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 | /* |
99 | * after copy_from_user, pages need to be dirtied and we need to make | |
100 | * sure holes are created between the current EOF and the start of | |
101 | * any next extents (if required). | |
102 | * | |
103 | * this also makes the decision about creating an inline extent vs | |
104 | * doing real data extents, marking pages dirty and delalloc as required. | |
105 | */ | |
98ed5174 | 106 | static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, |
39279cc3 CM |
107 | struct btrfs_root *root, |
108 | struct file *file, | |
109 | struct page **pages, | |
110 | size_t num_pages, | |
111 | loff_t pos, | |
112 | size_t write_bytes) | |
113 | { | |
39279cc3 | 114 | int err = 0; |
a52d9a80 | 115 | int i; |
6da6abae | 116 | struct inode *inode = fdentry(file)->d_inode; |
d1310b2e | 117 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
db94535d CM |
118 | u64 hint_byte; |
119 | u64 num_bytes; | |
a52d9a80 CM |
120 | u64 start_pos; |
121 | u64 end_of_last_block; | |
122 | u64 end_pos = pos + write_bytes; | |
123 | loff_t isize = i_size_read(inode); | |
39279cc3 | 124 | |
5f39d397 | 125 | start_pos = pos & ~((u64)root->sectorsize - 1); |
db94535d CM |
126 | num_bytes = (write_bytes + pos - start_pos + |
127 | root->sectorsize - 1) & ~((u64)root->sectorsize - 1); | |
39279cc3 | 128 | |
db94535d CM |
129 | end_of_last_block = start_pos + num_bytes - 1; |
130 | ||
d1310b2e | 131 | lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); |
37d1aeee | 132 | trans = btrfs_join_transaction(root, 1); |
a52d9a80 CM |
133 | if (!trans) { |
134 | err = -ENOMEM; | |
135 | goto out_unlock; | |
136 | } | |
137 | btrfs_set_trans_block_group(trans, inode); | |
db94535d | 138 | hint_byte = 0; |
a52d9a80 CM |
139 | |
140 | if ((end_of_last_block & 4095) == 0) { | |
9433063b | 141 | printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block); |
a52d9a80 | 142 | } |
d1310b2e | 143 | set_extent_uptodate(io_tree, start_pos, end_of_last_block, GFP_NOFS); |
a52d9a80 | 144 | |
c8b97818 CM |
145 | /* check for reserved extents on each page, we don't want |
146 | * to reset the delalloc bit on things that already have | |
147 | * extents reserved. | |
a52d9a80 | 148 | */ |
c8b97818 CM |
149 | btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block); |
150 | for (i = 0; i < num_pages; i++) { | |
151 | struct page *p = pages[i]; | |
152 | SetPageUptodate(p); | |
153 | ClearPageChecked(p); | |
154 | set_page_dirty(p); | |
a52d9a80 CM |
155 | } |
156 | if (end_pos > isize) { | |
157 | i_size_write(inode, end_pos); | |
158 | btrfs_update_inode(trans, root, inode); | |
39279cc3 | 159 | } |
017e5369 | 160 | err = btrfs_end_transaction(trans, root); |
a52d9a80 | 161 | out_unlock: |
d1310b2e | 162 | unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); |
39279cc3 CM |
163 | return err; |
164 | } | |
165 | ||
d352ac68 CM |
166 | /* |
167 | * this drops all the extents in the cache that intersect the range | |
168 | * [start, end]. Existing extents are split as required. | |
169 | */ | |
5b21f2ed ZY |
170 | int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, |
171 | int skip_pinned) | |
a52d9a80 CM |
172 | { |
173 | struct extent_map *em; | |
3b951516 CM |
174 | struct extent_map *split = NULL; |
175 | struct extent_map *split2 = NULL; | |
a52d9a80 | 176 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
39b5637f | 177 | u64 len = end - start + 1; |
3b951516 CM |
178 | int ret; |
179 | int testend = 1; | |
5b21f2ed | 180 | unsigned long flags; |
c8b97818 | 181 | int compressed = 0; |
a52d9a80 | 182 | |
e6dcd2dc | 183 | WARN_ON(end < start); |
3b951516 | 184 | if (end == (u64)-1) { |
39b5637f | 185 | len = (u64)-1; |
3b951516 CM |
186 | testend = 0; |
187 | } | |
a52d9a80 | 188 | while(1) { |
3b951516 CM |
189 | if (!split) |
190 | split = alloc_extent_map(GFP_NOFS); | |
191 | if (!split2) | |
192 | split2 = alloc_extent_map(GFP_NOFS); | |
193 | ||
d1310b2e | 194 | spin_lock(&em_tree->lock); |
39b5637f | 195 | em = lookup_extent_mapping(em_tree, start, len); |
d1310b2e CM |
196 | if (!em) { |
197 | spin_unlock(&em_tree->lock); | |
a52d9a80 | 198 | break; |
d1310b2e | 199 | } |
5b21f2ed ZY |
200 | flags = em->flags; |
201 | if (skip_pinned && test_bit(EXTENT_FLAG_PINNED, &em->flags)) { | |
202 | spin_unlock(&em_tree->lock); | |
203 | if (em->start <= start && | |
204 | (!testend || em->start + em->len >= start + len)) { | |
205 | free_extent_map(em); | |
206 | break; | |
207 | } | |
208 | if (start < em->start) { | |
209 | len = em->start - start; | |
210 | } else { | |
211 | len = start + len - (em->start + em->len); | |
212 | start = em->start + em->len; | |
213 | } | |
214 | free_extent_map(em); | |
215 | continue; | |
216 | } | |
c8b97818 | 217 | compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags); |
3ce7e67a | 218 | clear_bit(EXTENT_FLAG_PINNED, &em->flags); |
a52d9a80 | 219 | remove_extent_mapping(em_tree, em); |
3b951516 CM |
220 | |
221 | if (em->block_start < EXTENT_MAP_LAST_BYTE && | |
222 | em->start < start) { | |
223 | split->start = em->start; | |
224 | split->len = start - em->start; | |
225 | split->block_start = em->block_start; | |
c8b97818 CM |
226 | |
227 | if (compressed) | |
228 | split->block_len = em->block_len; | |
229 | else | |
230 | split->block_len = split->len; | |
231 | ||
3b951516 | 232 | split->bdev = em->bdev; |
5b21f2ed | 233 | split->flags = flags; |
3b951516 CM |
234 | ret = add_extent_mapping(em_tree, split); |
235 | BUG_ON(ret); | |
236 | free_extent_map(split); | |
237 | split = split2; | |
238 | split2 = NULL; | |
239 | } | |
240 | if (em->block_start < EXTENT_MAP_LAST_BYTE && | |
241 | testend && em->start + em->len > start + len) { | |
242 | u64 diff = start + len - em->start; | |
243 | ||
244 | split->start = start + len; | |
245 | split->len = em->start + em->len - (start + len); | |
246 | split->bdev = em->bdev; | |
5b21f2ed | 247 | split->flags = flags; |
3b951516 | 248 | |
c8b97818 CM |
249 | if (compressed) { |
250 | split->block_len = em->block_len; | |
251 | split->block_start = em->block_start; | |
252 | } else { | |
253 | split->block_len = split->len; | |
254 | split->block_start = em->block_start + diff; | |
255 | } | |
3b951516 CM |
256 | |
257 | ret = add_extent_mapping(em_tree, split); | |
258 | BUG_ON(ret); | |
259 | free_extent_map(split); | |
260 | split = NULL; | |
261 | } | |
d1310b2e CM |
262 | spin_unlock(&em_tree->lock); |
263 | ||
a52d9a80 CM |
264 | /* once for us */ |
265 | free_extent_map(em); | |
266 | /* once for the tree*/ | |
267 | free_extent_map(em); | |
268 | } | |
3b951516 CM |
269 | if (split) |
270 | free_extent_map(split); | |
271 | if (split2) | |
272 | free_extent_map(split2); | |
a52d9a80 CM |
273 | return 0; |
274 | } | |
275 | ||
5f56406a CM |
276 | int btrfs_check_file(struct btrfs_root *root, struct inode *inode) |
277 | { | |
278 | return 0; | |
279 | #if 0 | |
280 | struct btrfs_path *path; | |
281 | struct btrfs_key found_key; | |
282 | struct extent_buffer *leaf; | |
283 | struct btrfs_file_extent_item *extent; | |
284 | u64 last_offset = 0; | |
285 | int nritems; | |
286 | int slot; | |
287 | int found_type; | |
288 | int ret; | |
289 | int err = 0; | |
290 | u64 extent_end = 0; | |
291 | ||
292 | path = btrfs_alloc_path(); | |
293 | ret = btrfs_lookup_file_extent(NULL, root, path, inode->i_ino, | |
294 | last_offset, 0); | |
295 | while(1) { | |
296 | nritems = btrfs_header_nritems(path->nodes[0]); | |
297 | if (path->slots[0] >= nritems) { | |
298 | ret = btrfs_next_leaf(root, path); | |
299 | if (ret) | |
300 | goto out; | |
301 | nritems = btrfs_header_nritems(path->nodes[0]); | |
302 | } | |
303 | slot = path->slots[0]; | |
304 | leaf = path->nodes[0]; | |
305 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
306 | if (found_key.objectid != inode->i_ino) | |
307 | break; | |
308 | if (found_key.type != BTRFS_EXTENT_DATA_KEY) | |
309 | goto out; | |
310 | ||
9069218d | 311 | if (found_key.offset < last_offset) { |
5f56406a CM |
312 | WARN_ON(1); |
313 | btrfs_print_leaf(root, leaf); | |
314 | printk("inode %lu found offset %Lu expected %Lu\n", | |
315 | inode->i_ino, found_key.offset, last_offset); | |
316 | err = 1; | |
317 | goto out; | |
318 | } | |
319 | extent = btrfs_item_ptr(leaf, slot, | |
320 | struct btrfs_file_extent_item); | |
321 | found_type = btrfs_file_extent_type(leaf, extent); | |
322 | if (found_type == BTRFS_FILE_EXTENT_REG) { | |
323 | extent_end = found_key.offset + | |
324 | btrfs_file_extent_num_bytes(leaf, extent); | |
325 | } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { | |
326 | struct btrfs_item *item; | |
327 | item = btrfs_item_nr(leaf, slot); | |
328 | extent_end = found_key.offset + | |
c8b97818 | 329 | btrfs_file_extent_inline_len(leaf, extent); |
5f56406a CM |
330 | extent_end = (extent_end + root->sectorsize - 1) & |
331 | ~((u64)root->sectorsize -1 ); | |
332 | } | |
333 | last_offset = extent_end; | |
334 | path->slots[0]++; | |
335 | } | |
9069218d | 336 | if (0 && last_offset < inode->i_size) { |
5f56406a CM |
337 | WARN_ON(1); |
338 | btrfs_print_leaf(root, leaf); | |
339 | printk("inode %lu found offset %Lu size %Lu\n", inode->i_ino, | |
340 | last_offset, inode->i_size); | |
341 | err = 1; | |
342 | ||
343 | } | |
344 | out: | |
345 | btrfs_free_path(path); | |
346 | return err; | |
347 | #endif | |
348 | } | |
349 | ||
39279cc3 CM |
350 | /* |
351 | * this is very complex, but the basic idea is to drop all extents | |
352 | * in the range start - end. hint_block is filled in with a block number | |
353 | * that would be a good hint to the block allocator for this file. | |
354 | * | |
355 | * If an extent intersects the range but is not entirely inside the range | |
356 | * it is either truncated or split. Anything entirely inside the range | |
357 | * is deleted from the tree. | |
d352ac68 CM |
358 | * |
359 | * inline_limit is used to tell this code which offsets in the file to keep | |
360 | * if they contain inline extents. | |
39279cc3 | 361 | */ |
a1b32a59 | 362 | int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, |
39279cc3 | 363 | struct btrfs_root *root, struct inode *inode, |
00f5c795 | 364 | u64 start, u64 end, u64 inline_limit, u64 *hint_byte) |
39279cc3 | 365 | { |
00f5c795 CM |
366 | u64 extent_end = 0; |
367 | u64 search_start = start; | |
31840ae1 | 368 | u64 leaf_start; |
c8b97818 CM |
369 | u64 ram_bytes = 0; |
370 | u8 compression = 0; | |
371 | u8 encryption = 0; | |
372 | u16 other_encoding = 0; | |
31840ae1 ZY |
373 | u64 root_gen; |
374 | u64 root_owner; | |
5f39d397 | 375 | struct extent_buffer *leaf; |
39279cc3 | 376 | struct btrfs_file_extent_item *extent; |
39279cc3 | 377 | struct btrfs_path *path; |
00f5c795 CM |
378 | struct btrfs_key key; |
379 | struct btrfs_file_extent_item old; | |
380 | int keep; | |
381 | int slot; | |
39279cc3 CM |
382 | int bookend; |
383 | int found_type; | |
384 | int found_extent; | |
385 | int found_inline; | |
ccd467d6 | 386 | int recow; |
00f5c795 | 387 | int ret; |
39279cc3 | 388 | |
c8b97818 | 389 | inline_limit = 0; |
5b21f2ed | 390 | btrfs_drop_extent_cache(inode, start, end - 1, 0); |
a52d9a80 | 391 | |
39279cc3 CM |
392 | path = btrfs_alloc_path(); |
393 | if (!path) | |
394 | return -ENOMEM; | |
395 | while(1) { | |
ccd467d6 | 396 | recow = 0; |
39279cc3 CM |
397 | btrfs_release_path(root, path); |
398 | ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino, | |
399 | search_start, -1); | |
400 | if (ret < 0) | |
401 | goto out; | |
402 | if (ret > 0) { | |
403 | if (path->slots[0] == 0) { | |
404 | ret = 0; | |
405 | goto out; | |
406 | } | |
407 | path->slots[0]--; | |
408 | } | |
8c2383c3 | 409 | next_slot: |
39279cc3 CM |
410 | keep = 0; |
411 | bookend = 0; | |
412 | found_extent = 0; | |
413 | found_inline = 0; | |
31840ae1 ZY |
414 | leaf_start = 0; |
415 | root_gen = 0; | |
416 | root_owner = 0; | |
39279cc3 | 417 | extent = NULL; |
5f39d397 | 418 | leaf = path->nodes[0]; |
39279cc3 | 419 | slot = path->slots[0]; |
8c2383c3 | 420 | ret = 0; |
5f39d397 | 421 | btrfs_item_key_to_cpu(leaf, &key, slot); |
7261009c Y |
422 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY && |
423 | key.offset >= end) { | |
39279cc3 CM |
424 | goto out; |
425 | } | |
7261009c Y |
426 | if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || |
427 | key.objectid != inode->i_ino) { | |
39279cc3 CM |
428 | goto out; |
429 | } | |
ccd467d6 CM |
430 | if (recow) { |
431 | search_start = key.offset; | |
432 | continue; | |
433 | } | |
8c2383c3 CM |
434 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { |
435 | extent = btrfs_item_ptr(leaf, slot, | |
436 | struct btrfs_file_extent_item); | |
5f39d397 | 437 | found_type = btrfs_file_extent_type(leaf, extent); |
c8b97818 CM |
438 | compression = btrfs_file_extent_compression(leaf, |
439 | extent); | |
440 | encryption = btrfs_file_extent_encryption(leaf, | |
441 | extent); | |
442 | other_encoding = btrfs_file_extent_other_encoding(leaf, | |
443 | extent); | |
8c2383c3 | 444 | if (found_type == BTRFS_FILE_EXTENT_REG) { |
257d0ce3 CM |
445 | extent_end = |
446 | btrfs_file_extent_disk_bytenr(leaf, | |
447 | extent); | |
448 | if (extent_end) | |
449 | *hint_byte = extent_end; | |
450 | ||
8c2383c3 | 451 | extent_end = key.offset + |
db94535d | 452 | btrfs_file_extent_num_bytes(leaf, extent); |
c8b97818 CM |
453 | ram_bytes = btrfs_file_extent_ram_bytes(leaf, |
454 | extent); | |
8c2383c3 CM |
455 | found_extent = 1; |
456 | } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { | |
457 | found_inline = 1; | |
458 | extent_end = key.offset + | |
c8b97818 | 459 | btrfs_file_extent_inline_len(leaf, extent); |
8c2383c3 CM |
460 | } |
461 | } else { | |
462 | extent_end = search_start; | |
39279cc3 CM |
463 | } |
464 | ||
465 | /* we found nothing we can drop */ | |
8c2383c3 CM |
466 | if ((!found_extent && !found_inline) || |
467 | search_start >= extent_end) { | |
468 | int nextret; | |
469 | u32 nritems; | |
5f39d397 | 470 | nritems = btrfs_header_nritems(leaf); |
8c2383c3 CM |
471 | if (slot >= nritems - 1) { |
472 | nextret = btrfs_next_leaf(root, path); | |
473 | if (nextret) | |
474 | goto out; | |
ccd467d6 | 475 | recow = 1; |
8c2383c3 CM |
476 | } else { |
477 | path->slots[0]++; | |
478 | } | |
479 | goto next_slot; | |
39279cc3 CM |
480 | } |
481 | ||
39279cc3 | 482 | if (found_inline) { |
5f39d397 | 483 | u64 mask = root->sectorsize - 1; |
39279cc3 CM |
484 | search_start = (extent_end + mask) & ~mask; |
485 | } else | |
486 | search_start = extent_end; | |
c8b97818 CM |
487 | |
488 | if (end <= extent_end && start >= key.offset && found_inline) | |
179e29e4 | 489 | *hint_byte = EXTENT_MAP_INLINE; |
31840ae1 ZY |
490 | |
491 | if (found_extent) { | |
492 | read_extent_buffer(leaf, &old, (unsigned long)extent, | |
493 | sizeof(old)); | |
494 | root_gen = btrfs_header_generation(leaf); | |
495 | root_owner = btrfs_header_owner(leaf); | |
496 | leaf_start = leaf->start; | |
179e29e4 | 497 | } |
31840ae1 | 498 | |
39279cc3 | 499 | if (end < extent_end && end >= key.offset) { |
179e29e4 | 500 | bookend = 1; |
0181e58f | 501 | if (found_inline && start <= key.offset) |
179e29e4 | 502 | keep = 1; |
39279cc3 | 503 | } |
39279cc3 CM |
504 | /* truncate existing extent */ |
505 | if (start > key.offset) { | |
506 | u64 new_num; | |
507 | u64 old_num; | |
508 | keep = 1; | |
5f39d397 | 509 | WARN_ON(start & (root->sectorsize - 1)); |
39279cc3 | 510 | if (found_extent) { |
db94535d CM |
511 | new_num = start - key.offset; |
512 | old_num = btrfs_file_extent_num_bytes(leaf, | |
513 | extent); | |
514 | *hint_byte = | |
515 | btrfs_file_extent_disk_bytenr(leaf, | |
516 | extent); | |
517 | if (btrfs_file_extent_disk_bytenr(leaf, | |
518 | extent)) { | |
a76a3cd4 YZ |
519 | inode_sub_bytes(inode, old_num - |
520 | new_num); | |
39279cc3 | 521 | } |
db94535d CM |
522 | btrfs_set_file_extent_num_bytes(leaf, extent, |
523 | new_num); | |
5f39d397 | 524 | btrfs_mark_buffer_dirty(leaf); |
00f5c795 CM |
525 | } else if (key.offset < inline_limit && |
526 | (end > extent_end) && | |
527 | (inline_limit < extent_end)) { | |
3326d1b0 CM |
528 | u32 new_size; |
529 | new_size = btrfs_file_extent_calc_inline_size( | |
00f5c795 | 530 | inline_limit - key.offset); |
a76a3cd4 YZ |
531 | inode_sub_bytes(inode, extent_end - |
532 | inline_limit); | |
3326d1b0 | 533 | btrfs_truncate_item(trans, root, path, |
179e29e4 | 534 | new_size, 1); |
39279cc3 CM |
535 | } |
536 | } | |
537 | /* delete the entire extent */ | |
538 | if (!keep) { | |
a76a3cd4 YZ |
539 | if (found_inline) |
540 | inode_sub_bytes(inode, extent_end - | |
541 | key.offset); | |
39279cc3 | 542 | ret = btrfs_del_item(trans, root, path); |
54aa1f4d | 543 | /* TODO update progress marker and return */ |
39279cc3 | 544 | BUG_ON(ret); |
39279cc3 | 545 | extent = NULL; |
31840ae1 ZY |
546 | btrfs_release_path(root, path); |
547 | /* the extent will be freed later */ | |
39279cc3 | 548 | } |
0181e58f | 549 | if (bookend && found_inline && start <= key.offset) { |
179e29e4 CM |
550 | u32 new_size; |
551 | new_size = btrfs_file_extent_calc_inline_size( | |
0181e58f | 552 | extent_end - end); |
a76a3cd4 | 553 | inode_sub_bytes(inode, end - key.offset); |
31840ae1 ZY |
554 | ret = btrfs_truncate_item(trans, root, path, |
555 | new_size, 0); | |
556 | BUG_ON(ret); | |
179e29e4 | 557 | } |
39279cc3 CM |
558 | /* create bookend, splitting the extent in two */ |
559 | if (bookend && found_extent) { | |
31840ae1 | 560 | u64 disk_bytenr; |
39279cc3 CM |
561 | struct btrfs_key ins; |
562 | ins.objectid = inode->i_ino; | |
563 | ins.offset = end; | |
39279cc3 | 564 | btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY); |
39279cc3 CM |
565 | btrfs_release_path(root, path); |
566 | ret = btrfs_insert_empty_item(trans, root, path, &ins, | |
567 | sizeof(*extent)); | |
31840ae1 | 568 | BUG_ON(ret); |
8c2383c3 | 569 | |
5f39d397 | 570 | leaf = path->nodes[0]; |
5f39d397 CM |
571 | extent = btrfs_item_ptr(leaf, path->slots[0], |
572 | struct btrfs_file_extent_item); | |
573 | write_extent_buffer(leaf, &old, | |
574 | (unsigned long)extent, sizeof(old)); | |
575 | ||
c8b97818 CM |
576 | btrfs_set_file_extent_compression(leaf, extent, |
577 | compression); | |
578 | btrfs_set_file_extent_encryption(leaf, extent, | |
579 | encryption); | |
580 | btrfs_set_file_extent_other_encoding(leaf, extent, | |
581 | other_encoding); | |
5f39d397 | 582 | btrfs_set_file_extent_offset(leaf, extent, |
db94535d CM |
583 | le64_to_cpu(old.offset) + end - key.offset); |
584 | WARN_ON(le64_to_cpu(old.num_bytes) < | |
585 | (extent_end - end)); | |
586 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
587 | extent_end - end); | |
c8b97818 CM |
588 | |
589 | /* | |
590 | * set the ram bytes to the size of the full extent | |
591 | * before splitting. This is a worst case flag, | |
592 | * but its the best we can do because we don't know | |
593 | * how splitting affects compression | |
594 | */ | |
595 | btrfs_set_file_extent_ram_bytes(leaf, extent, | |
596 | ram_bytes); | |
5f39d397 | 597 | btrfs_set_file_extent_type(leaf, extent, |
39279cc3 | 598 | BTRFS_FILE_EXTENT_REG); |
db94535d | 599 | |
39279cc3 | 600 | btrfs_mark_buffer_dirty(path->nodes[0]); |
31840ae1 ZY |
601 | |
602 | disk_bytenr = le64_to_cpu(old.disk_bytenr); | |
603 | if (disk_bytenr != 0) { | |
604 | ret = btrfs_inc_extent_ref(trans, root, | |
605 | disk_bytenr, | |
606 | le64_to_cpu(old.disk_num_bytes), | |
607 | leaf->start, | |
608 | root->root_key.objectid, | |
3bb1a1bc | 609 | trans->transid, ins.objectid); |
31840ae1 ZY |
610 | BUG_ON(ret); |
611 | } | |
612 | btrfs_release_path(root, path); | |
613 | if (disk_bytenr != 0) { | |
a76a3cd4 | 614 | inode_add_bytes(inode, extent_end - end); |
39279cc3 | 615 | } |
31840ae1 ZY |
616 | } |
617 | ||
618 | if (found_extent && !keep) { | |
619 | u64 disk_bytenr = le64_to_cpu(old.disk_bytenr); | |
620 | ||
621 | if (disk_bytenr != 0) { | |
a76a3cd4 YZ |
622 | inode_sub_bytes(inode, |
623 | le64_to_cpu(old.num_bytes)); | |
31840ae1 ZY |
624 | ret = btrfs_free_extent(trans, root, |
625 | disk_bytenr, | |
626 | le64_to_cpu(old.disk_num_bytes), | |
627 | leaf_start, root_owner, | |
3bb1a1bc | 628 | root_gen, key.objectid, 0); |
31840ae1 ZY |
629 | BUG_ON(ret); |
630 | *hint_byte = disk_bytenr; | |
631 | } | |
632 | } | |
633 | ||
634 | if (search_start >= end) { | |
39279cc3 CM |
635 | ret = 0; |
636 | goto out; | |
637 | } | |
638 | } | |
639 | out: | |
640 | btrfs_free_path(path); | |
9069218d | 641 | btrfs_check_file(root, inode); |
39279cc3 CM |
642 | return ret; |
643 | } | |
644 | ||
645 | /* | |
d352ac68 CM |
646 | * this gets pages into the page cache and locks them down, it also properly |
647 | * waits for data=ordered extents to finish before allowing the pages to be | |
648 | * modified. | |
39279cc3 | 649 | */ |
a1b32a59 | 650 | static int noinline prepare_pages(struct btrfs_root *root, struct file *file, |
98ed5174 CM |
651 | struct page **pages, size_t num_pages, |
652 | loff_t pos, unsigned long first_index, | |
653 | unsigned long last_index, size_t write_bytes) | |
39279cc3 CM |
654 | { |
655 | int i; | |
656 | unsigned long index = pos >> PAGE_CACHE_SHIFT; | |
6da6abae | 657 | struct inode *inode = fdentry(file)->d_inode; |
39279cc3 | 658 | int err = 0; |
8c2383c3 | 659 | u64 start_pos; |
e6dcd2dc | 660 | u64 last_pos; |
8c2383c3 | 661 | |
5f39d397 | 662 | start_pos = pos & ~((u64)root->sectorsize - 1); |
e6dcd2dc | 663 | last_pos = ((u64)index + num_pages) << PAGE_CACHE_SHIFT; |
39279cc3 | 664 | |
9036c102 YZ |
665 | if (start_pos > inode->i_size) { |
666 | err = btrfs_cont_expand(inode, start_pos); | |
667 | if (err) | |
668 | return err; | |
669 | } | |
670 | ||
39279cc3 | 671 | memset(pages, 0, num_pages * sizeof(struct page *)); |
e6dcd2dc | 672 | again: |
39279cc3 CM |
673 | for (i = 0; i < num_pages; i++) { |
674 | pages[i] = grab_cache_page(inode->i_mapping, index + i); | |
675 | if (!pages[i]) { | |
676 | err = -ENOMEM; | |
a52d9a80 | 677 | BUG_ON(1); |
39279cc3 | 678 | } |
ccd467d6 | 679 | wait_on_page_writeback(pages[i]); |
39279cc3 | 680 | } |
0762704b | 681 | if (start_pos < inode->i_size) { |
e6dcd2dc | 682 | struct btrfs_ordered_extent *ordered; |
d99cb30a CM |
683 | lock_extent(&BTRFS_I(inode)->io_tree, |
684 | start_pos, last_pos - 1, GFP_NOFS); | |
e6dcd2dc CM |
685 | ordered = btrfs_lookup_first_ordered_extent(inode, last_pos -1); |
686 | if (ordered && | |
687 | ordered->file_offset + ordered->len > start_pos && | |
688 | ordered->file_offset < last_pos) { | |
689 | btrfs_put_ordered_extent(ordered); | |
690 | unlock_extent(&BTRFS_I(inode)->io_tree, | |
691 | start_pos, last_pos - 1, GFP_NOFS); | |
692 | for (i = 0; i < num_pages; i++) { | |
693 | unlock_page(pages[i]); | |
694 | page_cache_release(pages[i]); | |
695 | } | |
696 | btrfs_wait_ordered_range(inode, start_pos, | |
697 | last_pos - start_pos); | |
698 | goto again; | |
699 | } | |
700 | if (ordered) | |
701 | btrfs_put_ordered_extent(ordered); | |
702 | ||
0762704b CM |
703 | clear_extent_bits(&BTRFS_I(inode)->io_tree, start_pos, |
704 | last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC, | |
705 | GFP_NOFS); | |
d99cb30a CM |
706 | unlock_extent(&BTRFS_I(inode)->io_tree, |
707 | start_pos, last_pos - 1, GFP_NOFS); | |
0762704b | 708 | } |
e6dcd2dc | 709 | for (i = 0; i < num_pages; i++) { |
f87f057b | 710 | clear_page_dirty_for_io(pages[i]); |
e6dcd2dc CM |
711 | set_page_extent_mapped(pages[i]); |
712 | WARN_ON(!PageLocked(pages[i])); | |
713 | } | |
39279cc3 | 714 | return 0; |
39279cc3 CM |
715 | } |
716 | ||
717 | static ssize_t btrfs_file_write(struct file *file, const char __user *buf, | |
718 | size_t count, loff_t *ppos) | |
719 | { | |
720 | loff_t pos; | |
2ff3e9b6 CM |
721 | loff_t start_pos; |
722 | ssize_t num_written = 0; | |
723 | ssize_t err = 0; | |
39279cc3 | 724 | int ret = 0; |
6da6abae | 725 | struct inode *inode = fdentry(file)->d_inode; |
39279cc3 | 726 | struct btrfs_root *root = BTRFS_I(inode)->root; |
8c2383c3 CM |
727 | struct page **pages = NULL; |
728 | int nrptrs; | |
39279cc3 CM |
729 | struct page *pinned[2]; |
730 | unsigned long first_index; | |
731 | unsigned long last_index; | |
cb843a6f CM |
732 | int will_write; |
733 | ||
734 | will_write = ((file->f_flags & O_SYNC) || IS_SYNC(inode) || | |
735 | (file->f_flags & O_DIRECT)); | |
8c2383c3 CM |
736 | |
737 | nrptrs = min((count + PAGE_CACHE_SIZE - 1) / PAGE_CACHE_SIZE, | |
738 | PAGE_CACHE_SIZE / (sizeof(struct page *))); | |
39279cc3 CM |
739 | pinned[0] = NULL; |
740 | pinned[1] = NULL; | |
2ff3e9b6 | 741 | |
39279cc3 | 742 | pos = *ppos; |
2ff3e9b6 CM |
743 | start_pos = pos; |
744 | ||
39279cc3 CM |
745 | vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE); |
746 | current->backing_dev_info = inode->i_mapping->backing_dev_info; | |
747 | err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode)); | |
748 | if (err) | |
1832a6d5 | 749 | goto out_nolock; |
39279cc3 | 750 | if (count == 0) |
1832a6d5 | 751 | goto out_nolock; |
2b1f55b0 | 752 | |
0ee0fda0 | 753 | err = file_remove_suid(file); |
39279cc3 | 754 | if (err) |
1832a6d5 | 755 | goto out_nolock; |
39279cc3 CM |
756 | file_update_time(file); |
757 | ||
8c2383c3 | 758 | pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL); |
39279cc3 CM |
759 | |
760 | mutex_lock(&inode->i_mutex); | |
761 | first_index = pos >> PAGE_CACHE_SHIFT; | |
762 | last_index = (pos + count) >> PAGE_CACHE_SHIFT; | |
763 | ||
409c6118 CM |
764 | /* |
765 | * if this is a nodatasum mount, force summing off for the inode | |
766 | * all the time. That way a later mount with summing on won't | |
767 | * get confused | |
768 | */ | |
769 | if (btrfs_test_opt(root, NODATASUM)) | |
770 | btrfs_set_flag(inode, NODATASUM); | |
771 | ||
39279cc3 CM |
772 | /* |
773 | * there are lots of better ways to do this, but this code | |
774 | * makes sure the first and last page in the file range are | |
775 | * up to date and ready for cow | |
776 | */ | |
777 | if ((pos & (PAGE_CACHE_SIZE - 1))) { | |
778 | pinned[0] = grab_cache_page(inode->i_mapping, first_index); | |
779 | if (!PageUptodate(pinned[0])) { | |
9ebefb18 | 780 | ret = btrfs_readpage(NULL, pinned[0]); |
39279cc3 CM |
781 | BUG_ON(ret); |
782 | wait_on_page_locked(pinned[0]); | |
783 | } else { | |
784 | unlock_page(pinned[0]); | |
785 | } | |
786 | } | |
787 | if ((pos + count) & (PAGE_CACHE_SIZE - 1)) { | |
788 | pinned[1] = grab_cache_page(inode->i_mapping, last_index); | |
789 | if (!PageUptodate(pinned[1])) { | |
9ebefb18 | 790 | ret = btrfs_readpage(NULL, pinned[1]); |
39279cc3 CM |
791 | BUG_ON(ret); |
792 | wait_on_page_locked(pinned[1]); | |
793 | } else { | |
794 | unlock_page(pinned[1]); | |
795 | } | |
796 | } | |
797 | ||
39279cc3 CM |
798 | while(count > 0) { |
799 | size_t offset = pos & (PAGE_CACHE_SIZE - 1); | |
11bd143f CM |
800 | size_t write_bytes = min(count, nrptrs * |
801 | (size_t)PAGE_CACHE_SIZE - | |
8c2383c3 | 802 | offset); |
39279cc3 CM |
803 | size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >> |
804 | PAGE_CACHE_SHIFT; | |
805 | ||
8c2383c3 | 806 | WARN_ON(num_pages > nrptrs); |
39279cc3 | 807 | memset(pages, 0, sizeof(pages)); |
1832a6d5 | 808 | |
1832a6d5 | 809 | ret = btrfs_check_free_space(root, write_bytes, 0); |
1832a6d5 CM |
810 | if (ret) |
811 | goto out; | |
812 | ||
39279cc3 CM |
813 | ret = prepare_pages(root, file, pages, num_pages, |
814 | pos, first_index, last_index, | |
8c2383c3 | 815 | write_bytes); |
54aa1f4d CM |
816 | if (ret) |
817 | goto out; | |
39279cc3 | 818 | |
39279cc3 CM |
819 | ret = btrfs_copy_from_user(pos, num_pages, |
820 | write_bytes, pages, buf); | |
54aa1f4d CM |
821 | if (ret) { |
822 | btrfs_drop_pages(pages, num_pages); | |
823 | goto out; | |
824 | } | |
39279cc3 CM |
825 | |
826 | ret = dirty_and_release_pages(NULL, root, file, pages, | |
827 | num_pages, pos, write_bytes); | |
39279cc3 | 828 | btrfs_drop_pages(pages, num_pages); |
54aa1f4d CM |
829 | if (ret) |
830 | goto out; | |
39279cc3 | 831 | |
cb843a6f CM |
832 | if (will_write) { |
833 | btrfs_fdatawrite_range(inode->i_mapping, pos, | |
834 | pos + write_bytes - 1, | |
835 | WB_SYNC_NONE); | |
836 | } else { | |
837 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, | |
838 | num_pages); | |
839 | if (num_pages < | |
840 | (root->leafsize >> PAGE_CACHE_SHIFT) + 1) | |
841 | btrfs_btree_balance_dirty(root, 1); | |
842 | btrfs_throttle(root); | |
843 | } | |
844 | ||
39279cc3 CM |
845 | buf += write_bytes; |
846 | count -= write_bytes; | |
847 | pos += write_bytes; | |
848 | num_written += write_bytes; | |
849 | ||
39279cc3 CM |
850 | cond_resched(); |
851 | } | |
39279cc3 | 852 | out: |
1832a6d5 | 853 | mutex_unlock(&inode->i_mutex); |
5b92ee72 | 854 | |
1832a6d5 | 855 | out_nolock: |
8c2383c3 | 856 | kfree(pages); |
39279cc3 CM |
857 | if (pinned[0]) |
858 | page_cache_release(pinned[0]); | |
859 | if (pinned[1]) | |
860 | page_cache_release(pinned[1]); | |
861 | *ppos = pos; | |
2ff3e9b6 | 862 | |
cb843a6f | 863 | if (num_written > 0 && will_write) { |
e02119d5 CM |
864 | struct btrfs_trans_handle *trans; |
865 | ||
cb843a6f CM |
866 | err = btrfs_wait_ordered_range(inode, start_pos, num_written); |
867 | if (err) | |
2ff3e9b6 | 868 | num_written = err; |
e02119d5 | 869 | |
cb843a6f CM |
870 | if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) { |
871 | trans = btrfs_start_transaction(root, 1); | |
872 | ret = btrfs_log_dentry_safe(trans, root, | |
873 | file->f_dentry); | |
874 | if (ret == 0) { | |
875 | btrfs_sync_log(trans, root); | |
876 | btrfs_end_transaction(trans, root); | |
877 | } else { | |
878 | btrfs_commit_transaction(trans, root); | |
879 | } | |
880 | } | |
881 | if (file->f_flags & O_DIRECT) { | |
882 | invalidate_mapping_pages(inode->i_mapping, | |
883 | start_pos >> PAGE_CACHE_SHIFT, | |
884 | (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT); | |
e02119d5 | 885 | } |
2ff3e9b6 | 886 | } |
39279cc3 | 887 | current->backing_dev_info = NULL; |
39279cc3 CM |
888 | return num_written ? num_written : err; |
889 | } | |
890 | ||
6bf13c0c | 891 | int btrfs_release_file(struct inode * inode, struct file * filp) |
e1b81e67 | 892 | { |
6bf13c0c SW |
893 | if (filp->private_data) |
894 | btrfs_ioctl_trans_end(filp); | |
e1b81e67 M |
895 | return 0; |
896 | } | |
897 | ||
d352ac68 CM |
898 | /* |
899 | * fsync call for both files and directories. This logs the inode into | |
900 | * the tree log instead of forcing full commits whenever possible. | |
901 | * | |
902 | * It needs to call filemap_fdatawait so that all ordered extent updates are | |
903 | * in the metadata btree are up to date for copying to the log. | |
904 | * | |
905 | * It drops the inode mutex before doing the tree log commit. This is an | |
906 | * important optimization for directories because holding the mutex prevents | |
907 | * new operations on the dir while we write to disk. | |
908 | */ | |
e02119d5 | 909 | int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) |
39279cc3 CM |
910 | { |
911 | struct inode *inode = dentry->d_inode; | |
912 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
15ee9bc7 | 913 | int ret = 0; |
39279cc3 CM |
914 | struct btrfs_trans_handle *trans; |
915 | ||
916 | /* | |
15ee9bc7 JB |
917 | * check the transaction that last modified this inode |
918 | * and see if its already been committed | |
39279cc3 | 919 | */ |
15ee9bc7 JB |
920 | if (!BTRFS_I(inode)->last_trans) |
921 | goto out; | |
a2135011 | 922 | |
15ee9bc7 JB |
923 | mutex_lock(&root->fs_info->trans_mutex); |
924 | if (BTRFS_I(inode)->last_trans <= | |
925 | root->fs_info->last_trans_committed) { | |
926 | BTRFS_I(inode)->last_trans = 0; | |
927 | mutex_unlock(&root->fs_info->trans_mutex); | |
928 | goto out; | |
929 | } | |
930 | mutex_unlock(&root->fs_info->trans_mutex); | |
931 | ||
49eb7e46 | 932 | root->fs_info->tree_log_batch++; |
e02119d5 | 933 | filemap_fdatawait(inode->i_mapping); |
49eb7e46 | 934 | root->fs_info->tree_log_batch++; |
e02119d5 | 935 | |
15ee9bc7 | 936 | /* |
a52d9a80 CM |
937 | * ok we haven't committed the transaction yet, lets do a commit |
938 | */ | |
6bf13c0c SW |
939 | if (file->private_data) |
940 | btrfs_ioctl_trans_end(file); | |
941 | ||
39279cc3 CM |
942 | trans = btrfs_start_transaction(root, 1); |
943 | if (!trans) { | |
944 | ret = -ENOMEM; | |
945 | goto out; | |
946 | } | |
e02119d5 CM |
947 | |
948 | ret = btrfs_log_dentry_safe(trans, root, file->f_dentry); | |
49eb7e46 | 949 | if (ret < 0) { |
e02119d5 | 950 | goto out; |
49eb7e46 CM |
951 | } |
952 | ||
953 | /* we've logged all the items and now have a consistent | |
954 | * version of the file in the log. It is possible that | |
955 | * someone will come in and modify the file, but that's | |
956 | * fine because the log is consistent on disk, and we | |
957 | * have references to all of the file's extents | |
958 | * | |
959 | * It is possible that someone will come in and log the | |
960 | * file again, but that will end up using the synchronization | |
961 | * inside btrfs_sync_log to keep things safe. | |
962 | */ | |
963 | mutex_unlock(&file->f_dentry->d_inode->i_mutex); | |
964 | ||
e02119d5 CM |
965 | if (ret > 0) { |
966 | ret = btrfs_commit_transaction(trans, root); | |
967 | } else { | |
968 | btrfs_sync_log(trans, root); | |
969 | ret = btrfs_end_transaction(trans, root); | |
970 | } | |
49eb7e46 | 971 | mutex_lock(&file->f_dentry->d_inode->i_mutex); |
39279cc3 CM |
972 | out: |
973 | return ret > 0 ? EIO : ret; | |
974 | } | |
975 | ||
9ebefb18 | 976 | static struct vm_operations_struct btrfs_file_vm_ops = { |
92fee66d | 977 | .fault = filemap_fault, |
9ebefb18 CM |
978 | .page_mkwrite = btrfs_page_mkwrite, |
979 | }; | |
980 | ||
981 | static int btrfs_file_mmap(struct file *filp, struct vm_area_struct *vma) | |
982 | { | |
983 | vma->vm_ops = &btrfs_file_vm_ops; | |
984 | file_accessed(filp); | |
985 | return 0; | |
986 | } | |
987 | ||
39279cc3 CM |
988 | struct file_operations btrfs_file_operations = { |
989 | .llseek = generic_file_llseek, | |
990 | .read = do_sync_read, | |
9ebefb18 | 991 | .aio_read = generic_file_aio_read, |
e9906a98 | 992 | .splice_read = generic_file_splice_read, |
39279cc3 | 993 | .write = btrfs_file_write, |
9ebefb18 | 994 | .mmap = btrfs_file_mmap, |
39279cc3 | 995 | .open = generic_file_open, |
e1b81e67 | 996 | .release = btrfs_release_file, |
39279cc3 | 997 | .fsync = btrfs_sync_file, |
34287aa3 | 998 | .unlocked_ioctl = btrfs_ioctl, |
39279cc3 | 999 | #ifdef CONFIG_COMPAT |
34287aa3 | 1000 | .compat_ioctl = btrfs_ioctl, |
39279cc3 CM |
1001 | #endif |
1002 | }; |