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" | |
39 | ||
40 | ||
41 | static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes, | |
42 | struct page **prepared_pages, | |
43 | const char __user * buf) | |
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 | ||
71 | static 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 |
83 | static 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); |
54aa1f4d | 111 | |
3326d1b0 CM |
112 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); |
113 | if (ret < 0) { | |
54aa1f4d CM |
114 | err = ret; |
115 | goto fail; | |
116 | } | |
3326d1b0 | 117 | if (ret == 1) { |
179e29e4 CM |
118 | struct btrfs_key found_key; |
119 | ||
120 | if (path->slots[0] == 0) | |
121 | goto insert; | |
122 | ||
3326d1b0 CM |
123 | path->slots[0]--; |
124 | leaf = path->nodes[0]; | |
179e29e4 CM |
125 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); |
126 | ||
127 | if (found_key.objectid != inode->i_ino) | |
128 | goto insert; | |
129 | ||
130 | if (found_key.type != BTRFS_EXTENT_DATA_KEY) | |
131 | goto insert; | |
3326d1b0 CM |
132 | ei = btrfs_item_ptr(leaf, path->slots[0], |
133 | struct btrfs_file_extent_item); | |
134 | ||
135 | if (btrfs_file_extent_type(leaf, ei) != | |
136 | BTRFS_FILE_EXTENT_INLINE) { | |
137 | goto insert; | |
138 | } | |
139 | btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); | |
140 | ret = 0; | |
141 | } | |
142 | if (ret == 0) { | |
143 | u32 found_size; | |
18f16f7b | 144 | u64 found_end; |
3326d1b0 CM |
145 | |
146 | leaf = path->nodes[0]; | |
147 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
148 | struct btrfs_file_extent_item); | |
149 | ||
150 | if (btrfs_file_extent_type(leaf, ei) != | |
151 | BTRFS_FILE_EXTENT_INLINE) { | |
152 | err = ret; | |
153 | btrfs_print_leaf(root, leaf); | |
154 | printk("found wasn't inline offset %Lu inode %lu\n", | |
155 | offset, inode->i_ino); | |
156 | goto fail; | |
157 | } | |
3326d1b0 CM |
158 | found_size = btrfs_file_extent_inline_len(leaf, |
159 | btrfs_item_nr(leaf, path->slots[0])); | |
18f16f7b | 160 | found_end = key.offset + found_size; |
3326d1b0 | 161 | |
18f16f7b | 162 | if (found_end < offset + size) { |
3326d1b0 CM |
163 | btrfs_release_path(root, path); |
164 | ret = btrfs_search_slot(trans, root, &key, path, | |
18f16f7b | 165 | offset + size - found_end, 1); |
3326d1b0 | 166 | BUG_ON(ret != 0); |
179e29e4 | 167 | |
3326d1b0 | 168 | ret = btrfs_extend_item(trans, root, path, |
18f16f7b | 169 | offset + size - found_end); |
3326d1b0 CM |
170 | if (ret) { |
171 | err = ret; | |
172 | goto fail; | |
173 | } | |
174 | leaf = path->nodes[0]; | |
175 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
176 | struct btrfs_file_extent_item); | |
177 | } | |
18f16f7b Y |
178 | if (found_end < offset) { |
179 | ptr = btrfs_file_extent_inline_start(ei) + found_size; | |
180 | memset_extent_buffer(leaf, 0, ptr, offset - found_end); | |
181 | } | |
3326d1b0 CM |
182 | } else { |
183 | insert: | |
184 | btrfs_release_path(root, path); | |
18f16f7b Y |
185 | datasize = offset + size - key.offset; |
186 | datasize = btrfs_file_extent_calc_inline_size(datasize); | |
3326d1b0 CM |
187 | ret = btrfs_insert_empty_item(trans, root, path, &key, |
188 | datasize); | |
189 | if (ret) { | |
190 | err = ret; | |
191 | printk("got bad ret %d\n", ret); | |
192 | goto fail; | |
193 | } | |
194 | leaf = path->nodes[0]; | |
195 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
196 | struct btrfs_file_extent_item); | |
197 | btrfs_set_file_extent_generation(leaf, ei, trans->transid); | |
198 | btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); | |
199 | } | |
18f16f7b | 200 | ptr = btrfs_file_extent_inline_start(ei) + offset - key.offset; |
3326d1b0 CM |
201 | |
202 | cur_size = size; | |
203 | i = 0; | |
204 | while (size > 0) { | |
205 | page = pages[i]; | |
206 | kaddr = kmap_atomic(page, KM_USER0); | |
ae2f5411 | 207 | cur_size = min_t(size_t, PAGE_CACHE_SIZE - page_offset, size); |
3326d1b0 CM |
208 | write_extent_buffer(leaf, kaddr + page_offset, ptr, cur_size); |
209 | kunmap_atomic(kaddr, KM_USER0); | |
210 | page_offset = 0; | |
211 | ptr += cur_size; | |
212 | size -= cur_size; | |
213 | if (i >= num_pages) { | |
214 | printk("i %d num_pages %d\n", i, num_pages); | |
215 | } | |
216 | i++; | |
217 | } | |
5f39d397 | 218 | btrfs_mark_buffer_dirty(leaf); |
54aa1f4d CM |
219 | fail: |
220 | btrfs_free_path(path); | |
54aa1f4d CM |
221 | return err; |
222 | } | |
223 | ||
39279cc3 CM |
224 | static int dirty_and_release_pages(struct btrfs_trans_handle *trans, |
225 | struct btrfs_root *root, | |
226 | struct file *file, | |
227 | struct page **pages, | |
228 | size_t num_pages, | |
229 | loff_t pos, | |
230 | size_t write_bytes) | |
231 | { | |
39279cc3 | 232 | int err = 0; |
a52d9a80 | 233 | int i; |
39279cc3 | 234 | struct inode *inode = file->f_path.dentry->d_inode; |
a52d9a80 CM |
235 | struct extent_map *em; |
236 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
db94535d CM |
237 | u64 hint_byte; |
238 | u64 num_bytes; | |
a52d9a80 CM |
239 | u64 start_pos; |
240 | u64 end_of_last_block; | |
241 | u64 end_pos = pos + write_bytes; | |
db94535d | 242 | u32 inline_size; |
a52d9a80 | 243 | loff_t isize = i_size_read(inode); |
39279cc3 | 244 | |
a52d9a80 CM |
245 | em = alloc_extent_map(GFP_NOFS); |
246 | if (!em) | |
247 | return -ENOMEM; | |
39279cc3 | 248 | |
a52d9a80 | 249 | em->bdev = inode->i_sb->s_bdev; |
39279cc3 | 250 | |
5f39d397 | 251 | start_pos = pos & ~((u64)root->sectorsize - 1); |
db94535d CM |
252 | num_bytes = (write_bytes + pos - start_pos + |
253 | root->sectorsize - 1) & ~((u64)root->sectorsize - 1); | |
39279cc3 | 254 | |
011410bd | 255 | down_read(&BTRFS_I(inode)->root->snap_sem); |
db94535d CM |
256 | end_of_last_block = start_pos + num_bytes - 1; |
257 | ||
b888db2b | 258 | lock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); |
a52d9a80 CM |
259 | mutex_lock(&root->fs_info->fs_mutex); |
260 | trans = btrfs_start_transaction(root, 1); | |
261 | if (!trans) { | |
262 | err = -ENOMEM; | |
263 | goto out_unlock; | |
264 | } | |
265 | btrfs_set_trans_block_group(trans, inode); | |
db94535d CM |
266 | inode->i_blocks += num_bytes >> 9; |
267 | hint_byte = 0; | |
a52d9a80 CM |
268 | |
269 | if ((end_of_last_block & 4095) == 0) { | |
9433063b | 270 | printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block); |
a52d9a80 CM |
271 | } |
272 | set_extent_uptodate(em_tree, start_pos, end_of_last_block, GFP_NOFS); | |
273 | ||
274 | /* FIXME...EIEIO, ENOSPC and more */ | |
275 | ||
a52d9a80 CM |
276 | /* insert any holes we need to create */ |
277 | if (inode->i_size < start_pos) { | |
278 | u64 last_pos_in_file; | |
279 | u64 hole_size; | |
5f39d397 | 280 | u64 mask = root->sectorsize - 1; |
a52d9a80 CM |
281 | last_pos_in_file = (isize + mask) & ~mask; |
282 | hole_size = (start_pos - last_pos_in_file + mask) & ~mask; | |
2bf5a725 | 283 | |
a52d9a80 | 284 | if (last_pos_in_file < start_pos) { |
2bf5a725 CM |
285 | err = btrfs_drop_extents(trans, root, inode, |
286 | last_pos_in_file, | |
287 | last_pos_in_file + hole_size, | |
3326d1b0 | 288 | last_pos_in_file, |
db94535d | 289 | &hint_byte); |
2bf5a725 CM |
290 | if (err) |
291 | goto failed; | |
292 | ||
a52d9a80 CM |
293 | err = btrfs_insert_file_extent(trans, root, |
294 | inode->i_ino, | |
295 | last_pos_in_file, | |
296 | 0, 0, hole_size); | |
297 | } | |
298 | if (err) | |
39279cc3 | 299 | goto failed; |
a52d9a80 CM |
300 | } |
301 | ||
302 | /* | |
303 | * either allocate an extent for the new bytes or setup the key | |
304 | * to show we are doing inline data in the extent | |
305 | */ | |
3326d1b0 CM |
306 | inline_size = end_pos; |
307 | if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) || | |
179e29e4 | 308 | inline_size > 32768 || |
3326d1b0 | 309 | inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { |
b888db2b | 310 | u64 last_end; |
3326d1b0 | 311 | |
a52d9a80 CM |
312 | for (i = 0; i < num_pages; i++) { |
313 | struct page *p = pages[i]; | |
314 | SetPageUptodate(p); | |
b888db2b | 315 | set_page_dirty(p); |
39279cc3 | 316 | } |
35ebb934 CM |
317 | last_end = (u64)(pages[num_pages -1]->index) << |
318 | PAGE_CACHE_SHIFT; | |
b888db2b CM |
319 | last_end += PAGE_CACHE_SIZE - 1; |
320 | set_extent_delalloc(em_tree, start_pos, end_of_last_block, | |
321 | GFP_NOFS); | |
a52d9a80 | 322 | } else { |
3326d1b0 | 323 | u64 aligned_end; |
b888db2b | 324 | /* step one, delete the existing extents in this range */ |
3326d1b0 CM |
325 | aligned_end = (pos + write_bytes + root->sectorsize - 1) & |
326 | ~((u64)root->sectorsize - 1); | |
2bf5a725 | 327 | err = btrfs_drop_extents(trans, root, inode, start_pos, |
179e29e4 | 328 | aligned_end, aligned_end, &hint_byte); |
2bf5a725 CM |
329 | if (err) |
330 | goto failed; | |
a52d9a80 | 331 | err = insert_inline_extent(trans, root, inode, start_pos, |
3326d1b0 CM |
332 | end_pos - start_pos, pages, 0, |
333 | num_pages); | |
a52d9a80 | 334 | BUG_ON(err); |
a52d9a80 CM |
335 | } |
336 | if (end_pos > isize) { | |
337 | i_size_write(inode, end_pos); | |
338 | btrfs_update_inode(trans, root, inode); | |
39279cc3 CM |
339 | } |
340 | failed: | |
a52d9a80 CM |
341 | err = btrfs_end_transaction(trans, root); |
342 | out_unlock: | |
343 | mutex_unlock(&root->fs_info->fs_mutex); | |
b888db2b | 344 | unlock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); |
a52d9a80 | 345 | free_extent_map(em); |
011410bd | 346 | up_read(&BTRFS_I(inode)->root->snap_sem); |
39279cc3 CM |
347 | return err; |
348 | } | |
349 | ||
a52d9a80 CM |
350 | int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) |
351 | { | |
352 | struct extent_map *em; | |
353 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
354 | ||
355 | while(1) { | |
356 | em = lookup_extent_mapping(em_tree, start, end); | |
357 | if (!em) | |
358 | break; | |
359 | remove_extent_mapping(em_tree, em); | |
360 | /* once for us */ | |
361 | free_extent_map(em); | |
362 | /* once for the tree*/ | |
363 | free_extent_map(em); | |
364 | } | |
365 | return 0; | |
366 | } | |
367 | ||
39279cc3 CM |
368 | /* |
369 | * this is very complex, but the basic idea is to drop all extents | |
370 | * in the range start - end. hint_block is filled in with a block number | |
371 | * that would be a good hint to the block allocator for this file. | |
372 | * | |
373 | * If an extent intersects the range but is not entirely inside the range | |
374 | * it is either truncated or split. Anything entirely inside the range | |
375 | * is deleted from the tree. | |
376 | */ | |
377 | int btrfs_drop_extents(struct btrfs_trans_handle *trans, | |
378 | struct btrfs_root *root, struct inode *inode, | |
3326d1b0 | 379 | u64 start, u64 end, u64 inline_end, u64 *hint_byte) |
39279cc3 CM |
380 | { |
381 | int ret; | |
382 | struct btrfs_key key; | |
5f39d397 | 383 | struct extent_buffer *leaf; |
39279cc3 CM |
384 | int slot; |
385 | struct btrfs_file_extent_item *extent; | |
386 | u64 extent_end = 0; | |
387 | int keep; | |
388 | struct btrfs_file_extent_item old; | |
389 | struct btrfs_path *path; | |
390 | u64 search_start = start; | |
391 | int bookend; | |
392 | int found_type; | |
393 | int found_extent; | |
394 | int found_inline; | |
ccd467d6 | 395 | int recow; |
39279cc3 | 396 | |
a52d9a80 CM |
397 | btrfs_drop_extent_cache(inode, start, end - 1); |
398 | ||
39279cc3 CM |
399 | path = btrfs_alloc_path(); |
400 | if (!path) | |
401 | return -ENOMEM; | |
402 | while(1) { | |
ccd467d6 | 403 | recow = 0; |
39279cc3 CM |
404 | btrfs_release_path(root, path); |
405 | ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino, | |
406 | search_start, -1); | |
407 | if (ret < 0) | |
408 | goto out; | |
409 | if (ret > 0) { | |
410 | if (path->slots[0] == 0) { | |
411 | ret = 0; | |
412 | goto out; | |
413 | } | |
414 | path->slots[0]--; | |
415 | } | |
8c2383c3 | 416 | next_slot: |
39279cc3 CM |
417 | keep = 0; |
418 | bookend = 0; | |
419 | found_extent = 0; | |
420 | found_inline = 0; | |
421 | extent = NULL; | |
5f39d397 | 422 | leaf = path->nodes[0]; |
39279cc3 | 423 | slot = path->slots[0]; |
8c2383c3 | 424 | ret = 0; |
5f39d397 | 425 | btrfs_item_key_to_cpu(leaf, &key, slot); |
39279cc3 | 426 | if (key.offset >= end || key.objectid != inode->i_ino) { |
39279cc3 CM |
427 | goto out; |
428 | } | |
8c2383c3 | 429 | if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY) { |
39279cc3 CM |
430 | goto out; |
431 | } | |
ccd467d6 CM |
432 | if (recow) { |
433 | search_start = key.offset; | |
434 | continue; | |
435 | } | |
8c2383c3 CM |
436 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { |
437 | extent = btrfs_item_ptr(leaf, slot, | |
438 | struct btrfs_file_extent_item); | |
5f39d397 | 439 | found_type = btrfs_file_extent_type(leaf, extent); |
8c2383c3 CM |
440 | if (found_type == BTRFS_FILE_EXTENT_REG) { |
441 | extent_end = key.offset + | |
db94535d | 442 | btrfs_file_extent_num_bytes(leaf, extent); |
8c2383c3 CM |
443 | found_extent = 1; |
444 | } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { | |
5f39d397 CM |
445 | struct btrfs_item *item; |
446 | item = btrfs_item_nr(leaf, slot); | |
8c2383c3 CM |
447 | found_inline = 1; |
448 | extent_end = key.offset + | |
5f39d397 | 449 | btrfs_file_extent_inline_len(leaf, item); |
8c2383c3 CM |
450 | } |
451 | } else { | |
452 | extent_end = search_start; | |
39279cc3 CM |
453 | } |
454 | ||
455 | /* we found nothing we can drop */ | |
8c2383c3 CM |
456 | if ((!found_extent && !found_inline) || |
457 | search_start >= extent_end) { | |
458 | int nextret; | |
459 | u32 nritems; | |
5f39d397 | 460 | nritems = btrfs_header_nritems(leaf); |
8c2383c3 CM |
461 | if (slot >= nritems - 1) { |
462 | nextret = btrfs_next_leaf(root, path); | |
463 | if (nextret) | |
464 | goto out; | |
ccd467d6 | 465 | recow = 1; |
8c2383c3 CM |
466 | } else { |
467 | path->slots[0]++; | |
468 | } | |
469 | goto next_slot; | |
39279cc3 CM |
470 | } |
471 | ||
39279cc3 | 472 | if (found_inline) { |
5f39d397 | 473 | u64 mask = root->sectorsize - 1; |
39279cc3 CM |
474 | search_start = (extent_end + mask) & ~mask; |
475 | } else | |
476 | search_start = extent_end; | |
477 | ||
179e29e4 CM |
478 | if (end <= extent_end && start >= key.offset && found_inline) { |
479 | *hint_byte = EXTENT_MAP_INLINE; | |
480 | } | |
39279cc3 CM |
481 | if (end < extent_end && end >= key.offset) { |
482 | if (found_extent) { | |
db94535d CM |
483 | u64 disk_bytenr = |
484 | btrfs_file_extent_disk_bytenr(leaf, extent); | |
485 | u64 disk_num_bytes = | |
486 | btrfs_file_extent_disk_num_bytes(leaf, | |
5f39d397 CM |
487 | extent); |
488 | read_extent_buffer(leaf, &old, | |
489 | (unsigned long)extent, | |
490 | sizeof(old)); | |
db94535d | 491 | if (disk_bytenr != 0) { |
39279cc3 | 492 | ret = btrfs_inc_extent_ref(trans, root, |
db94535d | 493 | disk_bytenr, disk_num_bytes); |
39279cc3 CM |
494 | BUG_ON(ret); |
495 | } | |
496 | } | |
179e29e4 CM |
497 | bookend = 1; |
498 | if (found_inline && start <= key.offset && | |
499 | inline_end < extent_end) | |
500 | keep = 1; | |
39279cc3 | 501 | } |
39279cc3 CM |
502 | /* truncate existing extent */ |
503 | if (start > key.offset) { | |
504 | u64 new_num; | |
505 | u64 old_num; | |
506 | keep = 1; | |
5f39d397 | 507 | WARN_ON(start & (root->sectorsize - 1)); |
39279cc3 | 508 | if (found_extent) { |
db94535d CM |
509 | new_num = start - key.offset; |
510 | old_num = btrfs_file_extent_num_bytes(leaf, | |
511 | extent); | |
512 | *hint_byte = | |
513 | btrfs_file_extent_disk_bytenr(leaf, | |
514 | extent); | |
515 | if (btrfs_file_extent_disk_bytenr(leaf, | |
516 | extent)) { | |
39279cc3 | 517 | inode->i_blocks -= |
db94535d | 518 | (old_num - new_num) >> 9; |
39279cc3 | 519 | } |
db94535d CM |
520 | btrfs_set_file_extent_num_bytes(leaf, extent, |
521 | new_num); | |
5f39d397 | 522 | btrfs_mark_buffer_dirty(leaf); |
3326d1b0 CM |
523 | } else if (end > extent_end && |
524 | key.offset < inline_end && | |
525 | inline_end < extent_end) { | |
526 | u32 new_size; | |
527 | new_size = btrfs_file_extent_calc_inline_size( | |
528 | inline_end - key.offset); | |
529 | btrfs_truncate_item(trans, root, path, | |
179e29e4 | 530 | new_size, 1); |
39279cc3 CM |
531 | } |
532 | } | |
533 | /* delete the entire extent */ | |
534 | if (!keep) { | |
db94535d CM |
535 | u64 disk_bytenr = 0; |
536 | u64 disk_num_bytes = 0; | |
537 | u64 extent_num_bytes = 0; | |
39279cc3 | 538 | if (found_extent) { |
db94535d CM |
539 | disk_bytenr = |
540 | btrfs_file_extent_disk_bytenr(leaf, | |
5f39d397 | 541 | extent); |
db94535d CM |
542 | disk_num_bytes = |
543 | btrfs_file_extent_disk_num_bytes(leaf, | |
5f39d397 | 544 | extent); |
db94535d CM |
545 | extent_num_bytes = |
546 | btrfs_file_extent_num_bytes(leaf, extent); | |
547 | *hint_byte = | |
548 | btrfs_file_extent_disk_bytenr(leaf, | |
549 | extent); | |
39279cc3 CM |
550 | } |
551 | ret = btrfs_del_item(trans, root, path); | |
54aa1f4d | 552 | /* TODO update progress marker and return */ |
39279cc3 CM |
553 | BUG_ON(ret); |
554 | btrfs_release_path(root, path); | |
555 | extent = NULL; | |
db94535d CM |
556 | if (found_extent && disk_bytenr != 0) { |
557 | inode->i_blocks -= extent_num_bytes >> 9; | |
39279cc3 | 558 | ret = btrfs_free_extent(trans, root, |
db94535d CM |
559 | disk_bytenr, |
560 | disk_num_bytes, 0); | |
39279cc3 CM |
561 | } |
562 | ||
563 | BUG_ON(ret); | |
564 | if (!bookend && search_start >= end) { | |
565 | ret = 0; | |
566 | goto out; | |
567 | } | |
568 | if (!bookend) | |
569 | continue; | |
570 | } | |
179e29e4 CM |
571 | if (bookend && found_inline && start <= key.offset && |
572 | inline_end < extent_end) { | |
573 | u32 new_size; | |
574 | new_size = btrfs_file_extent_calc_inline_size( | |
575 | extent_end - inline_end); | |
576 | btrfs_truncate_item(trans, root, path, new_size, 0); | |
577 | } | |
39279cc3 CM |
578 | /* create bookend, splitting the extent in two */ |
579 | if (bookend && found_extent) { | |
580 | struct btrfs_key ins; | |
581 | ins.objectid = inode->i_ino; | |
582 | ins.offset = end; | |
39279cc3 | 583 | btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY); |
39279cc3 CM |
584 | btrfs_release_path(root, path); |
585 | ret = btrfs_insert_empty_item(trans, root, path, &ins, | |
586 | sizeof(*extent)); | |
8c2383c3 | 587 | |
5f39d397 | 588 | leaf = path->nodes[0]; |
8c2383c3 | 589 | if (ret) { |
5f39d397 CM |
590 | btrfs_print_leaf(root, leaf); |
591 | 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 | 592 | } |
39279cc3 | 593 | BUG_ON(ret); |
5f39d397 CM |
594 | extent = btrfs_item_ptr(leaf, path->slots[0], |
595 | struct btrfs_file_extent_item); | |
596 | write_extent_buffer(leaf, &old, | |
597 | (unsigned long)extent, sizeof(old)); | |
598 | ||
599 | btrfs_set_file_extent_offset(leaf, extent, | |
db94535d CM |
600 | le64_to_cpu(old.offset) + end - key.offset); |
601 | WARN_ON(le64_to_cpu(old.num_bytes) < | |
602 | (extent_end - end)); | |
603 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
604 | extent_end - end); | |
5f39d397 | 605 | btrfs_set_file_extent_type(leaf, extent, |
39279cc3 | 606 | BTRFS_FILE_EXTENT_REG); |
db94535d | 607 | |
39279cc3 | 608 | btrfs_mark_buffer_dirty(path->nodes[0]); |
db94535d | 609 | if (le64_to_cpu(old.disk_bytenr) != 0) { |
39279cc3 | 610 | inode->i_blocks += |
db94535d CM |
611 | btrfs_file_extent_num_bytes(leaf, |
612 | extent) >> 9; | |
39279cc3 CM |
613 | } |
614 | ret = 0; | |
615 | goto out; | |
616 | } | |
617 | } | |
618 | out: | |
619 | btrfs_free_path(path); | |
620 | return ret; | |
621 | } | |
622 | ||
623 | /* | |
624 | * this gets pages into the page cache and locks them down | |
625 | */ | |
626 | static int prepare_pages(struct btrfs_root *root, | |
627 | struct file *file, | |
628 | struct page **pages, | |
629 | size_t num_pages, | |
630 | loff_t pos, | |
631 | unsigned long first_index, | |
632 | unsigned long last_index, | |
8c2383c3 | 633 | size_t write_bytes) |
39279cc3 CM |
634 | { |
635 | int i; | |
636 | unsigned long index = pos >> PAGE_CACHE_SHIFT; | |
637 | struct inode *inode = file->f_path.dentry->d_inode; | |
39279cc3 | 638 | int err = 0; |
8c2383c3 | 639 | u64 start_pos; |
8c2383c3 | 640 | |
5f39d397 | 641 | start_pos = pos & ~((u64)root->sectorsize - 1); |
39279cc3 CM |
642 | |
643 | memset(pages, 0, num_pages * sizeof(struct page *)); | |
644 | ||
645 | for (i = 0; i < num_pages; i++) { | |
646 | pages[i] = grab_cache_page(inode->i_mapping, index + i); | |
647 | if (!pages[i]) { | |
648 | err = -ENOMEM; | |
a52d9a80 | 649 | BUG_ON(1); |
39279cc3 | 650 | } |
ccd467d6 CM |
651 | cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); |
652 | wait_on_page_writeback(pages[i]); | |
b3cfa35a | 653 | set_page_extent_mapped(pages[i]); |
b888db2b | 654 | WARN_ON(!PageLocked(pages[i])); |
39279cc3 CM |
655 | } |
656 | return 0; | |
39279cc3 CM |
657 | } |
658 | ||
659 | static ssize_t btrfs_file_write(struct file *file, const char __user *buf, | |
660 | size_t count, loff_t *ppos) | |
661 | { | |
662 | loff_t pos; | |
2ff3e9b6 CM |
663 | loff_t start_pos; |
664 | ssize_t num_written = 0; | |
665 | ssize_t err = 0; | |
39279cc3 CM |
666 | int ret = 0; |
667 | struct inode *inode = file->f_path.dentry->d_inode; | |
668 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
8c2383c3 CM |
669 | struct page **pages = NULL; |
670 | int nrptrs; | |
39279cc3 CM |
671 | struct page *pinned[2]; |
672 | unsigned long first_index; | |
673 | unsigned long last_index; | |
8c2383c3 CM |
674 | |
675 | nrptrs = min((count + PAGE_CACHE_SIZE - 1) / PAGE_CACHE_SIZE, | |
676 | PAGE_CACHE_SIZE / (sizeof(struct page *))); | |
39279cc3 CM |
677 | pinned[0] = NULL; |
678 | pinned[1] = NULL; | |
679 | if (file->f_flags & O_DIRECT) | |
680 | return -EINVAL; | |
2ff3e9b6 | 681 | |
39279cc3 | 682 | pos = *ppos; |
2ff3e9b6 CM |
683 | start_pos = pos; |
684 | ||
39279cc3 CM |
685 | vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE); |
686 | current->backing_dev_info = inode->i_mapping->backing_dev_info; | |
687 | err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode)); | |
688 | if (err) | |
689 | goto out; | |
690 | if (count == 0) | |
691 | goto out; | |
692 | err = remove_suid(file->f_path.dentry); | |
693 | if (err) | |
694 | goto out; | |
695 | file_update_time(file); | |
696 | ||
8c2383c3 | 697 | pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL); |
39279cc3 CM |
698 | |
699 | mutex_lock(&inode->i_mutex); | |
700 | first_index = pos >> PAGE_CACHE_SHIFT; | |
701 | last_index = (pos + count) >> PAGE_CACHE_SHIFT; | |
702 | ||
703 | /* | |
704 | * there are lots of better ways to do this, but this code | |
705 | * makes sure the first and last page in the file range are | |
706 | * up to date and ready for cow | |
707 | */ | |
708 | if ((pos & (PAGE_CACHE_SIZE - 1))) { | |
709 | pinned[0] = grab_cache_page(inode->i_mapping, first_index); | |
710 | if (!PageUptodate(pinned[0])) { | |
9ebefb18 | 711 | ret = btrfs_readpage(NULL, pinned[0]); |
39279cc3 CM |
712 | BUG_ON(ret); |
713 | wait_on_page_locked(pinned[0]); | |
714 | } else { | |
715 | unlock_page(pinned[0]); | |
716 | } | |
717 | } | |
718 | if ((pos + count) & (PAGE_CACHE_SIZE - 1)) { | |
719 | pinned[1] = grab_cache_page(inode->i_mapping, last_index); | |
720 | if (!PageUptodate(pinned[1])) { | |
9ebefb18 | 721 | ret = btrfs_readpage(NULL, pinned[1]); |
39279cc3 CM |
722 | BUG_ON(ret); |
723 | wait_on_page_locked(pinned[1]); | |
724 | } else { | |
725 | unlock_page(pinned[1]); | |
726 | } | |
727 | } | |
728 | ||
39279cc3 CM |
729 | while(count > 0) { |
730 | size_t offset = pos & (PAGE_CACHE_SIZE - 1); | |
11bd143f CM |
731 | size_t write_bytes = min(count, nrptrs * |
732 | (size_t)PAGE_CACHE_SIZE - | |
8c2383c3 | 733 | offset); |
39279cc3 CM |
734 | size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >> |
735 | PAGE_CACHE_SHIFT; | |
736 | ||
8c2383c3 | 737 | WARN_ON(num_pages > nrptrs); |
39279cc3 CM |
738 | memset(pages, 0, sizeof(pages)); |
739 | ret = prepare_pages(root, file, pages, num_pages, | |
740 | pos, first_index, last_index, | |
8c2383c3 | 741 | write_bytes); |
54aa1f4d CM |
742 | if (ret) |
743 | goto out; | |
39279cc3 | 744 | |
39279cc3 CM |
745 | ret = btrfs_copy_from_user(pos, num_pages, |
746 | write_bytes, pages, buf); | |
54aa1f4d CM |
747 | if (ret) { |
748 | btrfs_drop_pages(pages, num_pages); | |
749 | goto out; | |
750 | } | |
39279cc3 CM |
751 | |
752 | ret = dirty_and_release_pages(NULL, root, file, pages, | |
753 | num_pages, pos, write_bytes); | |
39279cc3 | 754 | btrfs_drop_pages(pages, num_pages); |
54aa1f4d CM |
755 | if (ret) |
756 | goto out; | |
39279cc3 CM |
757 | |
758 | buf += write_bytes; | |
759 | count -= write_bytes; | |
760 | pos += write_bytes; | |
761 | num_written += write_bytes; | |
762 | ||
8c2383c3 | 763 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages); |
d3c2fdcf | 764 | btrfs_btree_balance_dirty(root, 1); |
39279cc3 CM |
765 | cond_resched(); |
766 | } | |
39279cc3 CM |
767 | mutex_unlock(&inode->i_mutex); |
768 | out: | |
8c2383c3 | 769 | kfree(pages); |
39279cc3 CM |
770 | if (pinned[0]) |
771 | page_cache_release(pinned[0]); | |
772 | if (pinned[1]) | |
773 | page_cache_release(pinned[1]); | |
774 | *ppos = pos; | |
2ff3e9b6 CM |
775 | |
776 | if (num_written > 0 && ((file->f_flags & O_SYNC) || IS_SYNC(inode))) { | |
777 | err = sync_page_range(inode, inode->i_mapping, | |
778 | start_pos, num_written); | |
779 | if (err < 0) | |
780 | num_written = err; | |
781 | } | |
39279cc3 | 782 | current->backing_dev_info = NULL; |
39279cc3 CM |
783 | return num_written ? num_written : err; |
784 | } | |
785 | ||
39279cc3 CM |
786 | static int btrfs_sync_file(struct file *file, |
787 | struct dentry *dentry, int datasync) | |
788 | { | |
789 | struct inode *inode = dentry->d_inode; | |
790 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
15ee9bc7 | 791 | int ret = 0; |
39279cc3 CM |
792 | struct btrfs_trans_handle *trans; |
793 | ||
794 | /* | |
15ee9bc7 JB |
795 | * check the transaction that last modified this inode |
796 | * and see if its already been committed | |
39279cc3 CM |
797 | */ |
798 | mutex_lock(&root->fs_info->fs_mutex); | |
15ee9bc7 JB |
799 | if (!BTRFS_I(inode)->last_trans) |
800 | goto out; | |
801 | mutex_lock(&root->fs_info->trans_mutex); | |
802 | if (BTRFS_I(inode)->last_trans <= | |
803 | root->fs_info->last_trans_committed) { | |
804 | BTRFS_I(inode)->last_trans = 0; | |
805 | mutex_unlock(&root->fs_info->trans_mutex); | |
806 | goto out; | |
807 | } | |
808 | mutex_unlock(&root->fs_info->trans_mutex); | |
809 | ||
810 | /* | |
a52d9a80 CM |
811 | * ok we haven't committed the transaction yet, lets do a commit |
812 | */ | |
39279cc3 CM |
813 | trans = btrfs_start_transaction(root, 1); |
814 | if (!trans) { | |
815 | ret = -ENOMEM; | |
816 | goto out; | |
817 | } | |
818 | ret = btrfs_commit_transaction(trans, root); | |
39279cc3 | 819 | out: |
15ee9bc7 | 820 | mutex_unlock(&root->fs_info->fs_mutex); |
39279cc3 CM |
821 | return ret > 0 ? EIO : ret; |
822 | } | |
823 | ||
9ebefb18 | 824 | static struct vm_operations_struct btrfs_file_vm_ops = { |
92fee66d CM |
825 | #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) |
826 | .nopage = filemap_nopage, | |
827 | .populate = filemap_populate, | |
828 | #else | |
829 | .fault = filemap_fault, | |
830 | #endif | |
9ebefb18 CM |
831 | .page_mkwrite = btrfs_page_mkwrite, |
832 | }; | |
833 | ||
834 | static int btrfs_file_mmap(struct file *filp, struct vm_area_struct *vma) | |
835 | { | |
836 | vma->vm_ops = &btrfs_file_vm_ops; | |
837 | file_accessed(filp); | |
838 | return 0; | |
839 | } | |
840 | ||
39279cc3 CM |
841 | struct file_operations btrfs_file_operations = { |
842 | .llseek = generic_file_llseek, | |
843 | .read = do_sync_read, | |
9ebefb18 | 844 | .aio_read = generic_file_aio_read, |
39279cc3 | 845 | .write = btrfs_file_write, |
9ebefb18 | 846 | .mmap = btrfs_file_mmap, |
39279cc3 | 847 | .open = generic_file_open, |
39279cc3 | 848 | .fsync = btrfs_sync_file, |
34287aa3 | 849 | .unlocked_ioctl = btrfs_ioctl, |
39279cc3 | 850 | #ifdef CONFIG_COMPAT |
34287aa3 | 851 | .compat_ioctl = btrfs_ioctl, |
39279cc3 CM |
852 | #endif |
853 | }; | |
854 |