(trans, root, path, |
179e29e4 | 731 | new_size, 1); |
39279cc3 CM |
732 | } |
733 | } | |
734 | /* delete the entire extent */ | |
735 | if (!keep) { | |
a76a3cd4 YZ |
736 | if (found_inline) |
737 | inode_sub_bytes(inode, extent_end - | |
738 | key.offset); | |
39279cc3 | 739 | ret = btrfs_del_item(trans, root, path); |
54aa1f4d | 740 | /* TODO update progress marker and return */ |
39279cc3 | 741 | BUG_ON(ret); |
39279cc3 | 742 | extent = NULL; |
31840ae1 ZY |
743 | btrfs_release_path(root, path); |
744 | /* the extent will be freed later */ | |
39279cc3 | 745 | } |
0181e58f | 746 | if (bookend && found_inline && start <= key.offset) { |
179e29e4 CM |
747 | u32 new_size; |
748 | new_size = btrfs_file_extent_calc_inline_size( | |
0181e58f | 749 | extent_end - end); |
a76a3cd4 | 750 | inode_sub_bytes(inode, end - key.offset); |
31840ae1 ZY |
751 | ret = btrfs_truncate_item(trans, root, path, |
752 | new_size, 0); | |
753 | BUG_ON(ret); | |
179e29e4 | 754 | } |
39279cc3 CM |
755 | /* create bookend, splitting the extent in two */ |
756 | if (bookend && found_extent) { | |
31840ae1 | 757 | u64 disk_bytenr; |
39279cc3 CM |
758 | struct btrfs_key ins; |
759 | ins.objectid = inode->i_ino; | |
760 | ins.offset = end; | |
39279cc3 | 761 | btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY); |
39279cc3 CM |
762 | btrfs_release_path(root, path); |
763 | ret = btrfs_insert_empty_item(trans, root, path, &ins, | |
764 | sizeof(*extent)); | |
31840ae1 | 765 | BUG_ON(ret); |
8c2383c3 | 766 | |
5f39d397 | 767 | leaf = path->nodes[0]; |
5f39d397 CM |
768 | extent = btrfs_item_ptr(leaf, path->slots[0], |
769 | struct btrfs_file_extent_item); | |
770 | write_extent_buffer(leaf, &old, | |
771 | (unsigned long)extent, sizeof(old)); | |
772 | ||
773 | btrfs_set_file_extent_offset(leaf, extent, | |
db94535d CM |
774 | le64_to_cpu(old.offset) + end - key.offset); |
775 | WARN_ON(le64_to_cpu(old.num_bytes) < | |
776 | (extent_end - end)); | |
777 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
778 | extent_end - end); | |
5f39d397 | 779 | btrfs_set_file_extent_type(leaf, extent, |
39279cc3 | 780 | BTRFS_FILE_EXTENT_REG); |
db94535d | 781 | |
39279cc3 | 782 | btrfs_mark_buffer_dirty(path->nodes[0]); |
31840ae1 ZY |
783 | |
784 | disk_bytenr = le64_to_cpu(old.disk_bytenr); | |
785 | if (disk_bytenr != 0) { | |
786 | ret = btrfs_inc_extent_ref(trans, root, | |
787 | disk_bytenr, | |
788 | le64_to_cpu(old.disk_num_bytes), | |
789 | leaf->start, | |
790 | root->root_key.objectid, | |
3bb1a1bc | 791 | trans->transid, ins.objectid); |
31840ae1 ZY |
792 | BUG_ON(ret); |
793 | } | |
794 | btrfs_release_path(root, path); | |
795 | if (disk_bytenr != 0) { | |
a76a3cd4 | 796 | inode_add_bytes(inode, extent_end - end); |
39279cc3 | 797 | } |
31840ae1 ZY |
798 | } |
799 | ||
800 | if (found_extent && !keep) { | |
801 | u64 disk_bytenr = le64_to_cpu(old.disk_bytenr); | |
802 | ||
803 | if (disk_bytenr != 0) { | |
a76a3cd4 YZ |
804 | inode_sub_bytes(inode, |
805 | le64_to_cpu(old.num_bytes)); | |
31840ae1 ZY |
806 | ret = btrfs_free_extent(trans, root, |
807 | disk_bytenr, | |
808 | le64_to_cpu(old.disk_num_bytes), | |
809 | leaf_start, root_owner, | |
3bb1a1bc | 810 | root_gen, key.objectid, 0); |
31840ae1 ZY |
811 | BUG_ON(ret); |
812 | *hint_byte = disk_bytenr; | |
813 | } | |
814 | } | |
815 | ||
816 | if (search_start >= end) { | |
39279cc3 CM |
817 | ret = 0; |
818 | goto out; | |
819 | } | |
820 | } | |
821 | out: | |
822 | btrfs_free_path(path); | |
9069218d | 823 | btrfs_check_file(root, inode); |
39279cc3 CM |
824 | return ret; |
825 | } | |
826 | ||
827 | /* | |
d352ac68 CM |
828 | * this gets pages into the page cache and locks them down, it also properly |
829 | * waits for data=ordered extents to finish before allowing the pages to be | |
830 | * modified. | |
39279cc3 | 831 | */ |
a1b32a59 | 832 | static int noinline prepare_pages(struct btrfs_root *root, struct file *file, |
98ed5174 CM |
833 | struct page **pages, size_t num_pages, |
834 | loff_t pos, unsigned long first_index, | |
835 | unsigned long last_index, size_t write_bytes) | |
39279cc3 CM |
836 | { |
837 | int i; | |
838 | unsigned long index = pos >> PAGE_CACHE_SHIFT; | |
6da6abae | 839 | struct inode *inode = fdentry(file)->d_inode; |
39279cc3 | 840 | int err = 0; |
8c2383c3 | 841 | u64 start_pos; |
e6dcd2dc | 842 | u64 last_pos; |
8c2383c3 | 843 | |
5f39d397 | 844 | start_pos = pos & ~((u64)root->sectorsize - 1); |
e6dcd2dc | 845 | last_pos = ((u64)index + num_pages) << PAGE_CACHE_SHIFT; |
39279cc3 CM |
846 | |
847 | memset(pages, 0, num_pages * sizeof(struct page *)); | |
e6dcd2dc | 848 | again: |
39279cc3 CM |
849 | for (i = 0; i < num_pages; i++) { |
850 | pages[i] = grab_cache_page(inode->i_mapping, index + i); | |
851 | if (!pages[i]) { | |
852 | err = -ENOMEM; | |
a52d9a80 | 853 | BUG_ON(1); |
39279cc3 | 854 | } |
ccd467d6 | 855 | wait_on_page_writeback(pages[i]); |
39279cc3 | 856 | } |
0762704b | 857 | if (start_pos < inode->i_size) { |
e6dcd2dc | 858 | struct btrfs_ordered_extent *ordered; |
d99cb30a CM |
859 | lock_extent(&BTRFS_I(inode)->io_tree, |
860 | start_pos, last_pos - 1, GFP_NOFS); | |
e6dcd2dc CM |
861 | ordered = btrfs_lookup_first_ordered_extent(inode, last_pos -1); |
862 | if (ordered && | |
863 | ordered->file_offset + ordered->len > start_pos && | |
864 | ordered->file_offset < last_pos) { | |
865 | btrfs_put_ordered_extent(ordered); | |
866 | unlock_extent(&BTRFS_I(inode)->io_tree, | |
867 | start_pos, last_pos - 1, GFP_NOFS); | |
868 | for (i = 0; i < num_pages; i++) { | |
869 | unlock_page(pages[i]); | |
870 | page_cache_release(pages[i]); | |
871 | } | |
872 | btrfs_wait_ordered_range(inode, start_pos, | |
873 | last_pos - start_pos); | |
874 | goto again; | |
875 | } | |
876 | if (ordered) | |
877 | btrfs_put_ordered_extent(ordered); | |
878 | ||
0762704b CM |
879 | clear_extent_bits(&BTRFS_I(inode)->io_tree, start_pos, |
880 | last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC, | |
881 | GFP_NOFS); | |
d99cb30a CM |
882 | unlock_extent(&BTRFS_I(inode)->io_tree, |
883 | start_pos, last_pos - 1, GFP_NOFS); | |
0762704b | 884 | } |
e6dcd2dc | 885 | for (i = 0; i < num_pages; i++) { |
f87f057b | 886 | clear_page_dirty_for_io(pages[i]); |
e6dcd2dc CM |
887 | set_page_extent_mapped(pages[i]); |
888 | WARN_ON(!PageLocked(pages[i])); | |
889 | } | |
39279cc3 | 890 | return 0; |
39279cc3 CM |
891 | } |
892 | ||
893 | static ssize_t btrfs_file_write(struct file *file, const char __user *buf, | |
894 | size_t count, loff_t *ppos) | |
895 | { | |
896 | loff_t pos; | |
2ff3e9b6 CM |
897 | loff_t start_pos; |
898 | ssize_t num_written = 0; | |
899 | ssize_t err = 0; | |
39279cc3 | 900 | int ret = 0; |
6da6abae | 901 | struct inode *inode = fdentry(file)->d_inode; |
39279cc3 | 902 | struct btrfs_root *root = BTRFS_I(inode)->root; |
8c2383c3 CM |
903 | struct page **pages = NULL; |
904 | int nrptrs; | |
39279cc3 CM |
905 | struct page *pinned[2]; |
906 | unsigned long first_index; | |
907 | unsigned long last_index; | |
cb843a6f CM |
908 | int will_write; |
909 | ||
910 | will_write = ((file->f_flags & O_SYNC) || IS_SYNC(inode) || | |
911 | (file->f_flags & O_DIRECT)); | |
8c2383c3 CM |
912 | |
913 | nrptrs = min((count + PAGE_CACHE_SIZE - 1) / PAGE_CACHE_SIZE, | |
914 | PAGE_CACHE_SIZE / (sizeof(struct page *))); | |
39279cc3 CM |
915 | pinned[0] = NULL; |
916 | pinned[1] = NULL; | |
2ff3e9b6 | 917 | |
39279cc3 | 918 | pos = *ppos; |
2ff3e9b6 CM |
919 | start_pos = pos; |
920 | ||
39279cc3 CM |
921 | vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE); |
922 | current->backing_dev_info = inode->i_mapping->backing_dev_info; | |
923 | err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode)); | |
924 | if (err) | |
1832a6d5 | 925 | goto out_nolock; |
39279cc3 | 926 | if (count == 0) |
1832a6d5 | 927 | goto out_nolock; |
2b1f55b0 | 928 | |
0ee0fda0 | 929 | err = file_remove_suid(file); |
39279cc3 | 930 | if (err) |
1832a6d5 | 931 | goto out_nolock; |
39279cc3 CM |
932 | file_update_time(file); |
933 | ||
8c2383c3 | 934 | pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL); |
39279cc3 CM |
935 | |
936 | mutex_lock(&inode->i_mutex); | |
937 | first_index = pos >> PAGE_CACHE_SHIFT; | |
938 | last_index = (pos + count) >> PAGE_CACHE_SHIFT; | |
939 | ||
409c6118 CM |
940 | /* |
941 | * if this is a nodatasum mount, force summing off for the inode | |
942 | * all the time. That way a later mount with summing on won't | |
943 | * get confused | |
944 | */ | |
945 | if (btrfs_test_opt(root, NODATASUM)) | |
946 | btrfs_set_flag(inode, NODATASUM); | |
947 | ||
39279cc3 CM |
948 | /* |
949 | * there are lots of better ways to do this, but this code | |
950 | * makes sure the first and last page in the file range are | |
951 | * up to date and ready for cow | |
952 | */ | |
953 | if ((pos & (PAGE_CACHE_SIZE - 1))) { | |
954 | pinned[0] = grab_cache_page(inode->i_mapping, first_index); | |
955 | if (!PageUptodate(pinned[0])) { | |
9ebefb18 | 956 | ret = btrfs_readpage(NULL, pinned[0]); |
39279cc3 CM |
957 | BUG_ON(ret); |
958 | wait_on_page_locked(pinned[0]); | |
959 | } else { | |
960 | unlock_page(pinned[0]); | |
961 | } | |
962 | } | |
963 | if ((pos + count) & (PAGE_CACHE_SIZE - 1)) { | |
964 | pinned[1] = grab_cache_page(inode->i_mapping, last_index); | |
965 | if (!PageUptodate(pinned[1])) { | |
9ebefb18 | 966 | ret = btrfs_readpage(NULL, pinned[1]); |
39279cc3 CM |
967 | BUG_ON(ret); |
968 | wait_on_page_locked(pinned[1]); | |
969 | } else { | |
970 | unlock_page(pinned[1]); | |
971 | } | |
972 | } | |
973 | ||
39279cc3 CM |
974 | while(count > 0) { |
975 | size_t offset = pos & (PAGE_CACHE_SIZE - 1); | |
11bd143f CM |
976 | size_t write_bytes = min(count, nrptrs * |
977 | (size_t)PAGE_CACHE_SIZE - | |
8c2383c3 | 978 | offset); |
39279cc3 CM |
979 | size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >> |
980 | PAGE_CACHE_SHIFT; | |
981 | ||
8c2383c3 | 982 | WARN_ON(num_pages > nrptrs); |
39279cc3 | 983 | memset(pages, 0, sizeof(pages)); |
1832a6d5 | 984 | |
1832a6d5 | 985 | ret = btrfs_check_free_space(root, write_bytes, 0); |
1832a6d5 CM |
986 | if (ret) |
987 | goto out; | |
988 | ||
39279cc3 CM |
989 | ret = prepare_pages(root, file, pages, num_pages, |
990 | pos, first_index, last_index, | |
8c2383c3 | 991 | write_bytes); |
54aa1f4d CM |
992 | if (ret) |
993 | goto out; | |
39279cc3 | 994 | |
39279cc3 CM |
995 | ret = btrfs_copy_from_user(pos, num_pages, |
996 | write_bytes, pages, buf); | |
54aa1f4d CM |
997 | if (ret) { |
998 | btrfs_drop_pages(pages, num_pages); | |
999 | goto out; | |
1000 | } | |
39279cc3 CM |
1001 | |
1002 | ret = dirty_and_release_pages(NULL, root, file, pages, | |
1003 | num_pages, pos, write_bytes); | |
39279cc3 | 1004 | btrfs_drop_pages(pages, num_pages); |
54aa1f4d CM |
1005 | if (ret) |
1006 | goto out; | |
39279cc3 | 1007 | |
cb843a6f CM |
1008 | if (will_write) { |
1009 | btrfs_fdatawrite_range(inode->i_mapping, pos, | |
1010 | pos + write_bytes - 1, | |
1011 | WB_SYNC_NONE); | |
1012 | } else { | |
1013 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, | |
1014 | num_pages); | |
1015 | if (num_pages < | |
1016 | (root->leafsize >> PAGE_CACHE_SHIFT) + 1) | |
1017 | btrfs_btree_balance_dirty(root, 1); | |
1018 | btrfs_throttle(root); | |
1019 | } | |
1020 | ||
39279cc3 CM |
1021 | buf += write_bytes; |
1022 | count -= write_bytes; | |
1023 | pos += write_bytes; | |
1024 | num_written += write_bytes; | |
1025 | ||
39279cc3 CM |
1026 | cond_resched(); |
1027 | } | |
39279cc3 | 1028 | out: |
1832a6d5 | 1029 | mutex_unlock(&inode->i_mutex); |
5b92ee72 | 1030 | |
1832a6d5 | 1031 | out_nolock: |
8c2383c3 | 1032 | kfree(pages); |
39279cc3 CM |
1033 | if (pinned[0]) |
1034 | page_cache_release(pinned[0]); | |
1035 | if (pinned[1]) | |
1036 | page_cache_release(pinned[1]); | |
1037 | *ppos = pos; | |
2ff3e9b6 | 1038 | |
cb843a6f | 1039 | if (num_written > 0 && will_write) { |
e02119d5 CM |
1040 | struct btrfs_trans_handle *trans; |
1041 | ||
cb843a6f CM |
1042 | err = btrfs_wait_ordered_range(inode, start_pos, num_written); |
1043 | if (err) | |
2ff3e9b6 | 1044 | num_written = err; |
e02119d5 | 1045 | |
cb843a6f CM |
1046 | if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) { |
1047 | trans = btrfs_start_transaction(root, 1); | |
1048 | ret = btrfs_log_dentry_safe(trans, root, | |
1049 | file->f_dentry); | |
1050 | if (ret == 0) { | |
1051 | btrfs_sync_log(trans, root); | |
1052 | btrfs_end_transaction(trans, root); | |
1053 | } else { | |
1054 | btrfs_commit_transaction(trans, root); | |
1055 | } | |
1056 | } | |
1057 | if (file->f_flags & O_DIRECT) { | |
1058 | invalidate_mapping_pages(inode->i_mapping, | |
1059 | start_pos >> PAGE_CACHE_SHIFT, | |
1060 | (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT); | |
e02119d5 | 1061 | } |
2ff3e9b6 | 1062 | } |
39279cc3 | 1063 | current->backing_dev_info = NULL; |
39279cc3 CM |
1064 | return num_written ? num_written : err; |
1065 | } | |
1066 | ||
6bf13c0c | 1067 | int btrfs_release_file(struct inode * inode, struct file * filp) |
e1b81e67 | 1068 | { |
6bf13c0c SW |
1069 | if (filp->private_data) |
1070 | btrfs_ioctl_trans_end(filp); | |
e1b81e67 M |
1071 | return 0; |
1072 | } | |
1073 | ||
d352ac68 CM |
1074 | /* |
1075 | * fsync call for both files and directories. This logs the inode into | |
1076 | * the tree log instead of forcing full commits whenever possible. | |
1077 | * | |
1078 | * It needs to call filemap_fdatawait so that all ordered extent updates are | |
1079 | * in the metadata btree are up to date for copying to the log. | |
1080 | * | |
1081 | * It drops the inode mutex before doing the tree log commit. This is an | |
1082 | * important optimization for directories because holding the mutex prevents | |
1083 | * new operations on the dir while we write to disk. | |
1084 | */ | |
e02119d5 | 1085 | int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) |
39279cc3 CM |
1086 | { |
1087 | struct inode *inode = dentry->d_inode; | |
1088 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
15ee9bc7 | 1089 | int ret = 0; |
39279cc3 CM |
1090 | struct btrfs_trans_handle *trans; |
1091 | ||
1092 | /* | |
15ee9bc7 JB |
1093 | * check the transaction that last modified this inode |
1094 | * and see if its already been committed | |
39279cc3 | 1095 | */ |
15ee9bc7 JB |
1096 | if (!BTRFS_I(inode)->last_trans) |
1097 | goto out; | |
a2135011 | 1098 | |
15ee9bc7 JB |
1099 | mutex_lock(&root->fs_info->trans_mutex); |
1100 | if (BTRFS_I(inode)->last_trans <= | |
1101 | root->fs_info->last_trans_committed) { | |
1102 | BTRFS_I(inode)->last_trans = 0; | |
1103 | mutex_unlock(&root->fs_info->trans_mutex); | |
1104 | goto out; | |
1105 | } | |
1106 | mutex_unlock(&root->fs_info->trans_mutex); | |
1107 | ||
49eb7e46 | 1108 | root->fs_info->tree_log_batch++; |
e02119d5 | 1109 | filemap_fdatawait(inode->i_mapping); |
49eb7e46 | 1110 | root->fs_info->tree_log_batch++; |
e02119d5 | 1111 | |
15ee9bc7 | 1112 | /* |
a52d9a80 CM |
1113 | * ok we haven't committed the transaction yet, lets do a commit |
1114 | */ | |
6bf13c0c SW |
1115 | if (file->private_data) |
1116 | btrfs_ioctl_trans_end(file); | |
1117 | ||
39279cc3 CM |
1118 | trans = btrfs_start_transaction(root, 1); |
1119 | if (!trans) { | |
1120 | ret = -ENOMEM; | |
1121 | goto out; | |
1122 | } | |
e02119d5 CM |
1123 | |
1124 | ret = btrfs_log_dentry_safe(trans, root, file->f_dentry); | |
49eb7e46 | 1125 | if (ret < 0) { |
e02119d5 | 1126 | goto out; |
49eb7e46 CM |
1127 | } |
1128 | ||
1129 | /* we've logged all the items and now have a consistent | |
1130 | * version of the file in the log. It is possible that | |
1131 | * someone will come in and modify the file, but that's | |
1132 | * fine because the log is consistent on disk, and we | |
1133 | * have references to all of the file's extents | |
1134 | * | |
1135 | * It is possible that someone will come in and log the | |
1136 | * file again, but that will end up using the synchronization | |
1137 | * inside btrfs_sync_log to keep things safe. | |
1138 | */ | |
1139 | mutex_unlock(&file->f_dentry->d_inode->i_mutex); | |
1140 | ||
e02119d5 CM |
1141 | if (ret > 0) { |
1142 | ret = btrfs_commit_transaction(trans, root); | |
1143 | } else { | |
1144 | btrfs_sync_log(trans, root); | |
1145 | ret = btrfs_end_transaction(trans, root); | |
1146 | } | |
49eb7e46 | 1147 | mutex_lock(&file->f_dentry->d_inode->i_mutex); |
39279cc3 CM |
1148 | out: |
1149 | return ret > 0 ? EIO : ret; | |
1150 | } | |
1151 | ||
9ebefb18 | 1152 | static struct vm_operations_struct btrfs_file_vm_ops = { |
92fee66d | 1153 | .fault = filemap_fault, |
9ebefb18 CM |
1154 | .page_mkwrite = btrfs_page_mkwrite, |
1155 | }; | |
1156 | ||
1157 | static int btrfs_file_mmap(struct file *filp, struct vm_area_struct *vma) | |
1158 | { | |
1159 | vma->vm_ops = &btrfs_file_vm_ops; | |
1160 | file_accessed(filp); | |
1161 | return 0; | |
1162 | } | |
1163 | ||
39279cc3 CM |
1164 | struct file_operations btrfs_file_operations = { |
1165 | .llseek = generic_file_llseek, | |
1166 | .read = do_sync_read, | |
9ebefb18 | 1167 | .aio_read = generic_file_aio_read, |
e9906a98 | 1168 | .splice_read = generic_file_splice_read, |
39279cc3 | 1169 | .write = btrfs_file_write, |
9ebefb18 | 1170 | .mmap = btrfs_file_mmap, |
39279cc3 | 1171 | .open = generic_file_open, |
e1b81e67 | 1172 | .release = btrfs_release_file, |
39279cc3 | 1173 | .fsync = btrfs_sync_file, |
34287aa3 | 1174 | .unlocked_ioctl = btrfs_ioctl, |
39279cc3 | 1175 | #ifdef CONFIG_COMPAT |
34287aa3 | 1176 | .compat_ioctl = btrfs_ioctl, |
39279cc3 CM |
1177 | #endif |
1178 | }; |