]>
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; | |
dcfec0dc | 242 | u64 inline_size; |
a52d9a80 | 243 | loff_t isize = i_size_read(inode); |
a52d9a80 CM |
244 | em = alloc_extent_map(GFP_NOFS); |
245 | if (!em) | |
246 | return -ENOMEM; | |
39279cc3 | 247 | |
a52d9a80 | 248 | em->bdev = inode->i_sb->s_bdev; |
39279cc3 | 249 | |
5f39d397 | 250 | start_pos = pos & ~((u64)root->sectorsize - 1); |
db94535d CM |
251 | num_bytes = (write_bytes + pos - start_pos + |
252 | root->sectorsize - 1) & ~((u64)root->sectorsize - 1); | |
39279cc3 | 253 | |
011410bd | 254 | down_read(&BTRFS_I(inode)->root->snap_sem); |
db94535d CM |
255 | end_of_last_block = start_pos + num_bytes - 1; |
256 | ||
b888db2b | 257 | lock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); |
a52d9a80 CM |
258 | mutex_lock(&root->fs_info->fs_mutex); |
259 | trans = btrfs_start_transaction(root, 1); | |
260 | if (!trans) { | |
261 | err = -ENOMEM; | |
262 | goto out_unlock; | |
263 | } | |
264 | btrfs_set_trans_block_group(trans, inode); | |
db94535d CM |
265 | inode->i_blocks += num_bytes >> 9; |
266 | hint_byte = 0; | |
a52d9a80 CM |
267 | |
268 | if ((end_of_last_block & 4095) == 0) { | |
9433063b | 269 | printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block); |
a52d9a80 CM |
270 | } |
271 | set_extent_uptodate(em_tree, start_pos, end_of_last_block, GFP_NOFS); | |
272 | ||
273 | /* FIXME...EIEIO, ENOSPC and more */ | |
274 | ||
a52d9a80 CM |
275 | /* insert any holes we need to create */ |
276 | if (inode->i_size < start_pos) { | |
277 | u64 last_pos_in_file; | |
278 | u64 hole_size; | |
5f39d397 | 279 | u64 mask = root->sectorsize - 1; |
a52d9a80 CM |
280 | last_pos_in_file = (isize + mask) & ~mask; |
281 | hole_size = (start_pos - last_pos_in_file + mask) & ~mask; | |
2bf5a725 | 282 | |
a52d9a80 | 283 | if (last_pos_in_file < start_pos) { |
2bf5a725 CM |
284 | err = btrfs_drop_extents(trans, root, inode, |
285 | last_pos_in_file, | |
286 | last_pos_in_file + hole_size, | |
3326d1b0 | 287 | last_pos_in_file, |
db94535d | 288 | &hint_byte); |
2bf5a725 CM |
289 | if (err) |
290 | goto failed; | |
291 | ||
a52d9a80 CM |
292 | err = btrfs_insert_file_extent(trans, root, |
293 | inode->i_ino, | |
294 | last_pos_in_file, | |
295 | 0, 0, hole_size); | |
296 | } | |
297 | if (err) | |
39279cc3 | 298 | goto failed; |
a52d9a80 CM |
299 | } |
300 | ||
301 | /* | |
302 | * either allocate an extent for the new bytes or setup the key | |
303 | * to show we are doing inline data in the extent | |
304 | */ | |
3326d1b0 CM |
305 | inline_size = end_pos; |
306 | if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) || | |
179e29e4 | 307 | inline_size > 32768 || |
3326d1b0 | 308 | inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { |
b888db2b | 309 | u64 last_end; |
3326d1b0 | 310 | |
a52d9a80 CM |
311 | for (i = 0; i < num_pages; i++) { |
312 | struct page *p = pages[i]; | |
313 | SetPageUptodate(p); | |
b888db2b | 314 | set_page_dirty(p); |
39279cc3 | 315 | } |
35ebb934 CM |
316 | last_end = (u64)(pages[num_pages -1]->index) << |
317 | PAGE_CACHE_SHIFT; | |
b888db2b CM |
318 | last_end += PAGE_CACHE_SIZE - 1; |
319 | set_extent_delalloc(em_tree, start_pos, end_of_last_block, | |
320 | GFP_NOFS); | |
a52d9a80 | 321 | } else { |
3326d1b0 | 322 | u64 aligned_end; |
b888db2b | 323 | /* step one, delete the existing extents in this range */ |
3326d1b0 CM |
324 | aligned_end = (pos + write_bytes + root->sectorsize - 1) & |
325 | ~((u64)root->sectorsize - 1); | |
2bf5a725 | 326 | err = btrfs_drop_extents(trans, root, inode, start_pos, |
179e29e4 | 327 | aligned_end, aligned_end, &hint_byte); |
2bf5a725 CM |
328 | if (err) |
329 | goto failed; | |
dcfec0dc Y |
330 | if (isize > inline_size) |
331 | inline_size = min_t(u64, isize, aligned_end); | |
332 | inline_size -= start_pos; | |
a52d9a80 | 333 | err = insert_inline_extent(trans, root, inode, start_pos, |
dcfec0dc | 334 | inline_size, pages, 0, num_pages); |
a52d9a80 | 335 | BUG_ON(err); |
a52d9a80 CM |
336 | } |
337 | if (end_pos > isize) { | |
338 | i_size_write(inode, end_pos); | |
339 | btrfs_update_inode(trans, root, inode); | |
39279cc3 CM |
340 | } |
341 | failed: | |
a52d9a80 CM |
342 | err = btrfs_end_transaction(trans, root); |
343 | out_unlock: | |
344 | mutex_unlock(&root->fs_info->fs_mutex); | |
b888db2b | 345 | unlock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); |
a52d9a80 | 346 | free_extent_map(em); |
011410bd | 347 | up_read(&BTRFS_I(inode)->root->snap_sem); |
39279cc3 CM |
348 | return err; |
349 | } | |
350 | ||
a52d9a80 CM |
351 | int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) |
352 | { | |
353 | struct extent_map *em; | |
354 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
355 | ||
356 | while(1) { | |
357 | em = lookup_extent_mapping(em_tree, start, end); | |
358 | if (!em) | |
359 | break; | |
360 | remove_extent_mapping(em_tree, em); | |
361 | /* once for us */ | |
362 | free_extent_map(em); | |
363 | /* once for the tree*/ | |
364 | free_extent_map(em); | |
365 | } | |
366 | return 0; | |
367 | } | |
368 | ||
39279cc3 CM |
369 | /* |
370 | * this is very complex, but the basic idea is to drop all extents | |
371 | * in the range start - end. hint_block is filled in with a block number | |
372 | * that would be a good hint to the block allocator for this file. | |
373 | * | |
374 | * If an extent intersects the range but is not entirely inside the range | |
375 | * it is either truncated or split. Anything entirely inside the range | |
376 | * is deleted from the tree. | |
377 | */ | |
378 | int btrfs_drop_extents(struct btrfs_trans_handle *trans, | |
379 | struct btrfs_root *root, struct inode *inode, | |
00f5c795 | 380 | u64 start, u64 end, u64 inline_limit, u64 *hint_byte) |
39279cc3 | 381 | { |
00f5c795 CM |
382 | u64 extent_end = 0; |
383 | u64 search_start = start; | |
5f39d397 | 384 | struct extent_buffer *leaf; |
39279cc3 | 385 | struct btrfs_file_extent_item *extent; |
39279cc3 | 386 | struct btrfs_path *path; |
00f5c795 CM |
387 | struct btrfs_key key; |
388 | struct btrfs_file_extent_item old; | |
389 | int keep; | |
390 | int slot; | |
39279cc3 CM |
391 | int bookend; |
392 | int found_type; | |
393 | int found_extent; | |
394 | int found_inline; | |
ccd467d6 | 395 | int recow; |
00f5c795 | 396 | int ret; |
39279cc3 | 397 | |
a52d9a80 CM |
398 | btrfs_drop_extent_cache(inode, start, end - 1); |
399 | ||
39279cc3 CM |
400 | path = btrfs_alloc_path(); |
401 | if (!path) | |
402 | return -ENOMEM; | |
403 | while(1) { | |
ccd467d6 | 404 | recow = 0; |
39279cc3 CM |
405 | btrfs_release_path(root, path); |
406 | ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino, | |
407 | search_start, -1); | |
408 | if (ret < 0) | |
409 | goto out; | |
410 | if (ret > 0) { | |
411 | if (path->slots[0] == 0) { | |
412 | ret = 0; | |
413 | goto out; | |
414 | } | |
415 | path->slots[0]--; | |
416 | } | |
8c2383c3 | 417 | next_slot: |
39279cc3 CM |
418 | keep = 0; |
419 | bookend = 0; | |
420 | found_extent = 0; | |
421 | found_inline = 0; | |
422 | extent = NULL; | |
5f39d397 | 423 | leaf = path->nodes[0]; |
39279cc3 | 424 | slot = path->slots[0]; |
8c2383c3 | 425 | ret = 0; |
5f39d397 | 426 | btrfs_item_key_to_cpu(leaf, &key, slot); |
39279cc3 | 427 | if (key.offset >= end || key.objectid != inode->i_ino) { |
39279cc3 CM |
428 | goto out; |
429 | } | |
8c2383c3 | 430 | if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY) { |
39279cc3 CM |
431 | goto out; |
432 | } | |
ccd467d6 CM |
433 | if (recow) { |
434 | search_start = key.offset; | |
435 | continue; | |
436 | } | |
8c2383c3 CM |
437 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { |
438 | extent = btrfs_item_ptr(leaf, slot, | |
439 | struct btrfs_file_extent_item); | |
5f39d397 | 440 | found_type = btrfs_file_extent_type(leaf, extent); |
8c2383c3 | 441 | if (found_type == BTRFS_FILE_EXTENT_REG) { |
257d0ce3 CM |
442 | extent_end = |
443 | btrfs_file_extent_disk_bytenr(leaf, | |
444 | extent); | |
445 | if (extent_end) | |
446 | *hint_byte = extent_end; | |
447 | ||
8c2383c3 | 448 | extent_end = key.offset + |
db94535d | 449 | btrfs_file_extent_num_bytes(leaf, extent); |
8c2383c3 CM |
450 | found_extent = 1; |
451 | } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { | |
5f39d397 CM |
452 | struct btrfs_item *item; |
453 | item = btrfs_item_nr(leaf, slot); | |
8c2383c3 CM |
454 | found_inline = 1; |
455 | extent_end = key.offset + | |
5f39d397 | 456 | btrfs_file_extent_inline_len(leaf, item); |
8c2383c3 CM |
457 | } |
458 | } else { | |
459 | extent_end = search_start; | |
39279cc3 CM |
460 | } |
461 | ||
462 | /* we found nothing we can drop */ | |
8c2383c3 CM |
463 | if ((!found_extent && !found_inline) || |
464 | search_start >= extent_end) { | |
465 | int nextret; | |
466 | u32 nritems; | |
5f39d397 | 467 | nritems = btrfs_header_nritems(leaf); |
8c2383c3 CM |
468 | if (slot >= nritems - 1) { |
469 | nextret = btrfs_next_leaf(root, path); | |
470 | if (nextret) | |
471 | goto out; | |
ccd467d6 | 472 | recow = 1; |
8c2383c3 CM |
473 | } else { |
474 | path->slots[0]++; | |
475 | } | |
476 | goto next_slot; | |
39279cc3 CM |
477 | } |
478 | ||
39279cc3 | 479 | if (found_inline) { |
5f39d397 | 480 | u64 mask = root->sectorsize - 1; |
39279cc3 CM |
481 | search_start = (extent_end + mask) & ~mask; |
482 | } else | |
483 | search_start = extent_end; | |
a273208e | 484 | if (end < extent_end && start > key.offset && found_inline) { |
179e29e4 CM |
485 | *hint_byte = EXTENT_MAP_INLINE; |
486 | } | |
39279cc3 CM |
487 | if (end < extent_end && end >= key.offset) { |
488 | if (found_extent) { | |
db94535d CM |
489 | u64 disk_bytenr = |
490 | btrfs_file_extent_disk_bytenr(leaf, extent); | |
491 | u64 disk_num_bytes = | |
492 | btrfs_file_extent_disk_num_bytes(leaf, | |
5f39d397 CM |
493 | extent); |
494 | read_extent_buffer(leaf, &old, | |
495 | (unsigned long)extent, | |
496 | sizeof(old)); | |
db94535d | 497 | if (disk_bytenr != 0) { |
39279cc3 | 498 | ret = btrfs_inc_extent_ref(trans, root, |
db94535d | 499 | disk_bytenr, disk_num_bytes); |
39279cc3 CM |
500 | BUG_ON(ret); |
501 | } | |
502 | } | |
179e29e4 CM |
503 | bookend = 1; |
504 | if (found_inline && start <= key.offset && | |
00f5c795 | 505 | inline_limit < extent_end) |
179e29e4 | 506 | keep = 1; |
39279cc3 | 507 | } |
39279cc3 CM |
508 | /* truncate existing extent */ |
509 | if (start > key.offset) { | |
510 | u64 new_num; | |
511 | u64 old_num; | |
512 | keep = 1; | |
5f39d397 | 513 | WARN_ON(start & (root->sectorsize - 1)); |
39279cc3 | 514 | if (found_extent) { |
db94535d CM |
515 | new_num = start - key.offset; |
516 | old_num = btrfs_file_extent_num_bytes(leaf, | |
517 | extent); | |
518 | *hint_byte = | |
519 | btrfs_file_extent_disk_bytenr(leaf, | |
520 | extent); | |
521 | if (btrfs_file_extent_disk_bytenr(leaf, | |
522 | extent)) { | |
39279cc3 | 523 | inode->i_blocks -= |
db94535d | 524 | (old_num - new_num) >> 9; |
39279cc3 | 525 | } |
db94535d CM |
526 | btrfs_set_file_extent_num_bytes(leaf, extent, |
527 | new_num); | |
5f39d397 | 528 | btrfs_mark_buffer_dirty(leaf); |
00f5c795 CM |
529 | } else if (key.offset < inline_limit && |
530 | (end > extent_end) && | |
531 | (inline_limit < extent_end)) { | |
3326d1b0 CM |
532 | u32 new_size; |
533 | new_size = btrfs_file_extent_calc_inline_size( | |
00f5c795 | 534 | inline_limit - key.offset); |
3326d1b0 | 535 | btrfs_truncate_item(trans, root, path, |
179e29e4 | 536 | new_size, 1); |
39279cc3 CM |
537 | } |
538 | } | |
539 | /* delete the entire extent */ | |
540 | if (!keep) { | |
db94535d CM |
541 | u64 disk_bytenr = 0; |
542 | u64 disk_num_bytes = 0; | |
543 | u64 extent_num_bytes = 0; | |
39279cc3 | 544 | if (found_extent) { |
db94535d CM |
545 | disk_bytenr = |
546 | btrfs_file_extent_disk_bytenr(leaf, | |
5f39d397 | 547 | extent); |
db94535d CM |
548 | disk_num_bytes = |
549 | btrfs_file_extent_disk_num_bytes(leaf, | |
5f39d397 | 550 | extent); |
db94535d CM |
551 | extent_num_bytes = |
552 | btrfs_file_extent_num_bytes(leaf, extent); | |
553 | *hint_byte = | |
554 | btrfs_file_extent_disk_bytenr(leaf, | |
555 | extent); | |
39279cc3 CM |
556 | } |
557 | ret = btrfs_del_item(trans, root, path); | |
54aa1f4d | 558 | /* TODO update progress marker and return */ |
39279cc3 CM |
559 | BUG_ON(ret); |
560 | btrfs_release_path(root, path); | |
561 | extent = NULL; | |
db94535d CM |
562 | if (found_extent && disk_bytenr != 0) { |
563 | inode->i_blocks -= extent_num_bytes >> 9; | |
39279cc3 | 564 | ret = btrfs_free_extent(trans, root, |
db94535d CM |
565 | disk_bytenr, |
566 | disk_num_bytes, 0); | |
39279cc3 CM |
567 | } |
568 | ||
569 | BUG_ON(ret); | |
570 | if (!bookend && search_start >= end) { | |
571 | ret = 0; | |
572 | goto out; | |
573 | } | |
574 | if (!bookend) | |
575 | continue; | |
576 | } | |
179e29e4 | 577 | if (bookend && found_inline && start <= key.offset && |
00f5c795 | 578 | inline_limit < extent_end && key.offset <= inline_limit) { |
179e29e4 CM |
579 | u32 new_size; |
580 | new_size = btrfs_file_extent_calc_inline_size( | |
00f5c795 | 581 | extent_end - inline_limit); |
179e29e4 CM |
582 | btrfs_truncate_item(trans, root, path, new_size, 0); |
583 | } | |
39279cc3 CM |
584 | /* create bookend, splitting the extent in two */ |
585 | if (bookend && found_extent) { | |
586 | struct btrfs_key ins; | |
587 | ins.objectid = inode->i_ino; | |
588 | ins.offset = end; | |
39279cc3 | 589 | btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY); |
39279cc3 CM |
590 | btrfs_release_path(root, path); |
591 | ret = btrfs_insert_empty_item(trans, root, path, &ins, | |
592 | sizeof(*extent)); | |
8c2383c3 | 593 | |
5f39d397 | 594 | leaf = path->nodes[0]; |
8c2383c3 | 595 | if (ret) { |
5f39d397 CM |
596 | btrfs_print_leaf(root, leaf); |
597 | 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 | 598 | } |
39279cc3 | 599 | BUG_ON(ret); |
5f39d397 CM |
600 | extent = btrfs_item_ptr(leaf, path->slots[0], |
601 | struct btrfs_file_extent_item); | |
602 | write_extent_buffer(leaf, &old, | |
603 | (unsigned long)extent, sizeof(old)); | |
604 | ||
605 | btrfs_set_file_extent_offset(leaf, extent, | |
db94535d CM |
606 | le64_to_cpu(old.offset) + end - key.offset); |
607 | WARN_ON(le64_to_cpu(old.num_bytes) < | |
608 | (extent_end - end)); | |
609 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
610 | extent_end - end); | |
5f39d397 | 611 | btrfs_set_file_extent_type(leaf, extent, |
39279cc3 | 612 | BTRFS_FILE_EXTENT_REG); |
db94535d | 613 | |
39279cc3 | 614 | btrfs_mark_buffer_dirty(path->nodes[0]); |
db94535d | 615 | if (le64_to_cpu(old.disk_bytenr) != 0) { |
39279cc3 | 616 | inode->i_blocks += |
db94535d CM |
617 | btrfs_file_extent_num_bytes(leaf, |
618 | extent) >> 9; | |
39279cc3 CM |
619 | } |
620 | ret = 0; | |
621 | goto out; | |
622 | } | |
623 | } | |
624 | out: | |
625 | btrfs_free_path(path); | |
626 | return ret; | |
627 | } | |
628 | ||
629 | /* | |
630 | * this gets pages into the page cache and locks them down | |
631 | */ | |
632 | static int prepare_pages(struct btrfs_root *root, | |
633 | struct file *file, | |
634 | struct page **pages, | |
635 | size_t num_pages, | |
636 | loff_t pos, | |
637 | unsigned long first_index, | |
638 | unsigned long last_index, | |
8c2383c3 | 639 | size_t write_bytes) |
39279cc3 CM |
640 | { |
641 | int i; | |
642 | unsigned long index = pos >> PAGE_CACHE_SHIFT; | |
643 | struct inode *inode = file->f_path.dentry->d_inode; | |
39279cc3 | 644 | int err = 0; |
8c2383c3 | 645 | u64 start_pos; |
8c2383c3 | 646 | |
5f39d397 | 647 | start_pos = pos & ~((u64)root->sectorsize - 1); |
39279cc3 CM |
648 | |
649 | memset(pages, 0, num_pages * sizeof(struct page *)); | |
650 | ||
651 | for (i = 0; i < num_pages; i++) { | |
652 | pages[i] = grab_cache_page(inode->i_mapping, index + i); | |
653 | if (!pages[i]) { | |
654 | err = -ENOMEM; | |
a52d9a80 | 655 | BUG_ON(1); |
39279cc3 | 656 | } |
ccd467d6 CM |
657 | cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); |
658 | wait_on_page_writeback(pages[i]); | |
b3cfa35a | 659 | set_page_extent_mapped(pages[i]); |
b888db2b | 660 | WARN_ON(!PageLocked(pages[i])); |
39279cc3 CM |
661 | } |
662 | return 0; | |
39279cc3 CM |
663 | } |
664 | ||
665 | static ssize_t btrfs_file_write(struct file *file, const char __user *buf, | |
666 | size_t count, loff_t *ppos) | |
667 | { | |
668 | loff_t pos; | |
2ff3e9b6 CM |
669 | loff_t start_pos; |
670 | ssize_t num_written = 0; | |
671 | ssize_t err = 0; | |
39279cc3 CM |
672 | int ret = 0; |
673 | struct inode *inode = file->f_path.dentry->d_inode; | |
674 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
8c2383c3 CM |
675 | struct page **pages = NULL; |
676 | int nrptrs; | |
39279cc3 CM |
677 | struct page *pinned[2]; |
678 | unsigned long first_index; | |
679 | unsigned long last_index; | |
8c2383c3 CM |
680 | |
681 | nrptrs = min((count + PAGE_CACHE_SIZE - 1) / PAGE_CACHE_SIZE, | |
682 | PAGE_CACHE_SIZE / (sizeof(struct page *))); | |
39279cc3 CM |
683 | pinned[0] = NULL; |
684 | pinned[1] = NULL; | |
685 | if (file->f_flags & O_DIRECT) | |
686 | return -EINVAL; | |
2ff3e9b6 | 687 | |
39279cc3 | 688 | pos = *ppos; |
2ff3e9b6 CM |
689 | start_pos = pos; |
690 | ||
39279cc3 CM |
691 | vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE); |
692 | current->backing_dev_info = inode->i_mapping->backing_dev_info; | |
693 | err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode)); | |
694 | if (err) | |
695 | goto out; | |
696 | if (count == 0) | |
697 | goto out; | |
698 | err = remove_suid(file->f_path.dentry); | |
699 | if (err) | |
700 | goto out; | |
701 | file_update_time(file); | |
702 | ||
8c2383c3 | 703 | pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL); |
39279cc3 CM |
704 | |
705 | mutex_lock(&inode->i_mutex); | |
706 | first_index = pos >> PAGE_CACHE_SHIFT; | |
707 | last_index = (pos + count) >> PAGE_CACHE_SHIFT; | |
708 | ||
709 | /* | |
710 | * there are lots of better ways to do this, but this code | |
711 | * makes sure the first and last page in the file range are | |
712 | * up to date and ready for cow | |
713 | */ | |
714 | if ((pos & (PAGE_CACHE_SIZE - 1))) { | |
715 | pinned[0] = grab_cache_page(inode->i_mapping, first_index); | |
716 | if (!PageUptodate(pinned[0])) { | |
9ebefb18 | 717 | ret = btrfs_readpage(NULL, pinned[0]); |
39279cc3 CM |
718 | BUG_ON(ret); |
719 | wait_on_page_locked(pinned[0]); | |
720 | } else { | |
721 | unlock_page(pinned[0]); | |
722 | } | |
723 | } | |
724 | if ((pos + count) & (PAGE_CACHE_SIZE - 1)) { | |
725 | pinned[1] = grab_cache_page(inode->i_mapping, last_index); | |
726 | if (!PageUptodate(pinned[1])) { | |
9ebefb18 | 727 | ret = btrfs_readpage(NULL, pinned[1]); |
39279cc3 CM |
728 | BUG_ON(ret); |
729 | wait_on_page_locked(pinned[1]); | |
730 | } else { | |
731 | unlock_page(pinned[1]); | |
732 | } | |
733 | } | |
734 | ||
39279cc3 CM |
735 | while(count > 0) { |
736 | size_t offset = pos & (PAGE_CACHE_SIZE - 1); | |
11bd143f CM |
737 | size_t write_bytes = min(count, nrptrs * |
738 | (size_t)PAGE_CACHE_SIZE - | |
8c2383c3 | 739 | offset); |
39279cc3 CM |
740 | size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >> |
741 | PAGE_CACHE_SHIFT; | |
742 | ||
8c2383c3 | 743 | WARN_ON(num_pages > nrptrs); |
39279cc3 CM |
744 | memset(pages, 0, sizeof(pages)); |
745 | ret = prepare_pages(root, file, pages, num_pages, | |
746 | pos, first_index, last_index, | |
8c2383c3 | 747 | write_bytes); |
54aa1f4d CM |
748 | if (ret) |
749 | goto out; | |
39279cc3 | 750 | |
39279cc3 CM |
751 | ret = btrfs_copy_from_user(pos, num_pages, |
752 | write_bytes, pages, buf); | |
54aa1f4d CM |
753 | if (ret) { |
754 | btrfs_drop_pages(pages, num_pages); | |
755 | goto out; | |
756 | } | |
39279cc3 CM |
757 | |
758 | ret = dirty_and_release_pages(NULL, root, file, pages, | |
759 | num_pages, pos, write_bytes); | |
39279cc3 | 760 | btrfs_drop_pages(pages, num_pages); |
54aa1f4d CM |
761 | if (ret) |
762 | goto out; | |
39279cc3 CM |
763 | |
764 | buf += write_bytes; | |
765 | count -= write_bytes; | |
766 | pos += write_bytes; | |
767 | num_written += write_bytes; | |
768 | ||
8c2383c3 | 769 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages); |
448d640b CM |
770 | if (num_pages < (root->leafsize >> PAGE_CACHE_SHIFT) + 1) |
771 | btrfs_btree_balance_dirty(root, 1); | |
39279cc3 CM |
772 | cond_resched(); |
773 | } | |
39279cc3 CM |
774 | mutex_unlock(&inode->i_mutex); |
775 | out: | |
8c2383c3 | 776 | kfree(pages); |
39279cc3 CM |
777 | if (pinned[0]) |
778 | page_cache_release(pinned[0]); | |
779 | if (pinned[1]) | |
780 | page_cache_release(pinned[1]); | |
781 | *ppos = pos; | |
2ff3e9b6 CM |
782 | |
783 | if (num_written > 0 && ((file->f_flags & O_SYNC) || IS_SYNC(inode))) { | |
784 | err = sync_page_range(inode, inode->i_mapping, | |
785 | start_pos, num_written); | |
786 | if (err < 0) | |
787 | num_written = err; | |
788 | } | |
39279cc3 | 789 | current->backing_dev_info = NULL; |
39279cc3 CM |
790 | return num_written ? num_written : err; |
791 | } | |
792 | ||
39279cc3 CM |
793 | static int btrfs_sync_file(struct file *file, |
794 | struct dentry *dentry, int datasync) | |
795 | { | |
796 | struct inode *inode = dentry->d_inode; | |
797 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
15ee9bc7 | 798 | int ret = 0; |
39279cc3 CM |
799 | struct btrfs_trans_handle *trans; |
800 | ||
801 | /* | |
15ee9bc7 JB |
802 | * check the transaction that last modified this inode |
803 | * and see if its already been committed | |
39279cc3 CM |
804 | */ |
805 | mutex_lock(&root->fs_info->fs_mutex); | |
15ee9bc7 JB |
806 | if (!BTRFS_I(inode)->last_trans) |
807 | goto out; | |
808 | mutex_lock(&root->fs_info->trans_mutex); | |
809 | if (BTRFS_I(inode)->last_trans <= | |
810 | root->fs_info->last_trans_committed) { | |
811 | BTRFS_I(inode)->last_trans = 0; | |
812 | mutex_unlock(&root->fs_info->trans_mutex); | |
813 | goto out; | |
814 | } | |
815 | mutex_unlock(&root->fs_info->trans_mutex); | |
816 | ||
817 | /* | |
a52d9a80 CM |
818 | * ok we haven't committed the transaction yet, lets do a commit |
819 | */ | |
39279cc3 CM |
820 | trans = btrfs_start_transaction(root, 1); |
821 | if (!trans) { | |
822 | ret = -ENOMEM; | |
823 | goto out; | |
824 | } | |
825 | ret = btrfs_commit_transaction(trans, root); | |
39279cc3 | 826 | out: |
15ee9bc7 | 827 | mutex_unlock(&root->fs_info->fs_mutex); |
39279cc3 CM |
828 | return ret > 0 ? EIO : ret; |
829 | } | |
830 | ||
9ebefb18 | 831 | static struct vm_operations_struct btrfs_file_vm_ops = { |
92fee66d CM |
832 | #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) |
833 | .nopage = filemap_nopage, | |
834 | .populate = filemap_populate, | |
835 | #else | |
836 | .fault = filemap_fault, | |
837 | #endif | |
9ebefb18 CM |
838 | .page_mkwrite = btrfs_page_mkwrite, |
839 | }; | |
840 | ||
841 | static int btrfs_file_mmap(struct file *filp, struct vm_area_struct *vma) | |
842 | { | |
843 | vma->vm_ops = &btrfs_file_vm_ops; | |
844 | file_accessed(filp); | |
845 | return 0; | |
846 | } | |
847 | ||
39279cc3 CM |
848 | struct file_operations btrfs_file_operations = { |
849 | .llseek = generic_file_llseek, | |
850 | .read = do_sync_read, | |
9ebefb18 | 851 | .aio_read = generic_file_aio_read, |
39279cc3 | 852 | .write = btrfs_file_write, |
9ebefb18 | 853 | .mmap = btrfs_file_mmap, |
39279cc3 | 854 | .open = generic_file_open, |
39279cc3 | 855 | .fsync = btrfs_sync_file, |
34287aa3 | 856 | .unlocked_ioctl = btrfs_ioctl, |
39279cc3 | 857 | #ifdef CONFIG_COMPAT |
34287aa3 | 858 | .compat_ioctl = btrfs_ioctl, |
39279cc3 CM |
859 | #endif |
860 | }; | |
861 |