]>
Commit | Line | Data |
---|---|---|
a86c6181 AT |
1 | /* |
2 | * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com | |
3 | * Written by Alex Tomas <alex@clusterfs.com> | |
4 | * | |
5 | * Architecture independence: | |
6 | * Copyright (c) 2005, Bull S.A. | |
7 | * Written by Pierre Peiffer <pierre.peiffer@bull.net> | |
8 | * | |
9 | * This program is free software; you can redistribute it and/or modify | |
10 | * it under the terms of the GNU General Public License version 2 as | |
11 | * published by the Free Software Foundation. | |
12 | * | |
13 | * This program is distributed in the hope that it will be useful, | |
14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | * GNU General Public License for more details. | |
17 | * | |
18 | * You should have received a copy of the GNU General Public Licens | |
19 | * along with this program; if not, write to the Free Software | |
20 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111- | |
21 | */ | |
22 | ||
23 | /* | |
24 | * Extents support for EXT4 | |
25 | * | |
26 | * TODO: | |
27 | * - ext4*_error() should be used in some situations | |
28 | * - analyze all BUG()/BUG_ON(), use -EIO where appropriate | |
29 | * - smart tree reduction | |
30 | */ | |
31 | ||
32 | #include <linux/module.h> | |
33 | #include <linux/fs.h> | |
34 | #include <linux/time.h> | |
cd02ff0b | 35 | #include <linux/jbd2.h> |
a86c6181 AT |
36 | #include <linux/highuid.h> |
37 | #include <linux/pagemap.h> | |
38 | #include <linux/quotaops.h> | |
39 | #include <linux/string.h> | |
40 | #include <linux/slab.h> | |
a2df2a63 | 41 | #include <linux/falloc.h> |
a86c6181 | 42 | #include <asm/uaccess.h> |
6873fa0d | 43 | #include <linux/fiemap.h> |
3dcf5451 CH |
44 | #include "ext4_jbd2.h" |
45 | #include "ext4_extents.h" | |
a86c6181 AT |
46 | |
47 | ||
d0d856e8 RD |
48 | /* |
49 | * ext_pblock: | |
50 | * combine low and high parts of physical block number into ext4_fsblk_t | |
51 | */ | |
748de673 | 52 | ext4_fsblk_t ext_pblock(struct ext4_extent *ex) |
f65e6fba AT |
53 | { |
54 | ext4_fsblk_t block; | |
55 | ||
b377611d | 56 | block = le32_to_cpu(ex->ee_start_lo); |
9b8f1f01 | 57 | block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1; |
f65e6fba AT |
58 | return block; |
59 | } | |
60 | ||
d0d856e8 RD |
61 | /* |
62 | * idx_pblock: | |
63 | * combine low and high parts of a leaf physical block number into ext4_fsblk_t | |
64 | */ | |
c14c6fd5 | 65 | ext4_fsblk_t idx_pblock(struct ext4_extent_idx *ix) |
f65e6fba AT |
66 | { |
67 | ext4_fsblk_t block; | |
68 | ||
d8dd0b45 | 69 | block = le32_to_cpu(ix->ei_leaf_lo); |
9b8f1f01 | 70 | block |= ((ext4_fsblk_t) le16_to_cpu(ix->ei_leaf_hi) << 31) << 1; |
f65e6fba AT |
71 | return block; |
72 | } | |
73 | ||
d0d856e8 RD |
74 | /* |
75 | * ext4_ext_store_pblock: | |
76 | * stores a large physical block number into an extent struct, | |
77 | * breaking it into parts | |
78 | */ | |
c14c6fd5 | 79 | void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb) |
f65e6fba | 80 | { |
b377611d | 81 | ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff)); |
9b8f1f01 | 82 | ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff); |
f65e6fba AT |
83 | } |
84 | ||
d0d856e8 RD |
85 | /* |
86 | * ext4_idx_store_pblock: | |
87 | * stores a large physical block number into an index struct, | |
88 | * breaking it into parts | |
89 | */ | |
09b88252 | 90 | static void ext4_idx_store_pblock(struct ext4_extent_idx *ix, ext4_fsblk_t pb) |
f65e6fba | 91 | { |
d8dd0b45 | 92 | ix->ei_leaf_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff)); |
9b8f1f01 | 93 | ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff); |
f65e6fba AT |
94 | } |
95 | ||
487caeef JK |
96 | static int ext4_ext_truncate_extend_restart(handle_t *handle, |
97 | struct inode *inode, | |
98 | int needed) | |
a86c6181 AT |
99 | { |
100 | int err; | |
101 | ||
0390131b FM |
102 | if (!ext4_handle_valid(handle)) |
103 | return 0; | |
a86c6181 | 104 | if (handle->h_buffer_credits > needed) |
9102e4fa SF |
105 | return 0; |
106 | err = ext4_journal_extend(handle, needed); | |
0123c939 | 107 | if (err <= 0) |
9102e4fa | 108 | return err; |
487caeef JK |
109 | err = ext4_truncate_restart_trans(handle, inode, needed); |
110 | /* | |
111 | * We have dropped i_data_sem so someone might have cached again | |
112 | * an extent we are going to truncate. | |
113 | */ | |
114 | ext4_ext_invalidate_cache(inode); | |
115 | ||
116 | return err; | |
a86c6181 AT |
117 | } |
118 | ||
119 | /* | |
120 | * could return: | |
121 | * - EROFS | |
122 | * - ENOMEM | |
123 | */ | |
124 | static int ext4_ext_get_access(handle_t *handle, struct inode *inode, | |
125 | struct ext4_ext_path *path) | |
126 | { | |
127 | if (path->p_bh) { | |
128 | /* path points to block */ | |
129 | return ext4_journal_get_write_access(handle, path->p_bh); | |
130 | } | |
131 | /* path points to leaf/index in inode body */ | |
132 | /* we use in-core data, no need to protect them */ | |
133 | return 0; | |
134 | } | |
135 | ||
136 | /* | |
137 | * could return: | |
138 | * - EROFS | |
139 | * - ENOMEM | |
140 | * - EIO | |
141 | */ | |
142 | static int ext4_ext_dirty(handle_t *handle, struct inode *inode, | |
143 | struct ext4_ext_path *path) | |
144 | { | |
145 | int err; | |
146 | if (path->p_bh) { | |
147 | /* path points to block */ | |
0390131b | 148 | err = ext4_handle_dirty_metadata(handle, inode, path->p_bh); |
a86c6181 AT |
149 | } else { |
150 | /* path points to leaf/index in inode body */ | |
151 | err = ext4_mark_inode_dirty(handle, inode); | |
152 | } | |
153 | return err; | |
154 | } | |
155 | ||
f65e6fba | 156 | static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode, |
a86c6181 | 157 | struct ext4_ext_path *path, |
725d26d3 | 158 | ext4_lblk_t block) |
a86c6181 AT |
159 | { |
160 | struct ext4_inode_info *ei = EXT4_I(inode); | |
f65e6fba | 161 | ext4_fsblk_t bg_start; |
74d3487f | 162 | ext4_fsblk_t last_block; |
f65e6fba | 163 | ext4_grpblk_t colour; |
a4912123 TT |
164 | ext4_group_t block_group; |
165 | int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb)); | |
a86c6181 AT |
166 | int depth; |
167 | ||
168 | if (path) { | |
169 | struct ext4_extent *ex; | |
170 | depth = path->p_depth; | |
171 | ||
172 | /* try to predict block placement */ | |
7e028976 AM |
173 | ex = path[depth].p_ext; |
174 | if (ex) | |
f65e6fba | 175 | return ext_pblock(ex)+(block-le32_to_cpu(ex->ee_block)); |
a86c6181 | 176 | |
d0d856e8 RD |
177 | /* it looks like index is empty; |
178 | * try to find starting block from index itself */ | |
a86c6181 AT |
179 | if (path[depth].p_bh) |
180 | return path[depth].p_bh->b_blocknr; | |
181 | } | |
182 | ||
183 | /* OK. use inode's group */ | |
a4912123 TT |
184 | block_group = ei->i_block_group; |
185 | if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) { | |
186 | /* | |
187 | * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME | |
188 | * block groups per flexgroup, reserve the first block | |
189 | * group for directories and special files. Regular | |
190 | * files will start at the second block group. This | |
191 | * tends to speed up directory access and improves | |
192 | * fsck times. | |
193 | */ | |
194 | block_group &= ~(flex_size-1); | |
195 | if (S_ISREG(inode->i_mode)) | |
196 | block_group++; | |
197 | } | |
198 | bg_start = (block_group * EXT4_BLOCKS_PER_GROUP(inode->i_sb)) + | |
a86c6181 | 199 | le32_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_first_data_block); |
74d3487f VC |
200 | last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1; |
201 | ||
a4912123 TT |
202 | /* |
203 | * If we are doing delayed allocation, we don't need take | |
204 | * colour into account. | |
205 | */ | |
206 | if (test_opt(inode->i_sb, DELALLOC)) | |
207 | return bg_start; | |
208 | ||
74d3487f VC |
209 | if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block) |
210 | colour = (current->pid % 16) * | |
a86c6181 | 211 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); |
74d3487f VC |
212 | else |
213 | colour = (current->pid % 16) * ((last_block - bg_start) / 16); | |
a86c6181 AT |
214 | return bg_start + colour + block; |
215 | } | |
216 | ||
654b4908 AK |
217 | /* |
218 | * Allocation for a meta data block | |
219 | */ | |
f65e6fba | 220 | static ext4_fsblk_t |
654b4908 | 221 | ext4_ext_new_meta_block(handle_t *handle, struct inode *inode, |
a86c6181 AT |
222 | struct ext4_ext_path *path, |
223 | struct ext4_extent *ex, int *err) | |
224 | { | |
f65e6fba | 225 | ext4_fsblk_t goal, newblock; |
a86c6181 AT |
226 | |
227 | goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block)); | |
97df5d15 | 228 | newblock = ext4_new_meta_blocks(handle, inode, goal, NULL, err); |
a86c6181 AT |
229 | return newblock; |
230 | } | |
231 | ||
55ad63bf | 232 | static inline int ext4_ext_space_block(struct inode *inode, int check) |
a86c6181 AT |
233 | { |
234 | int size; | |
235 | ||
236 | size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header)) | |
237 | / sizeof(struct ext4_extent); | |
55ad63bf | 238 | if (!check) { |
bbf2f9fb | 239 | #ifdef AGGRESSIVE_TEST |
55ad63bf TT |
240 | if (size > 6) |
241 | size = 6; | |
a86c6181 | 242 | #endif |
55ad63bf | 243 | } |
a86c6181 AT |
244 | return size; |
245 | } | |
246 | ||
55ad63bf | 247 | static inline int ext4_ext_space_block_idx(struct inode *inode, int check) |
a86c6181 AT |
248 | { |
249 | int size; | |
250 | ||
251 | size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header)) | |
252 | / sizeof(struct ext4_extent_idx); | |
55ad63bf | 253 | if (!check) { |
bbf2f9fb | 254 | #ifdef AGGRESSIVE_TEST |
55ad63bf TT |
255 | if (size > 5) |
256 | size = 5; | |
a86c6181 | 257 | #endif |
55ad63bf | 258 | } |
a86c6181 AT |
259 | return size; |
260 | } | |
261 | ||
55ad63bf | 262 | static inline int ext4_ext_space_root(struct inode *inode, int check) |
a86c6181 AT |
263 | { |
264 | int size; | |
265 | ||
266 | size = sizeof(EXT4_I(inode)->i_data); | |
267 | size -= sizeof(struct ext4_extent_header); | |
268 | size /= sizeof(struct ext4_extent); | |
55ad63bf | 269 | if (!check) { |
bbf2f9fb | 270 | #ifdef AGGRESSIVE_TEST |
55ad63bf TT |
271 | if (size > 3) |
272 | size = 3; | |
a86c6181 | 273 | #endif |
55ad63bf | 274 | } |
a86c6181 AT |
275 | return size; |
276 | } | |
277 | ||
55ad63bf | 278 | static inline int ext4_ext_space_root_idx(struct inode *inode, int check) |
a86c6181 AT |
279 | { |
280 | int size; | |
281 | ||
282 | size = sizeof(EXT4_I(inode)->i_data); | |
283 | size -= sizeof(struct ext4_extent_header); | |
284 | size /= sizeof(struct ext4_extent_idx); | |
55ad63bf | 285 | if (!check) { |
bbf2f9fb | 286 | #ifdef AGGRESSIVE_TEST |
55ad63bf TT |
287 | if (size > 4) |
288 | size = 4; | |
a86c6181 | 289 | #endif |
55ad63bf | 290 | } |
a86c6181 AT |
291 | return size; |
292 | } | |
293 | ||
d2a17637 MC |
294 | /* |
295 | * Calculate the number of metadata blocks needed | |
296 | * to allocate @blocks | |
297 | * Worse case is one block per extent | |
298 | */ | |
9d0be502 | 299 | int ext4_ext_calc_metadata_amount(struct inode *inode, sector_t lblock) |
d2a17637 | 300 | { |
9d0be502 TT |
301 | struct ext4_inode_info *ei = EXT4_I(inode); |
302 | int idxs, num = 0; | |
d2a17637 | 303 | |
9d0be502 TT |
304 | idxs = ((inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header)) |
305 | / sizeof(struct ext4_extent_idx)); | |
d2a17637 MC |
306 | |
307 | /* | |
9d0be502 TT |
308 | * If the new delayed allocation block is contiguous with the |
309 | * previous da block, it can share index blocks with the | |
310 | * previous block, so we only need to allocate a new index | |
311 | * block every idxs leaf blocks. At ldxs**2 blocks, we need | |
312 | * an additional index block, and at ldxs**3 blocks, yet | |
313 | * another index blocks. | |
d2a17637 | 314 | */ |
9d0be502 TT |
315 | if (ei->i_da_metadata_calc_len && |
316 | ei->i_da_metadata_calc_last_lblock+1 == lblock) { | |
317 | if ((ei->i_da_metadata_calc_len % idxs) == 0) | |
318 | num++; | |
319 | if ((ei->i_da_metadata_calc_len % (idxs*idxs)) == 0) | |
320 | num++; | |
321 | if ((ei->i_da_metadata_calc_len % (idxs*idxs*idxs)) == 0) { | |
322 | num++; | |
323 | ei->i_da_metadata_calc_len = 0; | |
324 | } else | |
325 | ei->i_da_metadata_calc_len++; | |
326 | ei->i_da_metadata_calc_last_lblock++; | |
327 | return num; | |
328 | } | |
d2a17637 | 329 | |
9d0be502 TT |
330 | /* |
331 | * In the worst case we need a new set of index blocks at | |
332 | * every level of the inode's extent tree. | |
333 | */ | |
334 | ei->i_da_metadata_calc_len = 1; | |
335 | ei->i_da_metadata_calc_last_lblock = lblock; | |
336 | return ext_depth(inode) + 1; | |
d2a17637 MC |
337 | } |
338 | ||
c29c0ae7 AT |
339 | static int |
340 | ext4_ext_max_entries(struct inode *inode, int depth) | |
341 | { | |
342 | int max; | |
343 | ||
344 | if (depth == ext_depth(inode)) { | |
345 | if (depth == 0) | |
55ad63bf | 346 | max = ext4_ext_space_root(inode, 1); |
c29c0ae7 | 347 | else |
55ad63bf | 348 | max = ext4_ext_space_root_idx(inode, 1); |
c29c0ae7 AT |
349 | } else { |
350 | if (depth == 0) | |
55ad63bf | 351 | max = ext4_ext_space_block(inode, 1); |
c29c0ae7 | 352 | else |
55ad63bf | 353 | max = ext4_ext_space_block_idx(inode, 1); |
c29c0ae7 AT |
354 | } |
355 | ||
356 | return max; | |
357 | } | |
358 | ||
56b19868 AK |
359 | static int ext4_valid_extent(struct inode *inode, struct ext4_extent *ext) |
360 | { | |
6fd058f7 | 361 | ext4_fsblk_t block = ext_pblock(ext); |
56b19868 | 362 | int len = ext4_ext_get_actual_len(ext); |
e84a26ce | 363 | |
6fd058f7 | 364 | return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, len); |
56b19868 AK |
365 | } |
366 | ||
367 | static int ext4_valid_extent_idx(struct inode *inode, | |
368 | struct ext4_extent_idx *ext_idx) | |
369 | { | |
6fd058f7 | 370 | ext4_fsblk_t block = idx_pblock(ext_idx); |
e84a26ce | 371 | |
6fd058f7 | 372 | return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, 1); |
56b19868 AK |
373 | } |
374 | ||
375 | static int ext4_valid_extent_entries(struct inode *inode, | |
376 | struct ext4_extent_header *eh, | |
377 | int depth) | |
378 | { | |
379 | struct ext4_extent *ext; | |
380 | struct ext4_extent_idx *ext_idx; | |
381 | unsigned short entries; | |
382 | if (eh->eh_entries == 0) | |
383 | return 1; | |
384 | ||
385 | entries = le16_to_cpu(eh->eh_entries); | |
386 | ||
387 | if (depth == 0) { | |
388 | /* leaf entries */ | |
389 | ext = EXT_FIRST_EXTENT(eh); | |
390 | while (entries) { | |
391 | if (!ext4_valid_extent(inode, ext)) | |
392 | return 0; | |
393 | ext++; | |
394 | entries--; | |
395 | } | |
396 | } else { | |
397 | ext_idx = EXT_FIRST_INDEX(eh); | |
398 | while (entries) { | |
399 | if (!ext4_valid_extent_idx(inode, ext_idx)) | |
400 | return 0; | |
401 | ext_idx++; | |
402 | entries--; | |
403 | } | |
404 | } | |
405 | return 1; | |
406 | } | |
407 | ||
408 | static int __ext4_ext_check(const char *function, struct inode *inode, | |
c29c0ae7 AT |
409 | struct ext4_extent_header *eh, |
410 | int depth) | |
411 | { | |
412 | const char *error_msg; | |
413 | int max = 0; | |
414 | ||
415 | if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) { | |
416 | error_msg = "invalid magic"; | |
417 | goto corrupted; | |
418 | } | |
419 | if (unlikely(le16_to_cpu(eh->eh_depth) != depth)) { | |
420 | error_msg = "unexpected eh_depth"; | |
421 | goto corrupted; | |
422 | } | |
423 | if (unlikely(eh->eh_max == 0)) { | |
424 | error_msg = "invalid eh_max"; | |
425 | goto corrupted; | |
426 | } | |
427 | max = ext4_ext_max_entries(inode, depth); | |
428 | if (unlikely(le16_to_cpu(eh->eh_max) > max)) { | |
429 | error_msg = "too large eh_max"; | |
430 | goto corrupted; | |
431 | } | |
432 | if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) { | |
433 | error_msg = "invalid eh_entries"; | |
434 | goto corrupted; | |
435 | } | |
56b19868 AK |
436 | if (!ext4_valid_extent_entries(inode, eh, depth)) { |
437 | error_msg = "invalid extent entries"; | |
438 | goto corrupted; | |
439 | } | |
c29c0ae7 AT |
440 | return 0; |
441 | ||
442 | corrupted: | |
443 | ext4_error(inode->i_sb, function, | |
56b19868 | 444 | "bad header/extent in inode #%lu: %s - magic %x, " |
c29c0ae7 AT |
445 | "entries %u, max %u(%u), depth %u(%u)", |
446 | inode->i_ino, error_msg, le16_to_cpu(eh->eh_magic), | |
447 | le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max), | |
448 | max, le16_to_cpu(eh->eh_depth), depth); | |
449 | ||
450 | return -EIO; | |
451 | } | |
452 | ||
56b19868 AK |
453 | #define ext4_ext_check(inode, eh, depth) \ |
454 | __ext4_ext_check(__func__, inode, eh, depth) | |
c29c0ae7 | 455 | |
7a262f7c AK |
456 | int ext4_ext_check_inode(struct inode *inode) |
457 | { | |
458 | return ext4_ext_check(inode, ext_inode_hdr(inode), ext_depth(inode)); | |
459 | } | |
460 | ||
a86c6181 AT |
461 | #ifdef EXT_DEBUG |
462 | static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path) | |
463 | { | |
464 | int k, l = path->p_depth; | |
465 | ||
466 | ext_debug("path:"); | |
467 | for (k = 0; k <= l; k++, path++) { | |
468 | if (path->p_idx) { | |
2ae02107 | 469 | ext_debug(" %d->%llu", le32_to_cpu(path->p_idx->ei_block), |
f65e6fba | 470 | idx_pblock(path->p_idx)); |
a86c6181 | 471 | } else if (path->p_ext) { |
553f9008 | 472 | ext_debug(" %d:[%d]%d:%llu ", |
a86c6181 | 473 | le32_to_cpu(path->p_ext->ee_block), |
553f9008 | 474 | ext4_ext_is_uninitialized(path->p_ext), |
a2df2a63 | 475 | ext4_ext_get_actual_len(path->p_ext), |
f65e6fba | 476 | ext_pblock(path->p_ext)); |
a86c6181 AT |
477 | } else |
478 | ext_debug(" []"); | |
479 | } | |
480 | ext_debug("\n"); | |
481 | } | |
482 | ||
483 | static void ext4_ext_show_leaf(struct inode *inode, struct ext4_ext_path *path) | |
484 | { | |
485 | int depth = ext_depth(inode); | |
486 | struct ext4_extent_header *eh; | |
487 | struct ext4_extent *ex; | |
488 | int i; | |
489 | ||
490 | if (!path) | |
491 | return; | |
492 | ||
493 | eh = path[depth].p_hdr; | |
494 | ex = EXT_FIRST_EXTENT(eh); | |
495 | ||
553f9008 M |
496 | ext_debug("Displaying leaf extents for inode %lu\n", inode->i_ino); |
497 | ||
a86c6181 | 498 | for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) { |
553f9008 M |
499 | ext_debug("%d:[%d]%d:%llu ", le32_to_cpu(ex->ee_block), |
500 | ext4_ext_is_uninitialized(ex), | |
a2df2a63 | 501 | ext4_ext_get_actual_len(ex), ext_pblock(ex)); |
a86c6181 AT |
502 | } |
503 | ext_debug("\n"); | |
504 | } | |
505 | #else | |
af5bc92d TT |
506 | #define ext4_ext_show_path(inode, path) |
507 | #define ext4_ext_show_leaf(inode, path) | |
a86c6181 AT |
508 | #endif |
509 | ||
b35905c1 | 510 | void ext4_ext_drop_refs(struct ext4_ext_path *path) |
a86c6181 AT |
511 | { |
512 | int depth = path->p_depth; | |
513 | int i; | |
514 | ||
515 | for (i = 0; i <= depth; i++, path++) | |
516 | if (path->p_bh) { | |
517 | brelse(path->p_bh); | |
518 | path->p_bh = NULL; | |
519 | } | |
520 | } | |
521 | ||
522 | /* | |
d0d856e8 RD |
523 | * ext4_ext_binsearch_idx: |
524 | * binary search for the closest index of the given block | |
c29c0ae7 | 525 | * the header must be checked before calling this |
a86c6181 AT |
526 | */ |
527 | static void | |
725d26d3 AK |
528 | ext4_ext_binsearch_idx(struct inode *inode, |
529 | struct ext4_ext_path *path, ext4_lblk_t block) | |
a86c6181 AT |
530 | { |
531 | struct ext4_extent_header *eh = path->p_hdr; | |
532 | struct ext4_extent_idx *r, *l, *m; | |
533 | ||
a86c6181 | 534 | |
bba90743 | 535 | ext_debug("binsearch for %u(idx): ", block); |
a86c6181 AT |
536 | |
537 | l = EXT_FIRST_INDEX(eh) + 1; | |
e9f410b1 | 538 | r = EXT_LAST_INDEX(eh); |
a86c6181 AT |
539 | while (l <= r) { |
540 | m = l + (r - l) / 2; | |
541 | if (block < le32_to_cpu(m->ei_block)) | |
542 | r = m - 1; | |
543 | else | |
544 | l = m + 1; | |
26d535ed DM |
545 | ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ei_block), |
546 | m, le32_to_cpu(m->ei_block), | |
547 | r, le32_to_cpu(r->ei_block)); | |
a86c6181 AT |
548 | } |
549 | ||
550 | path->p_idx = l - 1; | |
f65e6fba | 551 | ext_debug(" -> %d->%lld ", le32_to_cpu(path->p_idx->ei_block), |
26d535ed | 552 | idx_pblock(path->p_idx)); |
a86c6181 AT |
553 | |
554 | #ifdef CHECK_BINSEARCH | |
555 | { | |
556 | struct ext4_extent_idx *chix, *ix; | |
557 | int k; | |
558 | ||
559 | chix = ix = EXT_FIRST_INDEX(eh); | |
560 | for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ix++) { | |
561 | if (k != 0 && | |
562 | le32_to_cpu(ix->ei_block) <= le32_to_cpu(ix[-1].ei_block)) { | |
4776004f TT |
563 | printk(KERN_DEBUG "k=%d, ix=0x%p, " |
564 | "first=0x%p\n", k, | |
565 | ix, EXT_FIRST_INDEX(eh)); | |
566 | printk(KERN_DEBUG "%u <= %u\n", | |
a86c6181 AT |
567 | le32_to_cpu(ix->ei_block), |
568 | le32_to_cpu(ix[-1].ei_block)); | |
569 | } | |
570 | BUG_ON(k && le32_to_cpu(ix->ei_block) | |
8c55e204 | 571 | <= le32_to_cpu(ix[-1].ei_block)); |
a86c6181 AT |
572 | if (block < le32_to_cpu(ix->ei_block)) |
573 | break; | |
574 | chix = ix; | |
575 | } | |
576 | BUG_ON(chix != path->p_idx); | |
577 | } | |
578 | #endif | |
579 | ||
580 | } | |
581 | ||
582 | /* | |
d0d856e8 RD |
583 | * ext4_ext_binsearch: |
584 | * binary search for closest extent of the given block | |
c29c0ae7 | 585 | * the header must be checked before calling this |
a86c6181 AT |
586 | */ |
587 | static void | |
725d26d3 AK |
588 | ext4_ext_binsearch(struct inode *inode, |
589 | struct ext4_ext_path *path, ext4_lblk_t block) | |
a86c6181 AT |
590 | { |
591 | struct ext4_extent_header *eh = path->p_hdr; | |
592 | struct ext4_extent *r, *l, *m; | |
593 | ||
a86c6181 AT |
594 | if (eh->eh_entries == 0) { |
595 | /* | |
d0d856e8 RD |
596 | * this leaf is empty: |
597 | * we get such a leaf in split/add case | |
a86c6181 AT |
598 | */ |
599 | return; | |
600 | } | |
601 | ||
bba90743 | 602 | ext_debug("binsearch for %u: ", block); |
a86c6181 AT |
603 | |
604 | l = EXT_FIRST_EXTENT(eh) + 1; | |
e9f410b1 | 605 | r = EXT_LAST_EXTENT(eh); |
a86c6181 AT |
606 | |
607 | while (l <= r) { | |
608 | m = l + (r - l) / 2; | |
609 | if (block < le32_to_cpu(m->ee_block)) | |
610 | r = m - 1; | |
611 | else | |
612 | l = m + 1; | |
26d535ed DM |
613 | ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ee_block), |
614 | m, le32_to_cpu(m->ee_block), | |
615 | r, le32_to_cpu(r->ee_block)); | |
a86c6181 AT |
616 | } |
617 | ||
618 | path->p_ext = l - 1; | |
553f9008 | 619 | ext_debug(" -> %d:%llu:[%d]%d ", |
8c55e204 DK |
620 | le32_to_cpu(path->p_ext->ee_block), |
621 | ext_pblock(path->p_ext), | |
553f9008 | 622 | ext4_ext_is_uninitialized(path->p_ext), |
a2df2a63 | 623 | ext4_ext_get_actual_len(path->p_ext)); |
a86c6181 AT |
624 | |
625 | #ifdef CHECK_BINSEARCH | |
626 | { | |
627 | struct ext4_extent *chex, *ex; | |
628 | int k; | |
629 | ||
630 | chex = ex = EXT_FIRST_EXTENT(eh); | |
631 | for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ex++) { | |
632 | BUG_ON(k && le32_to_cpu(ex->ee_block) | |
8c55e204 | 633 | <= le32_to_cpu(ex[-1].ee_block)); |
a86c6181 AT |
634 | if (block < le32_to_cpu(ex->ee_block)) |
635 | break; | |
636 | chex = ex; | |
637 | } | |
638 | BUG_ON(chex != path->p_ext); | |
639 | } | |
640 | #endif | |
641 | ||
642 | } | |
643 | ||
644 | int ext4_ext_tree_init(handle_t *handle, struct inode *inode) | |
645 | { | |
646 | struct ext4_extent_header *eh; | |
647 | ||
648 | eh = ext_inode_hdr(inode); | |
649 | eh->eh_depth = 0; | |
650 | eh->eh_entries = 0; | |
651 | eh->eh_magic = EXT4_EXT_MAGIC; | |
55ad63bf | 652 | eh->eh_max = cpu_to_le16(ext4_ext_space_root(inode, 0)); |
a86c6181 AT |
653 | ext4_mark_inode_dirty(handle, inode); |
654 | ext4_ext_invalidate_cache(inode); | |
655 | return 0; | |
656 | } | |
657 | ||
658 | struct ext4_ext_path * | |
725d26d3 AK |
659 | ext4_ext_find_extent(struct inode *inode, ext4_lblk_t block, |
660 | struct ext4_ext_path *path) | |
a86c6181 AT |
661 | { |
662 | struct ext4_extent_header *eh; | |
663 | struct buffer_head *bh; | |
664 | short int depth, i, ppos = 0, alloc = 0; | |
665 | ||
666 | eh = ext_inode_hdr(inode); | |
c29c0ae7 | 667 | depth = ext_depth(inode); |
a86c6181 AT |
668 | |
669 | /* account possible depth increase */ | |
670 | if (!path) { | |
5d4958f9 | 671 | path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 2), |
a86c6181 AT |
672 | GFP_NOFS); |
673 | if (!path) | |
674 | return ERR_PTR(-ENOMEM); | |
675 | alloc = 1; | |
676 | } | |
a86c6181 | 677 | path[0].p_hdr = eh; |
1973adcb | 678 | path[0].p_bh = NULL; |
a86c6181 | 679 | |
c29c0ae7 | 680 | i = depth; |
a86c6181 AT |
681 | /* walk through the tree */ |
682 | while (i) { | |
7a262f7c AK |
683 | int need_to_validate = 0; |
684 | ||
a86c6181 AT |
685 | ext_debug("depth %d: num %d, max %d\n", |
686 | ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max)); | |
c29c0ae7 | 687 | |
a86c6181 | 688 | ext4_ext_binsearch_idx(inode, path + ppos, block); |
f65e6fba | 689 | path[ppos].p_block = idx_pblock(path[ppos].p_idx); |
a86c6181 AT |
690 | path[ppos].p_depth = i; |
691 | path[ppos].p_ext = NULL; | |
692 | ||
7a262f7c AK |
693 | bh = sb_getblk(inode->i_sb, path[ppos].p_block); |
694 | if (unlikely(!bh)) | |
a86c6181 | 695 | goto err; |
7a262f7c AK |
696 | if (!bh_uptodate_or_lock(bh)) { |
697 | if (bh_submit_read(bh) < 0) { | |
698 | put_bh(bh); | |
699 | goto err; | |
700 | } | |
701 | /* validate the extent entries */ | |
702 | need_to_validate = 1; | |
703 | } | |
a86c6181 AT |
704 | eh = ext_block_hdr(bh); |
705 | ppos++; | |
706 | BUG_ON(ppos > depth); | |
707 | path[ppos].p_bh = bh; | |
708 | path[ppos].p_hdr = eh; | |
709 | i--; | |
710 | ||
7a262f7c | 711 | if (need_to_validate && ext4_ext_check(inode, eh, i)) |
a86c6181 AT |
712 | goto err; |
713 | } | |
714 | ||
715 | path[ppos].p_depth = i; | |
a86c6181 AT |
716 | path[ppos].p_ext = NULL; |
717 | path[ppos].p_idx = NULL; | |
718 | ||
a86c6181 AT |
719 | /* find extent */ |
720 | ext4_ext_binsearch(inode, path + ppos, block); | |
1973adcb SF |
721 | /* if not an empty leaf */ |
722 | if (path[ppos].p_ext) | |
723 | path[ppos].p_block = ext_pblock(path[ppos].p_ext); | |
a86c6181 AT |
724 | |
725 | ext4_ext_show_path(inode, path); | |
726 | ||
727 | return path; | |
728 | ||
729 | err: | |
730 | ext4_ext_drop_refs(path); | |
731 | if (alloc) | |
732 | kfree(path); | |
733 | return ERR_PTR(-EIO); | |
734 | } | |
735 | ||
736 | /* | |
d0d856e8 RD |
737 | * ext4_ext_insert_index: |
738 | * insert new index [@logical;@ptr] into the block at @curp; | |
739 | * check where to insert: before @curp or after @curp | |
a86c6181 | 740 | */ |
0031462b | 741 | int ext4_ext_insert_index(handle_t *handle, struct inode *inode, |
a86c6181 | 742 | struct ext4_ext_path *curp, |
f65e6fba | 743 | int logical, ext4_fsblk_t ptr) |
a86c6181 AT |
744 | { |
745 | struct ext4_extent_idx *ix; | |
746 | int len, err; | |
747 | ||
7e028976 AM |
748 | err = ext4_ext_get_access(handle, inode, curp); |
749 | if (err) | |
a86c6181 AT |
750 | return err; |
751 | ||
752 | BUG_ON(logical == le32_to_cpu(curp->p_idx->ei_block)); | |
753 | len = EXT_MAX_INDEX(curp->p_hdr) - curp->p_idx; | |
754 | if (logical > le32_to_cpu(curp->p_idx->ei_block)) { | |
755 | /* insert after */ | |
756 | if (curp->p_idx != EXT_LAST_INDEX(curp->p_hdr)) { | |
757 | len = (len - 1) * sizeof(struct ext4_extent_idx); | |
758 | len = len < 0 ? 0 : len; | |
26d535ed | 759 | ext_debug("insert new index %d after: %llu. " |
a86c6181 AT |
760 | "move %d from 0x%p to 0x%p\n", |
761 | logical, ptr, len, | |
762 | (curp->p_idx + 1), (curp->p_idx + 2)); | |
763 | memmove(curp->p_idx + 2, curp->p_idx + 1, len); | |
764 | } | |
765 | ix = curp->p_idx + 1; | |
766 | } else { | |
767 | /* insert before */ | |
768 | len = len * sizeof(struct ext4_extent_idx); | |
769 | len = len < 0 ? 0 : len; | |
26d535ed | 770 | ext_debug("insert new index %d before: %llu. " |
a86c6181 AT |
771 | "move %d from 0x%p to 0x%p\n", |
772 | logical, ptr, len, | |
773 | curp->p_idx, (curp->p_idx + 1)); | |
774 | memmove(curp->p_idx + 1, curp->p_idx, len); | |
775 | ix = curp->p_idx; | |
776 | } | |
777 | ||
778 | ix->ei_block = cpu_to_le32(logical); | |
f65e6fba | 779 | ext4_idx_store_pblock(ix, ptr); |
e8546d06 | 780 | le16_add_cpu(&curp->p_hdr->eh_entries, 1); |
a86c6181 AT |
781 | |
782 | BUG_ON(le16_to_cpu(curp->p_hdr->eh_entries) | |
8c55e204 | 783 | > le16_to_cpu(curp->p_hdr->eh_max)); |
a86c6181 AT |
784 | BUG_ON(ix > EXT_LAST_INDEX(curp->p_hdr)); |
785 | ||
786 | err = ext4_ext_dirty(handle, inode, curp); | |
787 | ext4_std_error(inode->i_sb, err); | |
788 | ||
789 | return err; | |
790 | } | |
791 | ||
792 | /* | |
d0d856e8 RD |
793 | * ext4_ext_split: |
794 | * inserts new subtree into the path, using free index entry | |
795 | * at depth @at: | |
796 | * - allocates all needed blocks (new leaf and all intermediate index blocks) | |
797 | * - makes decision where to split | |
798 | * - moves remaining extents and index entries (right to the split point) | |
799 | * into the newly allocated blocks | |
800 | * - initializes subtree | |
a86c6181 AT |
801 | */ |
802 | static int ext4_ext_split(handle_t *handle, struct inode *inode, | |
803 | struct ext4_ext_path *path, | |
804 | struct ext4_extent *newext, int at) | |
805 | { | |
806 | struct buffer_head *bh = NULL; | |
807 | int depth = ext_depth(inode); | |
808 | struct ext4_extent_header *neh; | |
809 | struct ext4_extent_idx *fidx; | |
810 | struct ext4_extent *ex; | |
811 | int i = at, k, m, a; | |
f65e6fba | 812 | ext4_fsblk_t newblock, oldblock; |
a86c6181 | 813 | __le32 border; |
f65e6fba | 814 | ext4_fsblk_t *ablocks = NULL; /* array of allocated blocks */ |
a86c6181 AT |
815 | int err = 0; |
816 | ||
817 | /* make decision: where to split? */ | |
d0d856e8 | 818 | /* FIXME: now decision is simplest: at current extent */ |
a86c6181 | 819 | |
d0d856e8 | 820 | /* if current leaf will be split, then we should use |
a86c6181 AT |
821 | * border from split point */ |
822 | BUG_ON(path[depth].p_ext > EXT_MAX_EXTENT(path[depth].p_hdr)); | |
823 | if (path[depth].p_ext != EXT_MAX_EXTENT(path[depth].p_hdr)) { | |
824 | border = path[depth].p_ext[1].ee_block; | |
d0d856e8 | 825 | ext_debug("leaf will be split." |
a86c6181 | 826 | " next leaf starts at %d\n", |
8c55e204 | 827 | le32_to_cpu(border)); |
a86c6181 AT |
828 | } else { |
829 | border = newext->ee_block; | |
830 | ext_debug("leaf will be added." | |
831 | " next leaf starts at %d\n", | |
8c55e204 | 832 | le32_to_cpu(border)); |
a86c6181 AT |
833 | } |
834 | ||
835 | /* | |
d0d856e8 RD |
836 | * If error occurs, then we break processing |
837 | * and mark filesystem read-only. index won't | |
a86c6181 | 838 | * be inserted and tree will be in consistent |
d0d856e8 | 839 | * state. Next mount will repair buffers too. |
a86c6181 AT |
840 | */ |
841 | ||
842 | /* | |
d0d856e8 RD |
843 | * Get array to track all allocated blocks. |
844 | * We need this to handle errors and free blocks | |
845 | * upon them. | |
a86c6181 | 846 | */ |
5d4958f9 | 847 | ablocks = kzalloc(sizeof(ext4_fsblk_t) * depth, GFP_NOFS); |
a86c6181 AT |
848 | if (!ablocks) |
849 | return -ENOMEM; | |
a86c6181 AT |
850 | |
851 | /* allocate all needed blocks */ | |
852 | ext_debug("allocate %d blocks for indexes/leaf\n", depth - at); | |
853 | for (a = 0; a < depth - at; a++) { | |
654b4908 AK |
854 | newblock = ext4_ext_new_meta_block(handle, inode, path, |
855 | newext, &err); | |
a86c6181 AT |
856 | if (newblock == 0) |
857 | goto cleanup; | |
858 | ablocks[a] = newblock; | |
859 | } | |
860 | ||
861 | /* initialize new leaf */ | |
862 | newblock = ablocks[--a]; | |
863 | BUG_ON(newblock == 0); | |
864 | bh = sb_getblk(inode->i_sb, newblock); | |
865 | if (!bh) { | |
866 | err = -EIO; | |
867 | goto cleanup; | |
868 | } | |
869 | lock_buffer(bh); | |
870 | ||
7e028976 AM |
871 | err = ext4_journal_get_create_access(handle, bh); |
872 | if (err) | |
a86c6181 AT |
873 | goto cleanup; |
874 | ||
875 | neh = ext_block_hdr(bh); | |
876 | neh->eh_entries = 0; | |
55ad63bf | 877 | neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0)); |
a86c6181 AT |
878 | neh->eh_magic = EXT4_EXT_MAGIC; |
879 | neh->eh_depth = 0; | |
880 | ex = EXT_FIRST_EXTENT(neh); | |
881 | ||
d0d856e8 | 882 | /* move remainder of path[depth] to the new leaf */ |
a86c6181 AT |
883 | BUG_ON(path[depth].p_hdr->eh_entries != path[depth].p_hdr->eh_max); |
884 | /* start copy from next extent */ | |
885 | /* TODO: we could do it by single memmove */ | |
886 | m = 0; | |
887 | path[depth].p_ext++; | |
888 | while (path[depth].p_ext <= | |
889 | EXT_MAX_EXTENT(path[depth].p_hdr)) { | |
553f9008 | 890 | ext_debug("move %d:%llu:[%d]%d in new leaf %llu\n", |
8c55e204 DK |
891 | le32_to_cpu(path[depth].p_ext->ee_block), |
892 | ext_pblock(path[depth].p_ext), | |
553f9008 | 893 | ext4_ext_is_uninitialized(path[depth].p_ext), |
a2df2a63 | 894 | ext4_ext_get_actual_len(path[depth].p_ext), |
a86c6181 AT |
895 | newblock); |
896 | /*memmove(ex++, path[depth].p_ext++, | |
897 | sizeof(struct ext4_extent)); | |
898 | neh->eh_entries++;*/ | |
899 | path[depth].p_ext++; | |
900 | m++; | |
901 | } | |
902 | if (m) { | |
903 | memmove(ex, path[depth].p_ext-m, sizeof(struct ext4_extent)*m); | |
e8546d06 | 904 | le16_add_cpu(&neh->eh_entries, m); |
a86c6181 AT |
905 | } |
906 | ||
907 | set_buffer_uptodate(bh); | |
908 | unlock_buffer(bh); | |
909 | ||
0390131b | 910 | err = ext4_handle_dirty_metadata(handle, inode, bh); |
7e028976 | 911 | if (err) |
a86c6181 AT |
912 | goto cleanup; |
913 | brelse(bh); | |
914 | bh = NULL; | |
915 | ||
916 | /* correct old leaf */ | |
917 | if (m) { | |
7e028976 AM |
918 | err = ext4_ext_get_access(handle, inode, path + depth); |
919 | if (err) | |
a86c6181 | 920 | goto cleanup; |
e8546d06 | 921 | le16_add_cpu(&path[depth].p_hdr->eh_entries, -m); |
7e028976 AM |
922 | err = ext4_ext_dirty(handle, inode, path + depth); |
923 | if (err) | |
a86c6181 AT |
924 | goto cleanup; |
925 | ||
926 | } | |
927 | ||
928 | /* create intermediate indexes */ | |
929 | k = depth - at - 1; | |
930 | BUG_ON(k < 0); | |
931 | if (k) | |
932 | ext_debug("create %d intermediate indices\n", k); | |
933 | /* insert new index into current index block */ | |
934 | /* current depth stored in i var */ | |
935 | i = depth - 1; | |
936 | while (k--) { | |
937 | oldblock = newblock; | |
938 | newblock = ablocks[--a]; | |
bba90743 | 939 | bh = sb_getblk(inode->i_sb, newblock); |
a86c6181 AT |
940 | if (!bh) { |
941 | err = -EIO; | |
942 | goto cleanup; | |
943 | } | |
944 | lock_buffer(bh); | |
945 | ||
7e028976 AM |
946 | err = ext4_journal_get_create_access(handle, bh); |
947 | if (err) | |
a86c6181 AT |
948 | goto cleanup; |
949 | ||
950 | neh = ext_block_hdr(bh); | |
951 | neh->eh_entries = cpu_to_le16(1); | |
952 | neh->eh_magic = EXT4_EXT_MAGIC; | |
55ad63bf | 953 | neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0)); |
a86c6181 AT |
954 | neh->eh_depth = cpu_to_le16(depth - i); |
955 | fidx = EXT_FIRST_INDEX(neh); | |
956 | fidx->ei_block = border; | |
f65e6fba | 957 | ext4_idx_store_pblock(fidx, oldblock); |
a86c6181 | 958 | |
bba90743 ES |
959 | ext_debug("int.index at %d (block %llu): %u -> %llu\n", |
960 | i, newblock, le32_to_cpu(border), oldblock); | |
a86c6181 AT |
961 | /* copy indexes */ |
962 | m = 0; | |
963 | path[i].p_idx++; | |
964 | ||
965 | ext_debug("cur 0x%p, last 0x%p\n", path[i].p_idx, | |
966 | EXT_MAX_INDEX(path[i].p_hdr)); | |
967 | BUG_ON(EXT_MAX_INDEX(path[i].p_hdr) != | |
968 | EXT_LAST_INDEX(path[i].p_hdr)); | |
969 | while (path[i].p_idx <= EXT_MAX_INDEX(path[i].p_hdr)) { | |
26d535ed | 970 | ext_debug("%d: move %d:%llu in new index %llu\n", i, |
8c55e204 DK |
971 | le32_to_cpu(path[i].p_idx->ei_block), |
972 | idx_pblock(path[i].p_idx), | |
973 | newblock); | |
a86c6181 AT |
974 | /*memmove(++fidx, path[i].p_idx++, |
975 | sizeof(struct ext4_extent_idx)); | |
976 | neh->eh_entries++; | |
977 | BUG_ON(neh->eh_entries > neh->eh_max);*/ | |
978 | path[i].p_idx++; | |
979 | m++; | |
980 | } | |
981 | if (m) { | |
982 | memmove(++fidx, path[i].p_idx - m, | |
983 | sizeof(struct ext4_extent_idx) * m); | |
e8546d06 | 984 | le16_add_cpu(&neh->eh_entries, m); |
a86c6181 AT |
985 | } |
986 | set_buffer_uptodate(bh); | |
987 | unlock_buffer(bh); | |
988 | ||
0390131b | 989 | err = ext4_handle_dirty_metadata(handle, inode, bh); |
7e028976 | 990 | if (err) |
a86c6181 AT |
991 | goto cleanup; |
992 | brelse(bh); | |
993 | bh = NULL; | |
994 | ||
995 | /* correct old index */ | |
996 | if (m) { | |
997 | err = ext4_ext_get_access(handle, inode, path + i); | |
998 | if (err) | |
999 | goto cleanup; | |
e8546d06 | 1000 | le16_add_cpu(&path[i].p_hdr->eh_entries, -m); |
a86c6181 AT |
1001 | err = ext4_ext_dirty(handle, inode, path + i); |
1002 | if (err) | |
1003 | goto cleanup; | |
1004 | } | |
1005 | ||
1006 | i--; | |
1007 | } | |
1008 | ||
1009 | /* insert new index */ | |
a86c6181 AT |
1010 | err = ext4_ext_insert_index(handle, inode, path + at, |
1011 | le32_to_cpu(border), newblock); | |
1012 | ||
1013 | cleanup: | |
1014 | if (bh) { | |
1015 | if (buffer_locked(bh)) | |
1016 | unlock_buffer(bh); | |
1017 | brelse(bh); | |
1018 | } | |
1019 | ||
1020 | if (err) { | |
1021 | /* free all allocated blocks in error case */ | |
1022 | for (i = 0; i < depth; i++) { | |
1023 | if (!ablocks[i]) | |
1024 | continue; | |
e6362609 TT |
1025 | ext4_free_blocks(handle, inode, 0, ablocks[i], 1, |
1026 | EXT4_FREE_BLOCKS_METADATA); | |
a86c6181 AT |
1027 | } |
1028 | } | |
1029 | kfree(ablocks); | |
1030 | ||
1031 | return err; | |
1032 | } | |
1033 | ||
1034 | /* | |
d0d856e8 RD |
1035 | * ext4_ext_grow_indepth: |
1036 | * implements tree growing procedure: | |
1037 | * - allocates new block | |
1038 | * - moves top-level data (index block or leaf) into the new block | |
1039 | * - initializes new top-level, creating index that points to the | |
1040 | * just created block | |
a86c6181 AT |
1041 | */ |
1042 | static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode, | |
1043 | struct ext4_ext_path *path, | |
1044 | struct ext4_extent *newext) | |
1045 | { | |
1046 | struct ext4_ext_path *curp = path; | |
1047 | struct ext4_extent_header *neh; | |
1048 | struct ext4_extent_idx *fidx; | |
1049 | struct buffer_head *bh; | |
f65e6fba | 1050 | ext4_fsblk_t newblock; |
a86c6181 AT |
1051 | int err = 0; |
1052 | ||
654b4908 | 1053 | newblock = ext4_ext_new_meta_block(handle, inode, path, newext, &err); |
a86c6181 AT |
1054 | if (newblock == 0) |
1055 | return err; | |
1056 | ||
1057 | bh = sb_getblk(inode->i_sb, newblock); | |
1058 | if (!bh) { | |
1059 | err = -EIO; | |
1060 | ext4_std_error(inode->i_sb, err); | |
1061 | return err; | |
1062 | } | |
1063 | lock_buffer(bh); | |
1064 | ||
7e028976 AM |
1065 | err = ext4_journal_get_create_access(handle, bh); |
1066 | if (err) { | |
a86c6181 AT |
1067 | unlock_buffer(bh); |
1068 | goto out; | |
1069 | } | |
1070 | ||
1071 | /* move top-level index/leaf into new block */ | |
1072 | memmove(bh->b_data, curp->p_hdr, sizeof(EXT4_I(inode)->i_data)); | |
1073 | ||
1074 | /* set size of new block */ | |
1075 | neh = ext_block_hdr(bh); | |
1076 | /* old root could have indexes or leaves | |
1077 | * so calculate e_max right way */ | |
1078 | if (ext_depth(inode)) | |
55ad63bf | 1079 | neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0)); |
a86c6181 | 1080 | else |
55ad63bf | 1081 | neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0)); |
a86c6181 AT |
1082 | neh->eh_magic = EXT4_EXT_MAGIC; |
1083 | set_buffer_uptodate(bh); | |
1084 | unlock_buffer(bh); | |
1085 | ||
0390131b | 1086 | err = ext4_handle_dirty_metadata(handle, inode, bh); |
7e028976 | 1087 | if (err) |
a86c6181 AT |
1088 | goto out; |
1089 | ||
1090 | /* create index in new top-level index: num,max,pointer */ | |
7e028976 AM |
1091 | err = ext4_ext_get_access(handle, inode, curp); |
1092 | if (err) | |
a86c6181 AT |
1093 | goto out; |
1094 | ||
1095 | curp->p_hdr->eh_magic = EXT4_EXT_MAGIC; | |
55ad63bf | 1096 | curp->p_hdr->eh_max = cpu_to_le16(ext4_ext_space_root_idx(inode, 0)); |
a86c6181 AT |
1097 | curp->p_hdr->eh_entries = cpu_to_le16(1); |
1098 | curp->p_idx = EXT_FIRST_INDEX(curp->p_hdr); | |
e9f410b1 DM |
1099 | |
1100 | if (path[0].p_hdr->eh_depth) | |
1101 | curp->p_idx->ei_block = | |
1102 | EXT_FIRST_INDEX(path[0].p_hdr)->ei_block; | |
1103 | else | |
1104 | curp->p_idx->ei_block = | |
1105 | EXT_FIRST_EXTENT(path[0].p_hdr)->ee_block; | |
f65e6fba | 1106 | ext4_idx_store_pblock(curp->p_idx, newblock); |
a86c6181 AT |
1107 | |
1108 | neh = ext_inode_hdr(inode); | |
1109 | fidx = EXT_FIRST_INDEX(neh); | |
2ae02107 | 1110 | ext_debug("new root: num %d(%d), lblock %d, ptr %llu\n", |
a86c6181 | 1111 | le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max), |
f65e6fba | 1112 | le32_to_cpu(fidx->ei_block), idx_pblock(fidx)); |
a86c6181 AT |
1113 | |
1114 | neh->eh_depth = cpu_to_le16(path->p_depth + 1); | |
1115 | err = ext4_ext_dirty(handle, inode, curp); | |
1116 | out: | |
1117 | brelse(bh); | |
1118 | ||
1119 | return err; | |
1120 | } | |
1121 | ||
1122 | /* | |
d0d856e8 RD |
1123 | * ext4_ext_create_new_leaf: |
1124 | * finds empty index and adds new leaf. | |
1125 | * if no free index is found, then it requests in-depth growing. | |
a86c6181 AT |
1126 | */ |
1127 | static int ext4_ext_create_new_leaf(handle_t *handle, struct inode *inode, | |
1128 | struct ext4_ext_path *path, | |
1129 | struct ext4_extent *newext) | |
1130 | { | |
1131 | struct ext4_ext_path *curp; | |
1132 | int depth, i, err = 0; | |
1133 | ||
1134 | repeat: | |
1135 | i = depth = ext_depth(inode); | |
1136 | ||
1137 | /* walk up to the tree and look for free index entry */ | |
1138 | curp = path + depth; | |
1139 | while (i > 0 && !EXT_HAS_FREE_INDEX(curp)) { | |
1140 | i--; | |
1141 | curp--; | |
1142 | } | |
1143 | ||
d0d856e8 RD |
1144 | /* we use already allocated block for index block, |
1145 | * so subsequent data blocks should be contiguous */ | |
a86c6181 AT |
1146 | if (EXT_HAS_FREE_INDEX(curp)) { |
1147 | /* if we found index with free entry, then use that | |
1148 | * entry: create all needed subtree and add new leaf */ | |
1149 | err = ext4_ext_split(handle, inode, path, newext, i); | |
787e0981 SF |
1150 | if (err) |
1151 | goto out; | |
a86c6181 AT |
1152 | |
1153 | /* refill path */ | |
1154 | ext4_ext_drop_refs(path); | |
1155 | path = ext4_ext_find_extent(inode, | |
725d26d3 AK |
1156 | (ext4_lblk_t)le32_to_cpu(newext->ee_block), |
1157 | path); | |
a86c6181 AT |
1158 | if (IS_ERR(path)) |
1159 | err = PTR_ERR(path); | |
1160 | } else { | |
1161 | /* tree is full, time to grow in depth */ | |
1162 | err = ext4_ext_grow_indepth(handle, inode, path, newext); | |
1163 | if (err) | |
1164 | goto out; | |
1165 | ||
1166 | /* refill path */ | |
1167 | ext4_ext_drop_refs(path); | |
1168 | path = ext4_ext_find_extent(inode, | |
725d26d3 AK |
1169 | (ext4_lblk_t)le32_to_cpu(newext->ee_block), |
1170 | path); | |
a86c6181 AT |
1171 | if (IS_ERR(path)) { |
1172 | err = PTR_ERR(path); | |
1173 | goto out; | |
1174 | } | |
1175 | ||
1176 | /* | |
d0d856e8 RD |
1177 | * only first (depth 0 -> 1) produces free space; |
1178 | * in all other cases we have to split the grown tree | |
a86c6181 AT |
1179 | */ |
1180 | depth = ext_depth(inode); | |
1181 | if (path[depth].p_hdr->eh_entries == path[depth].p_hdr->eh_max) { | |
d0d856e8 | 1182 | /* now we need to split */ |
a86c6181 AT |
1183 | goto repeat; |
1184 | } | |
1185 | } | |
1186 | ||
1187 | out: | |
1188 | return err; | |
1189 | } | |
1190 | ||
1988b51e AT |
1191 | /* |
1192 | * search the closest allocated block to the left for *logical | |
1193 | * and returns it at @logical + it's physical address at @phys | |
1194 | * if *logical is the smallest allocated block, the function | |
1195 | * returns 0 at @phys | |
1196 | * return value contains 0 (success) or error code | |
1197 | */ | |
1198 | int | |
1199 | ext4_ext_search_left(struct inode *inode, struct ext4_ext_path *path, | |
1200 | ext4_lblk_t *logical, ext4_fsblk_t *phys) | |
1201 | { | |
1202 | struct ext4_extent_idx *ix; | |
1203 | struct ext4_extent *ex; | |
b939e376 | 1204 | int depth, ee_len; |
1988b51e AT |
1205 | |
1206 | BUG_ON(path == NULL); | |
1207 | depth = path->p_depth; | |
1208 | *phys = 0; | |
1209 | ||
1210 | if (depth == 0 && path->p_ext == NULL) | |
1211 | return 0; | |
1212 | ||
1213 | /* usually extent in the path covers blocks smaller | |
1214 | * then *logical, but it can be that extent is the | |
1215 | * first one in the file */ | |
1216 | ||
1217 | ex = path[depth].p_ext; | |
b939e376 | 1218 | ee_len = ext4_ext_get_actual_len(ex); |
1988b51e AT |
1219 | if (*logical < le32_to_cpu(ex->ee_block)) { |
1220 | BUG_ON(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex); | |
1221 | while (--depth >= 0) { | |
1222 | ix = path[depth].p_idx; | |
1223 | BUG_ON(ix != EXT_FIRST_INDEX(path[depth].p_hdr)); | |
1224 | } | |
1225 | return 0; | |
1226 | } | |
1227 | ||
b939e376 | 1228 | BUG_ON(*logical < (le32_to_cpu(ex->ee_block) + ee_len)); |
1988b51e | 1229 | |
b939e376 AK |
1230 | *logical = le32_to_cpu(ex->ee_block) + ee_len - 1; |
1231 | *phys = ext_pblock(ex) + ee_len - 1; | |
1988b51e AT |
1232 | return 0; |
1233 | } | |
1234 | ||
1235 | /* | |
1236 | * search the closest allocated block to the right for *logical | |
1237 | * and returns it at @logical + it's physical address at @phys | |
1238 | * if *logical is the smallest allocated block, the function | |
1239 | * returns 0 at @phys | |
1240 | * return value contains 0 (success) or error code | |
1241 | */ | |
1242 | int | |
1243 | ext4_ext_search_right(struct inode *inode, struct ext4_ext_path *path, | |
1244 | ext4_lblk_t *logical, ext4_fsblk_t *phys) | |
1245 | { | |
1246 | struct buffer_head *bh = NULL; | |
1247 | struct ext4_extent_header *eh; | |
1248 | struct ext4_extent_idx *ix; | |
1249 | struct ext4_extent *ex; | |
1250 | ext4_fsblk_t block; | |
395a87bf ES |
1251 | int depth; /* Note, NOT eh_depth; depth from top of tree */ |
1252 | int ee_len; | |
1988b51e AT |
1253 | |
1254 | BUG_ON(path == NULL); | |
1255 | depth = path->p_depth; | |
1256 | *phys = 0; | |
1257 | ||
1258 | if (depth == 0 && path->p_ext == NULL) | |
1259 | return 0; | |
1260 | ||
1261 | /* usually extent in the path covers blocks smaller | |
1262 | * then *logical, but it can be that extent is the | |
1263 | * first one in the file */ | |
1264 | ||
1265 | ex = path[depth].p_ext; | |
b939e376 | 1266 | ee_len = ext4_ext_get_actual_len(ex); |
1988b51e AT |
1267 | if (*logical < le32_to_cpu(ex->ee_block)) { |
1268 | BUG_ON(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex); | |
1269 | while (--depth >= 0) { | |
1270 | ix = path[depth].p_idx; | |
1271 | BUG_ON(ix != EXT_FIRST_INDEX(path[depth].p_hdr)); | |
1272 | } | |
1273 | *logical = le32_to_cpu(ex->ee_block); | |
1274 | *phys = ext_pblock(ex); | |
1275 | return 0; | |
1276 | } | |
1277 | ||
b939e376 | 1278 | BUG_ON(*logical < (le32_to_cpu(ex->ee_block) + ee_len)); |
1988b51e AT |
1279 | |
1280 | if (ex != EXT_LAST_EXTENT(path[depth].p_hdr)) { | |
1281 | /* next allocated block in this leaf */ | |
1282 | ex++; | |
1283 | *logical = le32_to_cpu(ex->ee_block); | |
1284 | *phys = ext_pblock(ex); | |
1285 | return 0; | |
1286 | } | |
1287 | ||
1288 | /* go up and search for index to the right */ | |
1289 | while (--depth >= 0) { | |
1290 | ix = path[depth].p_idx; | |
1291 | if (ix != EXT_LAST_INDEX(path[depth].p_hdr)) | |
25f1ee3a | 1292 | goto got_index; |
1988b51e AT |
1293 | } |
1294 | ||
25f1ee3a WF |
1295 | /* we've gone up to the root and found no index to the right */ |
1296 | return 0; | |
1988b51e | 1297 | |
25f1ee3a | 1298 | got_index: |
1988b51e AT |
1299 | /* we've found index to the right, let's |
1300 | * follow it and find the closest allocated | |
1301 | * block to the right */ | |
1302 | ix++; | |
1303 | block = idx_pblock(ix); | |
1304 | while (++depth < path->p_depth) { | |
1305 | bh = sb_bread(inode->i_sb, block); | |
1306 | if (bh == NULL) | |
1307 | return -EIO; | |
1308 | eh = ext_block_hdr(bh); | |
395a87bf | 1309 | /* subtract from p_depth to get proper eh_depth */ |
56b19868 | 1310 | if (ext4_ext_check(inode, eh, path->p_depth - depth)) { |
1988b51e AT |
1311 | put_bh(bh); |
1312 | return -EIO; | |
1313 | } | |
1314 | ix = EXT_FIRST_INDEX(eh); | |
1315 | block = idx_pblock(ix); | |
1316 | put_bh(bh); | |
1317 | } | |
1318 | ||
1319 | bh = sb_bread(inode->i_sb, block); | |
1320 | if (bh == NULL) | |
1321 | return -EIO; | |
1322 | eh = ext_block_hdr(bh); | |
56b19868 | 1323 | if (ext4_ext_check(inode, eh, path->p_depth - depth)) { |
1988b51e AT |
1324 | put_bh(bh); |
1325 | return -EIO; | |
1326 | } | |
1327 | ex = EXT_FIRST_EXTENT(eh); | |
1328 | *logical = le32_to_cpu(ex->ee_block); | |
1329 | *phys = ext_pblock(ex); | |
1330 | put_bh(bh); | |
1331 | return 0; | |
1988b51e AT |
1332 | } |
1333 | ||
a86c6181 | 1334 | /* |
d0d856e8 RD |
1335 | * ext4_ext_next_allocated_block: |
1336 | * returns allocated block in subsequent extent or EXT_MAX_BLOCK. | |
1337 | * NOTE: it considers block number from index entry as | |
1338 | * allocated block. Thus, index entries have to be consistent | |
1339 | * with leaves. | |
a86c6181 | 1340 | */ |
725d26d3 | 1341 | static ext4_lblk_t |
a86c6181 AT |
1342 | ext4_ext_next_allocated_block(struct ext4_ext_path *path) |
1343 | { | |
1344 | int depth; | |
1345 | ||
1346 | BUG_ON(path == NULL); | |
1347 | depth = path->p_depth; | |
1348 | ||
1349 | if (depth == 0 && path->p_ext == NULL) | |
1350 | return EXT_MAX_BLOCK; | |
1351 | ||
1352 | while (depth >= 0) { | |
1353 | if (depth == path->p_depth) { | |
1354 | /* leaf */ | |
1355 | if (path[depth].p_ext != | |
1356 | EXT_LAST_EXTENT(path[depth].p_hdr)) | |
1357 | return le32_to_cpu(path[depth].p_ext[1].ee_block); | |
1358 | } else { | |
1359 | /* index */ | |
1360 | if (path[depth].p_idx != | |
1361 | EXT_LAST_INDEX(path[depth].p_hdr)) | |
1362 | return le32_to_cpu(path[depth].p_idx[1].ei_block); | |
1363 | } | |
1364 | depth--; | |
1365 | } | |
1366 | ||
1367 | return EXT_MAX_BLOCK; | |
1368 | } | |
1369 | ||
1370 | /* | |
d0d856e8 | 1371 | * ext4_ext_next_leaf_block: |
a86c6181 AT |
1372 | * returns first allocated block from next leaf or EXT_MAX_BLOCK |
1373 | */ | |
725d26d3 | 1374 | static ext4_lblk_t ext4_ext_next_leaf_block(struct inode *inode, |
63f57933 | 1375 | struct ext4_ext_path *path) |
a86c6181 AT |
1376 | { |
1377 | int depth; | |
1378 | ||
1379 | BUG_ON(path == NULL); | |
1380 | depth = path->p_depth; | |
1381 | ||
1382 | /* zero-tree has no leaf blocks at all */ | |
1383 | if (depth == 0) | |
1384 | return EXT_MAX_BLOCK; | |
1385 | ||
1386 | /* go to index block */ | |
1387 | depth--; | |
1388 | ||
1389 | while (depth >= 0) { | |
1390 | if (path[depth].p_idx != | |
1391 | EXT_LAST_INDEX(path[depth].p_hdr)) | |
725d26d3 AK |
1392 | return (ext4_lblk_t) |
1393 | le32_to_cpu(path[depth].p_idx[1].ei_block); | |
a86c6181 AT |
1394 | depth--; |
1395 | } | |
1396 | ||
1397 | return EXT_MAX_BLOCK; | |
1398 | } | |
1399 | ||
1400 | /* | |
d0d856e8 RD |
1401 | * ext4_ext_correct_indexes: |
1402 | * if leaf gets modified and modified extent is first in the leaf, | |
1403 | * then we have to correct all indexes above. | |
a86c6181 AT |
1404 | * TODO: do we need to correct tree in all cases? |
1405 | */ | |
1d03ec98 | 1406 | static int ext4_ext_correct_indexes(handle_t *handle, struct inode *inode, |
a86c6181 AT |
1407 | struct ext4_ext_path *path) |
1408 | { | |
1409 | struct ext4_extent_header *eh; | |
1410 | int depth = ext_depth(inode); | |
1411 | struct ext4_extent *ex; | |
1412 | __le32 border; | |
1413 | int k, err = 0; | |
1414 | ||
1415 | eh = path[depth].p_hdr; | |
1416 | ex = path[depth].p_ext; | |
1417 | BUG_ON(ex == NULL); | |
1418 | BUG_ON(eh == NULL); | |
1419 | ||
1420 | if (depth == 0) { | |
1421 | /* there is no tree at all */ | |
1422 | return 0; | |
1423 | } | |
1424 | ||
1425 | if (ex != EXT_FIRST_EXTENT(eh)) { | |
1426 | /* we correct tree if first leaf got modified only */ | |
1427 | return 0; | |
1428 | } | |
1429 | ||
1430 | /* | |
d0d856e8 | 1431 | * TODO: we need correction if border is smaller than current one |
a86c6181 AT |
1432 | */ |
1433 | k = depth - 1; | |
1434 | border = path[depth].p_ext->ee_block; | |
7e028976 AM |
1435 | err = ext4_ext_get_access(handle, inode, path + k); |
1436 | if (err) | |
a86c6181 AT |
1437 | return err; |
1438 | path[k].p_idx->ei_block = border; | |
7e028976 AM |
1439 | err = ext4_ext_dirty(handle, inode, path + k); |
1440 | if (err) | |
a86c6181 AT |
1441 | return err; |
1442 | ||
1443 | while (k--) { | |
1444 | /* change all left-side indexes */ | |
1445 | if (path[k+1].p_idx != EXT_FIRST_INDEX(path[k+1].p_hdr)) | |
1446 | break; | |
7e028976 AM |
1447 | err = ext4_ext_get_access(handle, inode, path + k); |
1448 | if (err) | |
a86c6181 AT |
1449 | break; |
1450 | path[k].p_idx->ei_block = border; | |
7e028976 AM |
1451 | err = ext4_ext_dirty(handle, inode, path + k); |
1452 | if (err) | |
a86c6181 AT |
1453 | break; |
1454 | } | |
1455 | ||
1456 | return err; | |
1457 | } | |
1458 | ||
748de673 | 1459 | int |
a86c6181 AT |
1460 | ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1, |
1461 | struct ext4_extent *ex2) | |
1462 | { | |
749269fa | 1463 | unsigned short ext1_ee_len, ext2_ee_len, max_len; |
a2df2a63 AA |
1464 | |
1465 | /* | |
1466 | * Make sure that either both extents are uninitialized, or | |
1467 | * both are _not_. | |
1468 | */ | |
1469 | if (ext4_ext_is_uninitialized(ex1) ^ ext4_ext_is_uninitialized(ex2)) | |
1470 | return 0; | |
1471 | ||
749269fa AA |
1472 | if (ext4_ext_is_uninitialized(ex1)) |
1473 | max_len = EXT_UNINIT_MAX_LEN; | |
1474 | else | |
1475 | max_len = EXT_INIT_MAX_LEN; | |
1476 | ||
a2df2a63 AA |
1477 | ext1_ee_len = ext4_ext_get_actual_len(ex1); |
1478 | ext2_ee_len = ext4_ext_get_actual_len(ex2); | |
1479 | ||
1480 | if (le32_to_cpu(ex1->ee_block) + ext1_ee_len != | |
63f57933 | 1481 | le32_to_cpu(ex2->ee_block)) |
a86c6181 AT |
1482 | return 0; |
1483 | ||
471d4011 SB |
1484 | /* |
1485 | * To allow future support for preallocated extents to be added | |
1486 | * as an RO_COMPAT feature, refuse to merge to extents if | |
d0d856e8 | 1487 | * this can result in the top bit of ee_len being set. |
471d4011 | 1488 | */ |
749269fa | 1489 | if (ext1_ee_len + ext2_ee_len > max_len) |
471d4011 | 1490 | return 0; |
bbf2f9fb | 1491 | #ifdef AGGRESSIVE_TEST |
b939e376 | 1492 | if (ext1_ee_len >= 4) |
a86c6181 AT |
1493 | return 0; |
1494 | #endif | |
1495 | ||
a2df2a63 | 1496 | if (ext_pblock(ex1) + ext1_ee_len == ext_pblock(ex2)) |
a86c6181 AT |
1497 | return 1; |
1498 | return 0; | |
1499 | } | |
1500 | ||
56055d3a AA |
1501 | /* |
1502 | * This function tries to merge the "ex" extent to the next extent in the tree. | |
1503 | * It always tries to merge towards right. If you want to merge towards | |
1504 | * left, pass "ex - 1" as argument instead of "ex". | |
1505 | * Returns 0 if the extents (ex and ex+1) were _not_ merged and returns | |
1506 | * 1 if they got merged. | |
1507 | */ | |
1508 | int ext4_ext_try_to_merge(struct inode *inode, | |
1509 | struct ext4_ext_path *path, | |
1510 | struct ext4_extent *ex) | |
1511 | { | |
1512 | struct ext4_extent_header *eh; | |
1513 | unsigned int depth, len; | |
1514 | int merge_done = 0; | |
1515 | int uninitialized = 0; | |
1516 | ||
1517 | depth = ext_depth(inode); | |
1518 | BUG_ON(path[depth].p_hdr == NULL); | |
1519 | eh = path[depth].p_hdr; | |
1520 | ||
1521 | while (ex < EXT_LAST_EXTENT(eh)) { | |
1522 | if (!ext4_can_extents_be_merged(inode, ex, ex + 1)) | |
1523 | break; | |
1524 | /* merge with next extent! */ | |
1525 | if (ext4_ext_is_uninitialized(ex)) | |
1526 | uninitialized = 1; | |
1527 | ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex) | |
1528 | + ext4_ext_get_actual_len(ex + 1)); | |
1529 | if (uninitialized) | |
1530 | ext4_ext_mark_uninitialized(ex); | |
1531 | ||
1532 | if (ex + 1 < EXT_LAST_EXTENT(eh)) { | |
1533 | len = (EXT_LAST_EXTENT(eh) - ex - 1) | |
1534 | * sizeof(struct ext4_extent); | |
1535 | memmove(ex + 1, ex + 2, len); | |
1536 | } | |
e8546d06 | 1537 | le16_add_cpu(&eh->eh_entries, -1); |
56055d3a AA |
1538 | merge_done = 1; |
1539 | WARN_ON(eh->eh_entries == 0); | |
1540 | if (!eh->eh_entries) | |
1541 | ext4_error(inode->i_sb, "ext4_ext_try_to_merge", | |
1542 | "inode#%lu, eh->eh_entries = 0!", inode->i_ino); | |
1543 | } | |
1544 | ||
1545 | return merge_done; | |
1546 | } | |
1547 | ||
25d14f98 AA |
1548 | /* |
1549 | * check if a portion of the "newext" extent overlaps with an | |
1550 | * existing extent. | |
1551 | * | |
1552 | * If there is an overlap discovered, it updates the length of the newext | |
1553 | * such that there will be no overlap, and then returns 1. | |
1554 | * If there is no overlap found, it returns 0. | |
1555 | */ | |
1556 | unsigned int ext4_ext_check_overlap(struct inode *inode, | |
1557 | struct ext4_extent *newext, | |
1558 | struct ext4_ext_path *path) | |
1559 | { | |
725d26d3 | 1560 | ext4_lblk_t b1, b2; |
25d14f98 AA |
1561 | unsigned int depth, len1; |
1562 | unsigned int ret = 0; | |
1563 | ||
1564 | b1 = le32_to_cpu(newext->ee_block); | |
a2df2a63 | 1565 | len1 = ext4_ext_get_actual_len(newext); |
25d14f98 AA |
1566 | depth = ext_depth(inode); |
1567 | if (!path[depth].p_ext) | |
1568 | goto out; | |
1569 | b2 = le32_to_cpu(path[depth].p_ext->ee_block); | |
1570 | ||
1571 | /* | |
1572 | * get the next allocated block if the extent in the path | |
2b2d6d01 | 1573 | * is before the requested block(s) |
25d14f98 AA |
1574 | */ |
1575 | if (b2 < b1) { | |
1576 | b2 = ext4_ext_next_allocated_block(path); | |
1577 | if (b2 == EXT_MAX_BLOCK) | |
1578 | goto out; | |
1579 | } | |
1580 | ||
725d26d3 | 1581 | /* check for wrap through zero on extent logical start block*/ |
25d14f98 AA |
1582 | if (b1 + len1 < b1) { |
1583 | len1 = EXT_MAX_BLOCK - b1; | |
1584 | newext->ee_len = cpu_to_le16(len1); | |
1585 | ret = 1; | |
1586 | } | |
1587 | ||
1588 | /* check for overlap */ | |
1589 | if (b1 + len1 > b2) { | |
1590 | newext->ee_len = cpu_to_le16(b2 - b1); | |
1591 | ret = 1; | |
1592 | } | |
1593 | out: | |
1594 | return ret; | |
1595 | } | |
1596 | ||
a86c6181 | 1597 | /* |
d0d856e8 RD |
1598 | * ext4_ext_insert_extent: |
1599 | * tries to merge requsted extent into the existing extent or | |
1600 | * inserts requested extent as new one into the tree, | |
1601 | * creating new leaf in the no-space case. | |
a86c6181 AT |
1602 | */ |
1603 | int ext4_ext_insert_extent(handle_t *handle, struct inode *inode, | |
1604 | struct ext4_ext_path *path, | |
0031462b | 1605 | struct ext4_extent *newext, int flag) |
a86c6181 | 1606 | { |
af5bc92d | 1607 | struct ext4_extent_header *eh; |
a86c6181 AT |
1608 | struct ext4_extent *ex, *fex; |
1609 | struct ext4_extent *nearex; /* nearest extent */ | |
1610 | struct ext4_ext_path *npath = NULL; | |
725d26d3 AK |
1611 | int depth, len, err; |
1612 | ext4_lblk_t next; | |
a2df2a63 | 1613 | unsigned uninitialized = 0; |
a86c6181 | 1614 | |
a2df2a63 | 1615 | BUG_ON(ext4_ext_get_actual_len(newext) == 0); |
a86c6181 AT |
1616 | depth = ext_depth(inode); |
1617 | ex = path[depth].p_ext; | |
1618 | BUG_ON(path[depth].p_hdr == NULL); | |
1619 | ||
1620 | /* try to insert block into found extent and return */ | |
0031462b MC |
1621 | if (ex && (flag != EXT4_GET_BLOCKS_DIO_CREATE_EXT) |
1622 | && ext4_can_extents_be_merged(inode, ex, newext)) { | |
553f9008 M |
1623 | ext_debug("append [%d]%d block to %d:[%d]%d (from %llu)\n", |
1624 | ext4_ext_is_uninitialized(newext), | |
a2df2a63 | 1625 | ext4_ext_get_actual_len(newext), |
a86c6181 | 1626 | le32_to_cpu(ex->ee_block), |
553f9008 | 1627 | ext4_ext_is_uninitialized(ex), |
a2df2a63 | 1628 | ext4_ext_get_actual_len(ex), ext_pblock(ex)); |
7e028976 AM |
1629 | err = ext4_ext_get_access(handle, inode, path + depth); |
1630 | if (err) | |
a86c6181 | 1631 | return err; |
a2df2a63 AA |
1632 | |
1633 | /* | |
1634 | * ext4_can_extents_be_merged should have checked that either | |
1635 | * both extents are uninitialized, or both aren't. Thus we | |
1636 | * need to check only one of them here. | |
1637 | */ | |
1638 | if (ext4_ext_is_uninitialized(ex)) | |
1639 | uninitialized = 1; | |
1640 | ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex) | |
1641 | + ext4_ext_get_actual_len(newext)); | |
1642 | if (uninitialized) | |
1643 | ext4_ext_mark_uninitialized(ex); | |
a86c6181 AT |
1644 | eh = path[depth].p_hdr; |
1645 | nearex = ex; | |
1646 | goto merge; | |
1647 | } | |
1648 | ||
1649 | repeat: | |
1650 | depth = ext_depth(inode); | |
1651 | eh = path[depth].p_hdr; | |
1652 | if (le16_to_cpu(eh->eh_entries) < le16_to_cpu(eh->eh_max)) | |
1653 | goto has_space; | |
1654 | ||
1655 | /* probably next leaf has space for us? */ | |
1656 | fex = EXT_LAST_EXTENT(eh); | |
1657 | next = ext4_ext_next_leaf_block(inode, path); | |
1658 | if (le32_to_cpu(newext->ee_block) > le32_to_cpu(fex->ee_block) | |
1659 | && next != EXT_MAX_BLOCK) { | |
1660 | ext_debug("next leaf block - %d\n", next); | |
1661 | BUG_ON(npath != NULL); | |
1662 | npath = ext4_ext_find_extent(inode, next, NULL); | |
1663 | if (IS_ERR(npath)) | |
1664 | return PTR_ERR(npath); | |
1665 | BUG_ON(npath->p_depth != path->p_depth); | |
1666 | eh = npath[depth].p_hdr; | |
1667 | if (le16_to_cpu(eh->eh_entries) < le16_to_cpu(eh->eh_max)) { | |
1668 | ext_debug("next leaf isnt full(%d)\n", | |
1669 | le16_to_cpu(eh->eh_entries)); | |
1670 | path = npath; | |
1671 | goto repeat; | |
1672 | } | |
1673 | ext_debug("next leaf has no free space(%d,%d)\n", | |
1674 | le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max)); | |
1675 | } | |
1676 | ||
1677 | /* | |
d0d856e8 RD |
1678 | * There is no free space in the found leaf. |
1679 | * We're gonna add a new leaf in the tree. | |
a86c6181 AT |
1680 | */ |
1681 | err = ext4_ext_create_new_leaf(handle, inode, path, newext); | |
1682 | if (err) | |
1683 | goto cleanup; | |
1684 | depth = ext_depth(inode); | |
1685 | eh = path[depth].p_hdr; | |
1686 | ||
1687 | has_space: | |
1688 | nearex = path[depth].p_ext; | |
1689 | ||
7e028976 AM |
1690 | err = ext4_ext_get_access(handle, inode, path + depth); |
1691 | if (err) | |
a86c6181 AT |
1692 | goto cleanup; |
1693 | ||
1694 | if (!nearex) { | |
1695 | /* there is no extent in this leaf, create first one */ | |
553f9008 | 1696 | ext_debug("first extent in the leaf: %d:%llu:[%d]%d\n", |
8c55e204 DK |
1697 | le32_to_cpu(newext->ee_block), |
1698 | ext_pblock(newext), | |
553f9008 | 1699 | ext4_ext_is_uninitialized(newext), |
a2df2a63 | 1700 | ext4_ext_get_actual_len(newext)); |
a86c6181 AT |
1701 | path[depth].p_ext = EXT_FIRST_EXTENT(eh); |
1702 | } else if (le32_to_cpu(newext->ee_block) | |
8c55e204 | 1703 | > le32_to_cpu(nearex->ee_block)) { |
a86c6181 AT |
1704 | /* BUG_ON(newext->ee_block == nearex->ee_block); */ |
1705 | if (nearex != EXT_LAST_EXTENT(eh)) { | |
1706 | len = EXT_MAX_EXTENT(eh) - nearex; | |
1707 | len = (len - 1) * sizeof(struct ext4_extent); | |
1708 | len = len < 0 ? 0 : len; | |
553f9008 | 1709 | ext_debug("insert %d:%llu:[%d]%d after: nearest 0x%p, " |
a86c6181 | 1710 | "move %d from 0x%p to 0x%p\n", |
8c55e204 DK |
1711 | le32_to_cpu(newext->ee_block), |
1712 | ext_pblock(newext), | |
553f9008 | 1713 | ext4_ext_is_uninitialized(newext), |
a2df2a63 | 1714 | ext4_ext_get_actual_len(newext), |
a86c6181 AT |
1715 | nearex, len, nearex + 1, nearex + 2); |
1716 | memmove(nearex + 2, nearex + 1, len); | |
1717 | } | |
1718 | path[depth].p_ext = nearex + 1; | |
1719 | } else { | |
1720 | BUG_ON(newext->ee_block == nearex->ee_block); | |
1721 | len = (EXT_MAX_EXTENT(eh) - nearex) * sizeof(struct ext4_extent); | |
1722 | len = len < 0 ? 0 : len; | |
553f9008 | 1723 | ext_debug("insert %d:%llu:[%d]%d before: nearest 0x%p, " |
a86c6181 AT |
1724 | "move %d from 0x%p to 0x%p\n", |
1725 | le32_to_cpu(newext->ee_block), | |
f65e6fba | 1726 | ext_pblock(newext), |
553f9008 | 1727 | ext4_ext_is_uninitialized(newext), |
a2df2a63 | 1728 | ext4_ext_get_actual_len(newext), |
a86c6181 AT |
1729 | nearex, len, nearex + 1, nearex + 2); |
1730 | memmove(nearex + 1, nearex, len); | |
1731 | path[depth].p_ext = nearex; | |
1732 | } | |
1733 | ||
e8546d06 | 1734 | le16_add_cpu(&eh->eh_entries, 1); |
a86c6181 AT |
1735 | nearex = path[depth].p_ext; |
1736 | nearex->ee_block = newext->ee_block; | |
b377611d | 1737 | ext4_ext_store_pblock(nearex, ext_pblock(newext)); |
a86c6181 | 1738 | nearex->ee_len = newext->ee_len; |
a86c6181 AT |
1739 | |
1740 | merge: | |
1741 | /* try to merge extents to the right */ | |
0031462b MC |
1742 | if (flag != EXT4_GET_BLOCKS_DIO_CREATE_EXT) |
1743 | ext4_ext_try_to_merge(inode, path, nearex); | |
a86c6181 AT |
1744 | |
1745 | /* try to merge extents to the left */ | |
1746 | ||
1747 | /* time to correct all indexes above */ | |
1748 | err = ext4_ext_correct_indexes(handle, inode, path); | |
1749 | if (err) | |
1750 | goto cleanup; | |
1751 | ||
1752 | err = ext4_ext_dirty(handle, inode, path + depth); | |
1753 | ||
1754 | cleanup: | |
1755 | if (npath) { | |
1756 | ext4_ext_drop_refs(npath); | |
1757 | kfree(npath); | |
1758 | } | |
a86c6181 AT |
1759 | ext4_ext_invalidate_cache(inode); |
1760 | return err; | |
1761 | } | |
1762 | ||
6873fa0d ES |
1763 | int ext4_ext_walk_space(struct inode *inode, ext4_lblk_t block, |
1764 | ext4_lblk_t num, ext_prepare_callback func, | |
1765 | void *cbdata) | |
1766 | { | |
1767 | struct ext4_ext_path *path = NULL; | |
1768 | struct ext4_ext_cache cbex; | |
1769 | struct ext4_extent *ex; | |
1770 | ext4_lblk_t next, start = 0, end = 0; | |
1771 | ext4_lblk_t last = block + num; | |
1772 | int depth, exists, err = 0; | |
1773 | ||
1774 | BUG_ON(func == NULL); | |
1775 | BUG_ON(inode == NULL); | |
1776 | ||
1777 | while (block < last && block != EXT_MAX_BLOCK) { | |
1778 | num = last - block; | |
1779 | /* find extent for this block */ | |
fab3a549 | 1780 | down_read(&EXT4_I(inode)->i_data_sem); |
6873fa0d | 1781 | path = ext4_ext_find_extent(inode, block, path); |
fab3a549 | 1782 | up_read(&EXT4_I(inode)->i_data_sem); |
6873fa0d ES |
1783 | if (IS_ERR(path)) { |
1784 | err = PTR_ERR(path); | |
1785 | path = NULL; | |
1786 | break; | |
1787 | } | |
1788 | ||
1789 | depth = ext_depth(inode); | |
1790 | BUG_ON(path[depth].p_hdr == NULL); | |
1791 | ex = path[depth].p_ext; | |
1792 | next = ext4_ext_next_allocated_block(path); | |
1793 | ||
1794 | exists = 0; | |
1795 | if (!ex) { | |
1796 | /* there is no extent yet, so try to allocate | |
1797 | * all requested space */ | |
1798 | start = block; | |
1799 | end = block + num; | |
1800 | } else if (le32_to_cpu(ex->ee_block) > block) { | |
1801 | /* need to allocate space before found extent */ | |
1802 | start = block; | |
1803 | end = le32_to_cpu(ex->ee_block); | |
1804 | if (block + num < end) | |
1805 | end = block + num; | |
1806 | } else if (block >= le32_to_cpu(ex->ee_block) | |
1807 | + ext4_ext_get_actual_len(ex)) { | |
1808 | /* need to allocate space after found extent */ | |
1809 | start = block; | |
1810 | end = block + num; | |
1811 | if (end >= next) | |
1812 | end = next; | |
1813 | } else if (block >= le32_to_cpu(ex->ee_block)) { | |
1814 | /* | |
1815 | * some part of requested space is covered | |
1816 | * by found extent | |
1817 | */ | |
1818 | start = block; | |
1819 | end = le32_to_cpu(ex->ee_block) | |
1820 | + ext4_ext_get_actual_len(ex); | |
1821 | if (block + num < end) | |
1822 | end = block + num; | |
1823 | exists = 1; | |
1824 | } else { | |
1825 | BUG(); | |
1826 | } | |
1827 | BUG_ON(end <= start); | |
1828 | ||
1829 | if (!exists) { | |
1830 | cbex.ec_block = start; | |
1831 | cbex.ec_len = end - start; | |
1832 | cbex.ec_start = 0; | |
1833 | cbex.ec_type = EXT4_EXT_CACHE_GAP; | |
1834 | } else { | |
1835 | cbex.ec_block = le32_to_cpu(ex->ee_block); | |
1836 | cbex.ec_len = ext4_ext_get_actual_len(ex); | |
1837 | cbex.ec_start = ext_pblock(ex); | |
1838 | cbex.ec_type = EXT4_EXT_CACHE_EXTENT; | |
1839 | } | |
1840 | ||
1841 | BUG_ON(cbex.ec_len == 0); | |
1842 | err = func(inode, path, &cbex, ex, cbdata); | |
1843 | ext4_ext_drop_refs(path); | |
1844 | ||
1845 | if (err < 0) | |
1846 | break; | |
1847 | ||
1848 | if (err == EXT_REPEAT) | |
1849 | continue; | |
1850 | else if (err == EXT_BREAK) { | |
1851 | err = 0; | |
1852 | break; | |
1853 | } | |
1854 | ||
1855 | if (ext_depth(inode) != depth) { | |
1856 | /* depth was changed. we have to realloc path */ | |
1857 | kfree(path); | |
1858 | path = NULL; | |
1859 | } | |
1860 | ||
1861 | block = cbex.ec_block + cbex.ec_len; | |
1862 | } | |
1863 | ||
1864 | if (path) { | |
1865 | ext4_ext_drop_refs(path); | |
1866 | kfree(path); | |
1867 | } | |
1868 | ||
1869 | return err; | |
1870 | } | |
1871 | ||
09b88252 | 1872 | static void |
725d26d3 | 1873 | ext4_ext_put_in_cache(struct inode *inode, ext4_lblk_t block, |
dd54567a | 1874 | __u32 len, ext4_fsblk_t start, int type) |
a86c6181 AT |
1875 | { |
1876 | struct ext4_ext_cache *cex; | |
1877 | BUG_ON(len == 0); | |
2ec0ae3a | 1878 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); |
a86c6181 AT |
1879 | cex = &EXT4_I(inode)->i_cached_extent; |
1880 | cex->ec_type = type; | |
1881 | cex->ec_block = block; | |
1882 | cex->ec_len = len; | |
1883 | cex->ec_start = start; | |
2ec0ae3a | 1884 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); |
a86c6181 AT |
1885 | } |
1886 | ||
1887 | /* | |
d0d856e8 RD |
1888 | * ext4_ext_put_gap_in_cache: |
1889 | * calculate boundaries of the gap that the requested block fits into | |
a86c6181 AT |
1890 | * and cache this gap |
1891 | */ | |
09b88252 | 1892 | static void |
a86c6181 | 1893 | ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path, |
725d26d3 | 1894 | ext4_lblk_t block) |
a86c6181 AT |
1895 | { |
1896 | int depth = ext_depth(inode); | |
725d26d3 AK |
1897 | unsigned long len; |
1898 | ext4_lblk_t lblock; | |
a86c6181 AT |
1899 | struct ext4_extent *ex; |
1900 | ||
1901 | ex = path[depth].p_ext; | |
1902 | if (ex == NULL) { | |
1903 | /* there is no extent yet, so gap is [0;-] */ | |
1904 | lblock = 0; | |
1905 | len = EXT_MAX_BLOCK; | |
1906 | ext_debug("cache gap(whole file):"); | |
1907 | } else if (block < le32_to_cpu(ex->ee_block)) { | |
1908 | lblock = block; | |
1909 | len = le32_to_cpu(ex->ee_block) - block; | |
bba90743 ES |
1910 | ext_debug("cache gap(before): %u [%u:%u]", |
1911 | block, | |
1912 | le32_to_cpu(ex->ee_block), | |
1913 | ext4_ext_get_actual_len(ex)); | |
a86c6181 | 1914 | } else if (block >= le32_to_cpu(ex->ee_block) |
a2df2a63 | 1915 | + ext4_ext_get_actual_len(ex)) { |
725d26d3 | 1916 | ext4_lblk_t next; |
8c55e204 | 1917 | lblock = le32_to_cpu(ex->ee_block) |
a2df2a63 | 1918 | + ext4_ext_get_actual_len(ex); |
725d26d3 AK |
1919 | |
1920 | next = ext4_ext_next_allocated_block(path); | |
bba90743 ES |
1921 | ext_debug("cache gap(after): [%u:%u] %u", |
1922 | le32_to_cpu(ex->ee_block), | |
1923 | ext4_ext_get_actual_len(ex), | |
1924 | block); | |
725d26d3 AK |
1925 | BUG_ON(next == lblock); |
1926 | len = next - lblock; | |
a86c6181 AT |
1927 | } else { |
1928 | lblock = len = 0; | |
1929 | BUG(); | |
1930 | } | |
1931 | ||
bba90743 | 1932 | ext_debug(" -> %u:%lu\n", lblock, len); |
a86c6181 AT |
1933 | ext4_ext_put_in_cache(inode, lblock, len, 0, EXT4_EXT_CACHE_GAP); |
1934 | } | |
1935 | ||
09b88252 | 1936 | static int |
725d26d3 | 1937 | ext4_ext_in_cache(struct inode *inode, ext4_lblk_t block, |
a86c6181 AT |
1938 | struct ext4_extent *ex) |
1939 | { | |
1940 | struct ext4_ext_cache *cex; | |
2ec0ae3a | 1941 | int ret = EXT4_EXT_CACHE_NO; |
a86c6181 | 1942 | |
2ec0ae3a TT |
1943 | /* |
1944 | * We borrow i_block_reservation_lock to protect i_cached_extent | |
1945 | */ | |
1946 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); | |
a86c6181 AT |
1947 | cex = &EXT4_I(inode)->i_cached_extent; |
1948 | ||
1949 | /* has cache valid data? */ | |
1950 | if (cex->ec_type == EXT4_EXT_CACHE_NO) | |
2ec0ae3a | 1951 | goto errout; |
a86c6181 AT |
1952 | |
1953 | BUG_ON(cex->ec_type != EXT4_EXT_CACHE_GAP && | |
1954 | cex->ec_type != EXT4_EXT_CACHE_EXTENT); | |
1955 | if (block >= cex->ec_block && block < cex->ec_block + cex->ec_len) { | |
8c55e204 | 1956 | ex->ee_block = cpu_to_le32(cex->ec_block); |
f65e6fba | 1957 | ext4_ext_store_pblock(ex, cex->ec_start); |
8c55e204 | 1958 | ex->ee_len = cpu_to_le16(cex->ec_len); |
bba90743 ES |
1959 | ext_debug("%u cached by %u:%u:%llu\n", |
1960 | block, | |
1961 | cex->ec_block, cex->ec_len, cex->ec_start); | |
2ec0ae3a | 1962 | ret = cex->ec_type; |
a86c6181 | 1963 | } |
2ec0ae3a TT |
1964 | errout: |
1965 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); | |
1966 | return ret; | |
a86c6181 AT |
1967 | } |
1968 | ||
1969 | /* | |
d0d856e8 RD |
1970 | * ext4_ext_rm_idx: |
1971 | * removes index from the index block. | |
1972 | * It's used in truncate case only, thus all requests are for | |
1973 | * last index in the block only. | |
a86c6181 | 1974 | */ |
1d03ec98 | 1975 | static int ext4_ext_rm_idx(handle_t *handle, struct inode *inode, |
a86c6181 AT |
1976 | struct ext4_ext_path *path) |
1977 | { | |
a86c6181 | 1978 | int err; |
f65e6fba | 1979 | ext4_fsblk_t leaf; |
a86c6181 AT |
1980 | |
1981 | /* free index block */ | |
1982 | path--; | |
f65e6fba | 1983 | leaf = idx_pblock(path->p_idx); |
a86c6181 | 1984 | BUG_ON(path->p_hdr->eh_entries == 0); |
7e028976 AM |
1985 | err = ext4_ext_get_access(handle, inode, path); |
1986 | if (err) | |
a86c6181 | 1987 | return err; |
e8546d06 | 1988 | le16_add_cpu(&path->p_hdr->eh_entries, -1); |
7e028976 AM |
1989 | err = ext4_ext_dirty(handle, inode, path); |
1990 | if (err) | |
a86c6181 | 1991 | return err; |
2ae02107 | 1992 | ext_debug("index is empty, remove it, free block %llu\n", leaf); |
e6362609 TT |
1993 | ext4_free_blocks(handle, inode, 0, leaf, 1, |
1994 | EXT4_FREE_BLOCKS_METADATA | EXT4_FREE_BLOCKS_FORGET); | |
a86c6181 AT |
1995 | return err; |
1996 | } | |
1997 | ||
1998 | /* | |
ee12b630 MC |
1999 | * ext4_ext_calc_credits_for_single_extent: |
2000 | * This routine returns max. credits that needed to insert an extent | |
2001 | * to the extent tree. | |
2002 | * When pass the actual path, the caller should calculate credits | |
2003 | * under i_data_sem. | |
a86c6181 | 2004 | */ |
525f4ed8 | 2005 | int ext4_ext_calc_credits_for_single_extent(struct inode *inode, int nrblocks, |
a86c6181 AT |
2006 | struct ext4_ext_path *path) |
2007 | { | |
a86c6181 | 2008 | if (path) { |
ee12b630 | 2009 | int depth = ext_depth(inode); |
f3bd1f3f | 2010 | int ret = 0; |
ee12b630 | 2011 | |
a86c6181 | 2012 | /* probably there is space in leaf? */ |
a86c6181 | 2013 | if (le16_to_cpu(path[depth].p_hdr->eh_entries) |
ee12b630 | 2014 | < le16_to_cpu(path[depth].p_hdr->eh_max)) { |
a86c6181 | 2015 | |
ee12b630 MC |
2016 | /* |
2017 | * There are some space in the leaf tree, no | |
2018 | * need to account for leaf block credit | |
2019 | * | |
2020 | * bitmaps and block group descriptor blocks | |
2021 | * and other metadat blocks still need to be | |
2022 | * accounted. | |
2023 | */ | |
525f4ed8 | 2024 | /* 1 bitmap, 1 block group descriptor */ |
ee12b630 | 2025 | ret = 2 + EXT4_META_TRANS_BLOCKS(inode->i_sb); |
5887e98b | 2026 | return ret; |
ee12b630 MC |
2027 | } |
2028 | } | |
a86c6181 | 2029 | |
525f4ed8 | 2030 | return ext4_chunk_trans_blocks(inode, nrblocks); |
ee12b630 | 2031 | } |
a86c6181 | 2032 | |
ee12b630 MC |
2033 | /* |
2034 | * How many index/leaf blocks need to change/allocate to modify nrblocks? | |
2035 | * | |
2036 | * if nrblocks are fit in a single extent (chunk flag is 1), then | |
2037 | * in the worse case, each tree level index/leaf need to be changed | |
2038 | * if the tree split due to insert a new extent, then the old tree | |
2039 | * index/leaf need to be updated too | |
2040 | * | |
2041 | * If the nrblocks are discontiguous, they could cause | |
2042 | * the whole tree split more than once, but this is really rare. | |
2043 | */ | |
525f4ed8 | 2044 | int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks, int chunk) |
ee12b630 MC |
2045 | { |
2046 | int index; | |
2047 | int depth = ext_depth(inode); | |
a86c6181 | 2048 | |
ee12b630 MC |
2049 | if (chunk) |
2050 | index = depth * 2; | |
2051 | else | |
2052 | index = depth * 3; | |
a86c6181 | 2053 | |
ee12b630 | 2054 | return index; |
a86c6181 AT |
2055 | } |
2056 | ||
2057 | static int ext4_remove_blocks(handle_t *handle, struct inode *inode, | |
2058 | struct ext4_extent *ex, | |
725d26d3 | 2059 | ext4_lblk_t from, ext4_lblk_t to) |
a86c6181 | 2060 | { |
a2df2a63 | 2061 | unsigned short ee_len = ext4_ext_get_actual_len(ex); |
e6362609 | 2062 | int flags = EXT4_FREE_BLOCKS_FORGET; |
a86c6181 | 2063 | |
c9de560d | 2064 | if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) |
e6362609 | 2065 | flags |= EXT4_FREE_BLOCKS_METADATA; |
a86c6181 AT |
2066 | #ifdef EXTENTS_STATS |
2067 | { | |
2068 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
a86c6181 AT |
2069 | spin_lock(&sbi->s_ext_stats_lock); |
2070 | sbi->s_ext_blocks += ee_len; | |
2071 | sbi->s_ext_extents++; | |
2072 | if (ee_len < sbi->s_ext_min) | |
2073 | sbi->s_ext_min = ee_len; | |
2074 | if (ee_len > sbi->s_ext_max) | |
2075 | sbi->s_ext_max = ee_len; | |
2076 | if (ext_depth(inode) > sbi->s_depth_max) | |
2077 | sbi->s_depth_max = ext_depth(inode); | |
2078 | spin_unlock(&sbi->s_ext_stats_lock); | |
2079 | } | |
2080 | #endif | |
2081 | if (from >= le32_to_cpu(ex->ee_block) | |
a2df2a63 | 2082 | && to == le32_to_cpu(ex->ee_block) + ee_len - 1) { |
a86c6181 | 2083 | /* tail removal */ |
725d26d3 | 2084 | ext4_lblk_t num; |
f65e6fba | 2085 | ext4_fsblk_t start; |
725d26d3 | 2086 | |
a2df2a63 AA |
2087 | num = le32_to_cpu(ex->ee_block) + ee_len - from; |
2088 | start = ext_pblock(ex) + ee_len - num; | |
725d26d3 | 2089 | ext_debug("free last %u blocks starting %llu\n", num, start); |
e6362609 | 2090 | ext4_free_blocks(handle, inode, 0, start, num, flags); |
a86c6181 | 2091 | } else if (from == le32_to_cpu(ex->ee_block) |
a2df2a63 | 2092 | && to <= le32_to_cpu(ex->ee_block) + ee_len - 1) { |
725d26d3 | 2093 | printk(KERN_INFO "strange request: removal %u-%u from %u:%u\n", |
a2df2a63 | 2094 | from, to, le32_to_cpu(ex->ee_block), ee_len); |
a86c6181 | 2095 | } else { |
725d26d3 AK |
2096 | printk(KERN_INFO "strange request: removal(2) " |
2097 | "%u-%u from %u:%u\n", | |
2098 | from, to, le32_to_cpu(ex->ee_block), ee_len); | |
a86c6181 AT |
2099 | } |
2100 | return 0; | |
2101 | } | |
2102 | ||
2103 | static int | |
2104 | ext4_ext_rm_leaf(handle_t *handle, struct inode *inode, | |
725d26d3 | 2105 | struct ext4_ext_path *path, ext4_lblk_t start) |
a86c6181 AT |
2106 | { |
2107 | int err = 0, correct_index = 0; | |
2108 | int depth = ext_depth(inode), credits; | |
2109 | struct ext4_extent_header *eh; | |
725d26d3 AK |
2110 | ext4_lblk_t a, b, block; |
2111 | unsigned num; | |
2112 | ext4_lblk_t ex_ee_block; | |
a86c6181 | 2113 | unsigned short ex_ee_len; |
a2df2a63 | 2114 | unsigned uninitialized = 0; |
a86c6181 AT |
2115 | struct ext4_extent *ex; |
2116 | ||
c29c0ae7 | 2117 | /* the header must be checked already in ext4_ext_remove_space() */ |
725d26d3 | 2118 | ext_debug("truncate since %u in leaf\n", start); |
a86c6181 AT |
2119 | if (!path[depth].p_hdr) |
2120 | path[depth].p_hdr = ext_block_hdr(path[depth].p_bh); | |
2121 | eh = path[depth].p_hdr; | |
2122 | BUG_ON(eh == NULL); | |
a86c6181 AT |
2123 | |
2124 | /* find where to start removing */ | |
2125 | ex = EXT_LAST_EXTENT(eh); | |
2126 | ||
2127 | ex_ee_block = le32_to_cpu(ex->ee_block); | |
a2df2a63 | 2128 | ex_ee_len = ext4_ext_get_actual_len(ex); |
a86c6181 AT |
2129 | |
2130 | while (ex >= EXT_FIRST_EXTENT(eh) && | |
2131 | ex_ee_block + ex_ee_len > start) { | |
a41f2071 AK |
2132 | |
2133 | if (ext4_ext_is_uninitialized(ex)) | |
2134 | uninitialized = 1; | |
2135 | else | |
2136 | uninitialized = 0; | |
2137 | ||
553f9008 M |
2138 | ext_debug("remove ext %u:[%d]%d\n", ex_ee_block, |
2139 | uninitialized, ex_ee_len); | |
a86c6181 AT |
2140 | path[depth].p_ext = ex; |
2141 | ||
2142 | a = ex_ee_block > start ? ex_ee_block : start; | |
2143 | b = ex_ee_block + ex_ee_len - 1 < EXT_MAX_BLOCK ? | |
2144 | ex_ee_block + ex_ee_len - 1 : EXT_MAX_BLOCK; | |
2145 | ||
2146 | ext_debug(" border %u:%u\n", a, b); | |
2147 | ||
2148 | if (a != ex_ee_block && b != ex_ee_block + ex_ee_len - 1) { | |
2149 | block = 0; | |
2150 | num = 0; | |
2151 | BUG(); | |
2152 | } else if (a != ex_ee_block) { | |
2153 | /* remove tail of the extent */ | |
2154 | block = ex_ee_block; | |
2155 | num = a - block; | |
2156 | } else if (b != ex_ee_block + ex_ee_len - 1) { | |
2157 | /* remove head of the extent */ | |
2158 | block = a; | |
2159 | num = b - a; | |
2160 | /* there is no "make a hole" API yet */ | |
2161 | BUG(); | |
2162 | } else { | |
2163 | /* remove whole extent: excellent! */ | |
2164 | block = ex_ee_block; | |
2165 | num = 0; | |
2166 | BUG_ON(a != ex_ee_block); | |
2167 | BUG_ON(b != ex_ee_block + ex_ee_len - 1); | |
2168 | } | |
2169 | ||
34071da7 TT |
2170 | /* |
2171 | * 3 for leaf, sb, and inode plus 2 (bmap and group | |
2172 | * descriptor) for each block group; assume two block | |
2173 | * groups plus ex_ee_len/blocks_per_block_group for | |
2174 | * the worst case | |
2175 | */ | |
2176 | credits = 7 + 2*(ex_ee_len/EXT4_BLOCKS_PER_GROUP(inode->i_sb)); | |
a86c6181 AT |
2177 | if (ex == EXT_FIRST_EXTENT(eh)) { |
2178 | correct_index = 1; | |
2179 | credits += (ext_depth(inode)) + 1; | |
2180 | } | |
5aca07eb | 2181 | credits += EXT4_MAXQUOTAS_TRANS_BLOCKS(inode->i_sb); |
a86c6181 | 2182 | |
487caeef | 2183 | err = ext4_ext_truncate_extend_restart(handle, inode, credits); |
9102e4fa | 2184 | if (err) |
a86c6181 | 2185 | goto out; |
a86c6181 AT |
2186 | |
2187 | err = ext4_ext_get_access(handle, inode, path + depth); | |
2188 | if (err) | |
2189 | goto out; | |
2190 | ||
2191 | err = ext4_remove_blocks(handle, inode, ex, a, b); | |
2192 | if (err) | |
2193 | goto out; | |
2194 | ||
2195 | if (num == 0) { | |
d0d856e8 | 2196 | /* this extent is removed; mark slot entirely unused */ |
f65e6fba | 2197 | ext4_ext_store_pblock(ex, 0); |
e8546d06 | 2198 | le16_add_cpu(&eh->eh_entries, -1); |
a86c6181 AT |
2199 | } |
2200 | ||
2201 | ex->ee_block = cpu_to_le32(block); | |
2202 | ex->ee_len = cpu_to_le16(num); | |
749269fa AA |
2203 | /* |
2204 | * Do not mark uninitialized if all the blocks in the | |
2205 | * extent have been removed. | |
2206 | */ | |
2207 | if (uninitialized && num) | |
a2df2a63 | 2208 | ext4_ext_mark_uninitialized(ex); |
a86c6181 AT |
2209 | |
2210 | err = ext4_ext_dirty(handle, inode, path + depth); | |
2211 | if (err) | |
2212 | goto out; | |
2213 | ||
2ae02107 | 2214 | ext_debug("new extent: %u:%u:%llu\n", block, num, |
f65e6fba | 2215 | ext_pblock(ex)); |
a86c6181 AT |
2216 | ex--; |
2217 | ex_ee_block = le32_to_cpu(ex->ee_block); | |
a2df2a63 | 2218 | ex_ee_len = ext4_ext_get_actual_len(ex); |
a86c6181 AT |
2219 | } |
2220 | ||
2221 | if (correct_index && eh->eh_entries) | |
2222 | err = ext4_ext_correct_indexes(handle, inode, path); | |
2223 | ||
2224 | /* if this leaf is free, then we should | |
2225 | * remove it from index block above */ | |
2226 | if (err == 0 && eh->eh_entries == 0 && path[depth].p_bh != NULL) | |
2227 | err = ext4_ext_rm_idx(handle, inode, path + depth); | |
2228 | ||
2229 | out: | |
2230 | return err; | |
2231 | } | |
2232 | ||
2233 | /* | |
d0d856e8 RD |
2234 | * ext4_ext_more_to_rm: |
2235 | * returns 1 if current index has to be freed (even partial) | |
a86c6181 | 2236 | */ |
09b88252 | 2237 | static int |
a86c6181 AT |
2238 | ext4_ext_more_to_rm(struct ext4_ext_path *path) |
2239 | { | |
2240 | BUG_ON(path->p_idx == NULL); | |
2241 | ||
2242 | if (path->p_idx < EXT_FIRST_INDEX(path->p_hdr)) | |
2243 | return 0; | |
2244 | ||
2245 | /* | |
d0d856e8 | 2246 | * if truncate on deeper level happened, it wasn't partial, |
a86c6181 AT |
2247 | * so we have to consider current index for truncation |
2248 | */ | |
2249 | if (le16_to_cpu(path->p_hdr->eh_entries) == path->p_block) | |
2250 | return 0; | |
2251 | return 1; | |
2252 | } | |
2253 | ||
1d03ec98 | 2254 | static int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start) |
a86c6181 AT |
2255 | { |
2256 | struct super_block *sb = inode->i_sb; | |
2257 | int depth = ext_depth(inode); | |
2258 | struct ext4_ext_path *path; | |
2259 | handle_t *handle; | |
2260 | int i = 0, err = 0; | |
2261 | ||
725d26d3 | 2262 | ext_debug("truncate since %u\n", start); |
a86c6181 AT |
2263 | |
2264 | /* probably first extent we're gonna free will be last in block */ | |
2265 | handle = ext4_journal_start(inode, depth + 1); | |
2266 | if (IS_ERR(handle)) | |
2267 | return PTR_ERR(handle); | |
2268 | ||
2269 | ext4_ext_invalidate_cache(inode); | |
2270 | ||
2271 | /* | |
d0d856e8 RD |
2272 | * We start scanning from right side, freeing all the blocks |
2273 | * after i_size and walking into the tree depth-wise. | |
a86c6181 | 2274 | */ |
216553c4 | 2275 | path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1), GFP_NOFS); |
a86c6181 AT |
2276 | if (path == NULL) { |
2277 | ext4_journal_stop(handle); | |
2278 | return -ENOMEM; | |
2279 | } | |
a86c6181 | 2280 | path[0].p_hdr = ext_inode_hdr(inode); |
56b19868 | 2281 | if (ext4_ext_check(inode, path[0].p_hdr, depth)) { |
a86c6181 AT |
2282 | err = -EIO; |
2283 | goto out; | |
2284 | } | |
2285 | path[0].p_depth = depth; | |
2286 | ||
2287 | while (i >= 0 && err == 0) { | |
2288 | if (i == depth) { | |
2289 | /* this is leaf block */ | |
2290 | err = ext4_ext_rm_leaf(handle, inode, path, start); | |
d0d856e8 | 2291 | /* root level has p_bh == NULL, brelse() eats this */ |
a86c6181 AT |
2292 | brelse(path[i].p_bh); |
2293 | path[i].p_bh = NULL; | |
2294 | i--; | |
2295 | continue; | |
2296 | } | |
2297 | ||
2298 | /* this is index block */ | |
2299 | if (!path[i].p_hdr) { | |
2300 | ext_debug("initialize header\n"); | |
2301 | path[i].p_hdr = ext_block_hdr(path[i].p_bh); | |
a86c6181 AT |
2302 | } |
2303 | ||
a86c6181 | 2304 | if (!path[i].p_idx) { |
d0d856e8 | 2305 | /* this level hasn't been touched yet */ |
a86c6181 AT |
2306 | path[i].p_idx = EXT_LAST_INDEX(path[i].p_hdr); |
2307 | path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries)+1; | |
2308 | ext_debug("init index ptr: hdr 0x%p, num %d\n", | |
2309 | path[i].p_hdr, | |
2310 | le16_to_cpu(path[i].p_hdr->eh_entries)); | |
2311 | } else { | |
d0d856e8 | 2312 | /* we were already here, see at next index */ |
a86c6181 AT |
2313 | path[i].p_idx--; |
2314 | } | |
2315 | ||
2316 | ext_debug("level %d - index, first 0x%p, cur 0x%p\n", | |
2317 | i, EXT_FIRST_INDEX(path[i].p_hdr), | |
2318 | path[i].p_idx); | |
2319 | if (ext4_ext_more_to_rm(path + i)) { | |
c29c0ae7 | 2320 | struct buffer_head *bh; |
a86c6181 | 2321 | /* go to the next level */ |
2ae02107 | 2322 | ext_debug("move to level %d (block %llu)\n", |
f65e6fba | 2323 | i + 1, idx_pblock(path[i].p_idx)); |
a86c6181 | 2324 | memset(path + i + 1, 0, sizeof(*path)); |
c29c0ae7 AT |
2325 | bh = sb_bread(sb, idx_pblock(path[i].p_idx)); |
2326 | if (!bh) { | |
a86c6181 AT |
2327 | /* should we reset i_size? */ |
2328 | err = -EIO; | |
2329 | break; | |
2330 | } | |
c29c0ae7 AT |
2331 | if (WARN_ON(i + 1 > depth)) { |
2332 | err = -EIO; | |
2333 | break; | |
2334 | } | |
56b19868 | 2335 | if (ext4_ext_check(inode, ext_block_hdr(bh), |
c29c0ae7 AT |
2336 | depth - i - 1)) { |
2337 | err = -EIO; | |
2338 | break; | |
2339 | } | |
2340 | path[i + 1].p_bh = bh; | |
a86c6181 | 2341 | |
d0d856e8 RD |
2342 | /* save actual number of indexes since this |
2343 | * number is changed at the next iteration */ | |
a86c6181 AT |
2344 | path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries); |
2345 | i++; | |
2346 | } else { | |
d0d856e8 | 2347 | /* we finished processing this index, go up */ |
a86c6181 | 2348 | if (path[i].p_hdr->eh_entries == 0 && i > 0) { |
d0d856e8 | 2349 | /* index is empty, remove it; |
a86c6181 AT |
2350 | * handle must be already prepared by the |
2351 | * truncatei_leaf() */ | |
2352 | err = ext4_ext_rm_idx(handle, inode, path + i); | |
2353 | } | |
d0d856e8 | 2354 | /* root level has p_bh == NULL, brelse() eats this */ |
a86c6181 AT |
2355 | brelse(path[i].p_bh); |
2356 | path[i].p_bh = NULL; | |
2357 | i--; | |
2358 | ext_debug("return to level %d\n", i); | |
2359 | } | |
2360 | } | |
2361 | ||
2362 | /* TODO: flexible tree reduction should be here */ | |
2363 | if (path->p_hdr->eh_entries == 0) { | |
2364 | /* | |
d0d856e8 RD |
2365 | * truncate to zero freed all the tree, |
2366 | * so we need to correct eh_depth | |
a86c6181 AT |
2367 | */ |
2368 | err = ext4_ext_get_access(handle, inode, path); | |
2369 | if (err == 0) { | |
2370 | ext_inode_hdr(inode)->eh_depth = 0; | |
2371 | ext_inode_hdr(inode)->eh_max = | |
55ad63bf | 2372 | cpu_to_le16(ext4_ext_space_root(inode, 0)); |
a86c6181 AT |
2373 | err = ext4_ext_dirty(handle, inode, path); |
2374 | } | |
2375 | } | |
2376 | out: | |
a86c6181 AT |
2377 | ext4_ext_drop_refs(path); |
2378 | kfree(path); | |
2379 | ext4_journal_stop(handle); | |
2380 | ||
2381 | return err; | |
2382 | } | |
2383 | ||
2384 | /* | |
2385 | * called at mount time | |
2386 | */ | |
2387 | void ext4_ext_init(struct super_block *sb) | |
2388 | { | |
2389 | /* | |
2390 | * possible initialization would be here | |
2391 | */ | |
2392 | ||
83982b6f | 2393 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) { |
90576c0b | 2394 | #if defined(AGGRESSIVE_TEST) || defined(CHECK_BINSEARCH) || defined(EXTENTS_STATS) |
4776004f | 2395 | printk(KERN_INFO "EXT4-fs: file extents enabled"); |
bbf2f9fb RD |
2396 | #ifdef AGGRESSIVE_TEST |
2397 | printk(", aggressive tests"); | |
a86c6181 AT |
2398 | #endif |
2399 | #ifdef CHECK_BINSEARCH | |
2400 | printk(", check binsearch"); | |
2401 | #endif | |
2402 | #ifdef EXTENTS_STATS | |
2403 | printk(", stats"); | |
2404 | #endif | |
2405 | printk("\n"); | |
90576c0b | 2406 | #endif |
a86c6181 AT |
2407 | #ifdef EXTENTS_STATS |
2408 | spin_lock_init(&EXT4_SB(sb)->s_ext_stats_lock); | |
2409 | EXT4_SB(sb)->s_ext_min = 1 << 30; | |
2410 | EXT4_SB(sb)->s_ext_max = 0; | |
2411 | #endif | |
2412 | } | |
2413 | } | |
2414 | ||
2415 | /* | |
2416 | * called at umount time | |
2417 | */ | |
2418 | void ext4_ext_release(struct super_block *sb) | |
2419 | { | |
83982b6f | 2420 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) |
a86c6181 AT |
2421 | return; |
2422 | ||
2423 | #ifdef EXTENTS_STATS | |
2424 | if (EXT4_SB(sb)->s_ext_blocks && EXT4_SB(sb)->s_ext_extents) { | |
2425 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2426 | printk(KERN_ERR "EXT4-fs: %lu blocks in %lu extents (%lu ave)\n", | |
2427 | sbi->s_ext_blocks, sbi->s_ext_extents, | |
2428 | sbi->s_ext_blocks / sbi->s_ext_extents); | |
2429 | printk(KERN_ERR "EXT4-fs: extents: %lu min, %lu max, max depth %lu\n", | |
2430 | sbi->s_ext_min, sbi->s_ext_max, sbi->s_depth_max); | |
2431 | } | |
2432 | #endif | |
2433 | } | |
2434 | ||
093a088b AK |
2435 | static void bi_complete(struct bio *bio, int error) |
2436 | { | |
2437 | complete((struct completion *)bio->bi_private); | |
2438 | } | |
2439 | ||
2440 | /* FIXME!! we need to try to merge to left or right after zero-out */ | |
2441 | static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex) | |
2442 | { | |
2443 | int ret = -EIO; | |
2444 | struct bio *bio; | |
2445 | int blkbits, blocksize; | |
2446 | sector_t ee_pblock; | |
2447 | struct completion event; | |
2448 | unsigned int ee_len, len, done, offset; | |
2449 | ||
2450 | ||
2451 | blkbits = inode->i_blkbits; | |
2452 | blocksize = inode->i_sb->s_blocksize; | |
2453 | ee_len = ext4_ext_get_actual_len(ex); | |
2454 | ee_pblock = ext_pblock(ex); | |
2455 | ||
2456 | /* convert ee_pblock to 512 byte sectors */ | |
2457 | ee_pblock = ee_pblock << (blkbits - 9); | |
2458 | ||
2459 | while (ee_len > 0) { | |
2460 | ||
2461 | if (ee_len > BIO_MAX_PAGES) | |
2462 | len = BIO_MAX_PAGES; | |
2463 | else | |
2464 | len = ee_len; | |
2465 | ||
2466 | bio = bio_alloc(GFP_NOIO, len); | |
093a088b AK |
2467 | bio->bi_sector = ee_pblock; |
2468 | bio->bi_bdev = inode->i_sb->s_bdev; | |
2469 | ||
2470 | done = 0; | |
2471 | offset = 0; | |
2472 | while (done < len) { | |
2473 | ret = bio_add_page(bio, ZERO_PAGE(0), | |
2474 | blocksize, offset); | |
2475 | if (ret != blocksize) { | |
2476 | /* | |
2477 | * We can't add any more pages because of | |
2478 | * hardware limitations. Start a new bio. | |
2479 | */ | |
2480 | break; | |
2481 | } | |
2482 | done++; | |
2483 | offset += blocksize; | |
2484 | if (offset >= PAGE_CACHE_SIZE) | |
2485 | offset = 0; | |
2486 | } | |
2487 | ||
2488 | init_completion(&event); | |
2489 | bio->bi_private = &event; | |
2490 | bio->bi_end_io = bi_complete; | |
2491 | submit_bio(WRITE, bio); | |
2492 | wait_for_completion(&event); | |
2493 | ||
2494 | if (test_bit(BIO_UPTODATE, &bio->bi_flags)) | |
2495 | ret = 0; | |
2496 | else { | |
2497 | ret = -EIO; | |
2498 | break; | |
2499 | } | |
2500 | bio_put(bio); | |
2501 | ee_len -= done; | |
2502 | ee_pblock += done << (blkbits - 9); | |
2503 | } | |
2504 | return ret; | |
2505 | } | |
2506 | ||
3977c965 | 2507 | #define EXT4_EXT_ZERO_LEN 7 |
56055d3a AA |
2508 | /* |
2509 | * This function is called by ext4_ext_get_blocks() if someone tries to write | |
2510 | * to an uninitialized extent. It may result in splitting the uninitialized | |
2511 | * extent into multiple extents (upto three - one initialized and two | |
2512 | * uninitialized). | |
2513 | * There are three possibilities: | |
2514 | * a> There is no split required: Entire extent should be initialized | |
2515 | * b> Splits in two extents: Write is happening at either end of the extent | |
2516 | * c> Splits in three extents: Somone is writing in middle of the extent | |
2517 | */ | |
725d26d3 AK |
2518 | static int ext4_ext_convert_to_initialized(handle_t *handle, |
2519 | struct inode *inode, | |
2520 | struct ext4_ext_path *path, | |
2521 | ext4_lblk_t iblock, | |
498e5f24 | 2522 | unsigned int max_blocks) |
56055d3a | 2523 | { |
95c3889c | 2524 | struct ext4_extent *ex, newex, orig_ex; |
56055d3a AA |
2525 | struct ext4_extent *ex1 = NULL; |
2526 | struct ext4_extent *ex2 = NULL; | |
2527 | struct ext4_extent *ex3 = NULL; | |
2528 | struct ext4_extent_header *eh; | |
725d26d3 AK |
2529 | ext4_lblk_t ee_block; |
2530 | unsigned int allocated, ee_len, depth; | |
56055d3a AA |
2531 | ext4_fsblk_t newblock; |
2532 | int err = 0; | |
2533 | int ret = 0; | |
2534 | ||
2535 | depth = ext_depth(inode); | |
2536 | eh = path[depth].p_hdr; | |
2537 | ex = path[depth].p_ext; | |
2538 | ee_block = le32_to_cpu(ex->ee_block); | |
2539 | ee_len = ext4_ext_get_actual_len(ex); | |
2540 | allocated = ee_len - (iblock - ee_block); | |
2541 | newblock = iblock - ee_block + ext_pblock(ex); | |
2542 | ex2 = ex; | |
95c3889c AK |
2543 | orig_ex.ee_block = ex->ee_block; |
2544 | orig_ex.ee_len = cpu_to_le16(ee_len); | |
2545 | ext4_ext_store_pblock(&orig_ex, ext_pblock(ex)); | |
56055d3a | 2546 | |
9df5643a AK |
2547 | err = ext4_ext_get_access(handle, inode, path + depth); |
2548 | if (err) | |
2549 | goto out; | |
3977c965 AK |
2550 | /* If extent has less than 2*EXT4_EXT_ZERO_LEN zerout directly */ |
2551 | if (ee_len <= 2*EXT4_EXT_ZERO_LEN) { | |
2552 | err = ext4_ext_zeroout(inode, &orig_ex); | |
2553 | if (err) | |
2554 | goto fix_extent_len; | |
2555 | /* update the extent length and mark as initialized */ | |
2556 | ex->ee_block = orig_ex.ee_block; | |
2557 | ex->ee_len = orig_ex.ee_len; | |
2558 | ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); | |
2559 | ext4_ext_dirty(handle, inode, path + depth); | |
161e7b7c AK |
2560 | /* zeroed the full extent */ |
2561 | return allocated; | |
3977c965 | 2562 | } |
9df5643a | 2563 | |
56055d3a AA |
2564 | /* ex1: ee_block to iblock - 1 : uninitialized */ |
2565 | if (iblock > ee_block) { | |
2566 | ex1 = ex; | |
2567 | ex1->ee_len = cpu_to_le16(iblock - ee_block); | |
2568 | ext4_ext_mark_uninitialized(ex1); | |
2569 | ex2 = &newex; | |
2570 | } | |
2571 | /* | |
2572 | * for sanity, update the length of the ex2 extent before | |
2573 | * we insert ex3, if ex1 is NULL. This is to avoid temporary | |
2574 | * overlap of blocks. | |
2575 | */ | |
2576 | if (!ex1 && allocated > max_blocks) | |
2577 | ex2->ee_len = cpu_to_le16(max_blocks); | |
2578 | /* ex3: to ee_block + ee_len : uninitialised */ | |
2579 | if (allocated > max_blocks) { | |
2580 | unsigned int newdepth; | |
3977c965 AK |
2581 | /* If extent has less than EXT4_EXT_ZERO_LEN zerout directly */ |
2582 | if (allocated <= EXT4_EXT_ZERO_LEN) { | |
d03856bd AK |
2583 | /* |
2584 | * iblock == ee_block is handled by the zerouout | |
2585 | * at the beginning. | |
2586 | * Mark first half uninitialized. | |
3977c965 AK |
2587 | * Mark second half initialized and zero out the |
2588 | * initialized extent | |
2589 | */ | |
2590 | ex->ee_block = orig_ex.ee_block; | |
2591 | ex->ee_len = cpu_to_le16(ee_len - allocated); | |
2592 | ext4_ext_mark_uninitialized(ex); | |
2593 | ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); | |
2594 | ext4_ext_dirty(handle, inode, path + depth); | |
2595 | ||
2596 | ex3 = &newex; | |
2597 | ex3->ee_block = cpu_to_le32(iblock); | |
2598 | ext4_ext_store_pblock(ex3, newblock); | |
2599 | ex3->ee_len = cpu_to_le16(allocated); | |
0031462b MC |
2600 | err = ext4_ext_insert_extent(handle, inode, path, |
2601 | ex3, 0); | |
3977c965 AK |
2602 | if (err == -ENOSPC) { |
2603 | err = ext4_ext_zeroout(inode, &orig_ex); | |
2604 | if (err) | |
2605 | goto fix_extent_len; | |
2606 | ex->ee_block = orig_ex.ee_block; | |
2607 | ex->ee_len = orig_ex.ee_len; | |
2608 | ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); | |
2609 | ext4_ext_dirty(handle, inode, path + depth); | |
d03856bd | 2610 | /* blocks available from iblock */ |
161e7b7c | 2611 | return allocated; |
3977c965 AK |
2612 | |
2613 | } else if (err) | |
2614 | goto fix_extent_len; | |
2615 | ||
161e7b7c AK |
2616 | /* |
2617 | * We need to zero out the second half because | |
2618 | * an fallocate request can update file size and | |
2619 | * converting the second half to initialized extent | |
2620 | * implies that we can leak some junk data to user | |
2621 | * space. | |
2622 | */ | |
2623 | err = ext4_ext_zeroout(inode, ex3); | |
2624 | if (err) { | |
2625 | /* | |
2626 | * We should actually mark the | |
2627 | * second half as uninit and return error | |
2628 | * Insert would have changed the extent | |
2629 | */ | |
2630 | depth = ext_depth(inode); | |
2631 | ext4_ext_drop_refs(path); | |
2632 | path = ext4_ext_find_extent(inode, | |
2633 | iblock, path); | |
2634 | if (IS_ERR(path)) { | |
2635 | err = PTR_ERR(path); | |
2636 | return err; | |
2637 | } | |
d03856bd | 2638 | /* get the second half extent details */ |
161e7b7c AK |
2639 | ex = path[depth].p_ext; |
2640 | err = ext4_ext_get_access(handle, inode, | |
2641 | path + depth); | |
2642 | if (err) | |
2643 | return err; | |
2644 | ext4_ext_mark_uninitialized(ex); | |
2645 | ext4_ext_dirty(handle, inode, path + depth); | |
2646 | return err; | |
2647 | } | |
2648 | ||
2649 | /* zeroed the second half */ | |
3977c965 AK |
2650 | return allocated; |
2651 | } | |
56055d3a AA |
2652 | ex3 = &newex; |
2653 | ex3->ee_block = cpu_to_le32(iblock + max_blocks); | |
2654 | ext4_ext_store_pblock(ex3, newblock + max_blocks); | |
2655 | ex3->ee_len = cpu_to_le16(allocated - max_blocks); | |
2656 | ext4_ext_mark_uninitialized(ex3); | |
0031462b | 2657 | err = ext4_ext_insert_extent(handle, inode, path, ex3, 0); |
093a088b AK |
2658 | if (err == -ENOSPC) { |
2659 | err = ext4_ext_zeroout(inode, &orig_ex); | |
2660 | if (err) | |
2661 | goto fix_extent_len; | |
2662 | /* update the extent length and mark as initialized */ | |
95c3889c AK |
2663 | ex->ee_block = orig_ex.ee_block; |
2664 | ex->ee_len = orig_ex.ee_len; | |
2665 | ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); | |
95c3889c | 2666 | ext4_ext_dirty(handle, inode, path + depth); |
161e7b7c | 2667 | /* zeroed the full extent */ |
d03856bd | 2668 | /* blocks available from iblock */ |
161e7b7c | 2669 | return allocated; |
093a088b AK |
2670 | |
2671 | } else if (err) | |
2672 | goto fix_extent_len; | |
56055d3a AA |
2673 | /* |
2674 | * The depth, and hence eh & ex might change | |
2675 | * as part of the insert above. | |
2676 | */ | |
2677 | newdepth = ext_depth(inode); | |
95c3889c | 2678 | /* |
73ac36ea | 2679 | * update the extent length after successful insert of the |
95c3889c AK |
2680 | * split extent |
2681 | */ | |
2682 | orig_ex.ee_len = cpu_to_le16(ee_len - | |
2683 | ext4_ext_get_actual_len(ex3)); | |
d03856bd AK |
2684 | depth = newdepth; |
2685 | ext4_ext_drop_refs(path); | |
2686 | path = ext4_ext_find_extent(inode, iblock, path); | |
2687 | if (IS_ERR(path)) { | |
2688 | err = PTR_ERR(path); | |
2689 | goto out; | |
56055d3a | 2690 | } |
d03856bd AK |
2691 | eh = path[depth].p_hdr; |
2692 | ex = path[depth].p_ext; | |
2693 | if (ex2 != &newex) | |
2694 | ex2 = ex; | |
2695 | ||
2696 | err = ext4_ext_get_access(handle, inode, path + depth); | |
2697 | if (err) | |
2698 | goto out; | |
2699 | ||
56055d3a | 2700 | allocated = max_blocks; |
3977c965 AK |
2701 | |
2702 | /* If extent has less than EXT4_EXT_ZERO_LEN and we are trying | |
2703 | * to insert a extent in the middle zerout directly | |
2704 | * otherwise give the extent a chance to merge to left | |
2705 | */ | |
2706 | if (le16_to_cpu(orig_ex.ee_len) <= EXT4_EXT_ZERO_LEN && | |
2707 | iblock != ee_block) { | |
2708 | err = ext4_ext_zeroout(inode, &orig_ex); | |
2709 | if (err) | |
2710 | goto fix_extent_len; | |
2711 | /* update the extent length and mark as initialized */ | |
2712 | ex->ee_block = orig_ex.ee_block; | |
2713 | ex->ee_len = orig_ex.ee_len; | |
2714 | ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); | |
2715 | ext4_ext_dirty(handle, inode, path + depth); | |
161e7b7c | 2716 | /* zero out the first half */ |
d03856bd | 2717 | /* blocks available from iblock */ |
161e7b7c | 2718 | return allocated; |
3977c965 | 2719 | } |
56055d3a AA |
2720 | } |
2721 | /* | |
2722 | * If there was a change of depth as part of the | |
2723 | * insertion of ex3 above, we need to update the length | |
2724 | * of the ex1 extent again here | |
2725 | */ | |
2726 | if (ex1 && ex1 != ex) { | |
2727 | ex1 = ex; | |
2728 | ex1->ee_len = cpu_to_le16(iblock - ee_block); | |
2729 | ext4_ext_mark_uninitialized(ex1); | |
2730 | ex2 = &newex; | |
2731 | } | |
2732 | /* ex2: iblock to iblock + maxblocks-1 : initialised */ | |
2733 | ex2->ee_block = cpu_to_le32(iblock); | |
56055d3a AA |
2734 | ext4_ext_store_pblock(ex2, newblock); |
2735 | ex2->ee_len = cpu_to_le16(allocated); | |
2736 | if (ex2 != ex) | |
2737 | goto insert; | |
56055d3a AA |
2738 | /* |
2739 | * New (initialized) extent starts from the first block | |
2740 | * in the current extent. i.e., ex2 == ex | |
2741 | * We have to see if it can be merged with the extent | |
2742 | * on the left. | |
2743 | */ | |
2744 | if (ex2 > EXT_FIRST_EXTENT(eh)) { | |
2745 | /* | |
2746 | * To merge left, pass "ex2 - 1" to try_to_merge(), | |
2747 | * since it merges towards right _only_. | |
2748 | */ | |
2749 | ret = ext4_ext_try_to_merge(inode, path, ex2 - 1); | |
2750 | if (ret) { | |
2751 | err = ext4_ext_correct_indexes(handle, inode, path); | |
2752 | if (err) | |
2753 | goto out; | |
2754 | depth = ext_depth(inode); | |
2755 | ex2--; | |
2756 | } | |
2757 | } | |
2758 | /* | |
2759 | * Try to Merge towards right. This might be required | |
2760 | * only when the whole extent is being written to. | |
2761 | * i.e. ex2 == ex and ex3 == NULL. | |
2762 | */ | |
2763 | if (!ex3) { | |
2764 | ret = ext4_ext_try_to_merge(inode, path, ex2); | |
2765 | if (ret) { | |
2766 | err = ext4_ext_correct_indexes(handle, inode, path); | |
2767 | if (err) | |
2768 | goto out; | |
2769 | } | |
2770 | } | |
2771 | /* Mark modified extent as dirty */ | |
2772 | err = ext4_ext_dirty(handle, inode, path + depth); | |
2773 | goto out; | |
2774 | insert: | |
0031462b | 2775 | err = ext4_ext_insert_extent(handle, inode, path, &newex, 0); |
093a088b AK |
2776 | if (err == -ENOSPC) { |
2777 | err = ext4_ext_zeroout(inode, &orig_ex); | |
2778 | if (err) | |
2779 | goto fix_extent_len; | |
2780 | /* update the extent length and mark as initialized */ | |
95c3889c AK |
2781 | ex->ee_block = orig_ex.ee_block; |
2782 | ex->ee_len = orig_ex.ee_len; | |
2783 | ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); | |
95c3889c | 2784 | ext4_ext_dirty(handle, inode, path + depth); |
161e7b7c AK |
2785 | /* zero out the first half */ |
2786 | return allocated; | |
093a088b AK |
2787 | } else if (err) |
2788 | goto fix_extent_len; | |
56055d3a | 2789 | out: |
553f9008 | 2790 | ext4_ext_show_leaf(inode, path); |
56055d3a | 2791 | return err ? err : allocated; |
093a088b AK |
2792 | |
2793 | fix_extent_len: | |
2794 | ex->ee_block = orig_ex.ee_block; | |
2795 | ex->ee_len = orig_ex.ee_len; | |
2796 | ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); | |
2797 | ext4_ext_mark_uninitialized(ex); | |
2798 | ext4_ext_dirty(handle, inode, path + depth); | |
2799 | return err; | |
56055d3a AA |
2800 | } |
2801 | ||
0031462b MC |
2802 | /* |
2803 | * This function is called by ext4_ext_get_blocks() from | |
2804 | * ext4_get_blocks_dio_write() when DIO to write | |
2805 | * to an uninitialized extent. | |
2806 | * | |
2807 | * Writing to an uninitized extent may result in splitting the uninitialized | |
2808 | * extent into multiple /intialized unintialized extents (up to three) | |
2809 | * There are three possibilities: | |
2810 | * a> There is no split required: Entire extent should be uninitialized | |
2811 | * b> Splits in two extents: Write is happening at either end of the extent | |
2812 | * c> Splits in three extents: Somone is writing in middle of the extent | |
2813 | * | |
2814 | * One of more index blocks maybe needed if the extent tree grow after | |
2815 | * the unintialized extent split. To prevent ENOSPC occur at the IO | |
2816 | * complete, we need to split the uninitialized extent before DIO submit | |
2817 | * the IO. The uninitilized extent called at this time will be split | |
2818 | * into three uninitialized extent(at most). After IO complete, the part | |
2819 | * being filled will be convert to initialized by the end_io callback function | |
2820 | * via ext4_convert_unwritten_extents(). | |
ba230c3f M |
2821 | * |
2822 | * Returns the size of uninitialized extent to be written on success. | |
0031462b MC |
2823 | */ |
2824 | static int ext4_split_unwritten_extents(handle_t *handle, | |
2825 | struct inode *inode, | |
2826 | struct ext4_ext_path *path, | |
2827 | ext4_lblk_t iblock, | |
2828 | unsigned int max_blocks, | |
2829 | int flags) | |
2830 | { | |
2831 | struct ext4_extent *ex, newex, orig_ex; | |
2832 | struct ext4_extent *ex1 = NULL; | |
2833 | struct ext4_extent *ex2 = NULL; | |
2834 | struct ext4_extent *ex3 = NULL; | |
2835 | struct ext4_extent_header *eh; | |
2836 | ext4_lblk_t ee_block; | |
2837 | unsigned int allocated, ee_len, depth; | |
2838 | ext4_fsblk_t newblock; | |
2839 | int err = 0; | |
0031462b MC |
2840 | |
2841 | ext_debug("ext4_split_unwritten_extents: inode %lu," | |
2842 | "iblock %llu, max_blocks %u\n", inode->i_ino, | |
2843 | (unsigned long long)iblock, max_blocks); | |
2844 | depth = ext_depth(inode); | |
2845 | eh = path[depth].p_hdr; | |
2846 | ex = path[depth].p_ext; | |
2847 | ee_block = le32_to_cpu(ex->ee_block); | |
2848 | ee_len = ext4_ext_get_actual_len(ex); | |
2849 | allocated = ee_len - (iblock - ee_block); | |
2850 | newblock = iblock - ee_block + ext_pblock(ex); | |
2851 | ex2 = ex; | |
2852 | orig_ex.ee_block = ex->ee_block; | |
2853 | orig_ex.ee_len = cpu_to_le16(ee_len); | |
2854 | ext4_ext_store_pblock(&orig_ex, ext_pblock(ex)); | |
2855 | ||
2856 | /* | |
ba230c3f M |
2857 | * If the uninitialized extent begins at the same logical |
2858 | * block where the write begins, and the write completely | |
2859 | * covers the extent, then we don't need to split it. | |
0031462b | 2860 | */ |
ba230c3f M |
2861 | if ((iblock == ee_block) && (allocated <= max_blocks)) |
2862 | return allocated; | |
0031462b MC |
2863 | |
2864 | err = ext4_ext_get_access(handle, inode, path + depth); | |
2865 | if (err) | |
2866 | goto out; | |
2867 | /* ex1: ee_block to iblock - 1 : uninitialized */ | |
2868 | if (iblock > ee_block) { | |
2869 | ex1 = ex; | |
2870 | ex1->ee_len = cpu_to_le16(iblock - ee_block); | |
2871 | ext4_ext_mark_uninitialized(ex1); | |
2872 | ex2 = &newex; | |
2873 | } | |
2874 | /* | |
2875 | * for sanity, update the length of the ex2 extent before | |
2876 | * we insert ex3, if ex1 is NULL. This is to avoid temporary | |
2877 | * overlap of blocks. | |
2878 | */ | |
2879 | if (!ex1 && allocated > max_blocks) | |
2880 | ex2->ee_len = cpu_to_le16(max_blocks); | |
2881 | /* ex3: to ee_block + ee_len : uninitialised */ | |
2882 | if (allocated > max_blocks) { | |
2883 | unsigned int newdepth; | |
2884 | ex3 = &newex; | |
2885 | ex3->ee_block = cpu_to_le32(iblock + max_blocks); | |
2886 | ext4_ext_store_pblock(ex3, newblock + max_blocks); | |
2887 | ex3->ee_len = cpu_to_le16(allocated - max_blocks); | |
2888 | ext4_ext_mark_uninitialized(ex3); | |
2889 | err = ext4_ext_insert_extent(handle, inode, path, ex3, flags); | |
2890 | if (err == -ENOSPC) { | |
2891 | err = ext4_ext_zeroout(inode, &orig_ex); | |
2892 | if (err) | |
2893 | goto fix_extent_len; | |
2894 | /* update the extent length and mark as initialized */ | |
2895 | ex->ee_block = orig_ex.ee_block; | |
2896 | ex->ee_len = orig_ex.ee_len; | |
2897 | ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); | |
2898 | ext4_ext_dirty(handle, inode, path + depth); | |
2899 | /* zeroed the full extent */ | |
2900 | /* blocks available from iblock */ | |
2901 | return allocated; | |
2902 | ||
2903 | } else if (err) | |
2904 | goto fix_extent_len; | |
2905 | /* | |
2906 | * The depth, and hence eh & ex might change | |
2907 | * as part of the insert above. | |
2908 | */ | |
2909 | newdepth = ext_depth(inode); | |
2910 | /* | |
2911 | * update the extent length after successful insert of the | |
2912 | * split extent | |
2913 | */ | |
2914 | orig_ex.ee_len = cpu_to_le16(ee_len - | |
2915 | ext4_ext_get_actual_len(ex3)); | |
2916 | depth = newdepth; | |
2917 | ext4_ext_drop_refs(path); | |
2918 | path = ext4_ext_find_extent(inode, iblock, path); | |
2919 | if (IS_ERR(path)) { | |
2920 | err = PTR_ERR(path); | |
2921 | goto out; | |
2922 | } | |
2923 | eh = path[depth].p_hdr; | |
2924 | ex = path[depth].p_ext; | |
2925 | if (ex2 != &newex) | |
2926 | ex2 = ex; | |
2927 | ||
2928 | err = ext4_ext_get_access(handle, inode, path + depth); | |
2929 | if (err) | |
2930 | goto out; | |
2931 | ||
2932 | allocated = max_blocks; | |
2933 | } | |
2934 | /* | |
2935 | * If there was a change of depth as part of the | |
2936 | * insertion of ex3 above, we need to update the length | |
2937 | * of the ex1 extent again here | |
2938 | */ | |
2939 | if (ex1 && ex1 != ex) { | |
2940 | ex1 = ex; | |
2941 | ex1->ee_len = cpu_to_le16(iblock - ee_block); | |
2942 | ext4_ext_mark_uninitialized(ex1); | |
2943 | ex2 = &newex; | |
2944 | } | |
2945 | /* | |
2946 | * ex2: iblock to iblock + maxblocks-1 : to be direct IO written, | |
2947 | * uninitialised still. | |
2948 | */ | |
2949 | ex2->ee_block = cpu_to_le32(iblock); | |
2950 | ext4_ext_store_pblock(ex2, newblock); | |
2951 | ex2->ee_len = cpu_to_le16(allocated); | |
2952 | ext4_ext_mark_uninitialized(ex2); | |
2953 | if (ex2 != ex) | |
2954 | goto insert; | |
2955 | /* Mark modified extent as dirty */ | |
2956 | err = ext4_ext_dirty(handle, inode, path + depth); | |
2957 | ext_debug("out here\n"); | |
2958 | goto out; | |
2959 | insert: | |
2960 | err = ext4_ext_insert_extent(handle, inode, path, &newex, flags); | |
2961 | if (err == -ENOSPC) { | |
2962 | err = ext4_ext_zeroout(inode, &orig_ex); | |
2963 | if (err) | |
2964 | goto fix_extent_len; | |
2965 | /* update the extent length and mark as initialized */ | |
2966 | ex->ee_block = orig_ex.ee_block; | |
2967 | ex->ee_len = orig_ex.ee_len; | |
2968 | ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); | |
2969 | ext4_ext_dirty(handle, inode, path + depth); | |
2970 | /* zero out the first half */ | |
2971 | return allocated; | |
2972 | } else if (err) | |
2973 | goto fix_extent_len; | |
2974 | out: | |
2975 | ext4_ext_show_leaf(inode, path); | |
2976 | return err ? err : allocated; | |
2977 | ||
2978 | fix_extent_len: | |
2979 | ex->ee_block = orig_ex.ee_block; | |
2980 | ex->ee_len = orig_ex.ee_len; | |
2981 | ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); | |
2982 | ext4_ext_mark_uninitialized(ex); | |
2983 | ext4_ext_dirty(handle, inode, path + depth); | |
2984 | return err; | |
2985 | } | |
2986 | static int ext4_convert_unwritten_extents_dio(handle_t *handle, | |
2987 | struct inode *inode, | |
2988 | struct ext4_ext_path *path) | |
2989 | { | |
2990 | struct ext4_extent *ex; | |
2991 | struct ext4_extent_header *eh; | |
2992 | int depth; | |
2993 | int err = 0; | |
2994 | int ret = 0; | |
2995 | ||
2996 | depth = ext_depth(inode); | |
2997 | eh = path[depth].p_hdr; | |
2998 | ex = path[depth].p_ext; | |
2999 | ||
3000 | err = ext4_ext_get_access(handle, inode, path + depth); | |
3001 | if (err) | |
3002 | goto out; | |
3003 | /* first mark the extent as initialized */ | |
3004 | ext4_ext_mark_initialized(ex); | |
3005 | ||
3006 | /* | |
3007 | * We have to see if it can be merged with the extent | |
3008 | * on the left. | |
3009 | */ | |
3010 | if (ex > EXT_FIRST_EXTENT(eh)) { | |
3011 | /* | |
3012 | * To merge left, pass "ex - 1" to try_to_merge(), | |
3013 | * since it merges towards right _only_. | |
3014 | */ | |
3015 | ret = ext4_ext_try_to_merge(inode, path, ex - 1); | |
3016 | if (ret) { | |
3017 | err = ext4_ext_correct_indexes(handle, inode, path); | |
3018 | if (err) | |
3019 | goto out; | |
3020 | depth = ext_depth(inode); | |
3021 | ex--; | |
3022 | } | |
3023 | } | |
3024 | /* | |
3025 | * Try to Merge towards right. | |
3026 | */ | |
3027 | ret = ext4_ext_try_to_merge(inode, path, ex); | |
3028 | if (ret) { | |
3029 | err = ext4_ext_correct_indexes(handle, inode, path); | |
3030 | if (err) | |
3031 | goto out; | |
3032 | depth = ext_depth(inode); | |
3033 | } | |
3034 | /* Mark modified extent as dirty */ | |
3035 | err = ext4_ext_dirty(handle, inode, path + depth); | |
3036 | out: | |
3037 | ext4_ext_show_leaf(inode, path); | |
3038 | return err; | |
3039 | } | |
3040 | ||
515f41c3 AK |
3041 | static void unmap_underlying_metadata_blocks(struct block_device *bdev, |
3042 | sector_t block, int count) | |
3043 | { | |
3044 | int i; | |
3045 | for (i = 0; i < count; i++) | |
3046 | unmap_underlying_metadata(bdev, block + i); | |
3047 | } | |
3048 | ||
0031462b MC |
3049 | static int |
3050 | ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode, | |
3051 | ext4_lblk_t iblock, unsigned int max_blocks, | |
3052 | struct ext4_ext_path *path, int flags, | |
3053 | unsigned int allocated, struct buffer_head *bh_result, | |
3054 | ext4_fsblk_t newblock) | |
3055 | { | |
3056 | int ret = 0; | |
3057 | int err = 0; | |
8d5d02e6 | 3058 | ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio; |
0031462b MC |
3059 | |
3060 | ext_debug("ext4_ext_handle_uninitialized_extents: inode %lu, logical" | |
3061 | "block %llu, max_blocks %u, flags %d, allocated %u", | |
3062 | inode->i_ino, (unsigned long long)iblock, max_blocks, | |
3063 | flags, allocated); | |
3064 | ext4_ext_show_leaf(inode, path); | |
3065 | ||
3066 | /* DIO get_block() before submit the IO, split the extent */ | |
3067 | if (flags == EXT4_GET_BLOCKS_DIO_CREATE_EXT) { | |
3068 | ret = ext4_split_unwritten_extents(handle, | |
3069 | inode, path, iblock, | |
3070 | max_blocks, flags); | |
5f524950 M |
3071 | /* |
3072 | * Flag the inode(non aio case) or end_io struct (aio case) | |
3073 | * that this IO needs to convertion to written when IO is | |
3074 | * completed | |
3075 | */ | |
8d5d02e6 MC |
3076 | if (io) |
3077 | io->flag = DIO_AIO_UNWRITTEN; | |
5f524950 | 3078 | else |
19f5fb7a | 3079 | ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN); |
0031462b MC |
3080 | goto out; |
3081 | } | |
5f524950 | 3082 | /* async DIO end_io complete, convert the filled extent to written */ |
0031462b MC |
3083 | if (flags == EXT4_GET_BLOCKS_DIO_CONVERT_EXT) { |
3084 | ret = ext4_convert_unwritten_extents_dio(handle, inode, | |
3085 | path); | |
b436b9be JK |
3086 | if (ret >= 0) |
3087 | ext4_update_inode_fsync_trans(handle, inode, 1); | |
0031462b MC |
3088 | goto out2; |
3089 | } | |
3090 | /* buffered IO case */ | |
3091 | /* | |
3092 | * repeat fallocate creation request | |
3093 | * we already have an unwritten extent | |
3094 | */ | |
3095 | if (flags & EXT4_GET_BLOCKS_UNINIT_EXT) | |
3096 | goto map_out; | |
3097 | ||
3098 | /* buffered READ or buffered write_begin() lookup */ | |
3099 | if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) { | |
3100 | /* | |
3101 | * We have blocks reserved already. We | |
3102 | * return allocated blocks so that delalloc | |
3103 | * won't do block reservation for us. But | |
3104 | * the buffer head will be unmapped so that | |
3105 | * a read from the block returns 0s. | |
3106 | */ | |
3107 | set_buffer_unwritten(bh_result); | |
3108 | goto out1; | |
3109 | } | |
3110 | ||
3111 | /* buffered write, writepage time, convert*/ | |
3112 | ret = ext4_ext_convert_to_initialized(handle, inode, | |
3113 | path, iblock, | |
3114 | max_blocks); | |
b436b9be JK |
3115 | if (ret >= 0) |
3116 | ext4_update_inode_fsync_trans(handle, inode, 1); | |
0031462b MC |
3117 | out: |
3118 | if (ret <= 0) { | |
3119 | err = ret; | |
3120 | goto out2; | |
3121 | } else | |
3122 | allocated = ret; | |
3123 | set_buffer_new(bh_result); | |
515f41c3 AK |
3124 | /* |
3125 | * if we allocated more blocks than requested | |
3126 | * we need to make sure we unmap the extra block | |
3127 | * allocated. The actual needed block will get | |
3128 | * unmapped later when we find the buffer_head marked | |
3129 | * new. | |
3130 | */ | |
3131 | if (allocated > max_blocks) { | |
3132 | unmap_underlying_metadata_blocks(inode->i_sb->s_bdev, | |
3133 | newblock + max_blocks, | |
3134 | allocated - max_blocks); | |
5f634d06 | 3135 | allocated = max_blocks; |
515f41c3 | 3136 | } |
5f634d06 AK |
3137 | |
3138 | /* | |
3139 | * If we have done fallocate with the offset that is already | |
3140 | * delayed allocated, we would have block reservation | |
3141 | * and quota reservation done in the delayed write path. | |
3142 | * But fallocate would have already updated quota and block | |
3143 | * count for this offset. So cancel these reservation | |
3144 | */ | |
1296cc85 | 3145 | if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) |
5f634d06 AK |
3146 | ext4_da_update_reserve_space(inode, allocated, 0); |
3147 | ||
0031462b MC |
3148 | map_out: |
3149 | set_buffer_mapped(bh_result); | |
3150 | out1: | |
3151 | if (allocated > max_blocks) | |
3152 | allocated = max_blocks; | |
3153 | ext4_ext_show_leaf(inode, path); | |
3154 | bh_result->b_bdev = inode->i_sb->s_bdev; | |
3155 | bh_result->b_blocknr = newblock; | |
3156 | out2: | |
3157 | if (path) { | |
3158 | ext4_ext_drop_refs(path); | |
3159 | kfree(path); | |
3160 | } | |
3161 | return err ? err : allocated; | |
3162 | } | |
c278bfec | 3163 | /* |
f5ab0d1f MC |
3164 | * Block allocation/map/preallocation routine for extents based files |
3165 | * | |
3166 | * | |
c278bfec | 3167 | * Need to be called with |
0e855ac8 AK |
3168 | * down_read(&EXT4_I(inode)->i_data_sem) if not allocating file system block |
3169 | * (ie, create is zero). Otherwise down_write(&EXT4_I(inode)->i_data_sem) | |
f5ab0d1f MC |
3170 | * |
3171 | * return > 0, number of of blocks already mapped/allocated | |
3172 | * if create == 0 and these are pre-allocated blocks | |
3173 | * buffer head is unmapped | |
3174 | * otherwise blocks are mapped | |
3175 | * | |
3176 | * return = 0, if plain look up failed (blocks have not been allocated) | |
3177 | * buffer head is unmapped | |
3178 | * | |
3179 | * return < 0, error case. | |
c278bfec | 3180 | */ |
f65e6fba | 3181 | int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, |
725d26d3 | 3182 | ext4_lblk_t iblock, |
498e5f24 | 3183 | unsigned int max_blocks, struct buffer_head *bh_result, |
c2177057 | 3184 | int flags) |
a86c6181 AT |
3185 | { |
3186 | struct ext4_ext_path *path = NULL; | |
56055d3a | 3187 | struct ext4_extent_header *eh; |
a86c6181 | 3188 | struct ext4_extent newex, *ex; |
498e5f24 TT |
3189 | ext4_fsblk_t newblock; |
3190 | int err = 0, depth, ret, cache_type; | |
3191 | unsigned int allocated = 0; | |
c9de560d | 3192 | struct ext4_allocation_request ar; |
8d5d02e6 | 3193 | ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio; |
a86c6181 AT |
3194 | |
3195 | __clear_bit(BH_New, &bh_result->b_state); | |
84fe3bef | 3196 | ext_debug("blocks %u/%u requested for inode %lu\n", |
bba90743 | 3197 | iblock, max_blocks, inode->i_ino); |
a86c6181 AT |
3198 | |
3199 | /* check in cache */ | |
498e5f24 TT |
3200 | cache_type = ext4_ext_in_cache(inode, iblock, &newex); |
3201 | if (cache_type) { | |
3202 | if (cache_type == EXT4_EXT_CACHE_GAP) { | |
c2177057 | 3203 | if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) { |
56055d3a AA |
3204 | /* |
3205 | * block isn't allocated yet and | |
3206 | * user doesn't want to allocate it | |
3207 | */ | |
a86c6181 AT |
3208 | goto out2; |
3209 | } | |
3210 | /* we should allocate requested block */ | |
498e5f24 | 3211 | } else if (cache_type == EXT4_EXT_CACHE_EXTENT) { |
a86c6181 | 3212 | /* block is already allocated */ |
8c55e204 DK |
3213 | newblock = iblock |
3214 | - le32_to_cpu(newex.ee_block) | |
3215 | + ext_pblock(&newex); | |
d0d856e8 | 3216 | /* number of remaining blocks in the extent */ |
b939e376 | 3217 | allocated = ext4_ext_get_actual_len(&newex) - |
a86c6181 AT |
3218 | (iblock - le32_to_cpu(newex.ee_block)); |
3219 | goto out; | |
3220 | } else { | |
3221 | BUG(); | |
3222 | } | |
3223 | } | |
3224 | ||
3225 | /* find extent for this block */ | |
3226 | path = ext4_ext_find_extent(inode, iblock, NULL); | |
3227 | if (IS_ERR(path)) { | |
3228 | err = PTR_ERR(path); | |
3229 | path = NULL; | |
3230 | goto out2; | |
3231 | } | |
3232 | ||
3233 | depth = ext_depth(inode); | |
3234 | ||
3235 | /* | |
d0d856e8 RD |
3236 | * consistent leaf must not be empty; |
3237 | * this situation is possible, though, _during_ tree modification; | |
a86c6181 AT |
3238 | * this is why assert can't be put in ext4_ext_find_extent() |
3239 | */ | |
034fb4c9 SP |
3240 | if (path[depth].p_ext == NULL && depth != 0) { |
3241 | ext4_error(inode->i_sb, __func__, "bad extent address " | |
3242 | "inode: %lu, iblock: %d, depth: %d", | |
3243 | inode->i_ino, iblock, depth); | |
3244 | err = -EIO; | |
3245 | goto out2; | |
3246 | } | |
56055d3a | 3247 | eh = path[depth].p_hdr; |
a86c6181 | 3248 | |
7e028976 AM |
3249 | ex = path[depth].p_ext; |
3250 | if (ex) { | |
725d26d3 | 3251 | ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block); |
f65e6fba | 3252 | ext4_fsblk_t ee_start = ext_pblock(ex); |
a2df2a63 | 3253 | unsigned short ee_len; |
471d4011 SB |
3254 | |
3255 | /* | |
471d4011 | 3256 | * Uninitialized extents are treated as holes, except that |
56055d3a | 3257 | * we split out initialized portions during a write. |
471d4011 | 3258 | */ |
a2df2a63 | 3259 | ee_len = ext4_ext_get_actual_len(ex); |
d0d856e8 | 3260 | /* if found extent covers block, simply return it */ |
8c55e204 | 3261 | if (iblock >= ee_block && iblock < ee_block + ee_len) { |
a86c6181 | 3262 | newblock = iblock - ee_block + ee_start; |
d0d856e8 | 3263 | /* number of remaining blocks in the extent */ |
a86c6181 | 3264 | allocated = ee_len - (iblock - ee_block); |
84fe3bef | 3265 | ext_debug("%u fit into %u:%d -> %llu\n", iblock, |
a86c6181 | 3266 | ee_block, ee_len, newblock); |
56055d3a | 3267 | |
a2df2a63 | 3268 | /* Do not put uninitialized extent in the cache */ |
56055d3a | 3269 | if (!ext4_ext_is_uninitialized(ex)) { |
a2df2a63 AA |
3270 | ext4_ext_put_in_cache(inode, ee_block, |
3271 | ee_len, ee_start, | |
3272 | EXT4_EXT_CACHE_EXTENT); | |
56055d3a AA |
3273 | goto out; |
3274 | } | |
0031462b MC |
3275 | ret = ext4_ext_handle_uninitialized_extents(handle, |
3276 | inode, iblock, max_blocks, path, | |
3277 | flags, allocated, bh_result, newblock); | |
3278 | return ret; | |
a86c6181 AT |
3279 | } |
3280 | } | |
3281 | ||
3282 | /* | |
d0d856e8 | 3283 | * requested block isn't allocated yet; |
a86c6181 AT |
3284 | * we couldn't try to create block if create flag is zero |
3285 | */ | |
c2177057 | 3286 | if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) { |
56055d3a AA |
3287 | /* |
3288 | * put just found gap into cache to speed up | |
3289 | * subsequent requests | |
3290 | */ | |
a86c6181 AT |
3291 | ext4_ext_put_gap_in_cache(inode, path, iblock); |
3292 | goto out2; | |
3293 | } | |
3294 | /* | |
c2ea3fde | 3295 | * Okay, we need to do block allocation. |
63f57933 | 3296 | */ |
a86c6181 | 3297 | |
c9de560d AT |
3298 | /* find neighbour allocated blocks */ |
3299 | ar.lleft = iblock; | |
3300 | err = ext4_ext_search_left(inode, path, &ar.lleft, &ar.pleft); | |
3301 | if (err) | |
3302 | goto out2; | |
3303 | ar.lright = iblock; | |
3304 | err = ext4_ext_search_right(inode, path, &ar.lright, &ar.pright); | |
3305 | if (err) | |
3306 | goto out2; | |
25d14f98 | 3307 | |
749269fa AA |
3308 | /* |
3309 | * See if request is beyond maximum number of blocks we can have in | |
3310 | * a single extent. For an initialized extent this limit is | |
3311 | * EXT_INIT_MAX_LEN and for an uninitialized extent this limit is | |
3312 | * EXT_UNINIT_MAX_LEN. | |
3313 | */ | |
3314 | if (max_blocks > EXT_INIT_MAX_LEN && | |
c2177057 | 3315 | !(flags & EXT4_GET_BLOCKS_UNINIT_EXT)) |
749269fa AA |
3316 | max_blocks = EXT_INIT_MAX_LEN; |
3317 | else if (max_blocks > EXT_UNINIT_MAX_LEN && | |
c2177057 | 3318 | (flags & EXT4_GET_BLOCKS_UNINIT_EXT)) |
749269fa AA |
3319 | max_blocks = EXT_UNINIT_MAX_LEN; |
3320 | ||
25d14f98 AA |
3321 | /* Check if we can really insert (iblock)::(iblock+max_blocks) extent */ |
3322 | newex.ee_block = cpu_to_le32(iblock); | |
3323 | newex.ee_len = cpu_to_le16(max_blocks); | |
3324 | err = ext4_ext_check_overlap(inode, &newex, path); | |
3325 | if (err) | |
b939e376 | 3326 | allocated = ext4_ext_get_actual_len(&newex); |
25d14f98 AA |
3327 | else |
3328 | allocated = max_blocks; | |
c9de560d AT |
3329 | |
3330 | /* allocate new block */ | |
3331 | ar.inode = inode; | |
3332 | ar.goal = ext4_ext_find_goal(inode, path, iblock); | |
3333 | ar.logical = iblock; | |
3334 | ar.len = allocated; | |
3335 | if (S_ISREG(inode->i_mode)) | |
3336 | ar.flags = EXT4_MB_HINT_DATA; | |
3337 | else | |
3338 | /* disable in-core preallocation for non-regular files */ | |
3339 | ar.flags = 0; | |
3340 | newblock = ext4_mb_new_blocks(handle, &ar, &err); | |
a86c6181 AT |
3341 | if (!newblock) |
3342 | goto out2; | |
84fe3bef | 3343 | ext_debug("allocate new block: goal %llu, found %llu/%u\n", |
498e5f24 | 3344 | ar.goal, newblock, allocated); |
a86c6181 AT |
3345 | |
3346 | /* try to insert new extent into found leaf and return */ | |
f65e6fba | 3347 | ext4_ext_store_pblock(&newex, newblock); |
c9de560d | 3348 | newex.ee_len = cpu_to_le16(ar.len); |
8d5d02e6 MC |
3349 | /* Mark uninitialized */ |
3350 | if (flags & EXT4_GET_BLOCKS_UNINIT_EXT){ | |
a2df2a63 | 3351 | ext4_ext_mark_uninitialized(&newex); |
8d5d02e6 MC |
3352 | /* |
3353 | * io_end structure was created for every async | |
3354 | * direct IO write to the middle of the file. | |
3355 | * To avoid unecessary convertion for every aio dio rewrite | |
3356 | * to the mid of file, here we flag the IO that is really | |
3357 | * need the convertion. | |
5f524950 M |
3358 | * For non asycn direct IO case, flag the inode state |
3359 | * that we need to perform convertion when IO is done. | |
8d5d02e6 | 3360 | */ |
5f524950 M |
3361 | if (flags == EXT4_GET_BLOCKS_DIO_CREATE_EXT) { |
3362 | if (io) | |
3363 | io->flag = DIO_AIO_UNWRITTEN; | |
3364 | else | |
19f5fb7a TT |
3365 | ext4_set_inode_state(inode, |
3366 | EXT4_STATE_DIO_UNWRITTEN); | |
5f524950 | 3367 | } |
8d5d02e6 | 3368 | } |
0031462b | 3369 | err = ext4_ext_insert_extent(handle, inode, path, &newex, flags); |
315054f0 AT |
3370 | if (err) { |
3371 | /* free data blocks we just allocated */ | |
c9de560d AT |
3372 | /* not a good idea to call discard here directly, |
3373 | * but otherwise we'd need to call it every free() */ | |
c2ea3fde | 3374 | ext4_discard_preallocations(inode); |
e6362609 TT |
3375 | ext4_free_blocks(handle, inode, 0, ext_pblock(&newex), |
3376 | ext4_ext_get_actual_len(&newex), 0); | |
a86c6181 | 3377 | goto out2; |
315054f0 | 3378 | } |
a86c6181 | 3379 | |
a86c6181 | 3380 | /* previous routine could use block we allocated */ |
f65e6fba | 3381 | newblock = ext_pblock(&newex); |
b939e376 | 3382 | allocated = ext4_ext_get_actual_len(&newex); |
5f634d06 AK |
3383 | if (allocated > max_blocks) |
3384 | allocated = max_blocks; | |
953e622b | 3385 | set_buffer_new(bh_result); |
a86c6181 | 3386 | |
5f634d06 AK |
3387 | /* |
3388 | * Update reserved blocks/metadata blocks after successful | |
3389 | * block allocation which had been deferred till now. | |
3390 | */ | |
1296cc85 | 3391 | if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) |
5f634d06 AK |
3392 | ext4_da_update_reserve_space(inode, allocated, 1); |
3393 | ||
b436b9be JK |
3394 | /* |
3395 | * Cache the extent and update transaction to commit on fdatasync only | |
3396 | * when it is _not_ an uninitialized extent. | |
3397 | */ | |
3398 | if ((flags & EXT4_GET_BLOCKS_UNINIT_EXT) == 0) { | |
a2df2a63 AA |
3399 | ext4_ext_put_in_cache(inode, iblock, allocated, newblock, |
3400 | EXT4_EXT_CACHE_EXTENT); | |
b436b9be JK |
3401 | ext4_update_inode_fsync_trans(handle, inode, 1); |
3402 | } else | |
3403 | ext4_update_inode_fsync_trans(handle, inode, 0); | |
a86c6181 AT |
3404 | out: |
3405 | if (allocated > max_blocks) | |
3406 | allocated = max_blocks; | |
3407 | ext4_ext_show_leaf(inode, path); | |
953e622b | 3408 | set_buffer_mapped(bh_result); |
a86c6181 AT |
3409 | bh_result->b_bdev = inode->i_sb->s_bdev; |
3410 | bh_result->b_blocknr = newblock; | |
3411 | out2: | |
3412 | if (path) { | |
3413 | ext4_ext_drop_refs(path); | |
3414 | kfree(path); | |
3415 | } | |
a86c6181 AT |
3416 | return err ? err : allocated; |
3417 | } | |
3418 | ||
cf108bca | 3419 | void ext4_ext_truncate(struct inode *inode) |
a86c6181 AT |
3420 | { |
3421 | struct address_space *mapping = inode->i_mapping; | |
3422 | struct super_block *sb = inode->i_sb; | |
725d26d3 | 3423 | ext4_lblk_t last_block; |
a86c6181 AT |
3424 | handle_t *handle; |
3425 | int err = 0; | |
3426 | ||
3427 | /* | |
3428 | * probably first extent we're gonna free will be last in block | |
3429 | */ | |
f3bd1f3f | 3430 | err = ext4_writepage_trans_blocks(inode); |
a86c6181 | 3431 | handle = ext4_journal_start(inode, err); |
cf108bca | 3432 | if (IS_ERR(handle)) |
a86c6181 | 3433 | return; |
a86c6181 | 3434 | |
cf108bca JK |
3435 | if (inode->i_size & (sb->s_blocksize - 1)) |
3436 | ext4_block_truncate_page(handle, mapping, inode->i_size); | |
a86c6181 | 3437 | |
9ddfc3dc JK |
3438 | if (ext4_orphan_add(handle, inode)) |
3439 | goto out_stop; | |
3440 | ||
0e855ac8 | 3441 | down_write(&EXT4_I(inode)->i_data_sem); |
a86c6181 AT |
3442 | ext4_ext_invalidate_cache(inode); |
3443 | ||
c2ea3fde | 3444 | ext4_discard_preallocations(inode); |
c9de560d | 3445 | |
a86c6181 | 3446 | /* |
d0d856e8 RD |
3447 | * TODO: optimization is possible here. |
3448 | * Probably we need not scan at all, | |
3449 | * because page truncation is enough. | |
a86c6181 | 3450 | */ |
a86c6181 AT |
3451 | |
3452 | /* we have to know where to truncate from in crash case */ | |
3453 | EXT4_I(inode)->i_disksize = inode->i_size; | |
3454 | ext4_mark_inode_dirty(handle, inode); | |
3455 | ||
3456 | last_block = (inode->i_size + sb->s_blocksize - 1) | |
3457 | >> EXT4_BLOCK_SIZE_BITS(sb); | |
3458 | err = ext4_ext_remove_space(inode, last_block); | |
3459 | ||
3460 | /* In a multi-transaction truncate, we only make the final | |
56055d3a AA |
3461 | * transaction synchronous. |
3462 | */ | |
a86c6181 | 3463 | if (IS_SYNC(inode)) |
0390131b | 3464 | ext4_handle_sync(handle); |
a86c6181 AT |
3465 | |
3466 | out_stop: | |
9ddfc3dc | 3467 | up_write(&EXT4_I(inode)->i_data_sem); |
a86c6181 | 3468 | /* |
d0d856e8 | 3469 | * If this was a simple ftruncate() and the file will remain alive, |
a86c6181 AT |
3470 | * then we need to clear up the orphan record which we created above. |
3471 | * However, if this was a real unlink then we were called by | |
3472 | * ext4_delete_inode(), and we allow that function to clean up the | |
3473 | * orphan info for us. | |
3474 | */ | |
3475 | if (inode->i_nlink) | |
3476 | ext4_orphan_del(handle, inode); | |
3477 | ||
ef737728 SR |
3478 | inode->i_mtime = inode->i_ctime = ext4_current_time(inode); |
3479 | ext4_mark_inode_dirty(handle, inode); | |
a86c6181 AT |
3480 | ext4_journal_stop(handle); |
3481 | } | |
3482 | ||
fd28784a AK |
3483 | static void ext4_falloc_update_inode(struct inode *inode, |
3484 | int mode, loff_t new_size, int update_ctime) | |
3485 | { | |
3486 | struct timespec now; | |
3487 | ||
3488 | if (update_ctime) { | |
3489 | now = current_fs_time(inode->i_sb); | |
3490 | if (!timespec_equal(&inode->i_ctime, &now)) | |
3491 | inode->i_ctime = now; | |
3492 | } | |
3493 | /* | |
3494 | * Update only when preallocation was requested beyond | |
3495 | * the file size. | |
3496 | */ | |
cf17fea6 AK |
3497 | if (!(mode & FALLOC_FL_KEEP_SIZE)) { |
3498 | if (new_size > i_size_read(inode)) | |
3499 | i_size_write(inode, new_size); | |
3500 | if (new_size > EXT4_I(inode)->i_disksize) | |
3501 | ext4_update_i_disksize(inode, new_size); | |
fd28784a AK |
3502 | } |
3503 | ||
3504 | } | |
3505 | ||
a2df2a63 AA |
3506 | /* |
3507 | * preallocate space for a file. This implements ext4's fallocate inode | |
3508 | * operation, which gets called from sys_fallocate system call. | |
3509 | * For block-mapped files, posix_fallocate should fall back to the method | |
3510 | * of writing zeroes to the required new blocks (the same behavior which is | |
3511 | * expected for file systems which do not support fallocate() system call). | |
3512 | */ | |
3513 | long ext4_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len) | |
3514 | { | |
3515 | handle_t *handle; | |
725d26d3 | 3516 | ext4_lblk_t block; |
fd28784a | 3517 | loff_t new_size; |
498e5f24 | 3518 | unsigned int max_blocks; |
a2df2a63 AA |
3519 | int ret = 0; |
3520 | int ret2 = 0; | |
3521 | int retries = 0; | |
3522 | struct buffer_head map_bh; | |
3523 | unsigned int credits, blkbits = inode->i_blkbits; | |
3524 | ||
3525 | /* | |
3526 | * currently supporting (pre)allocate mode for extent-based | |
3527 | * files _only_ | |
3528 | */ | |
3529 | if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL)) | |
3530 | return -EOPNOTSUPP; | |
3531 | ||
3532 | /* preallocation to directories is currently not supported */ | |
3533 | if (S_ISDIR(inode->i_mode)) | |
3534 | return -ENODEV; | |
3535 | ||
3536 | block = offset >> blkbits; | |
fd28784a AK |
3537 | /* |
3538 | * We can't just convert len to max_blocks because | |
3539 | * If blocksize = 4096 offset = 3072 and len = 2048 | |
3540 | */ | |
a2df2a63 | 3541 | max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) |
fd28784a | 3542 | - block; |
a2df2a63 | 3543 | /* |
f3bd1f3f | 3544 | * credits to insert 1 extent into extent tree |
a2df2a63 | 3545 | */ |
f3bd1f3f | 3546 | credits = ext4_chunk_trans_blocks(inode, max_blocks); |
55bd725a | 3547 | mutex_lock(&inode->i_mutex); |
a2df2a63 AA |
3548 | retry: |
3549 | while (ret >= 0 && ret < max_blocks) { | |
3550 | block = block + ret; | |
3551 | max_blocks = max_blocks - ret; | |
3552 | handle = ext4_journal_start(inode, credits); | |
3553 | if (IS_ERR(handle)) { | |
3554 | ret = PTR_ERR(handle); | |
3555 | break; | |
3556 | } | |
79ffab34 | 3557 | map_bh.b_state = 0; |
12b7ac17 TT |
3558 | ret = ext4_get_blocks(handle, inode, block, |
3559 | max_blocks, &map_bh, | |
c2177057 | 3560 | EXT4_GET_BLOCKS_CREATE_UNINIT_EXT); |
221879c9 | 3561 | if (ret <= 0) { |
2c98615d AK |
3562 | #ifdef EXT4FS_DEBUG |
3563 | WARN_ON(ret <= 0); | |
3564 | printk(KERN_ERR "%s: ext4_ext_get_blocks " | |
3565 | "returned error inode#%lu, block=%u, " | |
9fd9784c | 3566 | "max_blocks=%u", __func__, |
221879c9 | 3567 | inode->i_ino, block, max_blocks); |
2c98615d | 3568 | #endif |
a2df2a63 AA |
3569 | ext4_mark_inode_dirty(handle, inode); |
3570 | ret2 = ext4_journal_stop(handle); | |
3571 | break; | |
3572 | } | |
fd28784a AK |
3573 | if ((block + ret) >= (EXT4_BLOCK_ALIGN(offset + len, |
3574 | blkbits) >> blkbits)) | |
3575 | new_size = offset + len; | |
3576 | else | |
3577 | new_size = (block + ret) << blkbits; | |
a2df2a63 | 3578 | |
fd28784a AK |
3579 | ext4_falloc_update_inode(inode, mode, new_size, |
3580 | buffer_new(&map_bh)); | |
a2df2a63 AA |
3581 | ext4_mark_inode_dirty(handle, inode); |
3582 | ret2 = ext4_journal_stop(handle); | |
3583 | if (ret2) | |
3584 | break; | |
3585 | } | |
fd28784a AK |
3586 | if (ret == -ENOSPC && |
3587 | ext4_should_retry_alloc(inode->i_sb, &retries)) { | |
3588 | ret = 0; | |
a2df2a63 | 3589 | goto retry; |
a2df2a63 | 3590 | } |
55bd725a | 3591 | mutex_unlock(&inode->i_mutex); |
a2df2a63 AA |
3592 | return ret > 0 ? ret2 : ret; |
3593 | } | |
6873fa0d | 3594 | |
0031462b MC |
3595 | /* |
3596 | * This function convert a range of blocks to written extents | |
3597 | * The caller of this function will pass the start offset and the size. | |
3598 | * all unwritten extents within this range will be converted to | |
3599 | * written extents. | |
3600 | * | |
3601 | * This function is called from the direct IO end io call back | |
3602 | * function, to convert the fallocated extents after IO is completed. | |
109f5565 | 3603 | * Returns 0 on success. |
0031462b MC |
3604 | */ |
3605 | int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset, | |
a1de02dc | 3606 | ssize_t len) |
0031462b MC |
3607 | { |
3608 | handle_t *handle; | |
3609 | ext4_lblk_t block; | |
3610 | unsigned int max_blocks; | |
3611 | int ret = 0; | |
3612 | int ret2 = 0; | |
3613 | struct buffer_head map_bh; | |
3614 | unsigned int credits, blkbits = inode->i_blkbits; | |
3615 | ||
3616 | block = offset >> blkbits; | |
3617 | /* | |
3618 | * We can't just convert len to max_blocks because | |
3619 | * If blocksize = 4096 offset = 3072 and len = 2048 | |
3620 | */ | |
3621 | max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) | |
3622 | - block; | |
3623 | /* | |
3624 | * credits to insert 1 extent into extent tree | |
3625 | */ | |
3626 | credits = ext4_chunk_trans_blocks(inode, max_blocks); | |
3627 | while (ret >= 0 && ret < max_blocks) { | |
3628 | block = block + ret; | |
3629 | max_blocks = max_blocks - ret; | |
3630 | handle = ext4_journal_start(inode, credits); | |
3631 | if (IS_ERR(handle)) { | |
3632 | ret = PTR_ERR(handle); | |
3633 | break; | |
3634 | } | |
3635 | map_bh.b_state = 0; | |
3636 | ret = ext4_get_blocks(handle, inode, block, | |
3637 | max_blocks, &map_bh, | |
3638 | EXT4_GET_BLOCKS_DIO_CONVERT_EXT); | |
3639 | if (ret <= 0) { | |
3640 | WARN_ON(ret <= 0); | |
3641 | printk(KERN_ERR "%s: ext4_ext_get_blocks " | |
3642 | "returned error inode#%lu, block=%u, " | |
3643 | "max_blocks=%u", __func__, | |
3644 | inode->i_ino, block, max_blocks); | |
3645 | } | |
3646 | ext4_mark_inode_dirty(handle, inode); | |
3647 | ret2 = ext4_journal_stop(handle); | |
3648 | if (ret <= 0 || ret2 ) | |
3649 | break; | |
3650 | } | |
3651 | return ret > 0 ? ret2 : ret; | |
3652 | } | |
6873fa0d ES |
3653 | /* |
3654 | * Callback function called for each extent to gather FIEMAP information. | |
3655 | */ | |
3a06d778 | 3656 | static int ext4_ext_fiemap_cb(struct inode *inode, struct ext4_ext_path *path, |
6873fa0d ES |
3657 | struct ext4_ext_cache *newex, struct ext4_extent *ex, |
3658 | void *data) | |
3659 | { | |
3660 | struct fiemap_extent_info *fieinfo = data; | |
c9877b20 | 3661 | unsigned char blksize_bits = inode->i_sb->s_blocksize_bits; |
6873fa0d ES |
3662 | __u64 logical; |
3663 | __u64 physical; | |
3664 | __u64 length; | |
3665 | __u32 flags = 0; | |
3666 | int error; | |
3667 | ||
3668 | logical = (__u64)newex->ec_block << blksize_bits; | |
3669 | ||
3670 | if (newex->ec_type == EXT4_EXT_CACHE_GAP) { | |
3671 | pgoff_t offset; | |
3672 | struct page *page; | |
3673 | struct buffer_head *bh = NULL; | |
3674 | ||
3675 | offset = logical >> PAGE_SHIFT; | |
3676 | page = find_get_page(inode->i_mapping, offset); | |
3677 | if (!page || !page_has_buffers(page)) | |
3678 | return EXT_CONTINUE; | |
3679 | ||
3680 | bh = page_buffers(page); | |
3681 | ||
3682 | if (!bh) | |
3683 | return EXT_CONTINUE; | |
3684 | ||
3685 | if (buffer_delay(bh)) { | |
3686 | flags |= FIEMAP_EXTENT_DELALLOC; | |
3687 | page_cache_release(page); | |
3688 | } else { | |
3689 | page_cache_release(page); | |
3690 | return EXT_CONTINUE; | |
3691 | } | |
3692 | } | |
3693 | ||
3694 | physical = (__u64)newex->ec_start << blksize_bits; | |
3695 | length = (__u64)newex->ec_len << blksize_bits; | |
3696 | ||
3697 | if (ex && ext4_ext_is_uninitialized(ex)) | |
3698 | flags |= FIEMAP_EXTENT_UNWRITTEN; | |
3699 | ||
3700 | /* | |
3701 | * If this extent reaches EXT_MAX_BLOCK, it must be last. | |
3702 | * | |
3703 | * Or if ext4_ext_next_allocated_block is EXT_MAX_BLOCK, | |
3704 | * this also indicates no more allocated blocks. | |
3705 | * | |
3706 | * XXX this might miss a single-block extent at EXT_MAX_BLOCK | |
3707 | */ | |
c9877b20 | 3708 | if (ext4_ext_next_allocated_block(path) == EXT_MAX_BLOCK || |
eefd7f03 TT |
3709 | newex->ec_block + newex->ec_len - 1 == EXT_MAX_BLOCK) { |
3710 | loff_t size = i_size_read(inode); | |
3711 | loff_t bs = EXT4_BLOCK_SIZE(inode->i_sb); | |
3712 | ||
6873fa0d | 3713 | flags |= FIEMAP_EXTENT_LAST; |
eefd7f03 TT |
3714 | if ((flags & FIEMAP_EXTENT_DELALLOC) && |
3715 | logical+length > size) | |
3716 | length = (size - logical + bs - 1) & ~(bs-1); | |
3717 | } | |
6873fa0d ES |
3718 | |
3719 | error = fiemap_fill_next_extent(fieinfo, logical, physical, | |
3720 | length, flags); | |
3721 | if (error < 0) | |
3722 | return error; | |
3723 | if (error == 1) | |
3724 | return EXT_BREAK; | |
3725 | ||
3726 | return EXT_CONTINUE; | |
3727 | } | |
3728 | ||
3729 | /* fiemap flags we can handle specified here */ | |
3730 | #define EXT4_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR) | |
3731 | ||
3a06d778 AK |
3732 | static int ext4_xattr_fiemap(struct inode *inode, |
3733 | struct fiemap_extent_info *fieinfo) | |
6873fa0d ES |
3734 | { |
3735 | __u64 physical = 0; | |
3736 | __u64 length; | |
3737 | __u32 flags = FIEMAP_EXTENT_LAST; | |
3738 | int blockbits = inode->i_sb->s_blocksize_bits; | |
3739 | int error = 0; | |
3740 | ||
3741 | /* in-inode? */ | |
19f5fb7a | 3742 | if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) { |
6873fa0d ES |
3743 | struct ext4_iloc iloc; |
3744 | int offset; /* offset of xattr in inode */ | |
3745 | ||
3746 | error = ext4_get_inode_loc(inode, &iloc); | |
3747 | if (error) | |
3748 | return error; | |
3749 | physical = iloc.bh->b_blocknr << blockbits; | |
3750 | offset = EXT4_GOOD_OLD_INODE_SIZE + | |
3751 | EXT4_I(inode)->i_extra_isize; | |
3752 | physical += offset; | |
3753 | length = EXT4_SB(inode->i_sb)->s_inode_size - offset; | |
3754 | flags |= FIEMAP_EXTENT_DATA_INLINE; | |
3755 | } else { /* external block */ | |
3756 | physical = EXT4_I(inode)->i_file_acl << blockbits; | |
3757 | length = inode->i_sb->s_blocksize; | |
3758 | } | |
3759 | ||
3760 | if (physical) | |
3761 | error = fiemap_fill_next_extent(fieinfo, 0, physical, | |
3762 | length, flags); | |
3763 | return (error < 0 ? error : 0); | |
3764 | } | |
3765 | ||
3766 | int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, | |
3767 | __u64 start, __u64 len) | |
3768 | { | |
3769 | ext4_lblk_t start_blk; | |
3770 | ext4_lblk_t len_blks; | |
3771 | int error = 0; | |
3772 | ||
3773 | /* fallback to generic here if not in extents fmt */ | |
3774 | if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL)) | |
3775 | return generic_block_fiemap(inode, fieinfo, start, len, | |
3776 | ext4_get_block); | |
3777 | ||
3778 | if (fiemap_check_flags(fieinfo, EXT4_FIEMAP_FLAGS)) | |
3779 | return -EBADR; | |
3780 | ||
3781 | if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) { | |
3782 | error = ext4_xattr_fiemap(inode, fieinfo); | |
3783 | } else { | |
3784 | start_blk = start >> inode->i_sb->s_blocksize_bits; | |
3785 | len_blks = len >> inode->i_sb->s_blocksize_bits; | |
3786 | ||
3787 | /* | |
3788 | * Walk the extent tree gathering extent information. | |
3789 | * ext4_ext_fiemap_cb will push extents back to user. | |
3790 | */ | |
6873fa0d ES |
3791 | error = ext4_ext_walk_space(inode, start_blk, len_blks, |
3792 | ext4_ext_fiemap_cb, fieinfo); | |
6873fa0d ES |
3793 | } |
3794 | ||
3795 | return error; | |
3796 | } | |
3797 |