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" | |
dc17ff8f | 37 | #include "ordered-data.h" |
39279cc3 CM |
38 | #include "ioctl.h" |
39 | #include "print-tree.h" | |
40 | ||
41 | ||
42 | static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes, | |
43 | struct page **prepared_pages, | |
44 | const char __user * buf) | |
45 | { | |
46 | long page_fault = 0; | |
47 | int i; | |
48 | int offset = pos & (PAGE_CACHE_SIZE - 1); | |
49 | ||
50 | for (i = 0; i < num_pages && write_bytes > 0; i++, offset = 0) { | |
51 | size_t count = min_t(size_t, | |
52 | PAGE_CACHE_SIZE - offset, write_bytes); | |
53 | struct page *page = prepared_pages[i]; | |
54 | fault_in_pages_readable(buf, count); | |
55 | ||
56 | /* Copy data from userspace to the current page */ | |
57 | kmap(page); | |
58 | page_fault = __copy_from_user(page_address(page) + offset, | |
59 | buf, count); | |
60 | /* Flush processor's dcache for this page */ | |
61 | flush_dcache_page(page); | |
62 | kunmap(page); | |
63 | buf += count; | |
64 | write_bytes -= count; | |
65 | ||
66 | if (page_fault) | |
67 | break; | |
68 | } | |
69 | return page_fault ? -EFAULT : 0; | |
70 | } | |
71 | ||
72 | static void btrfs_drop_pages(struct page **pages, size_t num_pages) | |
73 | { | |
74 | size_t i; | |
75 | for (i = 0; i < num_pages; i++) { | |
76 | if (!pages[i]) | |
77 | break; | |
78 | unlock_page(pages[i]); | |
79 | mark_page_accessed(pages[i]); | |
80 | page_cache_release(pages[i]); | |
81 | } | |
82 | } | |
83 | ||
98ed5174 | 84 | static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, |
a52d9a80 | 85 | struct btrfs_root *root, struct inode *inode, |
3326d1b0 CM |
86 | u64 offset, size_t size, |
87 | struct page **pages, size_t page_offset, | |
88 | int num_pages) | |
54aa1f4d CM |
89 | { |
90 | struct btrfs_key key; | |
91 | struct btrfs_path *path; | |
5f39d397 CM |
92 | struct extent_buffer *leaf; |
93 | char *kaddr; | |
94 | unsigned long ptr; | |
54aa1f4d | 95 | struct btrfs_file_extent_item *ei; |
3326d1b0 | 96 | struct page *page; |
54aa1f4d CM |
97 | u32 datasize; |
98 | int err = 0; | |
99 | int ret; | |
3326d1b0 CM |
100 | int i; |
101 | ssize_t cur_size; | |
54aa1f4d CM |
102 | |
103 | path = btrfs_alloc_path(); | |
104 | if (!path) | |
105 | return -ENOMEM; | |
106 | ||
54aa1f4d CM |
107 | btrfs_set_trans_block_group(trans, inode); |
108 | ||
109 | key.objectid = inode->i_ino; | |
110 | key.offset = offset; | |
54aa1f4d | 111 | btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); |
54aa1f4d | 112 | |
3326d1b0 CM |
113 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); |
114 | if (ret < 0) { | |
54aa1f4d CM |
115 | err = ret; |
116 | goto fail; | |
117 | } | |
3326d1b0 | 118 | if (ret == 1) { |
179e29e4 CM |
119 | struct btrfs_key found_key; |
120 | ||
121 | if (path->slots[0] == 0) | |
122 | goto insert; | |
123 | ||
3326d1b0 CM |
124 | path->slots[0]--; |
125 | leaf = path->nodes[0]; | |
179e29e4 CM |
126 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); |
127 | ||
128 | if (found_key.objectid != inode->i_ino) | |
129 | goto insert; | |
130 | ||
131 | if (found_key.type != BTRFS_EXTENT_DATA_KEY) | |
132 | goto insert; | |
3326d1b0 CM |
133 | ei = btrfs_item_ptr(leaf, path->slots[0], |
134 | struct btrfs_file_extent_item); | |
135 | ||
136 | if (btrfs_file_extent_type(leaf, ei) != | |
137 | BTRFS_FILE_EXTENT_INLINE) { | |
138 | goto insert; | |
139 | } | |
140 | btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); | |
141 | ret = 0; | |
142 | } | |
143 | if (ret == 0) { | |
144 | u32 found_size; | |
18f16f7b | 145 | u64 found_end; |
3326d1b0 CM |
146 | |
147 | leaf = path->nodes[0]; | |
148 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
149 | struct btrfs_file_extent_item); | |
150 | ||
151 | if (btrfs_file_extent_type(leaf, ei) != | |
152 | BTRFS_FILE_EXTENT_INLINE) { | |
153 | err = ret; | |
154 | btrfs_print_leaf(root, leaf); | |
155 | printk("found wasn't inline offset %Lu inode %lu\n", | |
156 | offset, inode->i_ino); | |
157 | goto fail; | |
158 | } | |
3326d1b0 CM |
159 | found_size = btrfs_file_extent_inline_len(leaf, |
160 | btrfs_item_nr(leaf, path->slots[0])); | |
18f16f7b | 161 | found_end = key.offset + found_size; |
3326d1b0 | 162 | |
18f16f7b | 163 | if (found_end < offset + size) { |
3326d1b0 CM |
164 | btrfs_release_path(root, path); |
165 | ret = btrfs_search_slot(trans, root, &key, path, | |
18f16f7b | 166 | offset + size - found_end, 1); |
3326d1b0 | 167 | BUG_ON(ret != 0); |
179e29e4 | 168 | |
3326d1b0 | 169 | ret = btrfs_extend_item(trans, root, path, |
18f16f7b | 170 | offset + size - found_end); |
3326d1b0 CM |
171 | if (ret) { |
172 | err = ret; | |
173 | goto fail; | |
174 | } | |
175 | leaf = path->nodes[0]; | |
176 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
177 | struct btrfs_file_extent_item); | |
178 | } | |
18f16f7b Y |
179 | if (found_end < offset) { |
180 | ptr = btrfs_file_extent_inline_start(ei) + found_size; | |
181 | memset_extent_buffer(leaf, 0, ptr, offset - found_end); | |
182 | } | |
3326d1b0 CM |
183 | } else { |
184 | insert: | |
185 | btrfs_release_path(root, path); | |
18f16f7b Y |
186 | datasize = offset + size - key.offset; |
187 | datasize = btrfs_file_extent_calc_inline_size(datasize); | |
3326d1b0 CM |
188 | ret = btrfs_insert_empty_item(trans, root, path, &key, |
189 | datasize); | |
190 | if (ret) { | |
191 | err = ret; | |
192 | printk("got bad ret %d\n", ret); | |
193 | goto fail; | |
194 | } | |
195 | leaf = path->nodes[0]; | |
196 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
197 | struct btrfs_file_extent_item); | |
198 | btrfs_set_file_extent_generation(leaf, ei, trans->transid); | |
199 | btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); | |
200 | } | |
18f16f7b | 201 | ptr = btrfs_file_extent_inline_start(ei) + offset - key.offset; |
3326d1b0 CM |
202 | |
203 | cur_size = size; | |
204 | i = 0; | |
205 | while (size > 0) { | |
206 | page = pages[i]; | |
207 | kaddr = kmap_atomic(page, KM_USER0); | |
ae2f5411 | 208 | cur_size = min_t(size_t, PAGE_CACHE_SIZE - page_offset, size); |
3326d1b0 CM |
209 | write_extent_buffer(leaf, kaddr + page_offset, ptr, cur_size); |
210 | kunmap_atomic(kaddr, KM_USER0); | |
211 | page_offset = 0; | |
212 | ptr += cur_size; | |
213 | size -= cur_size; | |
214 | if (i >= num_pages) { | |
215 | printk("i %d num_pages %d\n", i, num_pages); | |
216 | } | |
217 | i++; | |
218 | } | |
5f39d397 | 219 | btrfs_mark_buffer_dirty(leaf); |
54aa1f4d CM |
220 | fail: |
221 | btrfs_free_path(path); | |
54aa1f4d CM |
222 | return err; |
223 | } | |
224 | ||
98ed5174 | 225 | static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, |
39279cc3 CM |
226 | struct btrfs_root *root, |
227 | struct file *file, | |
228 | struct page **pages, | |
229 | size_t num_pages, | |
230 | loff_t pos, | |
231 | size_t write_bytes) | |
232 | { | |
39279cc3 | 233 | int err = 0; |
a52d9a80 | 234 | int i; |
6da6abae | 235 | struct inode *inode = fdentry(file)->d_inode; |
d1310b2e | 236 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_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); |
39279cc3 | 244 | |
5f39d397 | 245 | start_pos = pos & ~((u64)root->sectorsize - 1); |
db94535d CM |
246 | num_bytes = (write_bytes + pos - start_pos + |
247 | root->sectorsize - 1) & ~((u64)root->sectorsize - 1); | |
39279cc3 | 248 | |
db94535d CM |
249 | end_of_last_block = start_pos + num_bytes - 1; |
250 | ||
d1310b2e | 251 | lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); |
a52d9a80 CM |
252 | mutex_lock(&root->fs_info->fs_mutex); |
253 | trans = btrfs_start_transaction(root, 1); | |
254 | if (!trans) { | |
255 | err = -ENOMEM; | |
256 | goto out_unlock; | |
257 | } | |
258 | btrfs_set_trans_block_group(trans, inode); | |
db94535d CM |
259 | inode->i_blocks += num_bytes >> 9; |
260 | hint_byte = 0; | |
a52d9a80 CM |
261 | |
262 | if ((end_of_last_block & 4095) == 0) { | |
9433063b | 263 | printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block); |
a52d9a80 | 264 | } |
d1310b2e | 265 | set_extent_uptodate(io_tree, start_pos, end_of_last_block, GFP_NOFS); |
a52d9a80 CM |
266 | |
267 | /* FIXME...EIEIO, ENOSPC and more */ | |
a52d9a80 | 268 | /* insert any holes we need to create */ |
dacb473a | 269 | if (isize < end_pos) { |
a52d9a80 CM |
270 | u64 last_pos_in_file; |
271 | u64 hole_size; | |
5f39d397 | 272 | u64 mask = root->sectorsize - 1; |
a52d9a80 | 273 | last_pos_in_file = (isize + mask) & ~mask; |
5f56406a | 274 | hole_size = (end_pos - last_pos_in_file + mask) & ~mask; |
dacb473a | 275 | if (last_pos_in_file < end_pos) { |
2bf5a725 CM |
276 | err = btrfs_drop_extents(trans, root, inode, |
277 | last_pos_in_file, | |
278 | last_pos_in_file + hole_size, | |
3326d1b0 | 279 | last_pos_in_file, |
db94535d | 280 | &hint_byte); |
2bf5a725 CM |
281 | if (err) |
282 | goto failed; | |
283 | ||
a52d9a80 CM |
284 | err = btrfs_insert_file_extent(trans, root, |
285 | inode->i_ino, | |
286 | last_pos_in_file, | |
287 | 0, 0, hole_size); | |
d1310b2e CM |
288 | btrfs_drop_extent_cache(inode, last_pos_in_file, |
289 | last_pos_in_file + hole_size -1); | |
5f56406a | 290 | btrfs_check_file(root, inode); |
a52d9a80 CM |
291 | } |
292 | if (err) | |
39279cc3 | 293 | goto failed; |
a52d9a80 CM |
294 | } |
295 | ||
296 | /* | |
297 | * either allocate an extent for the new bytes or setup the key | |
298 | * to show we are doing inline data in the extent | |
299 | */ | |
3326d1b0 CM |
300 | inline_size = end_pos; |
301 | if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) || | |
6f568d35 CM |
302 | inline_size > root->fs_info->max_inline || |
303 | (inline_size & (root->sectorsize -1)) == 0 || | |
3326d1b0 | 304 | inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { |
b888db2b | 305 | u64 last_end; |
1832a6d5 | 306 | u64 existing_delalloc = 0; |
3326d1b0 | 307 | |
a52d9a80 CM |
308 | for (i = 0; i < num_pages; i++) { |
309 | struct page *p = pages[i]; | |
310 | SetPageUptodate(p); | |
b888db2b | 311 | set_page_dirty(p); |
39279cc3 | 312 | } |
35ebb934 CM |
313 | last_end = (u64)(pages[num_pages -1]->index) << |
314 | PAGE_CACHE_SHIFT; | |
b888db2b | 315 | last_end += PAGE_CACHE_SIZE - 1; |
1832a6d5 CM |
316 | if (start_pos < isize) { |
317 | u64 delalloc_start = start_pos; | |
d1310b2e | 318 | existing_delalloc = count_range_bits(io_tree, |
1832a6d5 CM |
319 | &delalloc_start, |
320 | end_of_last_block, (u64)-1, | |
321 | EXTENT_DELALLOC); | |
322 | } | |
d1310b2e | 323 | set_extent_delalloc(io_tree, start_pos, end_of_last_block, |
b888db2b | 324 | GFP_NOFS); |
dc17ff8f | 325 | btrfs_add_ordered_inode(inode); |
a52d9a80 | 326 | } else { |
3326d1b0 | 327 | u64 aligned_end; |
b888db2b | 328 | /* step one, delete the existing extents in this range */ |
3326d1b0 CM |
329 | aligned_end = (pos + write_bytes + root->sectorsize - 1) & |
330 | ~((u64)root->sectorsize - 1); | |
2bf5a725 | 331 | err = btrfs_drop_extents(trans, root, inode, start_pos, |
179e29e4 | 332 | aligned_end, aligned_end, &hint_byte); |
2bf5a725 CM |
333 | if (err) |
334 | goto failed; | |
dcfec0dc Y |
335 | if (isize > inline_size) |
336 | inline_size = min_t(u64, isize, aligned_end); | |
337 | inline_size -= start_pos; | |
a52d9a80 | 338 | err = insert_inline_extent(trans, root, inode, start_pos, |
dcfec0dc | 339 | inline_size, pages, 0, num_pages); |
d1310b2e | 340 | btrfs_drop_extent_cache(inode, start_pos, aligned_end - 1); |
a52d9a80 | 341 | BUG_ON(err); |
a52d9a80 CM |
342 | } |
343 | if (end_pos > isize) { | |
344 | i_size_write(inode, end_pos); | |
345 | btrfs_update_inode(trans, root, inode); | |
39279cc3 CM |
346 | } |
347 | failed: | |
a52d9a80 CM |
348 | err = btrfs_end_transaction(trans, root); |
349 | out_unlock: | |
350 | mutex_unlock(&root->fs_info->fs_mutex); | |
d1310b2e | 351 | unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); |
39279cc3 CM |
352 | return err; |
353 | } | |
354 | ||
a52d9a80 CM |
355 | int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) |
356 | { | |
357 | struct extent_map *em; | |
358 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
359 | ||
360 | while(1) { | |
d1310b2e | 361 | spin_lock(&em_tree->lock); |
a52d9a80 | 362 | em = lookup_extent_mapping(em_tree, start, end); |
d1310b2e CM |
363 | if (!em) { |
364 | spin_unlock(&em_tree->lock); | |
a52d9a80 | 365 | break; |
d1310b2e | 366 | } |
a52d9a80 | 367 | remove_extent_mapping(em_tree, em); |
d1310b2e CM |
368 | spin_unlock(&em_tree->lock); |
369 | ||
a52d9a80 CM |
370 | /* once for us */ |
371 | free_extent_map(em); | |
372 | /* once for the tree*/ | |
373 | free_extent_map(em); | |
374 | } | |
375 | return 0; | |
376 | } | |
377 | ||
5f56406a CM |
378 | int btrfs_check_file(struct btrfs_root *root, struct inode *inode) |
379 | { | |
380 | return 0; | |
381 | #if 0 | |
382 | struct btrfs_path *path; | |
383 | struct btrfs_key found_key; | |
384 | struct extent_buffer *leaf; | |
385 | struct btrfs_file_extent_item *extent; | |
386 | u64 last_offset = 0; | |
387 | int nritems; | |
388 | int slot; | |
389 | int found_type; | |
390 | int ret; | |
391 | int err = 0; | |
392 | u64 extent_end = 0; | |
393 | ||
394 | path = btrfs_alloc_path(); | |
395 | ret = btrfs_lookup_file_extent(NULL, root, path, inode->i_ino, | |
396 | last_offset, 0); | |
397 | while(1) { | |
398 | nritems = btrfs_header_nritems(path->nodes[0]); | |
399 | if (path->slots[0] >= nritems) { | |
400 | ret = btrfs_next_leaf(root, path); | |
401 | if (ret) | |
402 | goto out; | |
403 | nritems = btrfs_header_nritems(path->nodes[0]); | |
404 | } | |
405 | slot = path->slots[0]; | |
406 | leaf = path->nodes[0]; | |
407 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
408 | if (found_key.objectid != inode->i_ino) | |
409 | break; | |
410 | if (found_key.type != BTRFS_EXTENT_DATA_KEY) | |
411 | goto out; | |
412 | ||
413 | if (found_key.offset != last_offset) { | |
414 | WARN_ON(1); | |
415 | btrfs_print_leaf(root, leaf); | |
416 | printk("inode %lu found offset %Lu expected %Lu\n", | |
417 | inode->i_ino, found_key.offset, last_offset); | |
418 | err = 1; | |
419 | goto out; | |
420 | } | |
421 | extent = btrfs_item_ptr(leaf, slot, | |
422 | struct btrfs_file_extent_item); | |
423 | found_type = btrfs_file_extent_type(leaf, extent); | |
424 | if (found_type == BTRFS_FILE_EXTENT_REG) { | |
425 | extent_end = found_key.offset + | |
426 | btrfs_file_extent_num_bytes(leaf, extent); | |
427 | } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { | |
428 | struct btrfs_item *item; | |
429 | item = btrfs_item_nr(leaf, slot); | |
430 | extent_end = found_key.offset + | |
431 | btrfs_file_extent_inline_len(leaf, item); | |
432 | extent_end = (extent_end + root->sectorsize - 1) & | |
433 | ~((u64)root->sectorsize -1 ); | |
434 | } | |
435 | last_offset = extent_end; | |
436 | path->slots[0]++; | |
437 | } | |
438 | if (last_offset < inode->i_size) { | |
439 | WARN_ON(1); | |
440 | btrfs_print_leaf(root, leaf); | |
441 | printk("inode %lu found offset %Lu size %Lu\n", inode->i_ino, | |
442 | last_offset, inode->i_size); | |
443 | err = 1; | |
444 | ||
445 | } | |
446 | out: | |
447 | btrfs_free_path(path); | |
448 | return err; | |
449 | #endif | |
450 | } | |
451 | ||
39279cc3 CM |
452 | /* |
453 | * this is very complex, but the basic idea is to drop all extents | |
454 | * in the range start - end. hint_block is filled in with a block number | |
455 | * that would be a good hint to the block allocator for this file. | |
456 | * | |
457 | * If an extent intersects the range but is not entirely inside the range | |
458 | * it is either truncated or split. Anything entirely inside the range | |
459 | * is deleted from the tree. | |
460 | */ | |
461 | int btrfs_drop_extents(struct btrfs_trans_handle *trans, | |
462 | struct btrfs_root *root, struct inode *inode, | |
00f5c795 | 463 | u64 start, u64 end, u64 inline_limit, u64 *hint_byte) |
39279cc3 | 464 | { |
00f5c795 CM |
465 | u64 extent_end = 0; |
466 | u64 search_start = start; | |
5f39d397 | 467 | struct extent_buffer *leaf; |
39279cc3 | 468 | struct btrfs_file_extent_item *extent; |
39279cc3 | 469 | struct btrfs_path *path; |
00f5c795 CM |
470 | struct btrfs_key key; |
471 | struct btrfs_file_extent_item old; | |
472 | int keep; | |
473 | int slot; | |
39279cc3 CM |
474 | int bookend; |
475 | int found_type; | |
476 | int found_extent; | |
477 | int found_inline; | |
ccd467d6 | 478 | int recow; |
00f5c795 | 479 | int ret; |
39279cc3 | 480 | |
a52d9a80 CM |
481 | btrfs_drop_extent_cache(inode, start, end - 1); |
482 | ||
39279cc3 CM |
483 | path = btrfs_alloc_path(); |
484 | if (!path) | |
485 | return -ENOMEM; | |
486 | while(1) { | |
ccd467d6 | 487 | recow = 0; |
39279cc3 CM |
488 | btrfs_release_path(root, path); |
489 | ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino, | |
490 | search_start, -1); | |
491 | if (ret < 0) | |
492 | goto out; | |
493 | if (ret > 0) { | |
494 | if (path->slots[0] == 0) { | |
495 | ret = 0; | |
496 | goto out; | |
497 | } | |
498 | path->slots[0]--; | |
499 | } | |
8c2383c3 | 500 | next_slot: |
39279cc3 CM |
501 | keep = 0; |
502 | bookend = 0; | |
503 | found_extent = 0; | |
504 | found_inline = 0; | |
505 | extent = NULL; | |
5f39d397 | 506 | leaf = path->nodes[0]; |
39279cc3 | 507 | slot = path->slots[0]; |
8c2383c3 | 508 | ret = 0; |
5f39d397 | 509 | btrfs_item_key_to_cpu(leaf, &key, slot); |
5f56406a | 510 | |
39279cc3 | 511 | if (key.offset >= end || key.objectid != inode->i_ino) { |
39279cc3 CM |
512 | goto out; |
513 | } | |
8c2383c3 | 514 | if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY) { |
39279cc3 CM |
515 | goto out; |
516 | } | |
ccd467d6 CM |
517 | if (recow) { |
518 | search_start = key.offset; | |
519 | continue; | |
520 | } | |
8c2383c3 CM |
521 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { |
522 | extent = btrfs_item_ptr(leaf, slot, | |
523 | struct btrfs_file_extent_item); | |
5f39d397 | 524 | found_type = btrfs_file_extent_type(leaf, extent); |
8c2383c3 | 525 | if (found_type == BTRFS_FILE_EXTENT_REG) { |
257d0ce3 CM |
526 | extent_end = |
527 | btrfs_file_extent_disk_bytenr(leaf, | |
528 | extent); | |
529 | if (extent_end) | |
530 | *hint_byte = extent_end; | |
531 | ||
8c2383c3 | 532 | extent_end = key.offset + |
db94535d | 533 | btrfs_file_extent_num_bytes(leaf, extent); |
8c2383c3 CM |
534 | found_extent = 1; |
535 | } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { | |
5f39d397 CM |
536 | struct btrfs_item *item; |
537 | item = btrfs_item_nr(leaf, slot); | |
8c2383c3 CM |
538 | found_inline = 1; |
539 | extent_end = key.offset + | |
5f39d397 | 540 | btrfs_file_extent_inline_len(leaf, item); |
8c2383c3 CM |
541 | } |
542 | } else { | |
543 | extent_end = search_start; | |
39279cc3 CM |
544 | } |
545 | ||
546 | /* we found nothing we can drop */ | |
8c2383c3 CM |
547 | if ((!found_extent && !found_inline) || |
548 | search_start >= extent_end) { | |
549 | int nextret; | |
550 | u32 nritems; | |
5f39d397 | 551 | nritems = btrfs_header_nritems(leaf); |
8c2383c3 CM |
552 | if (slot >= nritems - 1) { |
553 | nextret = btrfs_next_leaf(root, path); | |
554 | if (nextret) | |
555 | goto out; | |
ccd467d6 | 556 | recow = 1; |
8c2383c3 CM |
557 | } else { |
558 | path->slots[0]++; | |
559 | } | |
560 | goto next_slot; | |
39279cc3 CM |
561 | } |
562 | ||
39279cc3 | 563 | if (found_inline) { |
5f39d397 | 564 | u64 mask = root->sectorsize - 1; |
39279cc3 CM |
565 | search_start = (extent_end + mask) & ~mask; |
566 | } else | |
567 | search_start = extent_end; | |
6e3b9666 | 568 | if (end <= extent_end && start >= key.offset && found_inline) { |
179e29e4 | 569 | *hint_byte = EXTENT_MAP_INLINE; |
6e3b9666 | 570 | continue; |
179e29e4 | 571 | } |
39279cc3 CM |
572 | if (end < extent_end && end >= key.offset) { |
573 | if (found_extent) { | |
db94535d CM |
574 | u64 disk_bytenr = |
575 | btrfs_file_extent_disk_bytenr(leaf, extent); | |
576 | u64 disk_num_bytes = | |
577 | btrfs_file_extent_disk_num_bytes(leaf, | |
5f39d397 CM |
578 | extent); |
579 | read_extent_buffer(leaf, &old, | |
580 | (unsigned long)extent, | |
581 | sizeof(old)); | |
db94535d | 582 | if (disk_bytenr != 0) { |
39279cc3 | 583 | ret = btrfs_inc_extent_ref(trans, root, |
7bb86316 CM |
584 | disk_bytenr, disk_num_bytes, |
585 | root->root_key.objectid, | |
586 | trans->transid, | |
587 | key.objectid, end); | |
39279cc3 CM |
588 | BUG_ON(ret); |
589 | } | |
590 | } | |
179e29e4 | 591 | bookend = 1; |
0181e58f | 592 | if (found_inline && start <= key.offset) |
179e29e4 | 593 | keep = 1; |
39279cc3 | 594 | } |
39279cc3 CM |
595 | /* truncate existing extent */ |
596 | if (start > key.offset) { | |
597 | u64 new_num; | |
598 | u64 old_num; | |
599 | keep = 1; | |
5f39d397 | 600 | WARN_ON(start & (root->sectorsize - 1)); |
39279cc3 | 601 | if (found_extent) { |
db94535d CM |
602 | new_num = start - key.offset; |
603 | old_num = btrfs_file_extent_num_bytes(leaf, | |
604 | extent); | |
605 | *hint_byte = | |
606 | btrfs_file_extent_disk_bytenr(leaf, | |
607 | extent); | |
608 | if (btrfs_file_extent_disk_bytenr(leaf, | |
609 | extent)) { | |
39279cc3 | 610 | inode->i_blocks -= |
db94535d | 611 | (old_num - new_num) >> 9; |
39279cc3 | 612 | } |
db94535d CM |
613 | btrfs_set_file_extent_num_bytes(leaf, extent, |
614 | new_num); | |
5f39d397 | 615 | btrfs_mark_buffer_dirty(leaf); |
00f5c795 CM |
616 | } else if (key.offset < inline_limit && |
617 | (end > extent_end) && | |
618 | (inline_limit < extent_end)) { | |
3326d1b0 CM |
619 | u32 new_size; |
620 | new_size = btrfs_file_extent_calc_inline_size( | |
00f5c795 | 621 | inline_limit - key.offset); |
3326d1b0 | 622 | btrfs_truncate_item(trans, root, path, |
179e29e4 | 623 | new_size, 1); |
39279cc3 CM |
624 | } |
625 | } | |
626 | /* delete the entire extent */ | |
627 | if (!keep) { | |
db94535d CM |
628 | u64 disk_bytenr = 0; |
629 | u64 disk_num_bytes = 0; | |
630 | u64 extent_num_bytes = 0; | |
7bb86316 | 631 | u64 root_gen; |
d8d5f3e1 | 632 | u64 root_owner; |
7bb86316 | 633 | |
d8d5f3e1 CM |
634 | root_gen = btrfs_header_generation(leaf); |
635 | root_owner = btrfs_header_owner(leaf); | |
39279cc3 | 636 | if (found_extent) { |
db94535d CM |
637 | disk_bytenr = |
638 | btrfs_file_extent_disk_bytenr(leaf, | |
5f39d397 | 639 | extent); |
db94535d CM |
640 | disk_num_bytes = |
641 | btrfs_file_extent_disk_num_bytes(leaf, | |
5f39d397 | 642 | extent); |
db94535d CM |
643 | extent_num_bytes = |
644 | btrfs_file_extent_num_bytes(leaf, extent); | |
645 | *hint_byte = | |
646 | btrfs_file_extent_disk_bytenr(leaf, | |
647 | extent); | |
39279cc3 CM |
648 | } |
649 | ret = btrfs_del_item(trans, root, path); | |
54aa1f4d | 650 | /* TODO update progress marker and return */ |
39279cc3 CM |
651 | BUG_ON(ret); |
652 | btrfs_release_path(root, path); | |
653 | extent = NULL; | |
db94535d CM |
654 | if (found_extent && disk_bytenr != 0) { |
655 | inode->i_blocks -= extent_num_bytes >> 9; | |
39279cc3 | 656 | ret = btrfs_free_extent(trans, root, |
7bb86316 CM |
657 | disk_bytenr, |
658 | disk_num_bytes, | |
d8d5f3e1 | 659 | root_owner, |
7bb86316 CM |
660 | root_gen, inode->i_ino, |
661 | key.offset, 0); | |
39279cc3 CM |
662 | } |
663 | ||
664 | BUG_ON(ret); | |
665 | if (!bookend && search_start >= end) { | |
666 | ret = 0; | |
667 | goto out; | |
668 | } | |
669 | if (!bookend) | |
670 | continue; | |
671 | } | |
0181e58f | 672 | if (bookend && found_inline && start <= key.offset) { |
179e29e4 CM |
673 | u32 new_size; |
674 | new_size = btrfs_file_extent_calc_inline_size( | |
0181e58f | 675 | extent_end - end); |
179e29e4 CM |
676 | btrfs_truncate_item(trans, root, path, new_size, 0); |
677 | } | |
39279cc3 CM |
678 | /* create bookend, splitting the extent in two */ |
679 | if (bookend && found_extent) { | |
680 | struct btrfs_key ins; | |
681 | ins.objectid = inode->i_ino; | |
682 | ins.offset = end; | |
39279cc3 | 683 | btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY); |
39279cc3 CM |
684 | btrfs_release_path(root, path); |
685 | ret = btrfs_insert_empty_item(trans, root, path, &ins, | |
686 | sizeof(*extent)); | |
8c2383c3 | 687 | |
5f39d397 | 688 | leaf = path->nodes[0]; |
8c2383c3 | 689 | if (ret) { |
5f39d397 CM |
690 | btrfs_print_leaf(root, leaf); |
691 | 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 | 692 | } |
39279cc3 | 693 | BUG_ON(ret); |
5f39d397 CM |
694 | extent = btrfs_item_ptr(leaf, path->slots[0], |
695 | struct btrfs_file_extent_item); | |
696 | write_extent_buffer(leaf, &old, | |
697 | (unsigned long)extent, sizeof(old)); | |
698 | ||
699 | btrfs_set_file_extent_offset(leaf, extent, | |
db94535d CM |
700 | le64_to_cpu(old.offset) + end - key.offset); |
701 | WARN_ON(le64_to_cpu(old.num_bytes) < | |
702 | (extent_end - end)); | |
703 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
704 | extent_end - end); | |
5f39d397 | 705 | btrfs_set_file_extent_type(leaf, extent, |
39279cc3 | 706 | BTRFS_FILE_EXTENT_REG); |
db94535d | 707 | |
39279cc3 | 708 | btrfs_mark_buffer_dirty(path->nodes[0]); |
db94535d | 709 | if (le64_to_cpu(old.disk_bytenr) != 0) { |
39279cc3 | 710 | inode->i_blocks += |
db94535d CM |
711 | btrfs_file_extent_num_bytes(leaf, |
712 | extent) >> 9; | |
39279cc3 CM |
713 | } |
714 | ret = 0; | |
715 | goto out; | |
716 | } | |
717 | } | |
718 | out: | |
719 | btrfs_free_path(path); | |
720 | return ret; | |
721 | } | |
722 | ||
723 | /* | |
724 | * this gets pages into the page cache and locks them down | |
725 | */ | |
98ed5174 CM |
726 | static int prepare_pages(struct btrfs_root *root, struct file *file, |
727 | struct page **pages, size_t num_pages, | |
728 | loff_t pos, unsigned long first_index, | |
729 | unsigned long last_index, size_t write_bytes) | |
39279cc3 CM |
730 | { |
731 | int i; | |
732 | unsigned long index = pos >> PAGE_CACHE_SHIFT; | |
6da6abae | 733 | struct inode *inode = fdentry(file)->d_inode; |
39279cc3 | 734 | int err = 0; |
8c2383c3 | 735 | u64 start_pos; |
8c2383c3 | 736 | |
5f39d397 | 737 | start_pos = pos & ~((u64)root->sectorsize - 1); |
39279cc3 CM |
738 | |
739 | memset(pages, 0, num_pages * sizeof(struct page *)); | |
740 | ||
741 | for (i = 0; i < num_pages; i++) { | |
742 | pages[i] = grab_cache_page(inode->i_mapping, index + i); | |
743 | if (!pages[i]) { | |
744 | err = -ENOMEM; | |
a52d9a80 | 745 | BUG_ON(1); |
39279cc3 | 746 | } |
6da6abae CM |
747 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
748 | ClearPageDirty(pages[i]); | |
749 | #else | |
ccd467d6 | 750 | cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); |
6da6abae | 751 | #endif |
ccd467d6 | 752 | wait_on_page_writeback(pages[i]); |
b3cfa35a | 753 | set_page_extent_mapped(pages[i]); |
b888db2b | 754 | WARN_ON(!PageLocked(pages[i])); |
39279cc3 CM |
755 | } |
756 | return 0; | |
39279cc3 CM |
757 | } |
758 | ||
759 | static ssize_t btrfs_file_write(struct file *file, const char __user *buf, | |
760 | size_t count, loff_t *ppos) | |
761 | { | |
762 | loff_t pos; | |
2ff3e9b6 CM |
763 | loff_t start_pos; |
764 | ssize_t num_written = 0; | |
765 | ssize_t err = 0; | |
39279cc3 | 766 | int ret = 0; |
6da6abae | 767 | struct inode *inode = fdentry(file)->d_inode; |
39279cc3 | 768 | struct btrfs_root *root = BTRFS_I(inode)->root; |
8c2383c3 CM |
769 | struct page **pages = NULL; |
770 | int nrptrs; | |
39279cc3 CM |
771 | struct page *pinned[2]; |
772 | unsigned long first_index; | |
773 | unsigned long last_index; | |
8c2383c3 CM |
774 | |
775 | nrptrs = min((count + PAGE_CACHE_SIZE - 1) / PAGE_CACHE_SIZE, | |
776 | PAGE_CACHE_SIZE / (sizeof(struct page *))); | |
39279cc3 CM |
777 | pinned[0] = NULL; |
778 | pinned[1] = NULL; | |
779 | if (file->f_flags & O_DIRECT) | |
780 | return -EINVAL; | |
2ff3e9b6 | 781 | |
39279cc3 | 782 | pos = *ppos; |
2ff3e9b6 CM |
783 | start_pos = pos; |
784 | ||
39279cc3 CM |
785 | vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE); |
786 | current->backing_dev_info = inode->i_mapping->backing_dev_info; | |
787 | err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode)); | |
788 | if (err) | |
1832a6d5 | 789 | goto out_nolock; |
39279cc3 | 790 | if (count == 0) |
1832a6d5 | 791 | goto out_nolock; |
6da6abae | 792 | err = remove_suid(fdentry(file)); |
39279cc3 | 793 | if (err) |
1832a6d5 | 794 | goto out_nolock; |
39279cc3 CM |
795 | file_update_time(file); |
796 | ||
8c2383c3 | 797 | pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL); |
39279cc3 CM |
798 | |
799 | mutex_lock(&inode->i_mutex); | |
800 | first_index = pos >> PAGE_CACHE_SHIFT; | |
801 | last_index = (pos + count) >> PAGE_CACHE_SHIFT; | |
802 | ||
803 | /* | |
804 | * there are lots of better ways to do this, but this code | |
805 | * makes sure the first and last page in the file range are | |
806 | * up to date and ready for cow | |
807 | */ | |
808 | if ((pos & (PAGE_CACHE_SIZE - 1))) { | |
809 | pinned[0] = grab_cache_page(inode->i_mapping, first_index); | |
810 | if (!PageUptodate(pinned[0])) { | |
9ebefb18 | 811 | ret = btrfs_readpage(NULL, pinned[0]); |
39279cc3 CM |
812 | BUG_ON(ret); |
813 | wait_on_page_locked(pinned[0]); | |
814 | } else { | |
815 | unlock_page(pinned[0]); | |
816 | } | |
817 | } | |
818 | if ((pos + count) & (PAGE_CACHE_SIZE - 1)) { | |
819 | pinned[1] = grab_cache_page(inode->i_mapping, last_index); | |
820 | if (!PageUptodate(pinned[1])) { | |
9ebefb18 | 821 | ret = btrfs_readpage(NULL, pinned[1]); |
39279cc3 CM |
822 | BUG_ON(ret); |
823 | wait_on_page_locked(pinned[1]); | |
824 | } else { | |
825 | unlock_page(pinned[1]); | |
826 | } | |
827 | } | |
828 | ||
39279cc3 CM |
829 | while(count > 0) { |
830 | size_t offset = pos & (PAGE_CACHE_SIZE - 1); | |
11bd143f CM |
831 | size_t write_bytes = min(count, nrptrs * |
832 | (size_t)PAGE_CACHE_SIZE - | |
8c2383c3 | 833 | offset); |
39279cc3 CM |
834 | size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >> |
835 | PAGE_CACHE_SHIFT; | |
836 | ||
8c2383c3 | 837 | WARN_ON(num_pages > nrptrs); |
39279cc3 | 838 | memset(pages, 0, sizeof(pages)); |
1832a6d5 CM |
839 | |
840 | mutex_lock(&root->fs_info->fs_mutex); | |
841 | ret = btrfs_check_free_space(root, write_bytes, 0); | |
842 | mutex_unlock(&root->fs_info->fs_mutex); | |
843 | if (ret) | |
844 | goto out; | |
845 | ||
39279cc3 CM |
846 | ret = prepare_pages(root, file, pages, num_pages, |
847 | pos, first_index, last_index, | |
8c2383c3 | 848 | write_bytes); |
54aa1f4d CM |
849 | if (ret) |
850 | goto out; | |
39279cc3 | 851 | |
39279cc3 CM |
852 | ret = btrfs_copy_from_user(pos, num_pages, |
853 | write_bytes, pages, buf); | |
54aa1f4d CM |
854 | if (ret) { |
855 | btrfs_drop_pages(pages, num_pages); | |
856 | goto out; | |
857 | } | |
39279cc3 CM |
858 | |
859 | ret = dirty_and_release_pages(NULL, root, file, pages, | |
860 | num_pages, pos, write_bytes); | |
39279cc3 | 861 | btrfs_drop_pages(pages, num_pages); |
54aa1f4d CM |
862 | if (ret) |
863 | goto out; | |
39279cc3 CM |
864 | |
865 | buf += write_bytes; | |
866 | count -= write_bytes; | |
867 | pos += write_bytes; | |
868 | num_written += write_bytes; | |
869 | ||
8c2383c3 | 870 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages); |
448d640b CM |
871 | if (num_pages < (root->leafsize >> PAGE_CACHE_SHIFT) + 1) |
872 | btrfs_btree_balance_dirty(root, 1); | |
e2008b61 | 873 | btrfs_throttle(root); |
39279cc3 CM |
874 | cond_resched(); |
875 | } | |
39279cc3 | 876 | out: |
1832a6d5 | 877 | mutex_unlock(&inode->i_mutex); |
5b92ee72 | 878 | |
1832a6d5 | 879 | out_nolock: |
8c2383c3 | 880 | kfree(pages); |
39279cc3 CM |
881 | if (pinned[0]) |
882 | page_cache_release(pinned[0]); | |
883 | if (pinned[1]) | |
884 | page_cache_release(pinned[1]); | |
885 | *ppos = pos; | |
2ff3e9b6 CM |
886 | |
887 | if (num_written > 0 && ((file->f_flags & O_SYNC) || IS_SYNC(inode))) { | |
888 | err = sync_page_range(inode, inode->i_mapping, | |
889 | start_pos, num_written); | |
890 | if (err < 0) | |
891 | num_written = err; | |
892 | } | |
39279cc3 | 893 | current->backing_dev_info = NULL; |
39279cc3 CM |
894 | return num_written ? num_written : err; |
895 | } | |
896 | ||
39279cc3 CM |
897 | static int btrfs_sync_file(struct file *file, |
898 | struct dentry *dentry, int datasync) | |
899 | { | |
900 | struct inode *inode = dentry->d_inode; | |
901 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
15ee9bc7 | 902 | int ret = 0; |
39279cc3 CM |
903 | struct btrfs_trans_handle *trans; |
904 | ||
905 | /* | |
15ee9bc7 JB |
906 | * check the transaction that last modified this inode |
907 | * and see if its already been committed | |
39279cc3 CM |
908 | */ |
909 | mutex_lock(&root->fs_info->fs_mutex); | |
15ee9bc7 JB |
910 | if (!BTRFS_I(inode)->last_trans) |
911 | goto out; | |
912 | mutex_lock(&root->fs_info->trans_mutex); | |
913 | if (BTRFS_I(inode)->last_trans <= | |
914 | root->fs_info->last_trans_committed) { | |
915 | BTRFS_I(inode)->last_trans = 0; | |
916 | mutex_unlock(&root->fs_info->trans_mutex); | |
917 | goto out; | |
918 | } | |
919 | mutex_unlock(&root->fs_info->trans_mutex); | |
920 | ||
921 | /* | |
a52d9a80 CM |
922 | * ok we haven't committed the transaction yet, lets do a commit |
923 | */ | |
39279cc3 CM |
924 | trans = btrfs_start_transaction(root, 1); |
925 | if (!trans) { | |
926 | ret = -ENOMEM; | |
927 | goto out; | |
928 | } | |
929 | ret = btrfs_commit_transaction(trans, root); | |
39279cc3 | 930 | out: |
15ee9bc7 | 931 | mutex_unlock(&root->fs_info->fs_mutex); |
39279cc3 CM |
932 | return ret > 0 ? EIO : ret; |
933 | } | |
934 | ||
9ebefb18 | 935 | static struct vm_operations_struct btrfs_file_vm_ops = { |
92fee66d CM |
936 | #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) |
937 | .nopage = filemap_nopage, | |
938 | .populate = filemap_populate, | |
939 | #else | |
940 | .fault = filemap_fault, | |
941 | #endif | |
9ebefb18 CM |
942 | .page_mkwrite = btrfs_page_mkwrite, |
943 | }; | |
944 | ||
945 | static int btrfs_file_mmap(struct file *filp, struct vm_area_struct *vma) | |
946 | { | |
947 | vma->vm_ops = &btrfs_file_vm_ops; | |
948 | file_accessed(filp); | |
949 | return 0; | |
950 | } | |
951 | ||
39279cc3 CM |
952 | struct file_operations btrfs_file_operations = { |
953 | .llseek = generic_file_llseek, | |
954 | .read = do_sync_read, | |
9ebefb18 | 955 | .aio_read = generic_file_aio_read, |
e9906a98 | 956 | .splice_read = generic_file_splice_read, |
6da6abae CM |
957 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
958 | .sendfile = generic_file_sendfile, | |
959 | #endif | |
39279cc3 | 960 | .write = btrfs_file_write, |
9ebefb18 | 961 | .mmap = btrfs_file_mmap, |
39279cc3 | 962 | .open = generic_file_open, |
39279cc3 | 963 | .fsync = btrfs_sync_file, |
34287aa3 | 964 | .unlocked_ioctl = btrfs_ioctl, |
39279cc3 | 965 | #ifdef CONFIG_COMPAT |
34287aa3 | 966 | .compat_ioctl = btrfs_ioctl, |
39279cc3 CM |
967 | #endif |
968 | }; | |
969 |