]>
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 | 44 | #include "ext4_jbd2.h" |
a86c6181 | 45 | |
0562e0ba JZ |
46 | #include <trace/events/ext4.h> |
47 | ||
d583fb87 AH |
48 | static int ext4_split_extent(handle_t *handle, |
49 | struct inode *inode, | |
50 | struct ext4_ext_path *path, | |
51 | struct ext4_map_blocks *map, | |
52 | int split_flag, | |
53 | int flags); | |
54 | ||
487caeef JK |
55 | static int ext4_ext_truncate_extend_restart(handle_t *handle, |
56 | struct inode *inode, | |
57 | int needed) | |
a86c6181 AT |
58 | { |
59 | int err; | |
60 | ||
0390131b FM |
61 | if (!ext4_handle_valid(handle)) |
62 | return 0; | |
a86c6181 | 63 | if (handle->h_buffer_credits > needed) |
9102e4fa SF |
64 | return 0; |
65 | err = ext4_journal_extend(handle, needed); | |
0123c939 | 66 | if (err <= 0) |
9102e4fa | 67 | return err; |
487caeef | 68 | err = ext4_truncate_restart_trans(handle, inode, needed); |
0617b83f DM |
69 | if (err == 0) |
70 | err = -EAGAIN; | |
487caeef JK |
71 | |
72 | return err; | |
a86c6181 AT |
73 | } |
74 | ||
75 | /* | |
76 | * could return: | |
77 | * - EROFS | |
78 | * - ENOMEM | |
79 | */ | |
80 | static int ext4_ext_get_access(handle_t *handle, struct inode *inode, | |
81 | struct ext4_ext_path *path) | |
82 | { | |
83 | if (path->p_bh) { | |
84 | /* path points to block */ | |
85 | return ext4_journal_get_write_access(handle, path->p_bh); | |
86 | } | |
87 | /* path points to leaf/index in inode body */ | |
88 | /* we use in-core data, no need to protect them */ | |
89 | return 0; | |
90 | } | |
91 | ||
92 | /* | |
93 | * could return: | |
94 | * - EROFS | |
95 | * - ENOMEM | |
96 | * - EIO | |
97 | */ | |
9ea7a0df TT |
98 | #define ext4_ext_dirty(handle, inode, path) \ |
99 | __ext4_ext_dirty(__func__, __LINE__, (handle), (inode), (path)) | |
100 | static int __ext4_ext_dirty(const char *where, unsigned int line, | |
101 | handle_t *handle, struct inode *inode, | |
102 | struct ext4_ext_path *path) | |
a86c6181 AT |
103 | { |
104 | int err; | |
105 | if (path->p_bh) { | |
106 | /* path points to block */ | |
9ea7a0df TT |
107 | err = __ext4_handle_dirty_metadata(where, line, handle, |
108 | inode, path->p_bh); | |
a86c6181 AT |
109 | } else { |
110 | /* path points to leaf/index in inode body */ | |
111 | err = ext4_mark_inode_dirty(handle, inode); | |
112 | } | |
113 | return err; | |
114 | } | |
115 | ||
f65e6fba | 116 | static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode, |
a86c6181 | 117 | struct ext4_ext_path *path, |
725d26d3 | 118 | ext4_lblk_t block) |
a86c6181 | 119 | { |
a86c6181 | 120 | if (path) { |
81fdbb4a | 121 | int depth = path->p_depth; |
a86c6181 | 122 | struct ext4_extent *ex; |
a86c6181 | 123 | |
ad4fb9ca KM |
124 | /* |
125 | * Try to predict block placement assuming that we are | |
126 | * filling in a file which will eventually be | |
127 | * non-sparse --- i.e., in the case of libbfd writing | |
128 | * an ELF object sections out-of-order but in a way | |
129 | * the eventually results in a contiguous object or | |
130 | * executable file, or some database extending a table | |
131 | * space file. However, this is actually somewhat | |
132 | * non-ideal if we are writing a sparse file such as | |
133 | * qemu or KVM writing a raw image file that is going | |
134 | * to stay fairly sparse, since it will end up | |
135 | * fragmenting the file system's free space. Maybe we | |
136 | * should have some hueristics or some way to allow | |
137 | * userspace to pass a hint to file system, | |
b8d6568a | 138 | * especially if the latter case turns out to be |
ad4fb9ca KM |
139 | * common. |
140 | */ | |
7e028976 | 141 | ex = path[depth].p_ext; |
ad4fb9ca KM |
142 | if (ex) { |
143 | ext4_fsblk_t ext_pblk = ext4_ext_pblock(ex); | |
144 | ext4_lblk_t ext_block = le32_to_cpu(ex->ee_block); | |
145 | ||
146 | if (block > ext_block) | |
147 | return ext_pblk + (block - ext_block); | |
148 | else | |
149 | return ext_pblk - (ext_block - block); | |
150 | } | |
a86c6181 | 151 | |
d0d856e8 RD |
152 | /* it looks like index is empty; |
153 | * try to find starting block from index itself */ | |
a86c6181 AT |
154 | if (path[depth].p_bh) |
155 | return path[depth].p_bh->b_blocknr; | |
156 | } | |
157 | ||
158 | /* OK. use inode's group */ | |
f86186b4 | 159 | return ext4_inode_to_goal_block(inode); |
a86c6181 AT |
160 | } |
161 | ||
654b4908 AK |
162 | /* |
163 | * Allocation for a meta data block | |
164 | */ | |
f65e6fba | 165 | static ext4_fsblk_t |
654b4908 | 166 | ext4_ext_new_meta_block(handle_t *handle, struct inode *inode, |
a86c6181 | 167 | struct ext4_ext_path *path, |
55f020db | 168 | struct ext4_extent *ex, int *err, unsigned int flags) |
a86c6181 | 169 | { |
f65e6fba | 170 | ext4_fsblk_t goal, newblock; |
a86c6181 AT |
171 | |
172 | goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block)); | |
55f020db AH |
173 | newblock = ext4_new_meta_blocks(handle, inode, goal, flags, |
174 | NULL, err); | |
a86c6181 AT |
175 | return newblock; |
176 | } | |
177 | ||
55ad63bf | 178 | static inline int ext4_ext_space_block(struct inode *inode, int check) |
a86c6181 AT |
179 | { |
180 | int size; | |
181 | ||
182 | size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header)) | |
183 | / sizeof(struct ext4_extent); | |
55ad63bf | 184 | if (!check) { |
bbf2f9fb | 185 | #ifdef AGGRESSIVE_TEST |
55ad63bf TT |
186 | if (size > 6) |
187 | size = 6; | |
a86c6181 | 188 | #endif |
55ad63bf | 189 | } |
a86c6181 AT |
190 | return size; |
191 | } | |
192 | ||
55ad63bf | 193 | static inline int ext4_ext_space_block_idx(struct inode *inode, int check) |
a86c6181 AT |
194 | { |
195 | int size; | |
196 | ||
197 | size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header)) | |
198 | / sizeof(struct ext4_extent_idx); | |
55ad63bf | 199 | if (!check) { |
bbf2f9fb | 200 | #ifdef AGGRESSIVE_TEST |
55ad63bf TT |
201 | if (size > 5) |
202 | size = 5; | |
a86c6181 | 203 | #endif |
55ad63bf | 204 | } |
a86c6181 AT |
205 | return size; |
206 | } | |
207 | ||
55ad63bf | 208 | static inline int ext4_ext_space_root(struct inode *inode, int check) |
a86c6181 AT |
209 | { |
210 | int size; | |
211 | ||
212 | size = sizeof(EXT4_I(inode)->i_data); | |
213 | size -= sizeof(struct ext4_extent_header); | |
214 | size /= sizeof(struct ext4_extent); | |
55ad63bf | 215 | if (!check) { |
bbf2f9fb | 216 | #ifdef AGGRESSIVE_TEST |
55ad63bf TT |
217 | if (size > 3) |
218 | size = 3; | |
a86c6181 | 219 | #endif |
55ad63bf | 220 | } |
a86c6181 AT |
221 | return size; |
222 | } | |
223 | ||
55ad63bf | 224 | static inline int ext4_ext_space_root_idx(struct inode *inode, int check) |
a86c6181 AT |
225 | { |
226 | int size; | |
227 | ||
228 | size = sizeof(EXT4_I(inode)->i_data); | |
229 | size -= sizeof(struct ext4_extent_header); | |
230 | size /= sizeof(struct ext4_extent_idx); | |
55ad63bf | 231 | if (!check) { |
bbf2f9fb | 232 | #ifdef AGGRESSIVE_TEST |
55ad63bf TT |
233 | if (size > 4) |
234 | size = 4; | |
a86c6181 | 235 | #endif |
55ad63bf | 236 | } |
a86c6181 AT |
237 | return size; |
238 | } | |
239 | ||
d2a17637 MC |
240 | /* |
241 | * Calculate the number of metadata blocks needed | |
242 | * to allocate @blocks | |
243 | * Worse case is one block per extent | |
244 | */ | |
01f49d0b | 245 | int ext4_ext_calc_metadata_amount(struct inode *inode, ext4_lblk_t lblock) |
d2a17637 | 246 | { |
9d0be502 | 247 | struct ext4_inode_info *ei = EXT4_I(inode); |
81fdbb4a | 248 | int idxs; |
d2a17637 | 249 | |
9d0be502 TT |
250 | idxs = ((inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header)) |
251 | / sizeof(struct ext4_extent_idx)); | |
d2a17637 MC |
252 | |
253 | /* | |
9d0be502 TT |
254 | * If the new delayed allocation block is contiguous with the |
255 | * previous da block, it can share index blocks with the | |
256 | * previous block, so we only need to allocate a new index | |
257 | * block every idxs leaf blocks. At ldxs**2 blocks, we need | |
258 | * an additional index block, and at ldxs**3 blocks, yet | |
259 | * another index blocks. | |
d2a17637 | 260 | */ |
9d0be502 TT |
261 | if (ei->i_da_metadata_calc_len && |
262 | ei->i_da_metadata_calc_last_lblock+1 == lblock) { | |
81fdbb4a YY |
263 | int num = 0; |
264 | ||
9d0be502 TT |
265 | if ((ei->i_da_metadata_calc_len % idxs) == 0) |
266 | num++; | |
267 | if ((ei->i_da_metadata_calc_len % (idxs*idxs)) == 0) | |
268 | num++; | |
269 | if ((ei->i_da_metadata_calc_len % (idxs*idxs*idxs)) == 0) { | |
270 | num++; | |
271 | ei->i_da_metadata_calc_len = 0; | |
272 | } else | |
273 | ei->i_da_metadata_calc_len++; | |
274 | ei->i_da_metadata_calc_last_lblock++; | |
275 | return num; | |
276 | } | |
d2a17637 | 277 | |
9d0be502 TT |
278 | /* |
279 | * In the worst case we need a new set of index blocks at | |
280 | * every level of the inode's extent tree. | |
281 | */ | |
282 | ei->i_da_metadata_calc_len = 1; | |
283 | ei->i_da_metadata_calc_last_lblock = lblock; | |
284 | return ext_depth(inode) + 1; | |
d2a17637 MC |
285 | } |
286 | ||
c29c0ae7 AT |
287 | static int |
288 | ext4_ext_max_entries(struct inode *inode, int depth) | |
289 | { | |
290 | int max; | |
291 | ||
292 | if (depth == ext_depth(inode)) { | |
293 | if (depth == 0) | |
55ad63bf | 294 | max = ext4_ext_space_root(inode, 1); |
c29c0ae7 | 295 | else |
55ad63bf | 296 | max = ext4_ext_space_root_idx(inode, 1); |
c29c0ae7 AT |
297 | } else { |
298 | if (depth == 0) | |
55ad63bf | 299 | max = ext4_ext_space_block(inode, 1); |
c29c0ae7 | 300 | else |
55ad63bf | 301 | max = ext4_ext_space_block_idx(inode, 1); |
c29c0ae7 AT |
302 | } |
303 | ||
304 | return max; | |
305 | } | |
306 | ||
56b19868 AK |
307 | static int ext4_valid_extent(struct inode *inode, struct ext4_extent *ext) |
308 | { | |
bf89d16f | 309 | ext4_fsblk_t block = ext4_ext_pblock(ext); |
56b19868 | 310 | int len = ext4_ext_get_actual_len(ext); |
e84a26ce | 311 | |
6fd058f7 | 312 | return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, len); |
56b19868 AK |
313 | } |
314 | ||
315 | static int ext4_valid_extent_idx(struct inode *inode, | |
316 | struct ext4_extent_idx *ext_idx) | |
317 | { | |
bf89d16f | 318 | ext4_fsblk_t block = ext4_idx_pblock(ext_idx); |
e84a26ce | 319 | |
6fd058f7 | 320 | return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, 1); |
56b19868 AK |
321 | } |
322 | ||
323 | static int ext4_valid_extent_entries(struct inode *inode, | |
324 | struct ext4_extent_header *eh, | |
325 | int depth) | |
326 | { | |
56b19868 AK |
327 | unsigned short entries; |
328 | if (eh->eh_entries == 0) | |
329 | return 1; | |
330 | ||
331 | entries = le16_to_cpu(eh->eh_entries); | |
332 | ||
333 | if (depth == 0) { | |
334 | /* leaf entries */ | |
81fdbb4a | 335 | struct ext4_extent *ext = EXT_FIRST_EXTENT(eh); |
56b19868 AK |
336 | while (entries) { |
337 | if (!ext4_valid_extent(inode, ext)) | |
338 | return 0; | |
339 | ext++; | |
340 | entries--; | |
341 | } | |
342 | } else { | |
81fdbb4a | 343 | struct ext4_extent_idx *ext_idx = EXT_FIRST_INDEX(eh); |
56b19868 AK |
344 | while (entries) { |
345 | if (!ext4_valid_extent_idx(inode, ext_idx)) | |
346 | return 0; | |
347 | ext_idx++; | |
348 | entries--; | |
349 | } | |
350 | } | |
351 | return 1; | |
352 | } | |
353 | ||
c398eda0 TT |
354 | static int __ext4_ext_check(const char *function, unsigned int line, |
355 | struct inode *inode, struct ext4_extent_header *eh, | |
356 | int depth) | |
c29c0ae7 AT |
357 | { |
358 | const char *error_msg; | |
359 | int max = 0; | |
360 | ||
361 | if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) { | |
362 | error_msg = "invalid magic"; | |
363 | goto corrupted; | |
364 | } | |
365 | if (unlikely(le16_to_cpu(eh->eh_depth) != depth)) { | |
366 | error_msg = "unexpected eh_depth"; | |
367 | goto corrupted; | |
368 | } | |
369 | if (unlikely(eh->eh_max == 0)) { | |
370 | error_msg = "invalid eh_max"; | |
371 | goto corrupted; | |
372 | } | |
373 | max = ext4_ext_max_entries(inode, depth); | |
374 | if (unlikely(le16_to_cpu(eh->eh_max) > max)) { | |
375 | error_msg = "too large eh_max"; | |
376 | goto corrupted; | |
377 | } | |
378 | if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) { | |
379 | error_msg = "invalid eh_entries"; | |
380 | goto corrupted; | |
381 | } | |
56b19868 AK |
382 | if (!ext4_valid_extent_entries(inode, eh, depth)) { |
383 | error_msg = "invalid extent entries"; | |
384 | goto corrupted; | |
385 | } | |
c29c0ae7 AT |
386 | return 0; |
387 | ||
388 | corrupted: | |
c398eda0 | 389 | ext4_error_inode(inode, function, line, 0, |
24676da4 | 390 | "bad header/extent: %s - magic %x, " |
c29c0ae7 | 391 | "entries %u, max %u(%u), depth %u(%u)", |
24676da4 | 392 | error_msg, le16_to_cpu(eh->eh_magic), |
c29c0ae7 AT |
393 | le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max), |
394 | max, le16_to_cpu(eh->eh_depth), depth); | |
395 | ||
396 | return -EIO; | |
397 | } | |
398 | ||
56b19868 | 399 | #define ext4_ext_check(inode, eh, depth) \ |
c398eda0 | 400 | __ext4_ext_check(__func__, __LINE__, inode, eh, depth) |
c29c0ae7 | 401 | |
7a262f7c AK |
402 | int ext4_ext_check_inode(struct inode *inode) |
403 | { | |
404 | return ext4_ext_check(inode, ext_inode_hdr(inode), ext_depth(inode)); | |
405 | } | |
406 | ||
a86c6181 AT |
407 | #ifdef EXT_DEBUG |
408 | static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path) | |
409 | { | |
410 | int k, l = path->p_depth; | |
411 | ||
412 | ext_debug("path:"); | |
413 | for (k = 0; k <= l; k++, path++) { | |
414 | if (path->p_idx) { | |
2ae02107 | 415 | ext_debug(" %d->%llu", le32_to_cpu(path->p_idx->ei_block), |
bf89d16f | 416 | ext4_idx_pblock(path->p_idx)); |
a86c6181 | 417 | } else if (path->p_ext) { |
553f9008 | 418 | ext_debug(" %d:[%d]%d:%llu ", |
a86c6181 | 419 | le32_to_cpu(path->p_ext->ee_block), |
553f9008 | 420 | ext4_ext_is_uninitialized(path->p_ext), |
a2df2a63 | 421 | ext4_ext_get_actual_len(path->p_ext), |
bf89d16f | 422 | ext4_ext_pblock(path->p_ext)); |
a86c6181 AT |
423 | } else |
424 | ext_debug(" []"); | |
425 | } | |
426 | ext_debug("\n"); | |
427 | } | |
428 | ||
429 | static void ext4_ext_show_leaf(struct inode *inode, struct ext4_ext_path *path) | |
430 | { | |
431 | int depth = ext_depth(inode); | |
432 | struct ext4_extent_header *eh; | |
433 | struct ext4_extent *ex; | |
434 | int i; | |
435 | ||
436 | if (!path) | |
437 | return; | |
438 | ||
439 | eh = path[depth].p_hdr; | |
440 | ex = EXT_FIRST_EXTENT(eh); | |
441 | ||
553f9008 M |
442 | ext_debug("Displaying leaf extents for inode %lu\n", inode->i_ino); |
443 | ||
a86c6181 | 444 | for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) { |
553f9008 M |
445 | ext_debug("%d:[%d]%d:%llu ", le32_to_cpu(ex->ee_block), |
446 | ext4_ext_is_uninitialized(ex), | |
bf89d16f | 447 | ext4_ext_get_actual_len(ex), ext4_ext_pblock(ex)); |
a86c6181 AT |
448 | } |
449 | ext_debug("\n"); | |
450 | } | |
1b16da77 YY |
451 | |
452 | static void ext4_ext_show_move(struct inode *inode, struct ext4_ext_path *path, | |
453 | ext4_fsblk_t newblock, int level) | |
454 | { | |
455 | int depth = ext_depth(inode); | |
456 | struct ext4_extent *ex; | |
457 | ||
458 | if (depth != level) { | |
459 | struct ext4_extent_idx *idx; | |
460 | idx = path[level].p_idx; | |
461 | while (idx <= EXT_MAX_INDEX(path[level].p_hdr)) { | |
462 | ext_debug("%d: move %d:%llu in new index %llu\n", level, | |
463 | le32_to_cpu(idx->ei_block), | |
464 | ext4_idx_pblock(idx), | |
465 | newblock); | |
466 | idx++; | |
467 | } | |
468 | ||
469 | return; | |
470 | } | |
471 | ||
472 | ex = path[depth].p_ext; | |
473 | while (ex <= EXT_MAX_EXTENT(path[depth].p_hdr)) { | |
474 | ext_debug("move %d:%llu:[%d]%d in new leaf %llu\n", | |
475 | le32_to_cpu(ex->ee_block), | |
476 | ext4_ext_pblock(ex), | |
477 | ext4_ext_is_uninitialized(ex), | |
478 | ext4_ext_get_actual_len(ex), | |
479 | newblock); | |
480 | ex++; | |
481 | } | |
482 | } | |
483 | ||
a86c6181 | 484 | #else |
af5bc92d TT |
485 | #define ext4_ext_show_path(inode, path) |
486 | #define ext4_ext_show_leaf(inode, path) | |
1b16da77 | 487 | #define ext4_ext_show_move(inode, path, newblock, level) |
a86c6181 AT |
488 | #endif |
489 | ||
b35905c1 | 490 | void ext4_ext_drop_refs(struct ext4_ext_path *path) |
a86c6181 AT |
491 | { |
492 | int depth = path->p_depth; | |
493 | int i; | |
494 | ||
495 | for (i = 0; i <= depth; i++, path++) | |
496 | if (path->p_bh) { | |
497 | brelse(path->p_bh); | |
498 | path->p_bh = NULL; | |
499 | } | |
500 | } | |
501 | ||
502 | /* | |
d0d856e8 RD |
503 | * ext4_ext_binsearch_idx: |
504 | * binary search for the closest index of the given block | |
c29c0ae7 | 505 | * the header must be checked before calling this |
a86c6181 AT |
506 | */ |
507 | static void | |
725d26d3 AK |
508 | ext4_ext_binsearch_idx(struct inode *inode, |
509 | struct ext4_ext_path *path, ext4_lblk_t block) | |
a86c6181 AT |
510 | { |
511 | struct ext4_extent_header *eh = path->p_hdr; | |
512 | struct ext4_extent_idx *r, *l, *m; | |
513 | ||
a86c6181 | 514 | |
bba90743 | 515 | ext_debug("binsearch for %u(idx): ", block); |
a86c6181 AT |
516 | |
517 | l = EXT_FIRST_INDEX(eh) + 1; | |
e9f410b1 | 518 | r = EXT_LAST_INDEX(eh); |
a86c6181 AT |
519 | while (l <= r) { |
520 | m = l + (r - l) / 2; | |
521 | if (block < le32_to_cpu(m->ei_block)) | |
522 | r = m - 1; | |
523 | else | |
524 | l = m + 1; | |
26d535ed DM |
525 | ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ei_block), |
526 | m, le32_to_cpu(m->ei_block), | |
527 | r, le32_to_cpu(r->ei_block)); | |
a86c6181 AT |
528 | } |
529 | ||
530 | path->p_idx = l - 1; | |
f65e6fba | 531 | ext_debug(" -> %d->%lld ", le32_to_cpu(path->p_idx->ei_block), |
bf89d16f | 532 | ext4_idx_pblock(path->p_idx)); |
a86c6181 AT |
533 | |
534 | #ifdef CHECK_BINSEARCH | |
535 | { | |
536 | struct ext4_extent_idx *chix, *ix; | |
537 | int k; | |
538 | ||
539 | chix = ix = EXT_FIRST_INDEX(eh); | |
540 | for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ix++) { | |
541 | if (k != 0 && | |
542 | le32_to_cpu(ix->ei_block) <= le32_to_cpu(ix[-1].ei_block)) { | |
4776004f TT |
543 | printk(KERN_DEBUG "k=%d, ix=0x%p, " |
544 | "first=0x%p\n", k, | |
545 | ix, EXT_FIRST_INDEX(eh)); | |
546 | printk(KERN_DEBUG "%u <= %u\n", | |
a86c6181 AT |
547 | le32_to_cpu(ix->ei_block), |
548 | le32_to_cpu(ix[-1].ei_block)); | |
549 | } | |
550 | BUG_ON(k && le32_to_cpu(ix->ei_block) | |
8c55e204 | 551 | <= le32_to_cpu(ix[-1].ei_block)); |
a86c6181 AT |
552 | if (block < le32_to_cpu(ix->ei_block)) |
553 | break; | |
554 | chix = ix; | |
555 | } | |
556 | BUG_ON(chix != path->p_idx); | |
557 | } | |
558 | #endif | |
559 | ||
560 | } | |
561 | ||
562 | /* | |
d0d856e8 RD |
563 | * ext4_ext_binsearch: |
564 | * binary search for closest extent of the given block | |
c29c0ae7 | 565 | * the header must be checked before calling this |
a86c6181 AT |
566 | */ |
567 | static void | |
725d26d3 AK |
568 | ext4_ext_binsearch(struct inode *inode, |
569 | struct ext4_ext_path *path, ext4_lblk_t block) | |
a86c6181 AT |
570 | { |
571 | struct ext4_extent_header *eh = path->p_hdr; | |
572 | struct ext4_extent *r, *l, *m; | |
573 | ||
a86c6181 AT |
574 | if (eh->eh_entries == 0) { |
575 | /* | |
d0d856e8 RD |
576 | * this leaf is empty: |
577 | * we get such a leaf in split/add case | |
a86c6181 AT |
578 | */ |
579 | return; | |
580 | } | |
581 | ||
bba90743 | 582 | ext_debug("binsearch for %u: ", block); |
a86c6181 AT |
583 | |
584 | l = EXT_FIRST_EXTENT(eh) + 1; | |
e9f410b1 | 585 | r = EXT_LAST_EXTENT(eh); |
a86c6181 AT |
586 | |
587 | while (l <= r) { | |
588 | m = l + (r - l) / 2; | |
589 | if (block < le32_to_cpu(m->ee_block)) | |
590 | r = m - 1; | |
591 | else | |
592 | l = m + 1; | |
26d535ed DM |
593 | ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ee_block), |
594 | m, le32_to_cpu(m->ee_block), | |
595 | r, le32_to_cpu(r->ee_block)); | |
a86c6181 AT |
596 | } |
597 | ||
598 | path->p_ext = l - 1; | |
553f9008 | 599 | ext_debug(" -> %d:%llu:[%d]%d ", |
8c55e204 | 600 | le32_to_cpu(path->p_ext->ee_block), |
bf89d16f | 601 | ext4_ext_pblock(path->p_ext), |
553f9008 | 602 | ext4_ext_is_uninitialized(path->p_ext), |
a2df2a63 | 603 | ext4_ext_get_actual_len(path->p_ext)); |
a86c6181 AT |
604 | |
605 | #ifdef CHECK_BINSEARCH | |
606 | { | |
607 | struct ext4_extent *chex, *ex; | |
608 | int k; | |
609 | ||
610 | chex = ex = EXT_FIRST_EXTENT(eh); | |
611 | for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ex++) { | |
612 | BUG_ON(k && le32_to_cpu(ex->ee_block) | |
8c55e204 | 613 | <= le32_to_cpu(ex[-1].ee_block)); |
a86c6181 AT |
614 | if (block < le32_to_cpu(ex->ee_block)) |
615 | break; | |
616 | chex = ex; | |
617 | } | |
618 | BUG_ON(chex != path->p_ext); | |
619 | } | |
620 | #endif | |
621 | ||
622 | } | |
623 | ||
624 | int ext4_ext_tree_init(handle_t *handle, struct inode *inode) | |
625 | { | |
626 | struct ext4_extent_header *eh; | |
627 | ||
628 | eh = ext_inode_hdr(inode); | |
629 | eh->eh_depth = 0; | |
630 | eh->eh_entries = 0; | |
631 | eh->eh_magic = EXT4_EXT_MAGIC; | |
55ad63bf | 632 | eh->eh_max = cpu_to_le16(ext4_ext_space_root(inode, 0)); |
a86c6181 AT |
633 | ext4_mark_inode_dirty(handle, inode); |
634 | ext4_ext_invalidate_cache(inode); | |
635 | return 0; | |
636 | } | |
637 | ||
638 | struct ext4_ext_path * | |
725d26d3 AK |
639 | ext4_ext_find_extent(struct inode *inode, ext4_lblk_t block, |
640 | struct ext4_ext_path *path) | |
a86c6181 AT |
641 | { |
642 | struct ext4_extent_header *eh; | |
643 | struct buffer_head *bh; | |
644 | short int depth, i, ppos = 0, alloc = 0; | |
645 | ||
646 | eh = ext_inode_hdr(inode); | |
c29c0ae7 | 647 | depth = ext_depth(inode); |
a86c6181 AT |
648 | |
649 | /* account possible depth increase */ | |
650 | if (!path) { | |
5d4958f9 | 651 | path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 2), |
a86c6181 AT |
652 | GFP_NOFS); |
653 | if (!path) | |
654 | return ERR_PTR(-ENOMEM); | |
655 | alloc = 1; | |
656 | } | |
a86c6181 | 657 | path[0].p_hdr = eh; |
1973adcb | 658 | path[0].p_bh = NULL; |
a86c6181 | 659 | |
c29c0ae7 | 660 | i = depth; |
a86c6181 AT |
661 | /* walk through the tree */ |
662 | while (i) { | |
7a262f7c AK |
663 | int need_to_validate = 0; |
664 | ||
a86c6181 AT |
665 | ext_debug("depth %d: num %d, max %d\n", |
666 | ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max)); | |
c29c0ae7 | 667 | |
a86c6181 | 668 | ext4_ext_binsearch_idx(inode, path + ppos, block); |
bf89d16f | 669 | path[ppos].p_block = ext4_idx_pblock(path[ppos].p_idx); |
a86c6181 AT |
670 | path[ppos].p_depth = i; |
671 | path[ppos].p_ext = NULL; | |
672 | ||
7a262f7c AK |
673 | bh = sb_getblk(inode->i_sb, path[ppos].p_block); |
674 | if (unlikely(!bh)) | |
a86c6181 | 675 | goto err; |
7a262f7c | 676 | if (!bh_uptodate_or_lock(bh)) { |
0562e0ba JZ |
677 | trace_ext4_ext_load_extent(inode, block, |
678 | path[ppos].p_block); | |
7a262f7c AK |
679 | if (bh_submit_read(bh) < 0) { |
680 | put_bh(bh); | |
681 | goto err; | |
682 | } | |
683 | /* validate the extent entries */ | |
684 | need_to_validate = 1; | |
685 | } | |
a86c6181 AT |
686 | eh = ext_block_hdr(bh); |
687 | ppos++; | |
273df556 FM |
688 | if (unlikely(ppos > depth)) { |
689 | put_bh(bh); | |
690 | EXT4_ERROR_INODE(inode, | |
691 | "ppos %d > depth %d", ppos, depth); | |
692 | goto err; | |
693 | } | |
a86c6181 AT |
694 | path[ppos].p_bh = bh; |
695 | path[ppos].p_hdr = eh; | |
696 | i--; | |
697 | ||
7a262f7c | 698 | if (need_to_validate && ext4_ext_check(inode, eh, i)) |
a86c6181 AT |
699 | goto err; |
700 | } | |
701 | ||
702 | path[ppos].p_depth = i; | |
a86c6181 AT |
703 | path[ppos].p_ext = NULL; |
704 | path[ppos].p_idx = NULL; | |
705 | ||
a86c6181 AT |
706 | /* find extent */ |
707 | ext4_ext_binsearch(inode, path + ppos, block); | |
1973adcb SF |
708 | /* if not an empty leaf */ |
709 | if (path[ppos].p_ext) | |
bf89d16f | 710 | path[ppos].p_block = ext4_ext_pblock(path[ppos].p_ext); |
a86c6181 AT |
711 | |
712 | ext4_ext_show_path(inode, path); | |
713 | ||
714 | return path; | |
715 | ||
716 | err: | |
717 | ext4_ext_drop_refs(path); | |
718 | if (alloc) | |
719 | kfree(path); | |
720 | return ERR_PTR(-EIO); | |
721 | } | |
722 | ||
723 | /* | |
d0d856e8 RD |
724 | * ext4_ext_insert_index: |
725 | * insert new index [@logical;@ptr] into the block at @curp; | |
726 | * check where to insert: before @curp or after @curp | |
a86c6181 | 727 | */ |
1f109d5a TT |
728 | static int ext4_ext_insert_index(handle_t *handle, struct inode *inode, |
729 | struct ext4_ext_path *curp, | |
730 | int logical, ext4_fsblk_t ptr) | |
a86c6181 AT |
731 | { |
732 | struct ext4_extent_idx *ix; | |
733 | int len, err; | |
734 | ||
7e028976 AM |
735 | err = ext4_ext_get_access(handle, inode, curp); |
736 | if (err) | |
a86c6181 AT |
737 | return err; |
738 | ||
273df556 FM |
739 | if (unlikely(logical == le32_to_cpu(curp->p_idx->ei_block))) { |
740 | EXT4_ERROR_INODE(inode, | |
741 | "logical %d == ei_block %d!", | |
742 | logical, le32_to_cpu(curp->p_idx->ei_block)); | |
743 | return -EIO; | |
744 | } | |
d4620315 RD |
745 | |
746 | if (unlikely(le16_to_cpu(curp->p_hdr->eh_entries) | |
747 | >= le16_to_cpu(curp->p_hdr->eh_max))) { | |
748 | EXT4_ERROR_INODE(inode, | |
749 | "eh_entries %d >= eh_max %d!", | |
750 | le16_to_cpu(curp->p_hdr->eh_entries), | |
751 | le16_to_cpu(curp->p_hdr->eh_max)); | |
752 | return -EIO; | |
753 | } | |
754 | ||
a86c6181 AT |
755 | if (logical > le32_to_cpu(curp->p_idx->ei_block)) { |
756 | /* insert after */ | |
80e675f9 | 757 | ext_debug("insert new index %d after: %llu\n", logical, ptr); |
a86c6181 AT |
758 | ix = curp->p_idx + 1; |
759 | } else { | |
760 | /* insert before */ | |
80e675f9 | 761 | ext_debug("insert new index %d before: %llu\n", logical, ptr); |
a86c6181 AT |
762 | ix = curp->p_idx; |
763 | } | |
764 | ||
80e675f9 EG |
765 | len = EXT_LAST_INDEX(curp->p_hdr) - ix + 1; |
766 | BUG_ON(len < 0); | |
767 | if (len > 0) { | |
768 | ext_debug("insert new index %d: " | |
769 | "move %d indices from 0x%p to 0x%p\n", | |
770 | logical, len, ix, ix + 1); | |
771 | memmove(ix + 1, ix, len * sizeof(struct ext4_extent_idx)); | |
772 | } | |
773 | ||
f472e026 TM |
774 | if (unlikely(ix > EXT_MAX_INDEX(curp->p_hdr))) { |
775 | EXT4_ERROR_INODE(inode, "ix > EXT_MAX_INDEX!"); | |
776 | return -EIO; | |
777 | } | |
778 | ||
a86c6181 | 779 | ix->ei_block = cpu_to_le32(logical); |
f65e6fba | 780 | ext4_idx_store_pblock(ix, ptr); |
e8546d06 | 781 | le16_add_cpu(&curp->p_hdr->eh_entries, 1); |
a86c6181 | 782 | |
273df556 FM |
783 | if (unlikely(ix > EXT_LAST_INDEX(curp->p_hdr))) { |
784 | EXT4_ERROR_INODE(inode, "ix > EXT_LAST_INDEX!"); | |
785 | return -EIO; | |
786 | } | |
a86c6181 AT |
787 | |
788 | err = ext4_ext_dirty(handle, inode, curp); | |
789 | ext4_std_error(inode->i_sb, err); | |
790 | ||
791 | return err; | |
792 | } | |
793 | ||
794 | /* | |
d0d856e8 RD |
795 | * ext4_ext_split: |
796 | * inserts new subtree into the path, using free index entry | |
797 | * at depth @at: | |
798 | * - allocates all needed blocks (new leaf and all intermediate index blocks) | |
799 | * - makes decision where to split | |
800 | * - moves remaining extents and index entries (right to the split point) | |
801 | * into the newly allocated blocks | |
802 | * - initializes subtree | |
a86c6181 AT |
803 | */ |
804 | static int ext4_ext_split(handle_t *handle, struct inode *inode, | |
55f020db AH |
805 | unsigned int flags, |
806 | struct ext4_ext_path *path, | |
807 | struct ext4_extent *newext, int at) | |
a86c6181 AT |
808 | { |
809 | struct buffer_head *bh = NULL; | |
810 | int depth = ext_depth(inode); | |
811 | struct ext4_extent_header *neh; | |
812 | struct ext4_extent_idx *fidx; | |
a86c6181 | 813 | int i = at, k, m, a; |
f65e6fba | 814 | ext4_fsblk_t newblock, oldblock; |
a86c6181 | 815 | __le32 border; |
f65e6fba | 816 | ext4_fsblk_t *ablocks = NULL; /* array of allocated blocks */ |
a86c6181 AT |
817 | int err = 0; |
818 | ||
819 | /* make decision: where to split? */ | |
d0d856e8 | 820 | /* FIXME: now decision is simplest: at current extent */ |
a86c6181 | 821 | |
d0d856e8 | 822 | /* if current leaf will be split, then we should use |
a86c6181 | 823 | * border from split point */ |
273df556 FM |
824 | if (unlikely(path[depth].p_ext > EXT_MAX_EXTENT(path[depth].p_hdr))) { |
825 | EXT4_ERROR_INODE(inode, "p_ext > EXT_MAX_EXTENT!"); | |
826 | return -EIO; | |
827 | } | |
a86c6181 AT |
828 | if (path[depth].p_ext != EXT_MAX_EXTENT(path[depth].p_hdr)) { |
829 | border = path[depth].p_ext[1].ee_block; | |
d0d856e8 | 830 | ext_debug("leaf will be split." |
a86c6181 | 831 | " next leaf starts at %d\n", |
8c55e204 | 832 | le32_to_cpu(border)); |
a86c6181 AT |
833 | } else { |
834 | border = newext->ee_block; | |
835 | ext_debug("leaf will be added." | |
836 | " next leaf starts at %d\n", | |
8c55e204 | 837 | le32_to_cpu(border)); |
a86c6181 AT |
838 | } |
839 | ||
840 | /* | |
d0d856e8 RD |
841 | * If error occurs, then we break processing |
842 | * and mark filesystem read-only. index won't | |
a86c6181 | 843 | * be inserted and tree will be in consistent |
d0d856e8 | 844 | * state. Next mount will repair buffers too. |
a86c6181 AT |
845 | */ |
846 | ||
847 | /* | |
d0d856e8 RD |
848 | * Get array to track all allocated blocks. |
849 | * We need this to handle errors and free blocks | |
850 | * upon them. | |
a86c6181 | 851 | */ |
5d4958f9 | 852 | ablocks = kzalloc(sizeof(ext4_fsblk_t) * depth, GFP_NOFS); |
a86c6181 AT |
853 | if (!ablocks) |
854 | return -ENOMEM; | |
a86c6181 AT |
855 | |
856 | /* allocate all needed blocks */ | |
857 | ext_debug("allocate %d blocks for indexes/leaf\n", depth - at); | |
858 | for (a = 0; a < depth - at; a++) { | |
654b4908 | 859 | newblock = ext4_ext_new_meta_block(handle, inode, path, |
55f020db | 860 | newext, &err, flags); |
a86c6181 AT |
861 | if (newblock == 0) |
862 | goto cleanup; | |
863 | ablocks[a] = newblock; | |
864 | } | |
865 | ||
866 | /* initialize new leaf */ | |
867 | newblock = ablocks[--a]; | |
273df556 FM |
868 | if (unlikely(newblock == 0)) { |
869 | EXT4_ERROR_INODE(inode, "newblock == 0!"); | |
870 | err = -EIO; | |
871 | goto cleanup; | |
872 | } | |
a86c6181 AT |
873 | bh = sb_getblk(inode->i_sb, newblock); |
874 | if (!bh) { | |
875 | err = -EIO; | |
876 | goto cleanup; | |
877 | } | |
878 | lock_buffer(bh); | |
879 | ||
7e028976 AM |
880 | err = ext4_journal_get_create_access(handle, bh); |
881 | if (err) | |
a86c6181 AT |
882 | goto cleanup; |
883 | ||
884 | neh = ext_block_hdr(bh); | |
885 | neh->eh_entries = 0; | |
55ad63bf | 886 | neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0)); |
a86c6181 AT |
887 | neh->eh_magic = EXT4_EXT_MAGIC; |
888 | neh->eh_depth = 0; | |
a86c6181 | 889 | |
d0d856e8 | 890 | /* move remainder of path[depth] to the new leaf */ |
273df556 FM |
891 | if (unlikely(path[depth].p_hdr->eh_entries != |
892 | path[depth].p_hdr->eh_max)) { | |
893 | EXT4_ERROR_INODE(inode, "eh_entries %d != eh_max %d!", | |
894 | path[depth].p_hdr->eh_entries, | |
895 | path[depth].p_hdr->eh_max); | |
896 | err = -EIO; | |
897 | goto cleanup; | |
898 | } | |
a86c6181 | 899 | /* start copy from next extent */ |
1b16da77 YY |
900 | m = EXT_MAX_EXTENT(path[depth].p_hdr) - path[depth].p_ext++; |
901 | ext4_ext_show_move(inode, path, newblock, depth); | |
a86c6181 | 902 | if (m) { |
1b16da77 YY |
903 | struct ext4_extent *ex; |
904 | ex = EXT_FIRST_EXTENT(neh); | |
905 | memmove(ex, path[depth].p_ext, sizeof(struct ext4_extent) * m); | |
e8546d06 | 906 | le16_add_cpu(&neh->eh_entries, m); |
a86c6181 AT |
907 | } |
908 | ||
909 | set_buffer_uptodate(bh); | |
910 | unlock_buffer(bh); | |
911 | ||
0390131b | 912 | err = ext4_handle_dirty_metadata(handle, inode, bh); |
7e028976 | 913 | if (err) |
a86c6181 AT |
914 | goto cleanup; |
915 | brelse(bh); | |
916 | bh = NULL; | |
917 | ||
918 | /* correct old leaf */ | |
919 | if (m) { | |
7e028976 AM |
920 | err = ext4_ext_get_access(handle, inode, path + depth); |
921 | if (err) | |
a86c6181 | 922 | goto cleanup; |
e8546d06 | 923 | le16_add_cpu(&path[depth].p_hdr->eh_entries, -m); |
7e028976 AM |
924 | err = ext4_ext_dirty(handle, inode, path + depth); |
925 | if (err) | |
a86c6181 AT |
926 | goto cleanup; |
927 | ||
928 | } | |
929 | ||
930 | /* create intermediate indexes */ | |
931 | k = depth - at - 1; | |
273df556 FM |
932 | if (unlikely(k < 0)) { |
933 | EXT4_ERROR_INODE(inode, "k %d < 0!", k); | |
934 | err = -EIO; | |
935 | goto cleanup; | |
936 | } | |
a86c6181 AT |
937 | if (k) |
938 | ext_debug("create %d intermediate indices\n", k); | |
939 | /* insert new index into current index block */ | |
940 | /* current depth stored in i var */ | |
941 | i = depth - 1; | |
942 | while (k--) { | |
943 | oldblock = newblock; | |
944 | newblock = ablocks[--a]; | |
bba90743 | 945 | bh = sb_getblk(inode->i_sb, newblock); |
a86c6181 AT |
946 | if (!bh) { |
947 | err = -EIO; | |
948 | goto cleanup; | |
949 | } | |
950 | lock_buffer(bh); | |
951 | ||
7e028976 AM |
952 | err = ext4_journal_get_create_access(handle, bh); |
953 | if (err) | |
a86c6181 AT |
954 | goto cleanup; |
955 | ||
956 | neh = ext_block_hdr(bh); | |
957 | neh->eh_entries = cpu_to_le16(1); | |
958 | neh->eh_magic = EXT4_EXT_MAGIC; | |
55ad63bf | 959 | neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0)); |
a86c6181 AT |
960 | neh->eh_depth = cpu_to_le16(depth - i); |
961 | fidx = EXT_FIRST_INDEX(neh); | |
962 | fidx->ei_block = border; | |
f65e6fba | 963 | ext4_idx_store_pblock(fidx, oldblock); |
a86c6181 | 964 | |
bba90743 ES |
965 | ext_debug("int.index at %d (block %llu): %u -> %llu\n", |
966 | i, newblock, le32_to_cpu(border), oldblock); | |
a86c6181 | 967 | |
1b16da77 | 968 | /* move remainder of path[i] to the new index block */ |
273df556 FM |
969 | if (unlikely(EXT_MAX_INDEX(path[i].p_hdr) != |
970 | EXT_LAST_INDEX(path[i].p_hdr))) { | |
971 | EXT4_ERROR_INODE(inode, | |
972 | "EXT_MAX_INDEX != EXT_LAST_INDEX ee_block %d!", | |
973 | le32_to_cpu(path[i].p_ext->ee_block)); | |
974 | err = -EIO; | |
975 | goto cleanup; | |
976 | } | |
1b16da77 YY |
977 | /* start copy indexes */ |
978 | m = EXT_MAX_INDEX(path[i].p_hdr) - path[i].p_idx++; | |
979 | ext_debug("cur 0x%p, last 0x%p\n", path[i].p_idx, | |
980 | EXT_MAX_INDEX(path[i].p_hdr)); | |
981 | ext4_ext_show_move(inode, path, newblock, i); | |
a86c6181 | 982 | if (m) { |
1b16da77 | 983 | memmove(++fidx, path[i].p_idx, |
a86c6181 | 984 | sizeof(struct ext4_extent_idx) * m); |
e8546d06 | 985 | le16_add_cpu(&neh->eh_entries, m); |
a86c6181 AT |
986 | } |
987 | set_buffer_uptodate(bh); | |
988 | unlock_buffer(bh); | |
989 | ||
0390131b | 990 | err = ext4_handle_dirty_metadata(handle, inode, bh); |
7e028976 | 991 | if (err) |
a86c6181 AT |
992 | goto cleanup; |
993 | brelse(bh); | |
994 | bh = NULL; | |
995 | ||
996 | /* correct old index */ | |
997 | if (m) { | |
998 | err = ext4_ext_get_access(handle, inode, path + i); | |
999 | if (err) | |
1000 | goto cleanup; | |
e8546d06 | 1001 | le16_add_cpu(&path[i].p_hdr->eh_entries, -m); |
a86c6181 AT |
1002 | err = ext4_ext_dirty(handle, inode, path + i); |
1003 | if (err) | |
1004 | goto cleanup; | |
1005 | } | |
1006 | ||
1007 | i--; | |
1008 | } | |
1009 | ||
1010 | /* insert new index */ | |
a86c6181 AT |
1011 | err = ext4_ext_insert_index(handle, inode, path + at, |
1012 | le32_to_cpu(border), newblock); | |
1013 | ||
1014 | cleanup: | |
1015 | if (bh) { | |
1016 | if (buffer_locked(bh)) | |
1017 | unlock_buffer(bh); | |
1018 | brelse(bh); | |
1019 | } | |
1020 | ||
1021 | if (err) { | |
1022 | /* free all allocated blocks in error case */ | |
1023 | for (i = 0; i < depth; i++) { | |
1024 | if (!ablocks[i]) | |
1025 | continue; | |
7dc57615 | 1026 | ext4_free_blocks(handle, inode, NULL, ablocks[i], 1, |
e6362609 | 1027 | EXT4_FREE_BLOCKS_METADATA); |
a86c6181 AT |
1028 | } |
1029 | } | |
1030 | kfree(ablocks); | |
1031 | ||
1032 | return err; | |
1033 | } | |
1034 | ||
1035 | /* | |
d0d856e8 RD |
1036 | * ext4_ext_grow_indepth: |
1037 | * implements tree growing procedure: | |
1038 | * - allocates new block | |
1039 | * - moves top-level data (index block or leaf) into the new block | |
1040 | * - initializes new top-level, creating index that points to the | |
1041 | * just created block | |
a86c6181 AT |
1042 | */ |
1043 | static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode, | |
55f020db | 1044 | unsigned int flags, |
55f020db | 1045 | struct ext4_extent *newext) |
a86c6181 | 1046 | { |
a86c6181 | 1047 | struct ext4_extent_header *neh; |
a86c6181 | 1048 | struct buffer_head *bh; |
f65e6fba | 1049 | ext4_fsblk_t newblock; |
a86c6181 AT |
1050 | int err = 0; |
1051 | ||
1939dd84 | 1052 | newblock = ext4_ext_new_meta_block(handle, inode, NULL, |
55f020db | 1053 | newext, &err, flags); |
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 */ | |
1939dd84 DM |
1072 | memmove(bh->b_data, EXT4_I(inode)->i_data, |
1073 | sizeof(EXT4_I(inode)->i_data)); | |
a86c6181 AT |
1074 | |
1075 | /* set size of new block */ | |
1076 | neh = ext_block_hdr(bh); | |
1077 | /* old root could have indexes or leaves | |
1078 | * so calculate e_max right way */ | |
1079 | if (ext_depth(inode)) | |
55ad63bf | 1080 | neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0)); |
a86c6181 | 1081 | else |
55ad63bf | 1082 | neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0)); |
a86c6181 AT |
1083 | neh->eh_magic = EXT4_EXT_MAGIC; |
1084 | set_buffer_uptodate(bh); | |
1085 | unlock_buffer(bh); | |
1086 | ||
0390131b | 1087 | err = ext4_handle_dirty_metadata(handle, inode, bh); |
7e028976 | 1088 | if (err) |
a86c6181 AT |
1089 | goto out; |
1090 | ||
1939dd84 | 1091 | /* Update top-level index: num,max,pointer */ |
a86c6181 | 1092 | neh = ext_inode_hdr(inode); |
1939dd84 DM |
1093 | neh->eh_entries = cpu_to_le16(1); |
1094 | ext4_idx_store_pblock(EXT_FIRST_INDEX(neh), newblock); | |
1095 | if (neh->eh_depth == 0) { | |
1096 | /* Root extent block becomes index block */ | |
1097 | neh->eh_max = cpu_to_le16(ext4_ext_space_root_idx(inode, 0)); | |
1098 | EXT_FIRST_INDEX(neh)->ei_block = | |
1099 | EXT_FIRST_EXTENT(neh)->ee_block; | |
1100 | } | |
2ae02107 | 1101 | ext_debug("new root: num %d(%d), lblock %d, ptr %llu\n", |
a86c6181 | 1102 | le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max), |
5a0790c2 | 1103 | le32_to_cpu(EXT_FIRST_INDEX(neh)->ei_block), |
bf89d16f | 1104 | ext4_idx_pblock(EXT_FIRST_INDEX(neh))); |
a86c6181 | 1105 | |
1939dd84 DM |
1106 | neh->eh_depth = cpu_to_le16(neh->eh_depth + 1); |
1107 | ext4_mark_inode_dirty(handle, inode); | |
a86c6181 AT |
1108 | out: |
1109 | brelse(bh); | |
1110 | ||
1111 | return err; | |
1112 | } | |
1113 | ||
1114 | /* | |
d0d856e8 RD |
1115 | * ext4_ext_create_new_leaf: |
1116 | * finds empty index and adds new leaf. | |
1117 | * if no free index is found, then it requests in-depth growing. | |
a86c6181 AT |
1118 | */ |
1119 | static int ext4_ext_create_new_leaf(handle_t *handle, struct inode *inode, | |
55f020db AH |
1120 | unsigned int flags, |
1121 | struct ext4_ext_path *path, | |
1122 | struct ext4_extent *newext) | |
a86c6181 AT |
1123 | { |
1124 | struct ext4_ext_path *curp; | |
1125 | int depth, i, err = 0; | |
1126 | ||
1127 | repeat: | |
1128 | i = depth = ext_depth(inode); | |
1129 | ||
1130 | /* walk up to the tree and look for free index entry */ | |
1131 | curp = path + depth; | |
1132 | while (i > 0 && !EXT_HAS_FREE_INDEX(curp)) { | |
1133 | i--; | |
1134 | curp--; | |
1135 | } | |
1136 | ||
d0d856e8 RD |
1137 | /* we use already allocated block for index block, |
1138 | * so subsequent data blocks should be contiguous */ | |
a86c6181 AT |
1139 | if (EXT_HAS_FREE_INDEX(curp)) { |
1140 | /* if we found index with free entry, then use that | |
1141 | * entry: create all needed subtree and add new leaf */ | |
55f020db | 1142 | err = ext4_ext_split(handle, inode, flags, path, newext, i); |
787e0981 SF |
1143 | if (err) |
1144 | goto out; | |
a86c6181 AT |
1145 | |
1146 | /* refill path */ | |
1147 | ext4_ext_drop_refs(path); | |
1148 | path = ext4_ext_find_extent(inode, | |
725d26d3 AK |
1149 | (ext4_lblk_t)le32_to_cpu(newext->ee_block), |
1150 | path); | |
a86c6181 AT |
1151 | if (IS_ERR(path)) |
1152 | err = PTR_ERR(path); | |
1153 | } else { | |
1154 | /* tree is full, time to grow in depth */ | |
1939dd84 | 1155 | err = ext4_ext_grow_indepth(handle, inode, flags, newext); |
a86c6181 AT |
1156 | if (err) |
1157 | goto out; | |
1158 | ||
1159 | /* refill path */ | |
1160 | ext4_ext_drop_refs(path); | |
1161 | path = ext4_ext_find_extent(inode, | |
725d26d3 AK |
1162 | (ext4_lblk_t)le32_to_cpu(newext->ee_block), |
1163 | path); | |
a86c6181 AT |
1164 | if (IS_ERR(path)) { |
1165 | err = PTR_ERR(path); | |
1166 | goto out; | |
1167 | } | |
1168 | ||
1169 | /* | |
d0d856e8 RD |
1170 | * only first (depth 0 -> 1) produces free space; |
1171 | * in all other cases we have to split the grown tree | |
a86c6181 AT |
1172 | */ |
1173 | depth = ext_depth(inode); | |
1174 | if (path[depth].p_hdr->eh_entries == path[depth].p_hdr->eh_max) { | |
d0d856e8 | 1175 | /* now we need to split */ |
a86c6181 AT |
1176 | goto repeat; |
1177 | } | |
1178 | } | |
1179 | ||
1180 | out: | |
1181 | return err; | |
1182 | } | |
1183 | ||
1988b51e AT |
1184 | /* |
1185 | * search the closest allocated block to the left for *logical | |
1186 | * and returns it at @logical + it's physical address at @phys | |
1187 | * if *logical is the smallest allocated block, the function | |
1188 | * returns 0 at @phys | |
1189 | * return value contains 0 (success) or error code | |
1190 | */ | |
1f109d5a TT |
1191 | static int ext4_ext_search_left(struct inode *inode, |
1192 | struct ext4_ext_path *path, | |
1193 | ext4_lblk_t *logical, ext4_fsblk_t *phys) | |
1988b51e AT |
1194 | { |
1195 | struct ext4_extent_idx *ix; | |
1196 | struct ext4_extent *ex; | |
b939e376 | 1197 | int depth, ee_len; |
1988b51e | 1198 | |
273df556 FM |
1199 | if (unlikely(path == NULL)) { |
1200 | EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical); | |
1201 | return -EIO; | |
1202 | } | |
1988b51e AT |
1203 | depth = path->p_depth; |
1204 | *phys = 0; | |
1205 | ||
1206 | if (depth == 0 && path->p_ext == NULL) | |
1207 | return 0; | |
1208 | ||
1209 | /* usually extent in the path covers blocks smaller | |
1210 | * then *logical, but it can be that extent is the | |
1211 | * first one in the file */ | |
1212 | ||
1213 | ex = path[depth].p_ext; | |
b939e376 | 1214 | ee_len = ext4_ext_get_actual_len(ex); |
1988b51e | 1215 | if (*logical < le32_to_cpu(ex->ee_block)) { |
273df556 FM |
1216 | if (unlikely(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex)) { |
1217 | EXT4_ERROR_INODE(inode, | |
1218 | "EXT_FIRST_EXTENT != ex *logical %d ee_block %d!", | |
1219 | *logical, le32_to_cpu(ex->ee_block)); | |
1220 | return -EIO; | |
1221 | } | |
1988b51e AT |
1222 | while (--depth >= 0) { |
1223 | ix = path[depth].p_idx; | |
273df556 FM |
1224 | if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) { |
1225 | EXT4_ERROR_INODE(inode, | |
1226 | "ix (%d) != EXT_FIRST_INDEX (%d) (depth %d)!", | |
6ee3b212 | 1227 | ix != NULL ? le32_to_cpu(ix->ei_block) : 0, |
273df556 | 1228 | EXT_FIRST_INDEX(path[depth].p_hdr) != NULL ? |
6ee3b212 | 1229 | le32_to_cpu(EXT_FIRST_INDEX(path[depth].p_hdr)->ei_block) : 0, |
273df556 FM |
1230 | depth); |
1231 | return -EIO; | |
1232 | } | |
1988b51e AT |
1233 | } |
1234 | return 0; | |
1235 | } | |
1236 | ||
273df556 FM |
1237 | if (unlikely(*logical < (le32_to_cpu(ex->ee_block) + ee_len))) { |
1238 | EXT4_ERROR_INODE(inode, | |
1239 | "logical %d < ee_block %d + ee_len %d!", | |
1240 | *logical, le32_to_cpu(ex->ee_block), ee_len); | |
1241 | return -EIO; | |
1242 | } | |
1988b51e | 1243 | |
b939e376 | 1244 | *logical = le32_to_cpu(ex->ee_block) + ee_len - 1; |
bf89d16f | 1245 | *phys = ext4_ext_pblock(ex) + ee_len - 1; |
1988b51e AT |
1246 | return 0; |
1247 | } | |
1248 | ||
1249 | /* | |
1250 | * search the closest allocated block to the right for *logical | |
1251 | * and returns it at @logical + it's physical address at @phys | |
df3ab170 | 1252 | * if *logical is the largest allocated block, the function |
1988b51e AT |
1253 | * returns 0 at @phys |
1254 | * return value contains 0 (success) or error code | |
1255 | */ | |
1f109d5a TT |
1256 | static int ext4_ext_search_right(struct inode *inode, |
1257 | struct ext4_ext_path *path, | |
4d33b1ef TT |
1258 | ext4_lblk_t *logical, ext4_fsblk_t *phys, |
1259 | struct ext4_extent **ret_ex) | |
1988b51e AT |
1260 | { |
1261 | struct buffer_head *bh = NULL; | |
1262 | struct ext4_extent_header *eh; | |
1263 | struct ext4_extent_idx *ix; | |
1264 | struct ext4_extent *ex; | |
1265 | ext4_fsblk_t block; | |
395a87bf ES |
1266 | int depth; /* Note, NOT eh_depth; depth from top of tree */ |
1267 | int ee_len; | |
1988b51e | 1268 | |
273df556 FM |
1269 | if (unlikely(path == NULL)) { |
1270 | EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical); | |
1271 | return -EIO; | |
1272 | } | |
1988b51e AT |
1273 | depth = path->p_depth; |
1274 | *phys = 0; | |
1275 | ||
1276 | if (depth == 0 && path->p_ext == NULL) | |
1277 | return 0; | |
1278 | ||
1279 | /* usually extent in the path covers blocks smaller | |
1280 | * then *logical, but it can be that extent is the | |
1281 | * first one in the file */ | |
1282 | ||
1283 | ex = path[depth].p_ext; | |
b939e376 | 1284 | ee_len = ext4_ext_get_actual_len(ex); |
1988b51e | 1285 | if (*logical < le32_to_cpu(ex->ee_block)) { |
273df556 FM |
1286 | if (unlikely(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex)) { |
1287 | EXT4_ERROR_INODE(inode, | |
1288 | "first_extent(path[%d].p_hdr) != ex", | |
1289 | depth); | |
1290 | return -EIO; | |
1291 | } | |
1988b51e AT |
1292 | while (--depth >= 0) { |
1293 | ix = path[depth].p_idx; | |
273df556 FM |
1294 | if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) { |
1295 | EXT4_ERROR_INODE(inode, | |
1296 | "ix != EXT_FIRST_INDEX *logical %d!", | |
1297 | *logical); | |
1298 | return -EIO; | |
1299 | } | |
1988b51e | 1300 | } |
4d33b1ef | 1301 | goto found_extent; |
1988b51e AT |
1302 | } |
1303 | ||
273df556 FM |
1304 | if (unlikely(*logical < (le32_to_cpu(ex->ee_block) + ee_len))) { |
1305 | EXT4_ERROR_INODE(inode, | |
1306 | "logical %d < ee_block %d + ee_len %d!", | |
1307 | *logical, le32_to_cpu(ex->ee_block), ee_len); | |
1308 | return -EIO; | |
1309 | } | |
1988b51e AT |
1310 | |
1311 | if (ex != EXT_LAST_EXTENT(path[depth].p_hdr)) { | |
1312 | /* next allocated block in this leaf */ | |
1313 | ex++; | |
4d33b1ef | 1314 | goto found_extent; |
1988b51e AT |
1315 | } |
1316 | ||
1317 | /* go up and search for index to the right */ | |
1318 | while (--depth >= 0) { | |
1319 | ix = path[depth].p_idx; | |
1320 | if (ix != EXT_LAST_INDEX(path[depth].p_hdr)) | |
25f1ee3a | 1321 | goto got_index; |
1988b51e AT |
1322 | } |
1323 | ||
25f1ee3a WF |
1324 | /* we've gone up to the root and found no index to the right */ |
1325 | return 0; | |
1988b51e | 1326 | |
25f1ee3a | 1327 | got_index: |
1988b51e AT |
1328 | /* we've found index to the right, let's |
1329 | * follow it and find the closest allocated | |
1330 | * block to the right */ | |
1331 | ix++; | |
bf89d16f | 1332 | block = ext4_idx_pblock(ix); |
1988b51e AT |
1333 | while (++depth < path->p_depth) { |
1334 | bh = sb_bread(inode->i_sb, block); | |
1335 | if (bh == NULL) | |
1336 | return -EIO; | |
1337 | eh = ext_block_hdr(bh); | |
395a87bf | 1338 | /* subtract from p_depth to get proper eh_depth */ |
56b19868 | 1339 | if (ext4_ext_check(inode, eh, path->p_depth - depth)) { |
1988b51e AT |
1340 | put_bh(bh); |
1341 | return -EIO; | |
1342 | } | |
1343 | ix = EXT_FIRST_INDEX(eh); | |
bf89d16f | 1344 | block = ext4_idx_pblock(ix); |
1988b51e AT |
1345 | put_bh(bh); |
1346 | } | |
1347 | ||
1348 | bh = sb_bread(inode->i_sb, block); | |
1349 | if (bh == NULL) | |
1350 | return -EIO; | |
1351 | eh = ext_block_hdr(bh); | |
56b19868 | 1352 | if (ext4_ext_check(inode, eh, path->p_depth - depth)) { |
1988b51e AT |
1353 | put_bh(bh); |
1354 | return -EIO; | |
1355 | } | |
1356 | ex = EXT_FIRST_EXTENT(eh); | |
4d33b1ef | 1357 | found_extent: |
1988b51e | 1358 | *logical = le32_to_cpu(ex->ee_block); |
bf89d16f | 1359 | *phys = ext4_ext_pblock(ex); |
4d33b1ef TT |
1360 | *ret_ex = ex; |
1361 | if (bh) | |
1362 | put_bh(bh); | |
1988b51e | 1363 | return 0; |
1988b51e AT |
1364 | } |
1365 | ||
a86c6181 | 1366 | /* |
d0d856e8 | 1367 | * ext4_ext_next_allocated_block: |
f17722f9 | 1368 | * returns allocated block in subsequent extent or EXT_MAX_BLOCKS. |
d0d856e8 RD |
1369 | * NOTE: it considers block number from index entry as |
1370 | * allocated block. Thus, index entries have to be consistent | |
1371 | * with leaves. | |
a86c6181 | 1372 | */ |
725d26d3 | 1373 | static ext4_lblk_t |
a86c6181 AT |
1374 | ext4_ext_next_allocated_block(struct ext4_ext_path *path) |
1375 | { | |
1376 | int depth; | |
1377 | ||
1378 | BUG_ON(path == NULL); | |
1379 | depth = path->p_depth; | |
1380 | ||
1381 | if (depth == 0 && path->p_ext == NULL) | |
f17722f9 | 1382 | return EXT_MAX_BLOCKS; |
a86c6181 AT |
1383 | |
1384 | while (depth >= 0) { | |
1385 | if (depth == path->p_depth) { | |
1386 | /* leaf */ | |
6f8ff537 CW |
1387 | if (path[depth].p_ext && |
1388 | path[depth].p_ext != | |
a86c6181 AT |
1389 | EXT_LAST_EXTENT(path[depth].p_hdr)) |
1390 | return le32_to_cpu(path[depth].p_ext[1].ee_block); | |
1391 | } else { | |
1392 | /* index */ | |
1393 | if (path[depth].p_idx != | |
1394 | EXT_LAST_INDEX(path[depth].p_hdr)) | |
1395 | return le32_to_cpu(path[depth].p_idx[1].ei_block); | |
1396 | } | |
1397 | depth--; | |
1398 | } | |
1399 | ||
f17722f9 | 1400 | return EXT_MAX_BLOCKS; |
a86c6181 AT |
1401 | } |
1402 | ||
1403 | /* | |
d0d856e8 | 1404 | * ext4_ext_next_leaf_block: |
f17722f9 | 1405 | * returns first allocated block from next leaf or EXT_MAX_BLOCKS |
a86c6181 | 1406 | */ |
5718789d | 1407 | static ext4_lblk_t ext4_ext_next_leaf_block(struct ext4_ext_path *path) |
a86c6181 AT |
1408 | { |
1409 | int depth; | |
1410 | ||
1411 | BUG_ON(path == NULL); | |
1412 | depth = path->p_depth; | |
1413 | ||
1414 | /* zero-tree has no leaf blocks at all */ | |
1415 | if (depth == 0) | |
f17722f9 | 1416 | return EXT_MAX_BLOCKS; |
a86c6181 AT |
1417 | |
1418 | /* go to index block */ | |
1419 | depth--; | |
1420 | ||
1421 | while (depth >= 0) { | |
1422 | if (path[depth].p_idx != | |
1423 | EXT_LAST_INDEX(path[depth].p_hdr)) | |
725d26d3 AK |
1424 | return (ext4_lblk_t) |
1425 | le32_to_cpu(path[depth].p_idx[1].ei_block); | |
a86c6181 AT |
1426 | depth--; |
1427 | } | |
1428 | ||
f17722f9 | 1429 | return EXT_MAX_BLOCKS; |
a86c6181 AT |
1430 | } |
1431 | ||
1432 | /* | |
d0d856e8 RD |
1433 | * ext4_ext_correct_indexes: |
1434 | * if leaf gets modified and modified extent is first in the leaf, | |
1435 | * then we have to correct all indexes above. | |
a86c6181 AT |
1436 | * TODO: do we need to correct tree in all cases? |
1437 | */ | |
1d03ec98 | 1438 | static int ext4_ext_correct_indexes(handle_t *handle, struct inode *inode, |
a86c6181 AT |
1439 | struct ext4_ext_path *path) |
1440 | { | |
1441 | struct ext4_extent_header *eh; | |
1442 | int depth = ext_depth(inode); | |
1443 | struct ext4_extent *ex; | |
1444 | __le32 border; | |
1445 | int k, err = 0; | |
1446 | ||
1447 | eh = path[depth].p_hdr; | |
1448 | ex = path[depth].p_ext; | |
273df556 FM |
1449 | |
1450 | if (unlikely(ex == NULL || eh == NULL)) { | |
1451 | EXT4_ERROR_INODE(inode, | |
1452 | "ex %p == NULL or eh %p == NULL", ex, eh); | |
1453 | return -EIO; | |
1454 | } | |
a86c6181 AT |
1455 | |
1456 | if (depth == 0) { | |
1457 | /* there is no tree at all */ | |
1458 | return 0; | |
1459 | } | |
1460 | ||
1461 | if (ex != EXT_FIRST_EXTENT(eh)) { | |
1462 | /* we correct tree if first leaf got modified only */ | |
1463 | return 0; | |
1464 | } | |
1465 | ||
1466 | /* | |
d0d856e8 | 1467 | * TODO: we need correction if border is smaller than current one |
a86c6181 AT |
1468 | */ |
1469 | k = depth - 1; | |
1470 | border = path[depth].p_ext->ee_block; | |
7e028976 AM |
1471 | err = ext4_ext_get_access(handle, inode, path + k); |
1472 | if (err) | |
a86c6181 AT |
1473 | return err; |
1474 | path[k].p_idx->ei_block = border; | |
7e028976 AM |
1475 | err = ext4_ext_dirty(handle, inode, path + k); |
1476 | if (err) | |
a86c6181 AT |
1477 | return err; |
1478 | ||
1479 | while (k--) { | |
1480 | /* change all left-side indexes */ | |
1481 | if (path[k+1].p_idx != EXT_FIRST_INDEX(path[k+1].p_hdr)) | |
1482 | break; | |
7e028976 AM |
1483 | err = ext4_ext_get_access(handle, inode, path + k); |
1484 | if (err) | |
a86c6181 AT |
1485 | break; |
1486 | path[k].p_idx->ei_block = border; | |
7e028976 AM |
1487 | err = ext4_ext_dirty(handle, inode, path + k); |
1488 | if (err) | |
a86c6181 AT |
1489 | break; |
1490 | } | |
1491 | ||
1492 | return err; | |
1493 | } | |
1494 | ||
748de673 | 1495 | int |
a86c6181 AT |
1496 | ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1, |
1497 | struct ext4_extent *ex2) | |
1498 | { | |
749269fa | 1499 | unsigned short ext1_ee_len, ext2_ee_len, max_len; |
a2df2a63 AA |
1500 | |
1501 | /* | |
1502 | * Make sure that either both extents are uninitialized, or | |
1503 | * both are _not_. | |
1504 | */ | |
1505 | if (ext4_ext_is_uninitialized(ex1) ^ ext4_ext_is_uninitialized(ex2)) | |
1506 | return 0; | |
1507 | ||
749269fa AA |
1508 | if (ext4_ext_is_uninitialized(ex1)) |
1509 | max_len = EXT_UNINIT_MAX_LEN; | |
1510 | else | |
1511 | max_len = EXT_INIT_MAX_LEN; | |
1512 | ||
a2df2a63 AA |
1513 | ext1_ee_len = ext4_ext_get_actual_len(ex1); |
1514 | ext2_ee_len = ext4_ext_get_actual_len(ex2); | |
1515 | ||
1516 | if (le32_to_cpu(ex1->ee_block) + ext1_ee_len != | |
63f57933 | 1517 | le32_to_cpu(ex2->ee_block)) |
a86c6181 AT |
1518 | return 0; |
1519 | ||
471d4011 SB |
1520 | /* |
1521 | * To allow future support for preallocated extents to be added | |
1522 | * as an RO_COMPAT feature, refuse to merge to extents if | |
d0d856e8 | 1523 | * this can result in the top bit of ee_len being set. |
471d4011 | 1524 | */ |
749269fa | 1525 | if (ext1_ee_len + ext2_ee_len > max_len) |
471d4011 | 1526 | return 0; |
bbf2f9fb | 1527 | #ifdef AGGRESSIVE_TEST |
b939e376 | 1528 | if (ext1_ee_len >= 4) |
a86c6181 AT |
1529 | return 0; |
1530 | #endif | |
1531 | ||
bf89d16f | 1532 | if (ext4_ext_pblock(ex1) + ext1_ee_len == ext4_ext_pblock(ex2)) |
a86c6181 AT |
1533 | return 1; |
1534 | return 0; | |
1535 | } | |
1536 | ||
56055d3a AA |
1537 | /* |
1538 | * This function tries to merge the "ex" extent to the next extent in the tree. | |
1539 | * It always tries to merge towards right. If you want to merge towards | |
1540 | * left, pass "ex - 1" as argument instead of "ex". | |
1541 | * Returns 0 if the extents (ex and ex+1) were _not_ merged and returns | |
1542 | * 1 if they got merged. | |
1543 | */ | |
197217a5 | 1544 | static int ext4_ext_try_to_merge_right(struct inode *inode, |
1f109d5a TT |
1545 | struct ext4_ext_path *path, |
1546 | struct ext4_extent *ex) | |
56055d3a AA |
1547 | { |
1548 | struct ext4_extent_header *eh; | |
1549 | unsigned int depth, len; | |
1550 | int merge_done = 0; | |
1551 | int uninitialized = 0; | |
1552 | ||
1553 | depth = ext_depth(inode); | |
1554 | BUG_ON(path[depth].p_hdr == NULL); | |
1555 | eh = path[depth].p_hdr; | |
1556 | ||
1557 | while (ex < EXT_LAST_EXTENT(eh)) { | |
1558 | if (!ext4_can_extents_be_merged(inode, ex, ex + 1)) | |
1559 | break; | |
1560 | /* merge with next extent! */ | |
1561 | if (ext4_ext_is_uninitialized(ex)) | |
1562 | uninitialized = 1; | |
1563 | ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex) | |
1564 | + ext4_ext_get_actual_len(ex + 1)); | |
1565 | if (uninitialized) | |
1566 | ext4_ext_mark_uninitialized(ex); | |
1567 | ||
1568 | if (ex + 1 < EXT_LAST_EXTENT(eh)) { | |
1569 | len = (EXT_LAST_EXTENT(eh) - ex - 1) | |
1570 | * sizeof(struct ext4_extent); | |
1571 | memmove(ex + 1, ex + 2, len); | |
1572 | } | |
e8546d06 | 1573 | le16_add_cpu(&eh->eh_entries, -1); |
56055d3a AA |
1574 | merge_done = 1; |
1575 | WARN_ON(eh->eh_entries == 0); | |
1576 | if (!eh->eh_entries) | |
24676da4 | 1577 | EXT4_ERROR_INODE(inode, "eh->eh_entries = 0!"); |
56055d3a AA |
1578 | } |
1579 | ||
1580 | return merge_done; | |
1581 | } | |
1582 | ||
197217a5 YY |
1583 | /* |
1584 | * This function tries to merge the @ex extent to neighbours in the tree. | |
1585 | * return 1 if merge left else 0. | |
1586 | */ | |
1587 | static int ext4_ext_try_to_merge(struct inode *inode, | |
1588 | struct ext4_ext_path *path, | |
1589 | struct ext4_extent *ex) { | |
1590 | struct ext4_extent_header *eh; | |
1591 | unsigned int depth; | |
1592 | int merge_done = 0; | |
1593 | int ret = 0; | |
1594 | ||
1595 | depth = ext_depth(inode); | |
1596 | BUG_ON(path[depth].p_hdr == NULL); | |
1597 | eh = path[depth].p_hdr; | |
1598 | ||
1599 | if (ex > EXT_FIRST_EXTENT(eh)) | |
1600 | merge_done = ext4_ext_try_to_merge_right(inode, path, ex - 1); | |
1601 | ||
1602 | if (!merge_done) | |
1603 | ret = ext4_ext_try_to_merge_right(inode, path, ex); | |
1604 | ||
1605 | return ret; | |
1606 | } | |
1607 | ||
25d14f98 AA |
1608 | /* |
1609 | * check if a portion of the "newext" extent overlaps with an | |
1610 | * existing extent. | |
1611 | * | |
1612 | * If there is an overlap discovered, it updates the length of the newext | |
1613 | * such that there will be no overlap, and then returns 1. | |
1614 | * If there is no overlap found, it returns 0. | |
1615 | */ | |
4d33b1ef TT |
1616 | static unsigned int ext4_ext_check_overlap(struct ext4_sb_info *sbi, |
1617 | struct inode *inode, | |
1f109d5a TT |
1618 | struct ext4_extent *newext, |
1619 | struct ext4_ext_path *path) | |
25d14f98 | 1620 | { |
725d26d3 | 1621 | ext4_lblk_t b1, b2; |
25d14f98 AA |
1622 | unsigned int depth, len1; |
1623 | unsigned int ret = 0; | |
1624 | ||
1625 | b1 = le32_to_cpu(newext->ee_block); | |
a2df2a63 | 1626 | len1 = ext4_ext_get_actual_len(newext); |
25d14f98 AA |
1627 | depth = ext_depth(inode); |
1628 | if (!path[depth].p_ext) | |
1629 | goto out; | |
1630 | b2 = le32_to_cpu(path[depth].p_ext->ee_block); | |
4d33b1ef | 1631 | b2 &= ~(sbi->s_cluster_ratio - 1); |
25d14f98 AA |
1632 | |
1633 | /* | |
1634 | * get the next allocated block if the extent in the path | |
2b2d6d01 | 1635 | * is before the requested block(s) |
25d14f98 AA |
1636 | */ |
1637 | if (b2 < b1) { | |
1638 | b2 = ext4_ext_next_allocated_block(path); | |
f17722f9 | 1639 | if (b2 == EXT_MAX_BLOCKS) |
25d14f98 | 1640 | goto out; |
4d33b1ef | 1641 | b2 &= ~(sbi->s_cluster_ratio - 1); |
25d14f98 AA |
1642 | } |
1643 | ||
725d26d3 | 1644 | /* check for wrap through zero on extent logical start block*/ |
25d14f98 | 1645 | if (b1 + len1 < b1) { |
f17722f9 | 1646 | len1 = EXT_MAX_BLOCKS - b1; |
25d14f98 AA |
1647 | newext->ee_len = cpu_to_le16(len1); |
1648 | ret = 1; | |
1649 | } | |
1650 | ||
1651 | /* check for overlap */ | |
1652 | if (b1 + len1 > b2) { | |
1653 | newext->ee_len = cpu_to_le16(b2 - b1); | |
1654 | ret = 1; | |
1655 | } | |
1656 | out: | |
1657 | return ret; | |
1658 | } | |
1659 | ||
a86c6181 | 1660 | /* |
d0d856e8 RD |
1661 | * ext4_ext_insert_extent: |
1662 | * tries to merge requsted extent into the existing extent or | |
1663 | * inserts requested extent as new one into the tree, | |
1664 | * creating new leaf in the no-space case. | |
a86c6181 AT |
1665 | */ |
1666 | int ext4_ext_insert_extent(handle_t *handle, struct inode *inode, | |
1667 | struct ext4_ext_path *path, | |
0031462b | 1668 | struct ext4_extent *newext, int flag) |
a86c6181 | 1669 | { |
af5bc92d | 1670 | struct ext4_extent_header *eh; |
a86c6181 AT |
1671 | struct ext4_extent *ex, *fex; |
1672 | struct ext4_extent *nearex; /* nearest extent */ | |
1673 | struct ext4_ext_path *npath = NULL; | |
725d26d3 AK |
1674 | int depth, len, err; |
1675 | ext4_lblk_t next; | |
a2df2a63 | 1676 | unsigned uninitialized = 0; |
55f020db | 1677 | int flags = 0; |
a86c6181 | 1678 | |
273df556 FM |
1679 | if (unlikely(ext4_ext_get_actual_len(newext) == 0)) { |
1680 | EXT4_ERROR_INODE(inode, "ext4_ext_get_actual_len(newext) == 0"); | |
1681 | return -EIO; | |
1682 | } | |
a86c6181 AT |
1683 | depth = ext_depth(inode); |
1684 | ex = path[depth].p_ext; | |
273df556 FM |
1685 | if (unlikely(path[depth].p_hdr == NULL)) { |
1686 | EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth); | |
1687 | return -EIO; | |
1688 | } | |
a86c6181 AT |
1689 | |
1690 | /* try to insert block into found extent and return */ | |
744692dc | 1691 | if (ex && !(flag & EXT4_GET_BLOCKS_PRE_IO) |
0031462b | 1692 | && ext4_can_extents_be_merged(inode, ex, newext)) { |
553f9008 | 1693 | ext_debug("append [%d]%d block to %d:[%d]%d (from %llu)\n", |
bf89d16f TT |
1694 | ext4_ext_is_uninitialized(newext), |
1695 | ext4_ext_get_actual_len(newext), | |
1696 | le32_to_cpu(ex->ee_block), | |
1697 | ext4_ext_is_uninitialized(ex), | |
1698 | ext4_ext_get_actual_len(ex), | |
1699 | ext4_ext_pblock(ex)); | |
7e028976 AM |
1700 | err = ext4_ext_get_access(handle, inode, path + depth); |
1701 | if (err) | |
a86c6181 | 1702 | return err; |
a2df2a63 AA |
1703 | |
1704 | /* | |
1705 | * ext4_can_extents_be_merged should have checked that either | |
1706 | * both extents are uninitialized, or both aren't. Thus we | |
1707 | * need to check only one of them here. | |
1708 | */ | |
1709 | if (ext4_ext_is_uninitialized(ex)) | |
1710 | uninitialized = 1; | |
1711 | ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex) | |
1712 | + ext4_ext_get_actual_len(newext)); | |
1713 | if (uninitialized) | |
1714 | ext4_ext_mark_uninitialized(ex); | |
a86c6181 AT |
1715 | eh = path[depth].p_hdr; |
1716 | nearex = ex; | |
1717 | goto merge; | |
1718 | } | |
1719 | ||
a86c6181 AT |
1720 | depth = ext_depth(inode); |
1721 | eh = path[depth].p_hdr; | |
1722 | if (le16_to_cpu(eh->eh_entries) < le16_to_cpu(eh->eh_max)) | |
1723 | goto has_space; | |
1724 | ||
1725 | /* probably next leaf has space for us? */ | |
1726 | fex = EXT_LAST_EXTENT(eh); | |
598dbdf2 RD |
1727 | next = EXT_MAX_BLOCKS; |
1728 | if (le32_to_cpu(newext->ee_block) > le32_to_cpu(fex->ee_block)) | |
5718789d | 1729 | next = ext4_ext_next_leaf_block(path); |
598dbdf2 | 1730 | if (next != EXT_MAX_BLOCKS) { |
a86c6181 AT |
1731 | ext_debug("next leaf block - %d\n", next); |
1732 | BUG_ON(npath != NULL); | |
1733 | npath = ext4_ext_find_extent(inode, next, NULL); | |
1734 | if (IS_ERR(npath)) | |
1735 | return PTR_ERR(npath); | |
1736 | BUG_ON(npath->p_depth != path->p_depth); | |
1737 | eh = npath[depth].p_hdr; | |
1738 | if (le16_to_cpu(eh->eh_entries) < le16_to_cpu(eh->eh_max)) { | |
25985edc | 1739 | ext_debug("next leaf isn't full(%d)\n", |
a86c6181 AT |
1740 | le16_to_cpu(eh->eh_entries)); |
1741 | path = npath; | |
ffb505ff | 1742 | goto has_space; |
a86c6181 AT |
1743 | } |
1744 | ext_debug("next leaf has no free space(%d,%d)\n", | |
1745 | le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max)); | |
1746 | } | |
1747 | ||
1748 | /* | |
d0d856e8 RD |
1749 | * There is no free space in the found leaf. |
1750 | * We're gonna add a new leaf in the tree. | |
a86c6181 | 1751 | */ |
55f020db AH |
1752 | if (flag & EXT4_GET_BLOCKS_PUNCH_OUT_EXT) |
1753 | flags = EXT4_MB_USE_ROOT_BLOCKS; | |
1754 | err = ext4_ext_create_new_leaf(handle, inode, flags, path, newext); | |
a86c6181 AT |
1755 | if (err) |
1756 | goto cleanup; | |
1757 | depth = ext_depth(inode); | |
1758 | eh = path[depth].p_hdr; | |
1759 | ||
1760 | has_space: | |
1761 | nearex = path[depth].p_ext; | |
1762 | ||
7e028976 AM |
1763 | err = ext4_ext_get_access(handle, inode, path + depth); |
1764 | if (err) | |
a86c6181 AT |
1765 | goto cleanup; |
1766 | ||
1767 | if (!nearex) { | |
1768 | /* there is no extent in this leaf, create first one */ | |
553f9008 | 1769 | ext_debug("first extent in the leaf: %d:%llu:[%d]%d\n", |
8c55e204 | 1770 | le32_to_cpu(newext->ee_block), |
bf89d16f | 1771 | ext4_ext_pblock(newext), |
553f9008 | 1772 | ext4_ext_is_uninitialized(newext), |
a2df2a63 | 1773 | ext4_ext_get_actual_len(newext)); |
80e675f9 EG |
1774 | nearex = EXT_FIRST_EXTENT(eh); |
1775 | } else { | |
1776 | if (le32_to_cpu(newext->ee_block) | |
8c55e204 | 1777 | > le32_to_cpu(nearex->ee_block)) { |
80e675f9 EG |
1778 | /* Insert after */ |
1779 | ext_debug("insert %d:%llu:[%d]%d %s before: " | |
1780 | "nearest 0x%p\n" | |
1781 | le32_to_cpu(newext->ee_block), | |
1782 | ext4_ext_pblock(newext), | |
1783 | ext4_ext_is_uninitialized(newext), | |
1784 | ext4_ext_get_actual_len(newext), | |
1785 | nearex); | |
1786 | nearex++; | |
1787 | } else { | |
1788 | /* Insert before */ | |
1789 | BUG_ON(newext->ee_block == nearex->ee_block); | |
1790 | ext_debug("insert %d:%llu:[%d]%d %s after: " | |
1791 | "nearest 0x%p\n" | |
8c55e204 | 1792 | le32_to_cpu(newext->ee_block), |
bf89d16f | 1793 | ext4_ext_pblock(newext), |
553f9008 | 1794 | ext4_ext_is_uninitialized(newext), |
a2df2a63 | 1795 | ext4_ext_get_actual_len(newext), |
80e675f9 EG |
1796 | nearex); |
1797 | } | |
1798 | len = EXT_LAST_EXTENT(eh) - nearex + 1; | |
1799 | if (len > 0) { | |
1800 | ext_debug("insert %d:%llu:[%d]%d: " | |
1801 | "move %d extents from 0x%p to 0x%p\n", | |
1802 | le32_to_cpu(newext->ee_block), | |
1803 | ext4_ext_pblock(newext), | |
1804 | ext4_ext_is_uninitialized(newext), | |
1805 | ext4_ext_get_actual_len(newext), | |
1806 | len, nearex, nearex + 1); | |
1807 | memmove(nearex + 1, nearex, | |
1808 | len * sizeof(struct ext4_extent)); | |
a86c6181 | 1809 | } |
a86c6181 AT |
1810 | } |
1811 | ||
e8546d06 | 1812 | le16_add_cpu(&eh->eh_entries, 1); |
80e675f9 | 1813 | path[depth].p_ext = nearex; |
a86c6181 | 1814 | nearex->ee_block = newext->ee_block; |
bf89d16f | 1815 | ext4_ext_store_pblock(nearex, ext4_ext_pblock(newext)); |
a86c6181 | 1816 | nearex->ee_len = newext->ee_len; |
a86c6181 AT |
1817 | |
1818 | merge: | |
1819 | /* try to merge extents to the right */ | |
744692dc | 1820 | if (!(flag & EXT4_GET_BLOCKS_PRE_IO)) |
0031462b | 1821 | ext4_ext_try_to_merge(inode, path, nearex); |
a86c6181 AT |
1822 | |
1823 | /* try to merge extents to the left */ | |
1824 | ||
1825 | /* time to correct all indexes above */ | |
1826 | err = ext4_ext_correct_indexes(handle, inode, path); | |
1827 | if (err) | |
1828 | goto cleanup; | |
1829 | ||
1830 | err = ext4_ext_dirty(handle, inode, path + depth); | |
1831 | ||
1832 | cleanup: | |
1833 | if (npath) { | |
1834 | ext4_ext_drop_refs(npath); | |
1835 | kfree(npath); | |
1836 | } | |
a86c6181 AT |
1837 | ext4_ext_invalidate_cache(inode); |
1838 | return err; | |
1839 | } | |
1840 | ||
1f109d5a TT |
1841 | static int ext4_ext_walk_space(struct inode *inode, ext4_lblk_t block, |
1842 | ext4_lblk_t num, ext_prepare_callback func, | |
1843 | void *cbdata) | |
6873fa0d ES |
1844 | { |
1845 | struct ext4_ext_path *path = NULL; | |
1846 | struct ext4_ext_cache cbex; | |
1847 | struct ext4_extent *ex; | |
1848 | ext4_lblk_t next, start = 0, end = 0; | |
1849 | ext4_lblk_t last = block + num; | |
1850 | int depth, exists, err = 0; | |
1851 | ||
1852 | BUG_ON(func == NULL); | |
1853 | BUG_ON(inode == NULL); | |
1854 | ||
f17722f9 | 1855 | while (block < last && block != EXT_MAX_BLOCKS) { |
6873fa0d ES |
1856 | num = last - block; |
1857 | /* find extent for this block */ | |
fab3a549 | 1858 | down_read(&EXT4_I(inode)->i_data_sem); |
6873fa0d | 1859 | path = ext4_ext_find_extent(inode, block, path); |
fab3a549 | 1860 | up_read(&EXT4_I(inode)->i_data_sem); |
6873fa0d ES |
1861 | if (IS_ERR(path)) { |
1862 | err = PTR_ERR(path); | |
1863 | path = NULL; | |
1864 | break; | |
1865 | } | |
1866 | ||
1867 | depth = ext_depth(inode); | |
273df556 FM |
1868 | if (unlikely(path[depth].p_hdr == NULL)) { |
1869 | EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth); | |
1870 | err = -EIO; | |
1871 | break; | |
1872 | } | |
6873fa0d ES |
1873 | ex = path[depth].p_ext; |
1874 | next = ext4_ext_next_allocated_block(path); | |
1875 | ||
1876 | exists = 0; | |
1877 | if (!ex) { | |
1878 | /* there is no extent yet, so try to allocate | |
1879 | * all requested space */ | |
1880 | start = block; | |
1881 | end = block + num; | |
1882 | } else if (le32_to_cpu(ex->ee_block) > block) { | |
1883 | /* need to allocate space before found extent */ | |
1884 | start = block; | |
1885 | end = le32_to_cpu(ex->ee_block); | |
1886 | if (block + num < end) | |
1887 | end = block + num; | |
1888 | } else if (block >= le32_to_cpu(ex->ee_block) | |
1889 | + ext4_ext_get_actual_len(ex)) { | |
1890 | /* need to allocate space after found extent */ | |
1891 | start = block; | |
1892 | end = block + num; | |
1893 | if (end >= next) | |
1894 | end = next; | |
1895 | } else if (block >= le32_to_cpu(ex->ee_block)) { | |
1896 | /* | |
1897 | * some part of requested space is covered | |
1898 | * by found extent | |
1899 | */ | |
1900 | start = block; | |
1901 | end = le32_to_cpu(ex->ee_block) | |
1902 | + ext4_ext_get_actual_len(ex); | |
1903 | if (block + num < end) | |
1904 | end = block + num; | |
1905 | exists = 1; | |
1906 | } else { | |
1907 | BUG(); | |
1908 | } | |
1909 | BUG_ON(end <= start); | |
1910 | ||
1911 | if (!exists) { | |
1912 | cbex.ec_block = start; | |
1913 | cbex.ec_len = end - start; | |
1914 | cbex.ec_start = 0; | |
6873fa0d ES |
1915 | } else { |
1916 | cbex.ec_block = le32_to_cpu(ex->ee_block); | |
1917 | cbex.ec_len = ext4_ext_get_actual_len(ex); | |
bf89d16f | 1918 | cbex.ec_start = ext4_ext_pblock(ex); |
6873fa0d ES |
1919 | } |
1920 | ||
273df556 FM |
1921 | if (unlikely(cbex.ec_len == 0)) { |
1922 | EXT4_ERROR_INODE(inode, "cbex.ec_len == 0"); | |
1923 | err = -EIO; | |
1924 | break; | |
1925 | } | |
c03f8aa9 | 1926 | err = func(inode, next, &cbex, ex, cbdata); |
6873fa0d ES |
1927 | ext4_ext_drop_refs(path); |
1928 | ||
1929 | if (err < 0) | |
1930 | break; | |
1931 | ||
1932 | if (err == EXT_REPEAT) | |
1933 | continue; | |
1934 | else if (err == EXT_BREAK) { | |
1935 | err = 0; | |
1936 | break; | |
1937 | } | |
1938 | ||
1939 | if (ext_depth(inode) != depth) { | |
1940 | /* depth was changed. we have to realloc path */ | |
1941 | kfree(path); | |
1942 | path = NULL; | |
1943 | } | |
1944 | ||
1945 | block = cbex.ec_block + cbex.ec_len; | |
1946 | } | |
1947 | ||
1948 | if (path) { | |
1949 | ext4_ext_drop_refs(path); | |
1950 | kfree(path); | |
1951 | } | |
1952 | ||
1953 | return err; | |
1954 | } | |
1955 | ||
09b88252 | 1956 | static void |
725d26d3 | 1957 | ext4_ext_put_in_cache(struct inode *inode, ext4_lblk_t block, |
b05e6ae5 | 1958 | __u32 len, ext4_fsblk_t start) |
a86c6181 AT |
1959 | { |
1960 | struct ext4_ext_cache *cex; | |
1961 | BUG_ON(len == 0); | |
2ec0ae3a | 1962 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); |
d8990240 | 1963 | trace_ext4_ext_put_in_cache(inode, block, len, start); |
a86c6181 | 1964 | cex = &EXT4_I(inode)->i_cached_extent; |
a86c6181 AT |
1965 | cex->ec_block = block; |
1966 | cex->ec_len = len; | |
1967 | cex->ec_start = start; | |
2ec0ae3a | 1968 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); |
a86c6181 AT |
1969 | } |
1970 | ||
1971 | /* | |
d0d856e8 RD |
1972 | * ext4_ext_put_gap_in_cache: |
1973 | * calculate boundaries of the gap that the requested block fits into | |
a86c6181 AT |
1974 | * and cache this gap |
1975 | */ | |
09b88252 | 1976 | static void |
a86c6181 | 1977 | ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path, |
725d26d3 | 1978 | ext4_lblk_t block) |
a86c6181 AT |
1979 | { |
1980 | int depth = ext_depth(inode); | |
725d26d3 AK |
1981 | unsigned long len; |
1982 | ext4_lblk_t lblock; | |
a86c6181 AT |
1983 | struct ext4_extent *ex; |
1984 | ||
1985 | ex = path[depth].p_ext; | |
1986 | if (ex == NULL) { | |
1987 | /* there is no extent yet, so gap is [0;-] */ | |
1988 | lblock = 0; | |
f17722f9 | 1989 | len = EXT_MAX_BLOCKS; |
a86c6181 AT |
1990 | ext_debug("cache gap(whole file):"); |
1991 | } else if (block < le32_to_cpu(ex->ee_block)) { | |
1992 | lblock = block; | |
1993 | len = le32_to_cpu(ex->ee_block) - block; | |
bba90743 ES |
1994 | ext_debug("cache gap(before): %u [%u:%u]", |
1995 | block, | |
1996 | le32_to_cpu(ex->ee_block), | |
1997 | ext4_ext_get_actual_len(ex)); | |
a86c6181 | 1998 | } else if (block >= le32_to_cpu(ex->ee_block) |
a2df2a63 | 1999 | + ext4_ext_get_actual_len(ex)) { |
725d26d3 | 2000 | ext4_lblk_t next; |
8c55e204 | 2001 | lblock = le32_to_cpu(ex->ee_block) |
a2df2a63 | 2002 | + ext4_ext_get_actual_len(ex); |
725d26d3 AK |
2003 | |
2004 | next = ext4_ext_next_allocated_block(path); | |
bba90743 ES |
2005 | ext_debug("cache gap(after): [%u:%u] %u", |
2006 | le32_to_cpu(ex->ee_block), | |
2007 | ext4_ext_get_actual_len(ex), | |
2008 | block); | |
725d26d3 AK |
2009 | BUG_ON(next == lblock); |
2010 | len = next - lblock; | |
a86c6181 AT |
2011 | } else { |
2012 | lblock = len = 0; | |
2013 | BUG(); | |
2014 | } | |
2015 | ||
bba90743 | 2016 | ext_debug(" -> %u:%lu\n", lblock, len); |
b05e6ae5 | 2017 | ext4_ext_put_in_cache(inode, lblock, len, 0); |
a86c6181 AT |
2018 | } |
2019 | ||
b05e6ae5 | 2020 | /* |
b7ca1e8e | 2021 | * ext4_ext_check_cache() |
a4bb6b64 AH |
2022 | * Checks to see if the given block is in the cache. |
2023 | * If it is, the cached extent is stored in the given | |
2024 | * cache extent pointer. If the cached extent is a hole, | |
2025 | * this routine should be used instead of | |
2026 | * ext4_ext_in_cache if the calling function needs to | |
2027 | * know the size of the hole. | |
2028 | * | |
2029 | * @inode: The files inode | |
2030 | * @block: The block to look for in the cache | |
2031 | * @ex: Pointer where the cached extent will be stored | |
2032 | * if it contains block | |
2033 | * | |
b05e6ae5 TT |
2034 | * Return 0 if cache is invalid; 1 if the cache is valid |
2035 | */ | |
a4bb6b64 AH |
2036 | static int ext4_ext_check_cache(struct inode *inode, ext4_lblk_t block, |
2037 | struct ext4_ext_cache *ex){ | |
a86c6181 | 2038 | struct ext4_ext_cache *cex; |
77f4135f | 2039 | struct ext4_sb_info *sbi; |
b05e6ae5 | 2040 | int ret = 0; |
a86c6181 | 2041 | |
60e6679e | 2042 | /* |
2ec0ae3a TT |
2043 | * We borrow i_block_reservation_lock to protect i_cached_extent |
2044 | */ | |
2045 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); | |
a86c6181 | 2046 | cex = &EXT4_I(inode)->i_cached_extent; |
77f4135f | 2047 | sbi = EXT4_SB(inode->i_sb); |
a86c6181 AT |
2048 | |
2049 | /* has cache valid data? */ | |
b05e6ae5 | 2050 | if (cex->ec_len == 0) |
2ec0ae3a | 2051 | goto errout; |
a86c6181 | 2052 | |
731eb1a0 | 2053 | if (in_range(block, cex->ec_block, cex->ec_len)) { |
a4bb6b64 | 2054 | memcpy(ex, cex, sizeof(struct ext4_ext_cache)); |
bba90743 ES |
2055 | ext_debug("%u cached by %u:%u:%llu\n", |
2056 | block, | |
2057 | cex->ec_block, cex->ec_len, cex->ec_start); | |
b05e6ae5 | 2058 | ret = 1; |
a86c6181 | 2059 | } |
2ec0ae3a | 2060 | errout: |
77f4135f VH |
2061 | if (!ret) |
2062 | sbi->extent_cache_misses++; | |
2063 | else | |
2064 | sbi->extent_cache_hits++; | |
d8990240 | 2065 | trace_ext4_ext_in_cache(inode, block, ret); |
2ec0ae3a TT |
2066 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); |
2067 | return ret; | |
a86c6181 AT |
2068 | } |
2069 | ||
a4bb6b64 AH |
2070 | /* |
2071 | * ext4_ext_in_cache() | |
2072 | * Checks to see if the given block is in the cache. | |
2073 | * If it is, the cached extent is stored in the given | |
2074 | * extent pointer. | |
2075 | * | |
2076 | * @inode: The files inode | |
2077 | * @block: The block to look for in the cache | |
2078 | * @ex: Pointer where the cached extent will be stored | |
2079 | * if it contains block | |
2080 | * | |
2081 | * Return 0 if cache is invalid; 1 if the cache is valid | |
2082 | */ | |
2083 | static int | |
2084 | ext4_ext_in_cache(struct inode *inode, ext4_lblk_t block, | |
2085 | struct ext4_extent *ex) | |
2086 | { | |
2087 | struct ext4_ext_cache cex; | |
2088 | int ret = 0; | |
2089 | ||
2090 | if (ext4_ext_check_cache(inode, block, &cex)) { | |
2091 | ex->ee_block = cpu_to_le32(cex.ec_block); | |
2092 | ext4_ext_store_pblock(ex, cex.ec_start); | |
2093 | ex->ee_len = cpu_to_le16(cex.ec_len); | |
2094 | ret = 1; | |
2095 | } | |
2096 | ||
2097 | return ret; | |
2098 | } | |
2099 | ||
2100 | ||
a86c6181 | 2101 | /* |
d0d856e8 RD |
2102 | * ext4_ext_rm_idx: |
2103 | * removes index from the index block. | |
a86c6181 | 2104 | */ |
1d03ec98 | 2105 | static int ext4_ext_rm_idx(handle_t *handle, struct inode *inode, |
a86c6181 AT |
2106 | struct ext4_ext_path *path) |
2107 | { | |
a86c6181 | 2108 | int err; |
f65e6fba | 2109 | ext4_fsblk_t leaf; |
a86c6181 AT |
2110 | |
2111 | /* free index block */ | |
2112 | path--; | |
bf89d16f | 2113 | leaf = ext4_idx_pblock(path->p_idx); |
273df556 FM |
2114 | if (unlikely(path->p_hdr->eh_entries == 0)) { |
2115 | EXT4_ERROR_INODE(inode, "path->p_hdr->eh_entries == 0"); | |
2116 | return -EIO; | |
2117 | } | |
7e028976 AM |
2118 | err = ext4_ext_get_access(handle, inode, path); |
2119 | if (err) | |
a86c6181 | 2120 | return err; |
0e1147b0 RD |
2121 | |
2122 | if (path->p_idx != EXT_LAST_INDEX(path->p_hdr)) { | |
2123 | int len = EXT_LAST_INDEX(path->p_hdr) - path->p_idx; | |
2124 | len *= sizeof(struct ext4_extent_idx); | |
2125 | memmove(path->p_idx, path->p_idx + 1, len); | |
2126 | } | |
2127 | ||
e8546d06 | 2128 | le16_add_cpu(&path->p_hdr->eh_entries, -1); |
7e028976 AM |
2129 | err = ext4_ext_dirty(handle, inode, path); |
2130 | if (err) | |
a86c6181 | 2131 | return err; |
2ae02107 | 2132 | ext_debug("index is empty, remove it, free block %llu\n", leaf); |
d8990240 AK |
2133 | trace_ext4_ext_rm_idx(inode, leaf); |
2134 | ||
7dc57615 | 2135 | ext4_free_blocks(handle, inode, NULL, leaf, 1, |
e6362609 | 2136 | EXT4_FREE_BLOCKS_METADATA | EXT4_FREE_BLOCKS_FORGET); |
a86c6181 AT |
2137 | return err; |
2138 | } | |
2139 | ||
2140 | /* | |
ee12b630 MC |
2141 | * ext4_ext_calc_credits_for_single_extent: |
2142 | * This routine returns max. credits that needed to insert an extent | |
2143 | * to the extent tree. | |
2144 | * When pass the actual path, the caller should calculate credits | |
2145 | * under i_data_sem. | |
a86c6181 | 2146 | */ |
525f4ed8 | 2147 | int ext4_ext_calc_credits_for_single_extent(struct inode *inode, int nrblocks, |
a86c6181 AT |
2148 | struct ext4_ext_path *path) |
2149 | { | |
a86c6181 | 2150 | if (path) { |
ee12b630 | 2151 | int depth = ext_depth(inode); |
f3bd1f3f | 2152 | int ret = 0; |
ee12b630 | 2153 | |
a86c6181 | 2154 | /* probably there is space in leaf? */ |
a86c6181 | 2155 | if (le16_to_cpu(path[depth].p_hdr->eh_entries) |
ee12b630 | 2156 | < le16_to_cpu(path[depth].p_hdr->eh_max)) { |
a86c6181 | 2157 | |
ee12b630 MC |
2158 | /* |
2159 | * There are some space in the leaf tree, no | |
2160 | * need to account for leaf block credit | |
2161 | * | |
2162 | * bitmaps and block group descriptor blocks | |
df3ab170 | 2163 | * and other metadata blocks still need to be |
ee12b630 MC |
2164 | * accounted. |
2165 | */ | |
525f4ed8 | 2166 | /* 1 bitmap, 1 block group descriptor */ |
ee12b630 | 2167 | ret = 2 + EXT4_META_TRANS_BLOCKS(inode->i_sb); |
5887e98b | 2168 | return ret; |
ee12b630 MC |
2169 | } |
2170 | } | |
a86c6181 | 2171 | |
525f4ed8 | 2172 | return ext4_chunk_trans_blocks(inode, nrblocks); |
ee12b630 | 2173 | } |
a86c6181 | 2174 | |
ee12b630 MC |
2175 | /* |
2176 | * How many index/leaf blocks need to change/allocate to modify nrblocks? | |
2177 | * | |
2178 | * if nrblocks are fit in a single extent (chunk flag is 1), then | |
2179 | * in the worse case, each tree level index/leaf need to be changed | |
2180 | * if the tree split due to insert a new extent, then the old tree | |
2181 | * index/leaf need to be updated too | |
2182 | * | |
2183 | * If the nrblocks are discontiguous, they could cause | |
2184 | * the whole tree split more than once, but this is really rare. | |
2185 | */ | |
525f4ed8 | 2186 | int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks, int chunk) |
ee12b630 MC |
2187 | { |
2188 | int index; | |
2189 | int depth = ext_depth(inode); | |
a86c6181 | 2190 | |
ee12b630 MC |
2191 | if (chunk) |
2192 | index = depth * 2; | |
2193 | else | |
2194 | index = depth * 3; | |
a86c6181 | 2195 | |
ee12b630 | 2196 | return index; |
a86c6181 AT |
2197 | } |
2198 | ||
2199 | static int ext4_remove_blocks(handle_t *handle, struct inode *inode, | |
0aa06000 TT |
2200 | struct ext4_extent *ex, |
2201 | ext4_fsblk_t *partial_cluster, | |
2202 | ext4_lblk_t from, ext4_lblk_t to) | |
a86c6181 | 2203 | { |
0aa06000 | 2204 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); |
a2df2a63 | 2205 | unsigned short ee_len = ext4_ext_get_actual_len(ex); |
0aa06000 | 2206 | ext4_fsblk_t pblk; |
e6362609 | 2207 | int flags = EXT4_FREE_BLOCKS_FORGET; |
a86c6181 | 2208 | |
c9de560d | 2209 | if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) |
e6362609 | 2210 | flags |= EXT4_FREE_BLOCKS_METADATA; |
0aa06000 TT |
2211 | /* |
2212 | * For bigalloc file systems, we never free a partial cluster | |
2213 | * at the beginning of the extent. Instead, we make a note | |
2214 | * that we tried freeing the cluster, and check to see if we | |
2215 | * need to free it on a subsequent call to ext4_remove_blocks, | |
2216 | * or at the end of the ext4_truncate() operation. | |
2217 | */ | |
2218 | flags |= EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER; | |
2219 | ||
d8990240 | 2220 | trace_ext4_remove_blocks(inode, ex, from, to, *partial_cluster); |
0aa06000 TT |
2221 | /* |
2222 | * If we have a partial cluster, and it's different from the | |
2223 | * cluster of the last block, we need to explicitly free the | |
2224 | * partial cluster here. | |
2225 | */ | |
2226 | pblk = ext4_ext_pblock(ex) + ee_len - 1; | |
2227 | if (*partial_cluster && (EXT4_B2C(sbi, pblk) != *partial_cluster)) { | |
2228 | ext4_free_blocks(handle, inode, NULL, | |
2229 | EXT4_C2B(sbi, *partial_cluster), | |
2230 | sbi->s_cluster_ratio, flags); | |
2231 | *partial_cluster = 0; | |
2232 | } | |
2233 | ||
a86c6181 AT |
2234 | #ifdef EXTENTS_STATS |
2235 | { | |
2236 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
a86c6181 AT |
2237 | spin_lock(&sbi->s_ext_stats_lock); |
2238 | sbi->s_ext_blocks += ee_len; | |
2239 | sbi->s_ext_extents++; | |
2240 | if (ee_len < sbi->s_ext_min) | |
2241 | sbi->s_ext_min = ee_len; | |
2242 | if (ee_len > sbi->s_ext_max) | |
2243 | sbi->s_ext_max = ee_len; | |
2244 | if (ext_depth(inode) > sbi->s_depth_max) | |
2245 | sbi->s_depth_max = ext_depth(inode); | |
2246 | spin_unlock(&sbi->s_ext_stats_lock); | |
2247 | } | |
2248 | #endif | |
2249 | if (from >= le32_to_cpu(ex->ee_block) | |
a2df2a63 | 2250 | && to == le32_to_cpu(ex->ee_block) + ee_len - 1) { |
a86c6181 | 2251 | /* tail removal */ |
725d26d3 | 2252 | ext4_lblk_t num; |
725d26d3 | 2253 | |
a2df2a63 | 2254 | num = le32_to_cpu(ex->ee_block) + ee_len - from; |
0aa06000 TT |
2255 | pblk = ext4_ext_pblock(ex) + ee_len - num; |
2256 | ext_debug("free last %u blocks starting %llu\n", num, pblk); | |
2257 | ext4_free_blocks(handle, inode, NULL, pblk, num, flags); | |
2258 | /* | |
2259 | * If the block range to be freed didn't start at the | |
2260 | * beginning of a cluster, and we removed the entire | |
2261 | * extent, save the partial cluster here, since we | |
2262 | * might need to delete if we determine that the | |
2263 | * truncate operation has removed all of the blocks in | |
2264 | * the cluster. | |
2265 | */ | |
2266 | if (pblk & (sbi->s_cluster_ratio - 1) && | |
2267 | (ee_len == num)) | |
2268 | *partial_cluster = EXT4_B2C(sbi, pblk); | |
2269 | else | |
2270 | *partial_cluster = 0; | |
a86c6181 | 2271 | } else if (from == le32_to_cpu(ex->ee_block) |
a2df2a63 | 2272 | && to <= le32_to_cpu(ex->ee_block) + ee_len - 1) { |
d583fb87 AH |
2273 | /* head removal */ |
2274 | ext4_lblk_t num; | |
2275 | ext4_fsblk_t start; | |
2276 | ||
2277 | num = to - from; | |
2278 | start = ext4_ext_pblock(ex); | |
2279 | ||
2280 | ext_debug("free first %u blocks starting %llu\n", num, start); | |
ee90d57e | 2281 | ext4_free_blocks(handle, inode, NULL, start, num, flags); |
d583fb87 | 2282 | |
a86c6181 | 2283 | } else { |
725d26d3 AK |
2284 | printk(KERN_INFO "strange request: removal(2) " |
2285 | "%u-%u from %u:%u\n", | |
2286 | from, to, le32_to_cpu(ex->ee_block), ee_len); | |
a86c6181 AT |
2287 | } |
2288 | return 0; | |
2289 | } | |
2290 | ||
d583fb87 AH |
2291 | |
2292 | /* | |
2293 | * ext4_ext_rm_leaf() Removes the extents associated with the | |
2294 | * blocks appearing between "start" and "end", and splits the extents | |
2295 | * if "start" and "end" appear in the same extent | |
2296 | * | |
2297 | * @handle: The journal handle | |
2298 | * @inode: The files inode | |
2299 | * @path: The path to the leaf | |
2300 | * @start: The first block to remove | |
2301 | * @end: The last block to remove | |
2302 | */ | |
a86c6181 AT |
2303 | static int |
2304 | ext4_ext_rm_leaf(handle_t *handle, struct inode *inode, | |
0aa06000 TT |
2305 | struct ext4_ext_path *path, ext4_fsblk_t *partial_cluster, |
2306 | ext4_lblk_t start, ext4_lblk_t end) | |
a86c6181 | 2307 | { |
0aa06000 | 2308 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); |
a86c6181 AT |
2309 | int err = 0, correct_index = 0; |
2310 | int depth = ext_depth(inode), credits; | |
2311 | struct ext4_extent_header *eh; | |
750c9c47 | 2312 | ext4_lblk_t a, b; |
725d26d3 AK |
2313 | unsigned num; |
2314 | ext4_lblk_t ex_ee_block; | |
a86c6181 | 2315 | unsigned short ex_ee_len; |
a2df2a63 | 2316 | unsigned uninitialized = 0; |
a86c6181 AT |
2317 | struct ext4_extent *ex; |
2318 | ||
c29c0ae7 | 2319 | /* the header must be checked already in ext4_ext_remove_space() */ |
725d26d3 | 2320 | ext_debug("truncate since %u in leaf\n", start); |
a86c6181 AT |
2321 | if (!path[depth].p_hdr) |
2322 | path[depth].p_hdr = ext_block_hdr(path[depth].p_bh); | |
2323 | eh = path[depth].p_hdr; | |
273df556 FM |
2324 | if (unlikely(path[depth].p_hdr == NULL)) { |
2325 | EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth); | |
2326 | return -EIO; | |
2327 | } | |
a86c6181 AT |
2328 | /* find where to start removing */ |
2329 | ex = EXT_LAST_EXTENT(eh); | |
2330 | ||
2331 | ex_ee_block = le32_to_cpu(ex->ee_block); | |
a2df2a63 | 2332 | ex_ee_len = ext4_ext_get_actual_len(ex); |
a86c6181 | 2333 | |
d8990240 AK |
2334 | trace_ext4_ext_rm_leaf(inode, start, ex, *partial_cluster); |
2335 | ||
a86c6181 AT |
2336 | while (ex >= EXT_FIRST_EXTENT(eh) && |
2337 | ex_ee_block + ex_ee_len > start) { | |
a41f2071 AK |
2338 | |
2339 | if (ext4_ext_is_uninitialized(ex)) | |
2340 | uninitialized = 1; | |
2341 | else | |
2342 | uninitialized = 0; | |
2343 | ||
553f9008 M |
2344 | ext_debug("remove ext %u:[%d]%d\n", ex_ee_block, |
2345 | uninitialized, ex_ee_len); | |
a86c6181 AT |
2346 | path[depth].p_ext = ex; |
2347 | ||
2348 | a = ex_ee_block > start ? ex_ee_block : start; | |
d583fb87 AH |
2349 | b = ex_ee_block+ex_ee_len - 1 < end ? |
2350 | ex_ee_block+ex_ee_len - 1 : end; | |
a86c6181 AT |
2351 | |
2352 | ext_debug(" border %u:%u\n", a, b); | |
2353 | ||
d583fb87 AH |
2354 | /* If this extent is beyond the end of the hole, skip it */ |
2355 | if (end <= ex_ee_block) { | |
2356 | ex--; | |
2357 | ex_ee_block = le32_to_cpu(ex->ee_block); | |
2358 | ex_ee_len = ext4_ext_get_actual_len(ex); | |
2359 | continue; | |
750c9c47 DM |
2360 | } else if (b != ex_ee_block + ex_ee_len - 1) { |
2361 | EXT4_ERROR_INODE(inode," bad truncate %u:%u\n", | |
2362 | start, end); | |
2363 | err = -EIO; | |
2364 | goto out; | |
a86c6181 AT |
2365 | } else if (a != ex_ee_block) { |
2366 | /* remove tail of the extent */ | |
750c9c47 | 2367 | num = a - ex_ee_block; |
a86c6181 AT |
2368 | } else { |
2369 | /* remove whole extent: excellent! */ | |
a86c6181 | 2370 | num = 0; |
a86c6181 | 2371 | } |
34071da7 TT |
2372 | /* |
2373 | * 3 for leaf, sb, and inode plus 2 (bmap and group | |
2374 | * descriptor) for each block group; assume two block | |
2375 | * groups plus ex_ee_len/blocks_per_block_group for | |
2376 | * the worst case | |
2377 | */ | |
2378 | credits = 7 + 2*(ex_ee_len/EXT4_BLOCKS_PER_GROUP(inode->i_sb)); | |
a86c6181 AT |
2379 | if (ex == EXT_FIRST_EXTENT(eh)) { |
2380 | correct_index = 1; | |
2381 | credits += (ext_depth(inode)) + 1; | |
2382 | } | |
5aca07eb | 2383 | credits += EXT4_MAXQUOTAS_TRANS_BLOCKS(inode->i_sb); |
a86c6181 | 2384 | |
487caeef | 2385 | err = ext4_ext_truncate_extend_restart(handle, inode, credits); |
9102e4fa | 2386 | if (err) |
a86c6181 | 2387 | goto out; |
a86c6181 AT |
2388 | |
2389 | err = ext4_ext_get_access(handle, inode, path + depth); | |
2390 | if (err) | |
2391 | goto out; | |
2392 | ||
0aa06000 TT |
2393 | err = ext4_remove_blocks(handle, inode, ex, partial_cluster, |
2394 | a, b); | |
a86c6181 AT |
2395 | if (err) |
2396 | goto out; | |
2397 | ||
750c9c47 | 2398 | if (num == 0) |
d0d856e8 | 2399 | /* this extent is removed; mark slot entirely unused */ |
f65e6fba | 2400 | ext4_ext_store_pblock(ex, 0); |
a86c6181 | 2401 | |
a86c6181 | 2402 | ex->ee_len = cpu_to_le16(num); |
749269fa AA |
2403 | /* |
2404 | * Do not mark uninitialized if all the blocks in the | |
2405 | * extent have been removed. | |
2406 | */ | |
2407 | if (uninitialized && num) | |
a2df2a63 | 2408 | ext4_ext_mark_uninitialized(ex); |
d583fb87 AH |
2409 | /* |
2410 | * If the extent was completely released, | |
2411 | * we need to remove it from the leaf | |
2412 | */ | |
2413 | if (num == 0) { | |
f17722f9 | 2414 | if (end != EXT_MAX_BLOCKS - 1) { |
d583fb87 AH |
2415 | /* |
2416 | * For hole punching, we need to scoot all the | |
2417 | * extents up when an extent is removed so that | |
2418 | * we dont have blank extents in the middle | |
2419 | */ | |
2420 | memmove(ex, ex+1, (EXT_LAST_EXTENT(eh) - ex) * | |
2421 | sizeof(struct ext4_extent)); | |
2422 | ||
2423 | /* Now get rid of the one at the end */ | |
2424 | memset(EXT_LAST_EXTENT(eh), 0, | |
2425 | sizeof(struct ext4_extent)); | |
2426 | } | |
2427 | le16_add_cpu(&eh->eh_entries, -1); | |
0aa06000 TT |
2428 | } else |
2429 | *partial_cluster = 0; | |
d583fb87 | 2430 | |
750c9c47 DM |
2431 | err = ext4_ext_dirty(handle, inode, path + depth); |
2432 | if (err) | |
2433 | goto out; | |
2434 | ||
2ae02107 | 2435 | ext_debug("new extent: %u:%u:%llu\n", block, num, |
bf89d16f | 2436 | ext4_ext_pblock(ex)); |
a86c6181 AT |
2437 | ex--; |
2438 | ex_ee_block = le32_to_cpu(ex->ee_block); | |
a2df2a63 | 2439 | ex_ee_len = ext4_ext_get_actual_len(ex); |
a86c6181 AT |
2440 | } |
2441 | ||
2442 | if (correct_index && eh->eh_entries) | |
2443 | err = ext4_ext_correct_indexes(handle, inode, path); | |
2444 | ||
0aa06000 TT |
2445 | /* |
2446 | * If there is still a entry in the leaf node, check to see if | |
2447 | * it references the partial cluster. This is the only place | |
2448 | * where it could; if it doesn't, we can free the cluster. | |
2449 | */ | |
2450 | if (*partial_cluster && ex >= EXT_FIRST_EXTENT(eh) && | |
2451 | (EXT4_B2C(sbi, ext4_ext_pblock(ex) + ex_ee_len - 1) != | |
2452 | *partial_cluster)) { | |
2453 | int flags = EXT4_FREE_BLOCKS_FORGET; | |
2454 | ||
2455 | if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) | |
2456 | flags |= EXT4_FREE_BLOCKS_METADATA; | |
2457 | ||
2458 | ext4_free_blocks(handle, inode, NULL, | |
2459 | EXT4_C2B(sbi, *partial_cluster), | |
2460 | sbi->s_cluster_ratio, flags); | |
2461 | *partial_cluster = 0; | |
2462 | } | |
2463 | ||
a86c6181 AT |
2464 | /* if this leaf is free, then we should |
2465 | * remove it from index block above */ | |
2466 | if (err == 0 && eh->eh_entries == 0 && path[depth].p_bh != NULL) | |
2467 | err = ext4_ext_rm_idx(handle, inode, path + depth); | |
2468 | ||
2469 | out: | |
2470 | return err; | |
2471 | } | |
2472 | ||
2473 | /* | |
d0d856e8 RD |
2474 | * ext4_ext_more_to_rm: |
2475 | * returns 1 if current index has to be freed (even partial) | |
a86c6181 | 2476 | */ |
09b88252 | 2477 | static int |
a86c6181 AT |
2478 | ext4_ext_more_to_rm(struct ext4_ext_path *path) |
2479 | { | |
2480 | BUG_ON(path->p_idx == NULL); | |
2481 | ||
2482 | if (path->p_idx < EXT_FIRST_INDEX(path->p_hdr)) | |
2483 | return 0; | |
2484 | ||
2485 | /* | |
d0d856e8 | 2486 | * if truncate on deeper level happened, it wasn't partial, |
a86c6181 AT |
2487 | * so we have to consider current index for truncation |
2488 | */ | |
2489 | if (le16_to_cpu(path->p_hdr->eh_entries) == path->p_block) | |
2490 | return 0; | |
2491 | return 1; | |
2492 | } | |
2493 | ||
c6a0371c | 2494 | static int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start) |
a86c6181 AT |
2495 | { |
2496 | struct super_block *sb = inode->i_sb; | |
2497 | int depth = ext_depth(inode); | |
2498 | struct ext4_ext_path *path; | |
0aa06000 | 2499 | ext4_fsblk_t partial_cluster = 0; |
a86c6181 | 2500 | handle_t *handle; |
0617b83f | 2501 | int i, err; |
a86c6181 | 2502 | |
725d26d3 | 2503 | ext_debug("truncate since %u\n", start); |
a86c6181 AT |
2504 | |
2505 | /* probably first extent we're gonna free will be last in block */ | |
2506 | handle = ext4_journal_start(inode, depth + 1); | |
2507 | if (IS_ERR(handle)) | |
2508 | return PTR_ERR(handle); | |
2509 | ||
0617b83f | 2510 | again: |
a86c6181 AT |
2511 | ext4_ext_invalidate_cache(inode); |
2512 | ||
d8990240 AK |
2513 | trace_ext4_ext_remove_space(inode, start, depth); |
2514 | ||
a86c6181 | 2515 | /* |
d0d856e8 RD |
2516 | * We start scanning from right side, freeing all the blocks |
2517 | * after i_size and walking into the tree depth-wise. | |
a86c6181 | 2518 | */ |
0617b83f | 2519 | depth = ext_depth(inode); |
216553c4 | 2520 | path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1), GFP_NOFS); |
a86c6181 AT |
2521 | if (path == NULL) { |
2522 | ext4_journal_stop(handle); | |
2523 | return -ENOMEM; | |
2524 | } | |
0617b83f | 2525 | path[0].p_depth = depth; |
a86c6181 | 2526 | path[0].p_hdr = ext_inode_hdr(inode); |
56b19868 | 2527 | if (ext4_ext_check(inode, path[0].p_hdr, depth)) { |
a86c6181 AT |
2528 | err = -EIO; |
2529 | goto out; | |
2530 | } | |
0617b83f | 2531 | i = err = 0; |
a86c6181 AT |
2532 | |
2533 | while (i >= 0 && err == 0) { | |
2534 | if (i == depth) { | |
2535 | /* this is leaf block */ | |
d583fb87 | 2536 | err = ext4_ext_rm_leaf(handle, inode, path, |
0aa06000 TT |
2537 | &partial_cluster, start, |
2538 | EXT_MAX_BLOCKS - 1); | |
d0d856e8 | 2539 | /* root level has p_bh == NULL, brelse() eats this */ |
a86c6181 AT |
2540 | brelse(path[i].p_bh); |
2541 | path[i].p_bh = NULL; | |
2542 | i--; | |
2543 | continue; | |
2544 | } | |
2545 | ||
2546 | /* this is index block */ | |
2547 | if (!path[i].p_hdr) { | |
2548 | ext_debug("initialize header\n"); | |
2549 | path[i].p_hdr = ext_block_hdr(path[i].p_bh); | |
a86c6181 AT |
2550 | } |
2551 | ||
a86c6181 | 2552 | if (!path[i].p_idx) { |
d0d856e8 | 2553 | /* this level hasn't been touched yet */ |
a86c6181 AT |
2554 | path[i].p_idx = EXT_LAST_INDEX(path[i].p_hdr); |
2555 | path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries)+1; | |
2556 | ext_debug("init index ptr: hdr 0x%p, num %d\n", | |
2557 | path[i].p_hdr, | |
2558 | le16_to_cpu(path[i].p_hdr->eh_entries)); | |
2559 | } else { | |
d0d856e8 | 2560 | /* we were already here, see at next index */ |
a86c6181 AT |
2561 | path[i].p_idx--; |
2562 | } | |
2563 | ||
2564 | ext_debug("level %d - index, first 0x%p, cur 0x%p\n", | |
2565 | i, EXT_FIRST_INDEX(path[i].p_hdr), | |
2566 | path[i].p_idx); | |
2567 | if (ext4_ext_more_to_rm(path + i)) { | |
c29c0ae7 | 2568 | struct buffer_head *bh; |
a86c6181 | 2569 | /* go to the next level */ |
2ae02107 | 2570 | ext_debug("move to level %d (block %llu)\n", |
bf89d16f | 2571 | i + 1, ext4_idx_pblock(path[i].p_idx)); |
a86c6181 | 2572 | memset(path + i + 1, 0, sizeof(*path)); |
bf89d16f | 2573 | bh = sb_bread(sb, ext4_idx_pblock(path[i].p_idx)); |
c29c0ae7 | 2574 | if (!bh) { |
a86c6181 AT |
2575 | /* should we reset i_size? */ |
2576 | err = -EIO; | |
2577 | break; | |
2578 | } | |
c29c0ae7 AT |
2579 | if (WARN_ON(i + 1 > depth)) { |
2580 | err = -EIO; | |
2581 | break; | |
2582 | } | |
56b19868 | 2583 | if (ext4_ext_check(inode, ext_block_hdr(bh), |
c29c0ae7 AT |
2584 | depth - i - 1)) { |
2585 | err = -EIO; | |
2586 | break; | |
2587 | } | |
2588 | path[i + 1].p_bh = bh; | |
a86c6181 | 2589 | |
d0d856e8 RD |
2590 | /* save actual number of indexes since this |
2591 | * number is changed at the next iteration */ | |
a86c6181 AT |
2592 | path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries); |
2593 | i++; | |
2594 | } else { | |
d0d856e8 | 2595 | /* we finished processing this index, go up */ |
a86c6181 | 2596 | if (path[i].p_hdr->eh_entries == 0 && i > 0) { |
d0d856e8 | 2597 | /* index is empty, remove it; |
a86c6181 AT |
2598 | * handle must be already prepared by the |
2599 | * truncatei_leaf() */ | |
2600 | err = ext4_ext_rm_idx(handle, inode, path + i); | |
2601 | } | |
d0d856e8 | 2602 | /* root level has p_bh == NULL, brelse() eats this */ |
a86c6181 AT |
2603 | brelse(path[i].p_bh); |
2604 | path[i].p_bh = NULL; | |
2605 | i--; | |
2606 | ext_debug("return to level %d\n", i); | |
2607 | } | |
2608 | } | |
2609 | ||
d8990240 AK |
2610 | trace_ext4_ext_remove_space_done(inode, start, depth, partial_cluster, |
2611 | path->p_hdr->eh_entries); | |
2612 | ||
7b415bf6 AK |
2613 | /* If we still have something in the partial cluster and we have removed |
2614 | * even the first extent, then we should free the blocks in the partial | |
2615 | * cluster as well. */ | |
2616 | if (partial_cluster && path->p_hdr->eh_entries == 0) { | |
2617 | int flags = EXT4_FREE_BLOCKS_FORGET; | |
2618 | ||
2619 | if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) | |
2620 | flags |= EXT4_FREE_BLOCKS_METADATA; | |
2621 | ||
2622 | ext4_free_blocks(handle, inode, NULL, | |
2623 | EXT4_C2B(EXT4_SB(sb), partial_cluster), | |
2624 | EXT4_SB(sb)->s_cluster_ratio, flags); | |
2625 | partial_cluster = 0; | |
2626 | } | |
2627 | ||
a86c6181 AT |
2628 | /* TODO: flexible tree reduction should be here */ |
2629 | if (path->p_hdr->eh_entries == 0) { | |
2630 | /* | |
d0d856e8 RD |
2631 | * truncate to zero freed all the tree, |
2632 | * so we need to correct eh_depth | |
a86c6181 AT |
2633 | */ |
2634 | err = ext4_ext_get_access(handle, inode, path); | |
2635 | if (err == 0) { | |
2636 | ext_inode_hdr(inode)->eh_depth = 0; | |
2637 | ext_inode_hdr(inode)->eh_max = | |
55ad63bf | 2638 | cpu_to_le16(ext4_ext_space_root(inode, 0)); |
a86c6181 AT |
2639 | err = ext4_ext_dirty(handle, inode, path); |
2640 | } | |
2641 | } | |
2642 | out: | |
a86c6181 AT |
2643 | ext4_ext_drop_refs(path); |
2644 | kfree(path); | |
0617b83f DM |
2645 | if (err == -EAGAIN) |
2646 | goto again; | |
a86c6181 AT |
2647 | ext4_journal_stop(handle); |
2648 | ||
2649 | return err; | |
2650 | } | |
2651 | ||
2652 | /* | |
2653 | * called at mount time | |
2654 | */ | |
2655 | void ext4_ext_init(struct super_block *sb) | |
2656 | { | |
2657 | /* | |
2658 | * possible initialization would be here | |
2659 | */ | |
2660 | ||
83982b6f | 2661 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) { |
90576c0b | 2662 | #if defined(AGGRESSIVE_TEST) || defined(CHECK_BINSEARCH) || defined(EXTENTS_STATS) |
4776004f | 2663 | printk(KERN_INFO "EXT4-fs: file extents enabled"); |
bbf2f9fb RD |
2664 | #ifdef AGGRESSIVE_TEST |
2665 | printk(", aggressive tests"); | |
a86c6181 AT |
2666 | #endif |
2667 | #ifdef CHECK_BINSEARCH | |
2668 | printk(", check binsearch"); | |
2669 | #endif | |
2670 | #ifdef EXTENTS_STATS | |
2671 | printk(", stats"); | |
2672 | #endif | |
2673 | printk("\n"); | |
90576c0b | 2674 | #endif |
a86c6181 AT |
2675 | #ifdef EXTENTS_STATS |
2676 | spin_lock_init(&EXT4_SB(sb)->s_ext_stats_lock); | |
2677 | EXT4_SB(sb)->s_ext_min = 1 << 30; | |
2678 | EXT4_SB(sb)->s_ext_max = 0; | |
2679 | #endif | |
2680 | } | |
2681 | } | |
2682 | ||
2683 | /* | |
2684 | * called at umount time | |
2685 | */ | |
2686 | void ext4_ext_release(struct super_block *sb) | |
2687 | { | |
83982b6f | 2688 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) |
a86c6181 AT |
2689 | return; |
2690 | ||
2691 | #ifdef EXTENTS_STATS | |
2692 | if (EXT4_SB(sb)->s_ext_blocks && EXT4_SB(sb)->s_ext_extents) { | |
2693 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2694 | printk(KERN_ERR "EXT4-fs: %lu blocks in %lu extents (%lu ave)\n", | |
2695 | sbi->s_ext_blocks, sbi->s_ext_extents, | |
2696 | sbi->s_ext_blocks / sbi->s_ext_extents); | |
2697 | printk(KERN_ERR "EXT4-fs: extents: %lu min, %lu max, max depth %lu\n", | |
2698 | sbi->s_ext_min, sbi->s_ext_max, sbi->s_depth_max); | |
2699 | } | |
2700 | #endif | |
2701 | } | |
2702 | ||
093a088b AK |
2703 | /* FIXME!! we need to try to merge to left or right after zero-out */ |
2704 | static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex) | |
2705 | { | |
2407518d LC |
2706 | ext4_fsblk_t ee_pblock; |
2707 | unsigned int ee_len; | |
b720303d | 2708 | int ret; |
093a088b | 2709 | |
093a088b | 2710 | ee_len = ext4_ext_get_actual_len(ex); |
bf89d16f | 2711 | ee_pblock = ext4_ext_pblock(ex); |
b720303d | 2712 | |
a107e5a3 | 2713 | ret = sb_issue_zeroout(inode->i_sb, ee_pblock, ee_len, GFP_NOFS); |
2407518d LC |
2714 | if (ret > 0) |
2715 | ret = 0; | |
093a088b | 2716 | |
2407518d | 2717 | return ret; |
093a088b AK |
2718 | } |
2719 | ||
47ea3bb5 YY |
2720 | /* |
2721 | * used by extent splitting. | |
2722 | */ | |
2723 | #define EXT4_EXT_MAY_ZEROOUT 0x1 /* safe to zeroout if split fails \ | |
2724 | due to ENOSPC */ | |
2725 | #define EXT4_EXT_MARK_UNINIT1 0x2 /* mark first half uninitialized */ | |
2726 | #define EXT4_EXT_MARK_UNINIT2 0x4 /* mark second half uninitialized */ | |
2727 | ||
2728 | /* | |
2729 | * ext4_split_extent_at() splits an extent at given block. | |
2730 | * | |
2731 | * @handle: the journal handle | |
2732 | * @inode: the file inode | |
2733 | * @path: the path to the extent | |
2734 | * @split: the logical block where the extent is splitted. | |
2735 | * @split_flags: indicates if the extent could be zeroout if split fails, and | |
2736 | * the states(init or uninit) of new extents. | |
2737 | * @flags: flags used to insert new extent to extent tree. | |
2738 | * | |
2739 | * | |
2740 | * Splits extent [a, b] into two extents [a, @split) and [@split, b], states | |
2741 | * of which are deterimined by split_flag. | |
2742 | * | |
2743 | * There are two cases: | |
2744 | * a> the extent are splitted into two extent. | |
2745 | * b> split is not needed, and just mark the extent. | |
2746 | * | |
2747 | * return 0 on success. | |
2748 | */ | |
2749 | static int ext4_split_extent_at(handle_t *handle, | |
2750 | struct inode *inode, | |
2751 | struct ext4_ext_path *path, | |
2752 | ext4_lblk_t split, | |
2753 | int split_flag, | |
2754 | int flags) | |
2755 | { | |
2756 | ext4_fsblk_t newblock; | |
2757 | ext4_lblk_t ee_block; | |
2758 | struct ext4_extent *ex, newex, orig_ex; | |
2759 | struct ext4_extent *ex2 = NULL; | |
2760 | unsigned int ee_len, depth; | |
2761 | int err = 0; | |
2762 | ||
2763 | ext_debug("ext4_split_extents_at: inode %lu, logical" | |
2764 | "block %llu\n", inode->i_ino, (unsigned long long)split); | |
2765 | ||
2766 | ext4_ext_show_leaf(inode, path); | |
2767 | ||
2768 | depth = ext_depth(inode); | |
2769 | ex = path[depth].p_ext; | |
2770 | ee_block = le32_to_cpu(ex->ee_block); | |
2771 | ee_len = ext4_ext_get_actual_len(ex); | |
2772 | newblock = split - ee_block + ext4_ext_pblock(ex); | |
2773 | ||
2774 | BUG_ON(split < ee_block || split >= (ee_block + ee_len)); | |
2775 | ||
2776 | err = ext4_ext_get_access(handle, inode, path + depth); | |
2777 | if (err) | |
2778 | goto out; | |
2779 | ||
2780 | if (split == ee_block) { | |
2781 | /* | |
2782 | * case b: block @split is the block that the extent begins with | |
2783 | * then we just change the state of the extent, and splitting | |
2784 | * is not needed. | |
2785 | */ | |
2786 | if (split_flag & EXT4_EXT_MARK_UNINIT2) | |
2787 | ext4_ext_mark_uninitialized(ex); | |
2788 | else | |
2789 | ext4_ext_mark_initialized(ex); | |
2790 | ||
2791 | if (!(flags & EXT4_GET_BLOCKS_PRE_IO)) | |
2792 | ext4_ext_try_to_merge(inode, path, ex); | |
2793 | ||
2794 | err = ext4_ext_dirty(handle, inode, path + depth); | |
2795 | goto out; | |
2796 | } | |
2797 | ||
2798 | /* case a */ | |
2799 | memcpy(&orig_ex, ex, sizeof(orig_ex)); | |
2800 | ex->ee_len = cpu_to_le16(split - ee_block); | |
2801 | if (split_flag & EXT4_EXT_MARK_UNINIT1) | |
2802 | ext4_ext_mark_uninitialized(ex); | |
2803 | ||
2804 | /* | |
2805 | * path may lead to new leaf, not to original leaf any more | |
2806 | * after ext4_ext_insert_extent() returns, | |
2807 | */ | |
2808 | err = ext4_ext_dirty(handle, inode, path + depth); | |
2809 | if (err) | |
2810 | goto fix_extent_len; | |
2811 | ||
2812 | ex2 = &newex; | |
2813 | ex2->ee_block = cpu_to_le32(split); | |
2814 | ex2->ee_len = cpu_to_le16(ee_len - (split - ee_block)); | |
2815 | ext4_ext_store_pblock(ex2, newblock); | |
2816 | if (split_flag & EXT4_EXT_MARK_UNINIT2) | |
2817 | ext4_ext_mark_uninitialized(ex2); | |
2818 | ||
2819 | err = ext4_ext_insert_extent(handle, inode, path, &newex, flags); | |
2820 | if (err == -ENOSPC && (EXT4_EXT_MAY_ZEROOUT & split_flag)) { | |
2821 | err = ext4_ext_zeroout(inode, &orig_ex); | |
2822 | if (err) | |
2823 | goto fix_extent_len; | |
2824 | /* update the extent length and mark as initialized */ | |
2825 | ex->ee_len = cpu_to_le32(ee_len); | |
2826 | ext4_ext_try_to_merge(inode, path, ex); | |
2827 | err = ext4_ext_dirty(handle, inode, path + depth); | |
2828 | goto out; | |
2829 | } else if (err) | |
2830 | goto fix_extent_len; | |
2831 | ||
2832 | out: | |
2833 | ext4_ext_show_leaf(inode, path); | |
2834 | return err; | |
2835 | ||
2836 | fix_extent_len: | |
2837 | ex->ee_len = orig_ex.ee_len; | |
2838 | ext4_ext_dirty(handle, inode, path + depth); | |
2839 | return err; | |
2840 | } | |
2841 | ||
2842 | /* | |
2843 | * ext4_split_extents() splits an extent and mark extent which is covered | |
2844 | * by @map as split_flags indicates | |
2845 | * | |
2846 | * It may result in splitting the extent into multiple extents (upto three) | |
2847 | * There are three possibilities: | |
2848 | * a> There is no split required | |
2849 | * b> Splits in two extents: Split is happening at either end of the extent | |
2850 | * c> Splits in three extents: Somone is splitting in middle of the extent | |
2851 | * | |
2852 | */ | |
2853 | static int ext4_split_extent(handle_t *handle, | |
2854 | struct inode *inode, | |
2855 | struct ext4_ext_path *path, | |
2856 | struct ext4_map_blocks *map, | |
2857 | int split_flag, | |
2858 | int flags) | |
2859 | { | |
2860 | ext4_lblk_t ee_block; | |
2861 | struct ext4_extent *ex; | |
2862 | unsigned int ee_len, depth; | |
2863 | int err = 0; | |
2864 | int uninitialized; | |
2865 | int split_flag1, flags1; | |
2866 | ||
2867 | depth = ext_depth(inode); | |
2868 | ex = path[depth].p_ext; | |
2869 | ee_block = le32_to_cpu(ex->ee_block); | |
2870 | ee_len = ext4_ext_get_actual_len(ex); | |
2871 | uninitialized = ext4_ext_is_uninitialized(ex); | |
2872 | ||
2873 | if (map->m_lblk + map->m_len < ee_block + ee_len) { | |
2874 | split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT ? | |
2875 | EXT4_EXT_MAY_ZEROOUT : 0; | |
2876 | flags1 = flags | EXT4_GET_BLOCKS_PRE_IO; | |
2877 | if (uninitialized) | |
2878 | split_flag1 |= EXT4_EXT_MARK_UNINIT1 | | |
2879 | EXT4_EXT_MARK_UNINIT2; | |
2880 | err = ext4_split_extent_at(handle, inode, path, | |
2881 | map->m_lblk + map->m_len, split_flag1, flags1); | |
93917411 YY |
2882 | if (err) |
2883 | goto out; | |
47ea3bb5 YY |
2884 | } |
2885 | ||
2886 | ext4_ext_drop_refs(path); | |
2887 | path = ext4_ext_find_extent(inode, map->m_lblk, path); | |
2888 | if (IS_ERR(path)) | |
2889 | return PTR_ERR(path); | |
2890 | ||
2891 | if (map->m_lblk >= ee_block) { | |
2892 | split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT ? | |
2893 | EXT4_EXT_MAY_ZEROOUT : 0; | |
2894 | if (uninitialized) | |
2895 | split_flag1 |= EXT4_EXT_MARK_UNINIT1; | |
2896 | if (split_flag & EXT4_EXT_MARK_UNINIT2) | |
2897 | split_flag1 |= EXT4_EXT_MARK_UNINIT2; | |
2898 | err = ext4_split_extent_at(handle, inode, path, | |
2899 | map->m_lblk, split_flag1, flags); | |
2900 | if (err) | |
2901 | goto out; | |
2902 | } | |
2903 | ||
2904 | ext4_ext_show_leaf(inode, path); | |
2905 | out: | |
2906 | return err ? err : map->m_len; | |
2907 | } | |
2908 | ||
3977c965 | 2909 | #define EXT4_EXT_ZERO_LEN 7 |
56055d3a | 2910 | /* |
e35fd660 | 2911 | * This function is called by ext4_ext_map_blocks() if someone tries to write |
56055d3a | 2912 | * to an uninitialized extent. It may result in splitting the uninitialized |
25985edc | 2913 | * extent into multiple extents (up to three - one initialized and two |
56055d3a AA |
2914 | * uninitialized). |
2915 | * There are three possibilities: | |
2916 | * a> There is no split required: Entire extent should be initialized | |
2917 | * b> Splits in two extents: Write is happening at either end of the extent | |
2918 | * c> Splits in three extents: Somone is writing in middle of the extent | |
6f91bc5f EG |
2919 | * |
2920 | * Pre-conditions: | |
2921 | * - The extent pointed to by 'path' is uninitialized. | |
2922 | * - The extent pointed to by 'path' contains a superset | |
2923 | * of the logical span [map->m_lblk, map->m_lblk + map->m_len). | |
2924 | * | |
2925 | * Post-conditions on success: | |
2926 | * - the returned value is the number of blocks beyond map->l_lblk | |
2927 | * that are allocated and initialized. | |
2928 | * It is guaranteed to be >= map->m_len. | |
56055d3a | 2929 | */ |
725d26d3 | 2930 | static int ext4_ext_convert_to_initialized(handle_t *handle, |
e35fd660 TT |
2931 | struct inode *inode, |
2932 | struct ext4_map_blocks *map, | |
2933 | struct ext4_ext_path *path) | |
56055d3a | 2934 | { |
6f91bc5f | 2935 | struct ext4_extent_header *eh; |
667eff35 YY |
2936 | struct ext4_map_blocks split_map; |
2937 | struct ext4_extent zero_ex; | |
2938 | struct ext4_extent *ex; | |
21ca087a | 2939 | ext4_lblk_t ee_block, eof_block; |
f85b287a DC |
2940 | unsigned int ee_len, depth; |
2941 | int allocated; | |
56055d3a | 2942 | int err = 0; |
667eff35 | 2943 | int split_flag = 0; |
21ca087a DM |
2944 | |
2945 | ext_debug("ext4_ext_convert_to_initialized: inode %lu, logical" | |
2946 | "block %llu, max_blocks %u\n", inode->i_ino, | |
e35fd660 | 2947 | (unsigned long long)map->m_lblk, map->m_len); |
21ca087a DM |
2948 | |
2949 | eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >> | |
2950 | inode->i_sb->s_blocksize_bits; | |
e35fd660 TT |
2951 | if (eof_block < map->m_lblk + map->m_len) |
2952 | eof_block = map->m_lblk + map->m_len; | |
56055d3a AA |
2953 | |
2954 | depth = ext_depth(inode); | |
6f91bc5f | 2955 | eh = path[depth].p_hdr; |
56055d3a AA |
2956 | ex = path[depth].p_ext; |
2957 | ee_block = le32_to_cpu(ex->ee_block); | |
2958 | ee_len = ext4_ext_get_actual_len(ex); | |
e35fd660 | 2959 | allocated = ee_len - (map->m_lblk - ee_block); |
56055d3a | 2960 | |
6f91bc5f EG |
2961 | trace_ext4_ext_convert_to_initialized_enter(inode, map, ex); |
2962 | ||
2963 | /* Pre-conditions */ | |
2964 | BUG_ON(!ext4_ext_is_uninitialized(ex)); | |
2965 | BUG_ON(!in_range(map->m_lblk, ee_block, ee_len)); | |
2966 | BUG_ON(map->m_lblk + map->m_len > ee_block + ee_len); | |
2967 | ||
2968 | /* | |
2969 | * Attempt to transfer newly initialized blocks from the currently | |
2970 | * uninitialized extent to its left neighbor. This is much cheaper | |
2971 | * than an insertion followed by a merge as those involve costly | |
2972 | * memmove() calls. This is the common case in steady state for | |
2973 | * workloads doing fallocate(FALLOC_FL_KEEP_SIZE) followed by append | |
2974 | * writes. | |
2975 | * | |
2976 | * Limitations of the current logic: | |
2977 | * - L1: we only deal with writes at the start of the extent. | |
2978 | * The approach could be extended to writes at the end | |
2979 | * of the extent but this scenario was deemed less common. | |
2980 | * - L2: we do not deal with writes covering the whole extent. | |
2981 | * This would require removing the extent if the transfer | |
2982 | * is possible. | |
2983 | * - L3: we only attempt to merge with an extent stored in the | |
2984 | * same extent tree node. | |
2985 | */ | |
2986 | if ((map->m_lblk == ee_block) && /*L1*/ | |
2987 | (map->m_len < ee_len) && /*L2*/ | |
2988 | (ex > EXT_FIRST_EXTENT(eh))) { /*L3*/ | |
2989 | struct ext4_extent *prev_ex; | |
2990 | ext4_lblk_t prev_lblk; | |
2991 | ext4_fsblk_t prev_pblk, ee_pblk; | |
2992 | unsigned int prev_len, write_len; | |
2993 | ||
2994 | prev_ex = ex - 1; | |
2995 | prev_lblk = le32_to_cpu(prev_ex->ee_block); | |
2996 | prev_len = ext4_ext_get_actual_len(prev_ex); | |
2997 | prev_pblk = ext4_ext_pblock(prev_ex); | |
2998 | ee_pblk = ext4_ext_pblock(ex); | |
2999 | write_len = map->m_len; | |
3000 | ||
3001 | /* | |
3002 | * A transfer of blocks from 'ex' to 'prev_ex' is allowed | |
3003 | * upon those conditions: | |
3004 | * - C1: prev_ex is initialized, | |
3005 | * - C2: prev_ex is logically abutting ex, | |
3006 | * - C3: prev_ex is physically abutting ex, | |
3007 | * - C4: prev_ex can receive the additional blocks without | |
3008 | * overflowing the (initialized) length limit. | |
3009 | */ | |
3010 | if ((!ext4_ext_is_uninitialized(prev_ex)) && /*C1*/ | |
3011 | ((prev_lblk + prev_len) == ee_block) && /*C2*/ | |
3012 | ((prev_pblk + prev_len) == ee_pblk) && /*C3*/ | |
3013 | (prev_len < (EXT_INIT_MAX_LEN - write_len))) { /*C4*/ | |
3014 | err = ext4_ext_get_access(handle, inode, path + depth); | |
3015 | if (err) | |
3016 | goto out; | |
3017 | ||
3018 | trace_ext4_ext_convert_to_initialized_fastpath(inode, | |
3019 | map, ex, prev_ex); | |
3020 | ||
3021 | /* Shift the start of ex by 'write_len' blocks */ | |
3022 | ex->ee_block = cpu_to_le32(ee_block + write_len); | |
3023 | ext4_ext_store_pblock(ex, ee_pblk + write_len); | |
3024 | ex->ee_len = cpu_to_le16(ee_len - write_len); | |
3025 | ext4_ext_mark_uninitialized(ex); /* Restore the flag */ | |
3026 | ||
3027 | /* Extend prev_ex by 'write_len' blocks */ | |
3028 | prev_ex->ee_len = cpu_to_le16(prev_len + write_len); | |
3029 | ||
3030 | /* Mark the block containing both extents as dirty */ | |
3031 | ext4_ext_dirty(handle, inode, path + depth); | |
3032 | ||
3033 | /* Update path to point to the right extent */ | |
3034 | path[depth].p_ext = prev_ex; | |
3035 | ||
3036 | /* Result: number of initialized blocks past m_lblk */ | |
3037 | allocated = write_len; | |
3038 | goto out; | |
3039 | } | |
3040 | } | |
3041 | ||
667eff35 | 3042 | WARN_ON(map->m_lblk < ee_block); |
21ca087a DM |
3043 | /* |
3044 | * It is safe to convert extent to initialized via explicit | |
3045 | * zeroout only if extent is fully insde i_size or new_size. | |
3046 | */ | |
667eff35 | 3047 | split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0; |
21ca087a | 3048 | |
3977c965 | 3049 | /* If extent has less than 2*EXT4_EXT_ZERO_LEN zerout directly */ |
667eff35 YY |
3050 | if (ee_len <= 2*EXT4_EXT_ZERO_LEN && |
3051 | (EXT4_EXT_MAY_ZEROOUT & split_flag)) { | |
3052 | err = ext4_ext_zeroout(inode, ex); | |
3977c965 | 3053 | if (err) |
d03856bd | 3054 | goto out; |
d03856bd AK |
3055 | |
3056 | err = ext4_ext_get_access(handle, inode, path + depth); | |
3057 | if (err) | |
3058 | goto out; | |
667eff35 YY |
3059 | ext4_ext_mark_initialized(ex); |
3060 | ext4_ext_try_to_merge(inode, path, ex); | |
3061 | err = ext4_ext_dirty(handle, inode, path + depth); | |
3062 | goto out; | |
56055d3a | 3063 | } |
667eff35 | 3064 | |
56055d3a | 3065 | /* |
667eff35 YY |
3066 | * four cases: |
3067 | * 1. split the extent into three extents. | |
3068 | * 2. split the extent into two extents, zeroout the first half. | |
3069 | * 3. split the extent into two extents, zeroout the second half. | |
3070 | * 4. split the extent into two extents with out zeroout. | |
56055d3a | 3071 | */ |
667eff35 YY |
3072 | split_map.m_lblk = map->m_lblk; |
3073 | split_map.m_len = map->m_len; | |
3074 | ||
3075 | if (allocated > map->m_len) { | |
3076 | if (allocated <= EXT4_EXT_ZERO_LEN && | |
3077 | (EXT4_EXT_MAY_ZEROOUT & split_flag)) { | |
3078 | /* case 3 */ | |
3079 | zero_ex.ee_block = | |
9b940f8e AH |
3080 | cpu_to_le32(map->m_lblk); |
3081 | zero_ex.ee_len = cpu_to_le16(allocated); | |
667eff35 YY |
3082 | ext4_ext_store_pblock(&zero_ex, |
3083 | ext4_ext_pblock(ex) + map->m_lblk - ee_block); | |
3084 | err = ext4_ext_zeroout(inode, &zero_ex); | |
56055d3a AA |
3085 | if (err) |
3086 | goto out; | |
667eff35 YY |
3087 | split_map.m_lblk = map->m_lblk; |
3088 | split_map.m_len = allocated; | |
3089 | } else if ((map->m_lblk - ee_block + map->m_len < | |
3090 | EXT4_EXT_ZERO_LEN) && | |
3091 | (EXT4_EXT_MAY_ZEROOUT & split_flag)) { | |
3092 | /* case 2 */ | |
3093 | if (map->m_lblk != ee_block) { | |
3094 | zero_ex.ee_block = ex->ee_block; | |
3095 | zero_ex.ee_len = cpu_to_le16(map->m_lblk - | |
3096 | ee_block); | |
3097 | ext4_ext_store_pblock(&zero_ex, | |
3098 | ext4_ext_pblock(ex)); | |
3099 | err = ext4_ext_zeroout(inode, &zero_ex); | |
3100 | if (err) | |
3101 | goto out; | |
3102 | } | |
3103 | ||
667eff35 | 3104 | split_map.m_lblk = ee_block; |
9b940f8e AH |
3105 | split_map.m_len = map->m_lblk - ee_block + map->m_len; |
3106 | allocated = map->m_len; | |
56055d3a AA |
3107 | } |
3108 | } | |
667eff35 YY |
3109 | |
3110 | allocated = ext4_split_extent(handle, inode, path, | |
3111 | &split_map, split_flag, 0); | |
3112 | if (allocated < 0) | |
3113 | err = allocated; | |
3114 | ||
56055d3a AA |
3115 | out: |
3116 | return err ? err : allocated; | |
3117 | } | |
3118 | ||
0031462b | 3119 | /* |
e35fd660 | 3120 | * This function is called by ext4_ext_map_blocks() from |
0031462b MC |
3121 | * ext4_get_blocks_dio_write() when DIO to write |
3122 | * to an uninitialized extent. | |
3123 | * | |
fd018fe8 | 3124 | * Writing to an uninitialized extent may result in splitting the uninitialized |
b595076a | 3125 | * extent into multiple /initialized uninitialized extents (up to three) |
0031462b MC |
3126 | * There are three possibilities: |
3127 | * a> There is no split required: Entire extent should be uninitialized | |
3128 | * b> Splits in two extents: Write is happening at either end of the extent | |
3129 | * c> Splits in three extents: Somone is writing in middle of the extent | |
3130 | * | |
3131 | * One of more index blocks maybe needed if the extent tree grow after | |
b595076a | 3132 | * the uninitialized extent split. To prevent ENOSPC occur at the IO |
0031462b | 3133 | * complete, we need to split the uninitialized extent before DIO submit |
421f91d2 | 3134 | * the IO. The uninitialized extent called at this time will be split |
0031462b MC |
3135 | * into three uninitialized extent(at most). After IO complete, the part |
3136 | * being filled will be convert to initialized by the end_io callback function | |
3137 | * via ext4_convert_unwritten_extents(). | |
ba230c3f M |
3138 | * |
3139 | * Returns the size of uninitialized extent to be written on success. | |
0031462b MC |
3140 | */ |
3141 | static int ext4_split_unwritten_extents(handle_t *handle, | |
3142 | struct inode *inode, | |
e35fd660 | 3143 | struct ext4_map_blocks *map, |
0031462b | 3144 | struct ext4_ext_path *path, |
0031462b MC |
3145 | int flags) |
3146 | { | |
667eff35 YY |
3147 | ext4_lblk_t eof_block; |
3148 | ext4_lblk_t ee_block; | |
3149 | struct ext4_extent *ex; | |
3150 | unsigned int ee_len; | |
3151 | int split_flag = 0, depth; | |
21ca087a DM |
3152 | |
3153 | ext_debug("ext4_split_unwritten_extents: inode %lu, logical" | |
3154 | "block %llu, max_blocks %u\n", inode->i_ino, | |
e35fd660 | 3155 | (unsigned long long)map->m_lblk, map->m_len); |
21ca087a DM |
3156 | |
3157 | eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >> | |
3158 | inode->i_sb->s_blocksize_bits; | |
e35fd660 TT |
3159 | if (eof_block < map->m_lblk + map->m_len) |
3160 | eof_block = map->m_lblk + map->m_len; | |
21ca087a DM |
3161 | /* |
3162 | * It is safe to convert extent to initialized via explicit | |
3163 | * zeroout only if extent is fully insde i_size or new_size. | |
3164 | */ | |
667eff35 YY |
3165 | depth = ext_depth(inode); |
3166 | ex = path[depth].p_ext; | |
3167 | ee_block = le32_to_cpu(ex->ee_block); | |
3168 | ee_len = ext4_ext_get_actual_len(ex); | |
0031462b | 3169 | |
667eff35 YY |
3170 | split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0; |
3171 | split_flag |= EXT4_EXT_MARK_UNINIT2; | |
0031462b | 3172 | |
667eff35 YY |
3173 | flags |= EXT4_GET_BLOCKS_PRE_IO; |
3174 | return ext4_split_extent(handle, inode, path, map, split_flag, flags); | |
0031462b | 3175 | } |
197217a5 | 3176 | |
c7064ef1 | 3177 | static int ext4_convert_unwritten_extents_endio(handle_t *handle, |
0031462b MC |
3178 | struct inode *inode, |
3179 | struct ext4_ext_path *path) | |
3180 | { | |
3181 | struct ext4_extent *ex; | |
0031462b MC |
3182 | int depth; |
3183 | int err = 0; | |
0031462b MC |
3184 | |
3185 | depth = ext_depth(inode); | |
0031462b MC |
3186 | ex = path[depth].p_ext; |
3187 | ||
197217a5 YY |
3188 | ext_debug("ext4_convert_unwritten_extents_endio: inode %lu, logical" |
3189 | "block %llu, max_blocks %u\n", inode->i_ino, | |
3190 | (unsigned long long)le32_to_cpu(ex->ee_block), | |
3191 | ext4_ext_get_actual_len(ex)); | |
3192 | ||
0031462b MC |
3193 | err = ext4_ext_get_access(handle, inode, path + depth); |
3194 | if (err) | |
3195 | goto out; | |
3196 | /* first mark the extent as initialized */ | |
3197 | ext4_ext_mark_initialized(ex); | |
3198 | ||
197217a5 YY |
3199 | /* note: ext4_ext_correct_indexes() isn't needed here because |
3200 | * borders are not changed | |
0031462b | 3201 | */ |
197217a5 YY |
3202 | ext4_ext_try_to_merge(inode, path, ex); |
3203 | ||
0031462b MC |
3204 | /* Mark modified extent as dirty */ |
3205 | err = ext4_ext_dirty(handle, inode, path + depth); | |
3206 | out: | |
3207 | ext4_ext_show_leaf(inode, path); | |
3208 | return err; | |
3209 | } | |
3210 | ||
515f41c3 AK |
3211 | static void unmap_underlying_metadata_blocks(struct block_device *bdev, |
3212 | sector_t block, int count) | |
3213 | { | |
3214 | int i; | |
3215 | for (i = 0; i < count; i++) | |
3216 | unmap_underlying_metadata(bdev, block + i); | |
3217 | } | |
3218 | ||
58590b06 TT |
3219 | /* |
3220 | * Handle EOFBLOCKS_FL flag, clearing it if necessary | |
3221 | */ | |
3222 | static int check_eofblocks_fl(handle_t *handle, struct inode *inode, | |
d002ebf1 | 3223 | ext4_lblk_t lblk, |
58590b06 TT |
3224 | struct ext4_ext_path *path, |
3225 | unsigned int len) | |
3226 | { | |
3227 | int i, depth; | |
3228 | struct ext4_extent_header *eh; | |
65922cb5 | 3229 | struct ext4_extent *last_ex; |
58590b06 TT |
3230 | |
3231 | if (!ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS)) | |
3232 | return 0; | |
3233 | ||
3234 | depth = ext_depth(inode); | |
3235 | eh = path[depth].p_hdr; | |
58590b06 TT |
3236 | |
3237 | if (unlikely(!eh->eh_entries)) { | |
3238 | EXT4_ERROR_INODE(inode, "eh->eh_entries == 0 and " | |
3239 | "EOFBLOCKS_FL set"); | |
3240 | return -EIO; | |
3241 | } | |
3242 | last_ex = EXT_LAST_EXTENT(eh); | |
3243 | /* | |
3244 | * We should clear the EOFBLOCKS_FL flag if we are writing the | |
3245 | * last block in the last extent in the file. We test this by | |
3246 | * first checking to see if the caller to | |
3247 | * ext4_ext_get_blocks() was interested in the last block (or | |
3248 | * a block beyond the last block) in the current extent. If | |
3249 | * this turns out to be false, we can bail out from this | |
3250 | * function immediately. | |
3251 | */ | |
d002ebf1 | 3252 | if (lblk + len < le32_to_cpu(last_ex->ee_block) + |
58590b06 TT |
3253 | ext4_ext_get_actual_len(last_ex)) |
3254 | return 0; | |
3255 | /* | |
3256 | * If the caller does appear to be planning to write at or | |
3257 | * beyond the end of the current extent, we then test to see | |
3258 | * if the current extent is the last extent in the file, by | |
3259 | * checking to make sure it was reached via the rightmost node | |
3260 | * at each level of the tree. | |
3261 | */ | |
3262 | for (i = depth-1; i >= 0; i--) | |
3263 | if (path[i].p_idx != EXT_LAST_INDEX(path[i].p_hdr)) | |
3264 | return 0; | |
3265 | ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS); | |
3266 | return ext4_mark_inode_dirty(handle, inode); | |
3267 | } | |
3268 | ||
7b415bf6 AK |
3269 | /** |
3270 | * ext4_find_delalloc_range: find delayed allocated block in the given range. | |
3271 | * | |
3272 | * Goes through the buffer heads in the range [lblk_start, lblk_end] and returns | |
3273 | * whether there are any buffers marked for delayed allocation. It returns '1' | |
3274 | * on the first delalloc'ed buffer head found. If no buffer head in the given | |
3275 | * range is marked for delalloc, it returns 0. | |
3276 | * lblk_start should always be <= lblk_end. | |
3277 | * search_hint_reverse is to indicate that searching in reverse from lblk_end to | |
3278 | * lblk_start might be more efficient (i.e., we will likely hit the delalloc'ed | |
3279 | * block sooner). This is useful when blocks are truncated sequentially from | |
3280 | * lblk_start towards lblk_end. | |
3281 | */ | |
3282 | static int ext4_find_delalloc_range(struct inode *inode, | |
3283 | ext4_lblk_t lblk_start, | |
3284 | ext4_lblk_t lblk_end, | |
3285 | int search_hint_reverse) | |
3286 | { | |
3287 | struct address_space *mapping = inode->i_mapping; | |
3288 | struct buffer_head *head, *bh = NULL; | |
3289 | struct page *page; | |
3290 | ext4_lblk_t i, pg_lblk; | |
3291 | pgoff_t index; | |
3292 | ||
3293 | /* reverse search wont work if fs block size is less than page size */ | |
3294 | if (inode->i_blkbits < PAGE_CACHE_SHIFT) | |
3295 | search_hint_reverse = 0; | |
3296 | ||
3297 | if (search_hint_reverse) | |
3298 | i = lblk_end; | |
3299 | else | |
3300 | i = lblk_start; | |
3301 | ||
3302 | index = i >> (PAGE_CACHE_SHIFT - inode->i_blkbits); | |
3303 | ||
3304 | while ((i >= lblk_start) && (i <= lblk_end)) { | |
3305 | page = find_get_page(mapping, index); | |
5356f261 | 3306 | if (!page) |
7b415bf6 AK |
3307 | goto nextpage; |
3308 | ||
7b415bf6 AK |
3309 | if (!page_has_buffers(page)) |
3310 | goto nextpage; | |
3311 | ||
3312 | head = page_buffers(page); | |
3313 | if (!head) | |
3314 | goto nextpage; | |
3315 | ||
3316 | bh = head; | |
3317 | pg_lblk = index << (PAGE_CACHE_SHIFT - | |
3318 | inode->i_blkbits); | |
3319 | do { | |
3320 | if (unlikely(pg_lblk < lblk_start)) { | |
3321 | /* | |
3322 | * This is possible when fs block size is less | |
3323 | * than page size and our cluster starts/ends in | |
3324 | * middle of the page. So we need to skip the | |
3325 | * initial few blocks till we reach the 'lblk' | |
3326 | */ | |
3327 | pg_lblk++; | |
3328 | continue; | |
3329 | } | |
3330 | ||
5356f261 AK |
3331 | /* Check if the buffer is delayed allocated and that it |
3332 | * is not yet mapped. (when da-buffers are mapped during | |
3333 | * their writeout, their da_mapped bit is set.) | |
3334 | */ | |
3335 | if (buffer_delay(bh) && !buffer_da_mapped(bh)) { | |
7b415bf6 | 3336 | page_cache_release(page); |
d8990240 AK |
3337 | trace_ext4_find_delalloc_range(inode, |
3338 | lblk_start, lblk_end, | |
3339 | search_hint_reverse, | |
3340 | 1, i); | |
7b415bf6 AK |
3341 | return 1; |
3342 | } | |
3343 | if (search_hint_reverse) | |
3344 | i--; | |
3345 | else | |
3346 | i++; | |
3347 | } while ((i >= lblk_start) && (i <= lblk_end) && | |
3348 | ((bh = bh->b_this_page) != head)); | |
3349 | nextpage: | |
3350 | if (page) | |
3351 | page_cache_release(page); | |
3352 | /* | |
3353 | * Move to next page. 'i' will be the first lblk in the next | |
3354 | * page. | |
3355 | */ | |
3356 | if (search_hint_reverse) | |
3357 | index--; | |
3358 | else | |
3359 | index++; | |
3360 | i = index << (PAGE_CACHE_SHIFT - inode->i_blkbits); | |
3361 | } | |
3362 | ||
d8990240 AK |
3363 | trace_ext4_find_delalloc_range(inode, lblk_start, lblk_end, |
3364 | search_hint_reverse, 0, 0); | |
7b415bf6 AK |
3365 | return 0; |
3366 | } | |
3367 | ||
3368 | int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk, | |
3369 | int search_hint_reverse) | |
3370 | { | |
3371 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
3372 | ext4_lblk_t lblk_start, lblk_end; | |
3373 | lblk_start = lblk & (~(sbi->s_cluster_ratio - 1)); | |
3374 | lblk_end = lblk_start + sbi->s_cluster_ratio - 1; | |
3375 | ||
3376 | return ext4_find_delalloc_range(inode, lblk_start, lblk_end, | |
3377 | search_hint_reverse); | |
3378 | } | |
3379 | ||
3380 | /** | |
3381 | * Determines how many complete clusters (out of those specified by the 'map') | |
3382 | * are under delalloc and were reserved quota for. | |
3383 | * This function is called when we are writing out the blocks that were | |
3384 | * originally written with their allocation delayed, but then the space was | |
3385 | * allocated using fallocate() before the delayed allocation could be resolved. | |
3386 | * The cases to look for are: | |
3387 | * ('=' indicated delayed allocated blocks | |
3388 | * '-' indicates non-delayed allocated blocks) | |
3389 | * (a) partial clusters towards beginning and/or end outside of allocated range | |
3390 | * are not delalloc'ed. | |
3391 | * Ex: | |
3392 | * |----c---=|====c====|====c====|===-c----| | |
3393 | * |++++++ allocated ++++++| | |
3394 | * ==> 4 complete clusters in above example | |
3395 | * | |
3396 | * (b) partial cluster (outside of allocated range) towards either end is | |
3397 | * marked for delayed allocation. In this case, we will exclude that | |
3398 | * cluster. | |
3399 | * Ex: | |
3400 | * |----====c========|========c========| | |
3401 | * |++++++ allocated ++++++| | |
3402 | * ==> 1 complete clusters in above example | |
3403 | * | |
3404 | * Ex: | |
3405 | * |================c================| | |
3406 | * |++++++ allocated ++++++| | |
3407 | * ==> 0 complete clusters in above example | |
3408 | * | |
3409 | * The ext4_da_update_reserve_space will be called only if we | |
3410 | * determine here that there were some "entire" clusters that span | |
3411 | * this 'allocated' range. | |
3412 | * In the non-bigalloc case, this function will just end up returning num_blks | |
3413 | * without ever calling ext4_find_delalloc_range. | |
3414 | */ | |
3415 | static unsigned int | |
3416 | get_reserved_cluster_alloc(struct inode *inode, ext4_lblk_t lblk_start, | |
3417 | unsigned int num_blks) | |
3418 | { | |
3419 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
3420 | ext4_lblk_t alloc_cluster_start, alloc_cluster_end; | |
3421 | ext4_lblk_t lblk_from, lblk_to, c_offset; | |
3422 | unsigned int allocated_clusters = 0; | |
3423 | ||
3424 | alloc_cluster_start = EXT4_B2C(sbi, lblk_start); | |
3425 | alloc_cluster_end = EXT4_B2C(sbi, lblk_start + num_blks - 1); | |
3426 | ||
3427 | /* max possible clusters for this allocation */ | |
3428 | allocated_clusters = alloc_cluster_end - alloc_cluster_start + 1; | |
3429 | ||
d8990240 AK |
3430 | trace_ext4_get_reserved_cluster_alloc(inode, lblk_start, num_blks); |
3431 | ||
7b415bf6 AK |
3432 | /* Check towards left side */ |
3433 | c_offset = lblk_start & (sbi->s_cluster_ratio - 1); | |
3434 | if (c_offset) { | |
3435 | lblk_from = lblk_start & (~(sbi->s_cluster_ratio - 1)); | |
3436 | lblk_to = lblk_from + c_offset - 1; | |
3437 | ||
3438 | if (ext4_find_delalloc_range(inode, lblk_from, lblk_to, 0)) | |
3439 | allocated_clusters--; | |
3440 | } | |
3441 | ||
3442 | /* Now check towards right. */ | |
3443 | c_offset = (lblk_start + num_blks) & (sbi->s_cluster_ratio - 1); | |
3444 | if (allocated_clusters && c_offset) { | |
3445 | lblk_from = lblk_start + num_blks; | |
3446 | lblk_to = lblk_from + (sbi->s_cluster_ratio - c_offset) - 1; | |
3447 | ||
3448 | if (ext4_find_delalloc_range(inode, lblk_from, lblk_to, 0)) | |
3449 | allocated_clusters--; | |
3450 | } | |
3451 | ||
3452 | return allocated_clusters; | |
3453 | } | |
3454 | ||
0031462b MC |
3455 | static int |
3456 | ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode, | |
e35fd660 | 3457 | struct ext4_map_blocks *map, |
0031462b | 3458 | struct ext4_ext_path *path, int flags, |
e35fd660 | 3459 | unsigned int allocated, ext4_fsblk_t newblock) |
0031462b MC |
3460 | { |
3461 | int ret = 0; | |
3462 | int err = 0; | |
8d5d02e6 | 3463 | ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio; |
0031462b MC |
3464 | |
3465 | ext_debug("ext4_ext_handle_uninitialized_extents: inode %lu, logical" | |
3466 | "block %llu, max_blocks %u, flags %d, allocated %u", | |
e35fd660 | 3467 | inode->i_ino, (unsigned long long)map->m_lblk, map->m_len, |
0031462b MC |
3468 | flags, allocated); |
3469 | ext4_ext_show_leaf(inode, path); | |
3470 | ||
d8990240 AK |
3471 | trace_ext4_ext_handle_uninitialized_extents(inode, map, allocated, |
3472 | newblock); | |
3473 | ||
c7064ef1 | 3474 | /* get_block() before submit the IO, split the extent */ |
744692dc | 3475 | if ((flags & EXT4_GET_BLOCKS_PRE_IO)) { |
e35fd660 TT |
3476 | ret = ext4_split_unwritten_extents(handle, inode, map, |
3477 | path, flags); | |
5f524950 M |
3478 | /* |
3479 | * Flag the inode(non aio case) or end_io struct (aio case) | |
25985edc | 3480 | * that this IO needs to conversion to written when IO is |
5f524950 M |
3481 | * completed |
3482 | */ | |
b3ff0569 TM |
3483 | if (io) { |
3484 | if (!(io->flag & EXT4_IO_END_UNWRITTEN)) { | |
3485 | io->flag = EXT4_IO_END_UNWRITTEN; | |
3486 | atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten); | |
3487 | } | |
e9e3bcec | 3488 | } else |
19f5fb7a | 3489 | ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN); |
744692dc | 3490 | if (ext4_should_dioread_nolock(inode)) |
e35fd660 | 3491 | map->m_flags |= EXT4_MAP_UNINIT; |
0031462b MC |
3492 | goto out; |
3493 | } | |
c7064ef1 | 3494 | /* IO end_io complete, convert the filled extent to written */ |
744692dc | 3495 | if ((flags & EXT4_GET_BLOCKS_CONVERT)) { |
c7064ef1 | 3496 | ret = ext4_convert_unwritten_extents_endio(handle, inode, |
0031462b | 3497 | path); |
58590b06 | 3498 | if (ret >= 0) { |
b436b9be | 3499 | ext4_update_inode_fsync_trans(handle, inode, 1); |
d002ebf1 ES |
3500 | err = check_eofblocks_fl(handle, inode, map->m_lblk, |
3501 | path, map->m_len); | |
58590b06 TT |
3502 | } else |
3503 | err = ret; | |
0031462b MC |
3504 | goto out2; |
3505 | } | |
3506 | /* buffered IO case */ | |
3507 | /* | |
3508 | * repeat fallocate creation request | |
3509 | * we already have an unwritten extent | |
3510 | */ | |
3511 | if (flags & EXT4_GET_BLOCKS_UNINIT_EXT) | |
3512 | goto map_out; | |
3513 | ||
3514 | /* buffered READ or buffered write_begin() lookup */ | |
3515 | if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) { | |
3516 | /* | |
3517 | * We have blocks reserved already. We | |
3518 | * return allocated blocks so that delalloc | |
3519 | * won't do block reservation for us. But | |
3520 | * the buffer head will be unmapped so that | |
3521 | * a read from the block returns 0s. | |
3522 | */ | |
e35fd660 | 3523 | map->m_flags |= EXT4_MAP_UNWRITTEN; |
0031462b MC |
3524 | goto out1; |
3525 | } | |
3526 | ||
3527 | /* buffered write, writepage time, convert*/ | |
e35fd660 | 3528 | ret = ext4_ext_convert_to_initialized(handle, inode, map, path); |
a4e5d88b | 3529 | if (ret >= 0) |
b436b9be | 3530 | ext4_update_inode_fsync_trans(handle, inode, 1); |
0031462b MC |
3531 | out: |
3532 | if (ret <= 0) { | |
3533 | err = ret; | |
3534 | goto out2; | |
3535 | } else | |
3536 | allocated = ret; | |
e35fd660 | 3537 | map->m_flags |= EXT4_MAP_NEW; |
515f41c3 AK |
3538 | /* |
3539 | * if we allocated more blocks than requested | |
3540 | * we need to make sure we unmap the extra block | |
3541 | * allocated. The actual needed block will get | |
3542 | * unmapped later when we find the buffer_head marked | |
3543 | * new. | |
3544 | */ | |
e35fd660 | 3545 | if (allocated > map->m_len) { |
515f41c3 | 3546 | unmap_underlying_metadata_blocks(inode->i_sb->s_bdev, |
e35fd660 TT |
3547 | newblock + map->m_len, |
3548 | allocated - map->m_len); | |
3549 | allocated = map->m_len; | |
515f41c3 | 3550 | } |
5f634d06 AK |
3551 | |
3552 | /* | |
3553 | * If we have done fallocate with the offset that is already | |
3554 | * delayed allocated, we would have block reservation | |
3555 | * and quota reservation done in the delayed write path. | |
3556 | * But fallocate would have already updated quota and block | |
3557 | * count for this offset. So cancel these reservation | |
3558 | */ | |
7b415bf6 AK |
3559 | if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) { |
3560 | unsigned int reserved_clusters; | |
3561 | reserved_clusters = get_reserved_cluster_alloc(inode, | |
3562 | map->m_lblk, map->m_len); | |
3563 | if (reserved_clusters) | |
3564 | ext4_da_update_reserve_space(inode, | |
3565 | reserved_clusters, | |
3566 | 0); | |
3567 | } | |
5f634d06 | 3568 | |
0031462b | 3569 | map_out: |
e35fd660 | 3570 | map->m_flags |= EXT4_MAP_MAPPED; |
a4e5d88b DM |
3571 | if ((flags & EXT4_GET_BLOCKS_KEEP_SIZE) == 0) { |
3572 | err = check_eofblocks_fl(handle, inode, map->m_lblk, path, | |
3573 | map->m_len); | |
3574 | if (err < 0) | |
3575 | goto out2; | |
3576 | } | |
0031462b | 3577 | out1: |
e35fd660 TT |
3578 | if (allocated > map->m_len) |
3579 | allocated = map->m_len; | |
0031462b | 3580 | ext4_ext_show_leaf(inode, path); |
e35fd660 TT |
3581 | map->m_pblk = newblock; |
3582 | map->m_len = allocated; | |
0031462b MC |
3583 | out2: |
3584 | if (path) { | |
3585 | ext4_ext_drop_refs(path); | |
3586 | kfree(path); | |
3587 | } | |
3588 | return err ? err : allocated; | |
3589 | } | |
58590b06 | 3590 | |
4d33b1ef TT |
3591 | /* |
3592 | * get_implied_cluster_alloc - check to see if the requested | |
3593 | * allocation (in the map structure) overlaps with a cluster already | |
3594 | * allocated in an extent. | |
d8990240 | 3595 | * @sb The filesystem superblock structure |
4d33b1ef TT |
3596 | * @map The requested lblk->pblk mapping |
3597 | * @ex The extent structure which might contain an implied | |
3598 | * cluster allocation | |
3599 | * | |
3600 | * This function is called by ext4_ext_map_blocks() after we failed to | |
3601 | * find blocks that were already in the inode's extent tree. Hence, | |
3602 | * we know that the beginning of the requested region cannot overlap | |
3603 | * the extent from the inode's extent tree. There are three cases we | |
3604 | * want to catch. The first is this case: | |
3605 | * | |
3606 | * |--- cluster # N--| | |
3607 | * |--- extent ---| |---- requested region ---| | |
3608 | * |==========| | |
3609 | * | |
3610 | * The second case that we need to test for is this one: | |
3611 | * | |
3612 | * |--------- cluster # N ----------------| | |
3613 | * |--- requested region --| |------- extent ----| | |
3614 | * |=======================| | |
3615 | * | |
3616 | * The third case is when the requested region lies between two extents | |
3617 | * within the same cluster: | |
3618 | * |------------- cluster # N-------------| | |
3619 | * |----- ex -----| |---- ex_right ----| | |
3620 | * |------ requested region ------| | |
3621 | * |================| | |
3622 | * | |
3623 | * In each of the above cases, we need to set the map->m_pblk and | |
3624 | * map->m_len so it corresponds to the return the extent labelled as | |
3625 | * "|====|" from cluster #N, since it is already in use for data in | |
3626 | * cluster EXT4_B2C(sbi, map->m_lblk). We will then return 1 to | |
3627 | * signal to ext4_ext_map_blocks() that map->m_pblk should be treated | |
3628 | * as a new "allocated" block region. Otherwise, we will return 0 and | |
3629 | * ext4_ext_map_blocks() will then allocate one or more new clusters | |
3630 | * by calling ext4_mb_new_blocks(). | |
3631 | */ | |
d8990240 | 3632 | static int get_implied_cluster_alloc(struct super_block *sb, |
4d33b1ef TT |
3633 | struct ext4_map_blocks *map, |
3634 | struct ext4_extent *ex, | |
3635 | struct ext4_ext_path *path) | |
3636 | { | |
d8990240 | 3637 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
4d33b1ef TT |
3638 | ext4_lblk_t c_offset = map->m_lblk & (sbi->s_cluster_ratio-1); |
3639 | ext4_lblk_t ex_cluster_start, ex_cluster_end; | |
3640 | ext4_lblk_t rr_cluster_start, rr_cluster_end; | |
3641 | ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block); | |
3642 | ext4_fsblk_t ee_start = ext4_ext_pblock(ex); | |
3643 | unsigned short ee_len = ext4_ext_get_actual_len(ex); | |
3644 | ||
3645 | /* The extent passed in that we are trying to match */ | |
3646 | ex_cluster_start = EXT4_B2C(sbi, ee_block); | |
3647 | ex_cluster_end = EXT4_B2C(sbi, ee_block + ee_len - 1); | |
3648 | ||
3649 | /* The requested region passed into ext4_map_blocks() */ | |
3650 | rr_cluster_start = EXT4_B2C(sbi, map->m_lblk); | |
3651 | rr_cluster_end = EXT4_B2C(sbi, map->m_lblk + map->m_len - 1); | |
3652 | ||
3653 | if ((rr_cluster_start == ex_cluster_end) || | |
3654 | (rr_cluster_start == ex_cluster_start)) { | |
3655 | if (rr_cluster_start == ex_cluster_end) | |
3656 | ee_start += ee_len - 1; | |
3657 | map->m_pblk = (ee_start & ~(sbi->s_cluster_ratio - 1)) + | |
3658 | c_offset; | |
3659 | map->m_len = min(map->m_len, | |
3660 | (unsigned) sbi->s_cluster_ratio - c_offset); | |
3661 | /* | |
3662 | * Check for and handle this case: | |
3663 | * | |
3664 | * |--------- cluster # N-------------| | |
3665 | * |------- extent ----| | |
3666 | * |--- requested region ---| | |
3667 | * |===========| | |
3668 | */ | |
3669 | ||
3670 | if (map->m_lblk < ee_block) | |
3671 | map->m_len = min(map->m_len, ee_block - map->m_lblk); | |
3672 | ||
3673 | /* | |
3674 | * Check for the case where there is already another allocated | |
3675 | * block to the right of 'ex' but before the end of the cluster. | |
3676 | * | |
3677 | * |------------- cluster # N-------------| | |
3678 | * |----- ex -----| |---- ex_right ----| | |
3679 | * |------ requested region ------| | |
3680 | * |================| | |
3681 | */ | |
3682 | if (map->m_lblk > ee_block) { | |
3683 | ext4_lblk_t next = ext4_ext_next_allocated_block(path); | |
3684 | map->m_len = min(map->m_len, next - map->m_lblk); | |
3685 | } | |
d8990240 AK |
3686 | |
3687 | trace_ext4_get_implied_cluster_alloc_exit(sb, map, 1); | |
4d33b1ef TT |
3688 | return 1; |
3689 | } | |
d8990240 AK |
3690 | |
3691 | trace_ext4_get_implied_cluster_alloc_exit(sb, map, 0); | |
4d33b1ef TT |
3692 | return 0; |
3693 | } | |
3694 | ||
3695 | ||
c278bfec | 3696 | /* |
f5ab0d1f MC |
3697 | * Block allocation/map/preallocation routine for extents based files |
3698 | * | |
3699 | * | |
c278bfec | 3700 | * Need to be called with |
0e855ac8 AK |
3701 | * down_read(&EXT4_I(inode)->i_data_sem) if not allocating file system block |
3702 | * (ie, create is zero). Otherwise down_write(&EXT4_I(inode)->i_data_sem) | |
f5ab0d1f MC |
3703 | * |
3704 | * return > 0, number of of blocks already mapped/allocated | |
3705 | * if create == 0 and these are pre-allocated blocks | |
3706 | * buffer head is unmapped | |
3707 | * otherwise blocks are mapped | |
3708 | * | |
3709 | * return = 0, if plain look up failed (blocks have not been allocated) | |
3710 | * buffer head is unmapped | |
3711 | * | |
3712 | * return < 0, error case. | |
c278bfec | 3713 | */ |
e35fd660 TT |
3714 | int ext4_ext_map_blocks(handle_t *handle, struct inode *inode, |
3715 | struct ext4_map_blocks *map, int flags) | |
a86c6181 AT |
3716 | { |
3717 | struct ext4_ext_path *path = NULL; | |
4d33b1ef TT |
3718 | struct ext4_extent newex, *ex, *ex2; |
3719 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
0562e0ba | 3720 | ext4_fsblk_t newblock = 0; |
4d33b1ef TT |
3721 | int free_on_err = 0, err = 0, depth, ret; |
3722 | unsigned int allocated = 0, offset = 0; | |
81fdbb4a | 3723 | unsigned int allocated_clusters = 0; |
e861304b AH |
3724 | unsigned int punched_out = 0; |
3725 | unsigned int result = 0; | |
c9de560d | 3726 | struct ext4_allocation_request ar; |
8d5d02e6 | 3727 | ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio; |
4d33b1ef | 3728 | ext4_lblk_t cluster_offset; |
a86c6181 | 3729 | |
84fe3bef | 3730 | ext_debug("blocks %u/%u requested for inode %lu\n", |
e35fd660 | 3731 | map->m_lblk, map->m_len, inode->i_ino); |
0562e0ba | 3732 | trace_ext4_ext_map_blocks_enter(inode, map->m_lblk, map->m_len, flags); |
a86c6181 AT |
3733 | |
3734 | /* check in cache */ | |
015861ba RD |
3735 | if (!(flags & EXT4_GET_BLOCKS_PUNCH_OUT_EXT) && |
3736 | ext4_ext_in_cache(inode, map->m_lblk, &newex)) { | |
b05e6ae5 | 3737 | if (!newex.ee_start_lo && !newex.ee_start_hi) { |
7b415bf6 AK |
3738 | if ((sbi->s_cluster_ratio > 1) && |
3739 | ext4_find_delalloc_cluster(inode, map->m_lblk, 0)) | |
3740 | map->m_flags |= EXT4_MAP_FROM_CLUSTER; | |
3741 | ||
c2177057 | 3742 | if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) { |
56055d3a AA |
3743 | /* |
3744 | * block isn't allocated yet and | |
3745 | * user doesn't want to allocate it | |
3746 | */ | |
a86c6181 AT |
3747 | goto out2; |
3748 | } | |
3749 | /* we should allocate requested block */ | |
b05e6ae5 | 3750 | } else { |
a86c6181 | 3751 | /* block is already allocated */ |
7b415bf6 AK |
3752 | if (sbi->s_cluster_ratio > 1) |
3753 | map->m_flags |= EXT4_MAP_FROM_CLUSTER; | |
e35fd660 | 3754 | newblock = map->m_lblk |
8c55e204 | 3755 | - le32_to_cpu(newex.ee_block) |
bf89d16f | 3756 | + ext4_ext_pblock(&newex); |
d0d856e8 | 3757 | /* number of remaining blocks in the extent */ |
b939e376 | 3758 | allocated = ext4_ext_get_actual_len(&newex) - |
e35fd660 | 3759 | (map->m_lblk - le32_to_cpu(newex.ee_block)); |
a86c6181 | 3760 | goto out; |
a86c6181 AT |
3761 | } |
3762 | } | |
3763 | ||
3764 | /* find extent for this block */ | |
e35fd660 | 3765 | path = ext4_ext_find_extent(inode, map->m_lblk, NULL); |
a86c6181 AT |
3766 | if (IS_ERR(path)) { |
3767 | err = PTR_ERR(path); | |
3768 | path = NULL; | |
3769 | goto out2; | |
3770 | } | |
3771 | ||
3772 | depth = ext_depth(inode); | |
3773 | ||
3774 | /* | |
d0d856e8 RD |
3775 | * consistent leaf must not be empty; |
3776 | * this situation is possible, though, _during_ tree modification; | |
a86c6181 AT |
3777 | * this is why assert can't be put in ext4_ext_find_extent() |
3778 | */ | |
273df556 FM |
3779 | if (unlikely(path[depth].p_ext == NULL && depth != 0)) { |
3780 | EXT4_ERROR_INODE(inode, "bad extent address " | |
f70f362b TT |
3781 | "lblock: %lu, depth: %d pblock %lld", |
3782 | (unsigned long) map->m_lblk, depth, | |
3783 | path[depth].p_block); | |
034fb4c9 SP |
3784 | err = -EIO; |
3785 | goto out2; | |
3786 | } | |
a86c6181 | 3787 | |
7e028976 AM |
3788 | ex = path[depth].p_ext; |
3789 | if (ex) { | |
725d26d3 | 3790 | ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block); |
bf89d16f | 3791 | ext4_fsblk_t ee_start = ext4_ext_pblock(ex); |
a2df2a63 | 3792 | unsigned short ee_len; |
471d4011 SB |
3793 | |
3794 | /* | |
471d4011 | 3795 | * Uninitialized extents are treated as holes, except that |
56055d3a | 3796 | * we split out initialized portions during a write. |
471d4011 | 3797 | */ |
a2df2a63 | 3798 | ee_len = ext4_ext_get_actual_len(ex); |
d8990240 AK |
3799 | |
3800 | trace_ext4_ext_show_extent(inode, ee_block, ee_start, ee_len); | |
3801 | ||
d0d856e8 | 3802 | /* if found extent covers block, simply return it */ |
e35fd660 | 3803 | if (in_range(map->m_lblk, ee_block, ee_len)) { |
81fdbb4a | 3804 | struct ext4_map_blocks punch_map; |
0aa06000 TT |
3805 | ext4_fsblk_t partial_cluster = 0; |
3806 | ||
e35fd660 | 3807 | newblock = map->m_lblk - ee_block + ee_start; |
d0d856e8 | 3808 | /* number of remaining blocks in the extent */ |
e35fd660 TT |
3809 | allocated = ee_len - (map->m_lblk - ee_block); |
3810 | ext_debug("%u fit into %u:%d -> %llu\n", map->m_lblk, | |
3811 | ee_block, ee_len, newblock); | |
56055d3a | 3812 | |
e861304b AH |
3813 | if ((flags & EXT4_GET_BLOCKS_PUNCH_OUT_EXT) == 0) { |
3814 | /* | |
3815 | * Do not put uninitialized extent | |
3816 | * in the cache | |
3817 | */ | |
3818 | if (!ext4_ext_is_uninitialized(ex)) { | |
3819 | ext4_ext_put_in_cache(inode, ee_block, | |
3820 | ee_len, ee_start); | |
3821 | goto out; | |
3822 | } | |
3823 | ret = ext4_ext_handle_uninitialized_extents( | |
3824 | handle, inode, map, path, flags, | |
3825 | allocated, newblock); | |
3826 | return ret; | |
56055d3a | 3827 | } |
e861304b AH |
3828 | |
3829 | /* | |
3830 | * Punch out the map length, but only to the | |
3831 | * end of the extent | |
3832 | */ | |
3833 | punched_out = allocated < map->m_len ? | |
3834 | allocated : map->m_len; | |
3835 | ||
3836 | /* | |
3837 | * Sense extents need to be converted to | |
3838 | * uninitialized, they must fit in an | |
3839 | * uninitialized extent | |
3840 | */ | |
3841 | if (punched_out > EXT_UNINIT_MAX_LEN) | |
3842 | punched_out = EXT_UNINIT_MAX_LEN; | |
3843 | ||
3844 | punch_map.m_lblk = map->m_lblk; | |
3845 | punch_map.m_pblk = newblock; | |
3846 | punch_map.m_len = punched_out; | |
3847 | punch_map.m_flags = 0; | |
3848 | ||
3849 | /* Check to see if the extent needs to be split */ | |
3850 | if (punch_map.m_len != ee_len || | |
3851 | punch_map.m_lblk != ee_block) { | |
3852 | ||
3853 | ret = ext4_split_extent(handle, inode, | |
3854 | path, &punch_map, 0, | |
3855 | EXT4_GET_BLOCKS_PUNCH_OUT_EXT | | |
3856 | EXT4_GET_BLOCKS_PRE_IO); | |
3857 | ||
3858 | if (ret < 0) { | |
3859 | err = ret; | |
3860 | goto out2; | |
3861 | } | |
3862 | /* | |
3863 | * find extent for the block at | |
3864 | * the start of the hole | |
3865 | */ | |
3866 | ext4_ext_drop_refs(path); | |
3867 | kfree(path); | |
3868 | ||
3869 | path = ext4_ext_find_extent(inode, | |
3870 | map->m_lblk, NULL); | |
3871 | if (IS_ERR(path)) { | |
3872 | err = PTR_ERR(path); | |
3873 | path = NULL; | |
3874 | goto out2; | |
3875 | } | |
3876 | ||
3877 | depth = ext_depth(inode); | |
3878 | ex = path[depth].p_ext; | |
3879 | ee_len = ext4_ext_get_actual_len(ex); | |
3880 | ee_block = le32_to_cpu(ex->ee_block); | |
3881 | ee_start = ext4_ext_pblock(ex); | |
3882 | ||
3883 | } | |
3884 | ||
3885 | ext4_ext_mark_uninitialized(ex); | |
3886 | ||
f7d0d379 AH |
3887 | ext4_ext_invalidate_cache(inode); |
3888 | ||
3889 | err = ext4_ext_rm_leaf(handle, inode, path, | |
0aa06000 TT |
3890 | &partial_cluster, map->m_lblk, |
3891 | map->m_lblk + punched_out); | |
f7d0d379 AH |
3892 | |
3893 | if (!err && path->p_hdr->eh_entries == 0) { | |
3894 | /* | |
3895 | * Punch hole freed all of this sub tree, | |
3896 | * so we need to correct eh_depth | |
3897 | */ | |
3898 | err = ext4_ext_get_access(handle, inode, path); | |
3899 | if (err == 0) { | |
3900 | ext_inode_hdr(inode)->eh_depth = 0; | |
3901 | ext_inode_hdr(inode)->eh_max = | |
3902 | cpu_to_le16(ext4_ext_space_root( | |
3903 | inode, 0)); | |
3904 | ||
3905 | err = ext4_ext_dirty( | |
3906 | handle, inode, path); | |
3907 | } | |
3908 | } | |
e861304b AH |
3909 | |
3910 | goto out2; | |
a86c6181 AT |
3911 | } |
3912 | } | |
3913 | ||
7b415bf6 AK |
3914 | if ((sbi->s_cluster_ratio > 1) && |
3915 | ext4_find_delalloc_cluster(inode, map->m_lblk, 0)) | |
3916 | map->m_flags |= EXT4_MAP_FROM_CLUSTER; | |
3917 | ||
a86c6181 | 3918 | /* |
d0d856e8 | 3919 | * requested block isn't allocated yet; |
a86c6181 AT |
3920 | * we couldn't try to create block if create flag is zero |
3921 | */ | |
c2177057 | 3922 | if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) { |
56055d3a AA |
3923 | /* |
3924 | * put just found gap into cache to speed up | |
3925 | * subsequent requests | |
3926 | */ | |
e35fd660 | 3927 | ext4_ext_put_gap_in_cache(inode, path, map->m_lblk); |
a86c6181 AT |
3928 | goto out2; |
3929 | } | |
4d33b1ef | 3930 | |
a86c6181 | 3931 | /* |
c2ea3fde | 3932 | * Okay, we need to do block allocation. |
63f57933 | 3933 | */ |
7b415bf6 | 3934 | map->m_flags &= ~EXT4_MAP_FROM_CLUSTER; |
4d33b1ef TT |
3935 | newex.ee_block = cpu_to_le32(map->m_lblk); |
3936 | cluster_offset = map->m_lblk & (sbi->s_cluster_ratio-1); | |
3937 | ||
3938 | /* | |
3939 | * If we are doing bigalloc, check to see if the extent returned | |
3940 | * by ext4_ext_find_extent() implies a cluster we can use. | |
3941 | */ | |
3942 | if (cluster_offset && ex && | |
d8990240 | 3943 | get_implied_cluster_alloc(inode->i_sb, map, ex, path)) { |
4d33b1ef TT |
3944 | ar.len = allocated = map->m_len; |
3945 | newblock = map->m_pblk; | |
7b415bf6 | 3946 | map->m_flags |= EXT4_MAP_FROM_CLUSTER; |
4d33b1ef TT |
3947 | goto got_allocated_blocks; |
3948 | } | |
a86c6181 | 3949 | |
c9de560d | 3950 | /* find neighbour allocated blocks */ |
e35fd660 | 3951 | ar.lleft = map->m_lblk; |
c9de560d AT |
3952 | err = ext4_ext_search_left(inode, path, &ar.lleft, &ar.pleft); |
3953 | if (err) | |
3954 | goto out2; | |
e35fd660 | 3955 | ar.lright = map->m_lblk; |
4d33b1ef TT |
3956 | ex2 = NULL; |
3957 | err = ext4_ext_search_right(inode, path, &ar.lright, &ar.pright, &ex2); | |
c9de560d AT |
3958 | if (err) |
3959 | goto out2; | |
25d14f98 | 3960 | |
4d33b1ef TT |
3961 | /* Check if the extent after searching to the right implies a |
3962 | * cluster we can use. */ | |
3963 | if ((sbi->s_cluster_ratio > 1) && ex2 && | |
d8990240 | 3964 | get_implied_cluster_alloc(inode->i_sb, map, ex2, path)) { |
4d33b1ef TT |
3965 | ar.len = allocated = map->m_len; |
3966 | newblock = map->m_pblk; | |
7b415bf6 | 3967 | map->m_flags |= EXT4_MAP_FROM_CLUSTER; |
4d33b1ef TT |
3968 | goto got_allocated_blocks; |
3969 | } | |
3970 | ||
749269fa AA |
3971 | /* |
3972 | * See if request is beyond maximum number of blocks we can have in | |
3973 | * a single extent. For an initialized extent this limit is | |
3974 | * EXT_INIT_MAX_LEN and for an uninitialized extent this limit is | |
3975 | * EXT_UNINIT_MAX_LEN. | |
3976 | */ | |
e35fd660 | 3977 | if (map->m_len > EXT_INIT_MAX_LEN && |
c2177057 | 3978 | !(flags & EXT4_GET_BLOCKS_UNINIT_EXT)) |
e35fd660 TT |
3979 | map->m_len = EXT_INIT_MAX_LEN; |
3980 | else if (map->m_len > EXT_UNINIT_MAX_LEN && | |
c2177057 | 3981 | (flags & EXT4_GET_BLOCKS_UNINIT_EXT)) |
e35fd660 | 3982 | map->m_len = EXT_UNINIT_MAX_LEN; |
749269fa | 3983 | |
e35fd660 | 3984 | /* Check if we can really insert (m_lblk)::(m_lblk + m_len) extent */ |
e35fd660 | 3985 | newex.ee_len = cpu_to_le16(map->m_len); |
4d33b1ef | 3986 | err = ext4_ext_check_overlap(sbi, inode, &newex, path); |
25d14f98 | 3987 | if (err) |
b939e376 | 3988 | allocated = ext4_ext_get_actual_len(&newex); |
25d14f98 | 3989 | else |
e35fd660 | 3990 | allocated = map->m_len; |
c9de560d AT |
3991 | |
3992 | /* allocate new block */ | |
3993 | ar.inode = inode; | |
e35fd660 TT |
3994 | ar.goal = ext4_ext_find_goal(inode, path, map->m_lblk); |
3995 | ar.logical = map->m_lblk; | |
4d33b1ef TT |
3996 | /* |
3997 | * We calculate the offset from the beginning of the cluster | |
3998 | * for the logical block number, since when we allocate a | |
3999 | * physical cluster, the physical block should start at the | |
4000 | * same offset from the beginning of the cluster. This is | |
4001 | * needed so that future calls to get_implied_cluster_alloc() | |
4002 | * work correctly. | |
4003 | */ | |
4004 | offset = map->m_lblk & (sbi->s_cluster_ratio - 1); | |
4005 | ar.len = EXT4_NUM_B2C(sbi, offset+allocated); | |
4006 | ar.goal -= offset; | |
4007 | ar.logical -= offset; | |
c9de560d AT |
4008 | if (S_ISREG(inode->i_mode)) |
4009 | ar.flags = EXT4_MB_HINT_DATA; | |
4010 | else | |
4011 | /* disable in-core preallocation for non-regular files */ | |
4012 | ar.flags = 0; | |
556b27ab VH |
4013 | if (flags & EXT4_GET_BLOCKS_NO_NORMALIZE) |
4014 | ar.flags |= EXT4_MB_HINT_NOPREALLOC; | |
c9de560d | 4015 | newblock = ext4_mb_new_blocks(handle, &ar, &err); |
a86c6181 AT |
4016 | if (!newblock) |
4017 | goto out2; | |
84fe3bef | 4018 | ext_debug("allocate new block: goal %llu, found %llu/%u\n", |
498e5f24 | 4019 | ar.goal, newblock, allocated); |
4d33b1ef | 4020 | free_on_err = 1; |
7b415bf6 | 4021 | allocated_clusters = ar.len; |
4d33b1ef TT |
4022 | ar.len = EXT4_C2B(sbi, ar.len) - offset; |
4023 | if (ar.len > allocated) | |
4024 | ar.len = allocated; | |
a86c6181 | 4025 | |
4d33b1ef | 4026 | got_allocated_blocks: |
a86c6181 | 4027 | /* try to insert new extent into found leaf and return */ |
4d33b1ef | 4028 | ext4_ext_store_pblock(&newex, newblock + offset); |
c9de560d | 4029 | newex.ee_len = cpu_to_le16(ar.len); |
8d5d02e6 MC |
4030 | /* Mark uninitialized */ |
4031 | if (flags & EXT4_GET_BLOCKS_UNINIT_EXT){ | |
a2df2a63 | 4032 | ext4_ext_mark_uninitialized(&newex); |
8d5d02e6 | 4033 | /* |
744692dc | 4034 | * io_end structure was created for every IO write to an |
25985edc | 4035 | * uninitialized extent. To avoid unnecessary conversion, |
744692dc | 4036 | * here we flag the IO that really needs the conversion. |
5f524950 | 4037 | * For non asycn direct IO case, flag the inode state |
25985edc | 4038 | * that we need to perform conversion when IO is done. |
8d5d02e6 | 4039 | */ |
744692dc | 4040 | if ((flags & EXT4_GET_BLOCKS_PRE_IO)) { |
b3ff0569 TM |
4041 | if (io) { |
4042 | if (!(io->flag & EXT4_IO_END_UNWRITTEN)) { | |
4043 | io->flag = EXT4_IO_END_UNWRITTEN; | |
4044 | atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten); | |
4045 | } | |
e9e3bcec | 4046 | } else |
19f5fb7a TT |
4047 | ext4_set_inode_state(inode, |
4048 | EXT4_STATE_DIO_UNWRITTEN); | |
5f524950 | 4049 | } |
744692dc | 4050 | if (ext4_should_dioread_nolock(inode)) |
e35fd660 | 4051 | map->m_flags |= EXT4_MAP_UNINIT; |
8d5d02e6 | 4052 | } |
c8d46e41 | 4053 | |
a4e5d88b DM |
4054 | err = 0; |
4055 | if ((flags & EXT4_GET_BLOCKS_KEEP_SIZE) == 0) | |
4056 | err = check_eofblocks_fl(handle, inode, map->m_lblk, | |
4057 | path, ar.len); | |
575a1d4b JZ |
4058 | if (!err) |
4059 | err = ext4_ext_insert_extent(handle, inode, path, | |
4060 | &newex, flags); | |
4d33b1ef | 4061 | if (err && free_on_err) { |
7132de74 MP |
4062 | int fb_flags = flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE ? |
4063 | EXT4_FREE_BLOCKS_NO_QUOT_UPDATE : 0; | |
315054f0 | 4064 | /* free data blocks we just allocated */ |
c9de560d AT |
4065 | /* not a good idea to call discard here directly, |
4066 | * but otherwise we'd need to call it every free() */ | |
c2ea3fde | 4067 | ext4_discard_preallocations(inode); |
7dc57615 | 4068 | ext4_free_blocks(handle, inode, NULL, ext4_ext_pblock(&newex), |
7132de74 | 4069 | ext4_ext_get_actual_len(&newex), fb_flags); |
a86c6181 | 4070 | goto out2; |
315054f0 | 4071 | } |
a86c6181 | 4072 | |
a86c6181 | 4073 | /* previous routine could use block we allocated */ |
bf89d16f | 4074 | newblock = ext4_ext_pblock(&newex); |
b939e376 | 4075 | allocated = ext4_ext_get_actual_len(&newex); |
e35fd660 TT |
4076 | if (allocated > map->m_len) |
4077 | allocated = map->m_len; | |
4078 | map->m_flags |= EXT4_MAP_NEW; | |
a86c6181 | 4079 | |
5f634d06 AK |
4080 | /* |
4081 | * Update reserved blocks/metadata blocks after successful | |
4082 | * block allocation which had been deferred till now. | |
4083 | */ | |
7b415bf6 | 4084 | if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) { |
81fdbb4a | 4085 | unsigned int reserved_clusters; |
7b415bf6 | 4086 | /* |
81fdbb4a | 4087 | * Check how many clusters we had reserved this allocated range |
7b415bf6 AK |
4088 | */ |
4089 | reserved_clusters = get_reserved_cluster_alloc(inode, | |
4090 | map->m_lblk, allocated); | |
4091 | if (map->m_flags & EXT4_MAP_FROM_CLUSTER) { | |
4092 | if (reserved_clusters) { | |
4093 | /* | |
4094 | * We have clusters reserved for this range. | |
4095 | * But since we are not doing actual allocation | |
4096 | * and are simply using blocks from previously | |
4097 | * allocated cluster, we should release the | |
4098 | * reservation and not claim quota. | |
4099 | */ | |
4100 | ext4_da_update_reserve_space(inode, | |
4101 | reserved_clusters, 0); | |
4102 | } | |
4103 | } else { | |
4104 | BUG_ON(allocated_clusters < reserved_clusters); | |
4105 | /* We will claim quota for all newly allocated blocks.*/ | |
4106 | ext4_da_update_reserve_space(inode, allocated_clusters, | |
4107 | 1); | |
4108 | if (reserved_clusters < allocated_clusters) { | |
5356f261 | 4109 | struct ext4_inode_info *ei = EXT4_I(inode); |
7b415bf6 AK |
4110 | int reservation = allocated_clusters - |
4111 | reserved_clusters; | |
4112 | /* | |
4113 | * It seems we claimed few clusters outside of | |
4114 | * the range of this allocation. We should give | |
4115 | * it back to the reservation pool. This can | |
4116 | * happen in the following case: | |
4117 | * | |
4118 | * * Suppose s_cluster_ratio is 4 (i.e., each | |
4119 | * cluster has 4 blocks. Thus, the clusters | |
4120 | * are [0-3],[4-7],[8-11]... | |
4121 | * * First comes delayed allocation write for | |
4122 | * logical blocks 10 & 11. Since there were no | |
4123 | * previous delayed allocated blocks in the | |
4124 | * range [8-11], we would reserve 1 cluster | |
4125 | * for this write. | |
4126 | * * Next comes write for logical blocks 3 to 8. | |
4127 | * In this case, we will reserve 2 clusters | |
4128 | * (for [0-3] and [4-7]; and not for [8-11] as | |
4129 | * that range has a delayed allocated blocks. | |
4130 | * Thus total reserved clusters now becomes 3. | |
4131 | * * Now, during the delayed allocation writeout | |
4132 | * time, we will first write blocks [3-8] and | |
4133 | * allocate 3 clusters for writing these | |
4134 | * blocks. Also, we would claim all these | |
4135 | * three clusters above. | |
4136 | * * Now when we come here to writeout the | |
4137 | * blocks [10-11], we would expect to claim | |
4138 | * the reservation of 1 cluster we had made | |
4139 | * (and we would claim it since there are no | |
4140 | * more delayed allocated blocks in the range | |
4141 | * [8-11]. But our reserved cluster count had | |
4142 | * already gone to 0. | |
4143 | * | |
4144 | * Thus, at the step 4 above when we determine | |
4145 | * that there are still some unwritten delayed | |
4146 | * allocated blocks outside of our current | |
4147 | * block range, we should increment the | |
4148 | * reserved clusters count so that when the | |
4149 | * remaining blocks finally gets written, we | |
4150 | * could claim them. | |
4151 | */ | |
5356f261 AK |
4152 | dquot_reserve_block(inode, |
4153 | EXT4_C2B(sbi, reservation)); | |
4154 | spin_lock(&ei->i_block_reservation_lock); | |
4155 | ei->i_reserved_data_blocks += reservation; | |
4156 | spin_unlock(&ei->i_block_reservation_lock); | |
7b415bf6 AK |
4157 | } |
4158 | } | |
4159 | } | |
5f634d06 | 4160 | |
b436b9be JK |
4161 | /* |
4162 | * Cache the extent and update transaction to commit on fdatasync only | |
4163 | * when it is _not_ an uninitialized extent. | |
4164 | */ | |
4165 | if ((flags & EXT4_GET_BLOCKS_UNINIT_EXT) == 0) { | |
b05e6ae5 | 4166 | ext4_ext_put_in_cache(inode, map->m_lblk, allocated, newblock); |
b436b9be JK |
4167 | ext4_update_inode_fsync_trans(handle, inode, 1); |
4168 | } else | |
4169 | ext4_update_inode_fsync_trans(handle, inode, 0); | |
a86c6181 | 4170 | out: |
e35fd660 TT |
4171 | if (allocated > map->m_len) |
4172 | allocated = map->m_len; | |
a86c6181 | 4173 | ext4_ext_show_leaf(inode, path); |
e35fd660 TT |
4174 | map->m_flags |= EXT4_MAP_MAPPED; |
4175 | map->m_pblk = newblock; | |
4176 | map->m_len = allocated; | |
a86c6181 AT |
4177 | out2: |
4178 | if (path) { | |
4179 | ext4_ext_drop_refs(path); | |
4180 | kfree(path); | |
4181 | } | |
0562e0ba JZ |
4182 | trace_ext4_ext_map_blocks_exit(inode, map->m_lblk, |
4183 | newblock, map->m_len, err ? err : allocated); | |
e861304b AH |
4184 | |
4185 | result = (flags & EXT4_GET_BLOCKS_PUNCH_OUT_EXT) ? | |
4186 | punched_out : allocated; | |
4187 | ||
4188 | return err ? err : result; | |
a86c6181 AT |
4189 | } |
4190 | ||
cf108bca | 4191 | void ext4_ext_truncate(struct inode *inode) |
a86c6181 AT |
4192 | { |
4193 | struct address_space *mapping = inode->i_mapping; | |
4194 | struct super_block *sb = inode->i_sb; | |
725d26d3 | 4195 | ext4_lblk_t last_block; |
a86c6181 | 4196 | handle_t *handle; |
189e868f | 4197 | loff_t page_len; |
a86c6181 AT |
4198 | int err = 0; |
4199 | ||
3889fd57 JZ |
4200 | /* |
4201 | * finish any pending end_io work so we won't run the risk of | |
4202 | * converting any truncated blocks to initialized later | |
4203 | */ | |
4204 | ext4_flush_completed_IO(inode); | |
4205 | ||
a86c6181 AT |
4206 | /* |
4207 | * probably first extent we're gonna free will be last in block | |
4208 | */ | |
f3bd1f3f | 4209 | err = ext4_writepage_trans_blocks(inode); |
a86c6181 | 4210 | handle = ext4_journal_start(inode, err); |
cf108bca | 4211 | if (IS_ERR(handle)) |
a86c6181 | 4212 | return; |
a86c6181 | 4213 | |
189e868f AH |
4214 | if (inode->i_size % PAGE_CACHE_SIZE != 0) { |
4215 | page_len = PAGE_CACHE_SIZE - | |
4216 | (inode->i_size & (PAGE_CACHE_SIZE - 1)); | |
4217 | ||
4218 | err = ext4_discard_partial_page_buffers(handle, | |
4219 | mapping, inode->i_size, page_len, 0); | |
4220 | ||
4221 | if (err) | |
4222 | goto out_stop; | |
4223 | } | |
a86c6181 | 4224 | |
9ddfc3dc JK |
4225 | if (ext4_orphan_add(handle, inode)) |
4226 | goto out_stop; | |
4227 | ||
0e855ac8 | 4228 | down_write(&EXT4_I(inode)->i_data_sem); |
a86c6181 AT |
4229 | ext4_ext_invalidate_cache(inode); |
4230 | ||
c2ea3fde | 4231 | ext4_discard_preallocations(inode); |
c9de560d | 4232 | |
a86c6181 | 4233 | /* |
d0d856e8 RD |
4234 | * TODO: optimization is possible here. |
4235 | * Probably we need not scan at all, | |
4236 | * because page truncation is enough. | |
a86c6181 | 4237 | */ |
a86c6181 AT |
4238 | |
4239 | /* we have to know where to truncate from in crash case */ | |
4240 | EXT4_I(inode)->i_disksize = inode->i_size; | |
4241 | ext4_mark_inode_dirty(handle, inode); | |
4242 | ||
4243 | last_block = (inode->i_size + sb->s_blocksize - 1) | |
4244 | >> EXT4_BLOCK_SIZE_BITS(sb); | |
c6a0371c | 4245 | err = ext4_ext_remove_space(inode, last_block); |
a86c6181 AT |
4246 | |
4247 | /* In a multi-transaction truncate, we only make the final | |
56055d3a AA |
4248 | * transaction synchronous. |
4249 | */ | |
a86c6181 | 4250 | if (IS_SYNC(inode)) |
0390131b | 4251 | ext4_handle_sync(handle); |
a86c6181 | 4252 | |
9ddfc3dc | 4253 | up_write(&EXT4_I(inode)->i_data_sem); |
f6d2f6b3 EG |
4254 | |
4255 | out_stop: | |
a86c6181 | 4256 | /* |
d0d856e8 | 4257 | * If this was a simple ftruncate() and the file will remain alive, |
a86c6181 AT |
4258 | * then we need to clear up the orphan record which we created above. |
4259 | * However, if this was a real unlink then we were called by | |
4260 | * ext4_delete_inode(), and we allow that function to clean up the | |
4261 | * orphan info for us. | |
4262 | */ | |
4263 | if (inode->i_nlink) | |
4264 | ext4_orphan_del(handle, inode); | |
4265 | ||
ef737728 SR |
4266 | inode->i_mtime = inode->i_ctime = ext4_current_time(inode); |
4267 | ext4_mark_inode_dirty(handle, inode); | |
a86c6181 AT |
4268 | ext4_journal_stop(handle); |
4269 | } | |
4270 | ||
fd28784a AK |
4271 | static void ext4_falloc_update_inode(struct inode *inode, |
4272 | int mode, loff_t new_size, int update_ctime) | |
4273 | { | |
4274 | struct timespec now; | |
4275 | ||
4276 | if (update_ctime) { | |
4277 | now = current_fs_time(inode->i_sb); | |
4278 | if (!timespec_equal(&inode->i_ctime, &now)) | |
4279 | inode->i_ctime = now; | |
4280 | } | |
4281 | /* | |
4282 | * Update only when preallocation was requested beyond | |
4283 | * the file size. | |
4284 | */ | |
cf17fea6 AK |
4285 | if (!(mode & FALLOC_FL_KEEP_SIZE)) { |
4286 | if (new_size > i_size_read(inode)) | |
4287 | i_size_write(inode, new_size); | |
4288 | if (new_size > EXT4_I(inode)->i_disksize) | |
4289 | ext4_update_i_disksize(inode, new_size); | |
c8d46e41 JZ |
4290 | } else { |
4291 | /* | |
4292 | * Mark that we allocate beyond EOF so the subsequent truncate | |
4293 | * can proceed even if the new size is the same as i_size. | |
4294 | */ | |
4295 | if (new_size > i_size_read(inode)) | |
12e9b892 | 4296 | ext4_set_inode_flag(inode, EXT4_INODE_EOFBLOCKS); |
fd28784a AK |
4297 | } |
4298 | ||
4299 | } | |
4300 | ||
a2df2a63 | 4301 | /* |
2fe17c10 | 4302 | * preallocate space for a file. This implements ext4's fallocate file |
a2df2a63 AA |
4303 | * operation, which gets called from sys_fallocate system call. |
4304 | * For block-mapped files, posix_fallocate should fall back to the method | |
4305 | * of writing zeroes to the required new blocks (the same behavior which is | |
4306 | * expected for file systems which do not support fallocate() system call). | |
4307 | */ | |
2fe17c10 | 4308 | long ext4_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
a2df2a63 | 4309 | { |
2fe17c10 | 4310 | struct inode *inode = file->f_path.dentry->d_inode; |
a2df2a63 | 4311 | handle_t *handle; |
fd28784a | 4312 | loff_t new_size; |
498e5f24 | 4313 | unsigned int max_blocks; |
a2df2a63 AA |
4314 | int ret = 0; |
4315 | int ret2 = 0; | |
4316 | int retries = 0; | |
a4e5d88b | 4317 | int flags; |
2ed88685 | 4318 | struct ext4_map_blocks map; |
a2df2a63 AA |
4319 | unsigned int credits, blkbits = inode->i_blkbits; |
4320 | ||
4321 | /* | |
4322 | * currently supporting (pre)allocate mode for extent-based | |
4323 | * files _only_ | |
4324 | */ | |
12e9b892 | 4325 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) |
a2df2a63 AA |
4326 | return -EOPNOTSUPP; |
4327 | ||
a4bb6b64 AH |
4328 | /* Return error if mode is not supported */ |
4329 | if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE)) | |
4330 | return -EOPNOTSUPP; | |
4331 | ||
4332 | if (mode & FALLOC_FL_PUNCH_HOLE) | |
4333 | return ext4_punch_hole(file, offset, len); | |
4334 | ||
0562e0ba | 4335 | trace_ext4_fallocate_enter(inode, offset, len, mode); |
2ed88685 | 4336 | map.m_lblk = offset >> blkbits; |
fd28784a AK |
4337 | /* |
4338 | * We can't just convert len to max_blocks because | |
4339 | * If blocksize = 4096 offset = 3072 and len = 2048 | |
4340 | */ | |
a2df2a63 | 4341 | max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) |
2ed88685 | 4342 | - map.m_lblk; |
a2df2a63 | 4343 | /* |
f3bd1f3f | 4344 | * credits to insert 1 extent into extent tree |
a2df2a63 | 4345 | */ |
f3bd1f3f | 4346 | credits = ext4_chunk_trans_blocks(inode, max_blocks); |
55bd725a | 4347 | mutex_lock(&inode->i_mutex); |
6d19c42b NK |
4348 | ret = inode_newsize_ok(inode, (len + offset)); |
4349 | if (ret) { | |
4350 | mutex_unlock(&inode->i_mutex); | |
0562e0ba | 4351 | trace_ext4_fallocate_exit(inode, offset, max_blocks, ret); |
6d19c42b NK |
4352 | return ret; |
4353 | } | |
a4e5d88b DM |
4354 | flags = EXT4_GET_BLOCKS_CREATE_UNINIT_EXT | |
4355 | EXT4_GET_BLOCKS_NO_NORMALIZE; | |
4356 | if (mode & FALLOC_FL_KEEP_SIZE) | |
4357 | flags |= EXT4_GET_BLOCKS_KEEP_SIZE; | |
a2df2a63 AA |
4358 | retry: |
4359 | while (ret >= 0 && ret < max_blocks) { | |
2ed88685 TT |
4360 | map.m_lblk = map.m_lblk + ret; |
4361 | map.m_len = max_blocks = max_blocks - ret; | |
a2df2a63 AA |
4362 | handle = ext4_journal_start(inode, credits); |
4363 | if (IS_ERR(handle)) { | |
4364 | ret = PTR_ERR(handle); | |
4365 | break; | |
4366 | } | |
a4e5d88b | 4367 | ret = ext4_map_blocks(handle, inode, &map, flags); |
221879c9 | 4368 | if (ret <= 0) { |
2c98615d AK |
4369 | #ifdef EXT4FS_DEBUG |
4370 | WARN_ON(ret <= 0); | |
e35fd660 | 4371 | printk(KERN_ERR "%s: ext4_ext_map_blocks " |
2c98615d | 4372 | "returned error inode#%lu, block=%u, " |
9fd9784c | 4373 | "max_blocks=%u", __func__, |
a6371b63 | 4374 | inode->i_ino, map.m_lblk, max_blocks); |
2c98615d | 4375 | #endif |
a2df2a63 AA |
4376 | ext4_mark_inode_dirty(handle, inode); |
4377 | ret2 = ext4_journal_stop(handle); | |
4378 | break; | |
4379 | } | |
2ed88685 | 4380 | if ((map.m_lblk + ret) >= (EXT4_BLOCK_ALIGN(offset + len, |
fd28784a AK |
4381 | blkbits) >> blkbits)) |
4382 | new_size = offset + len; | |
4383 | else | |
29ae07b7 | 4384 | new_size = ((loff_t) map.m_lblk + ret) << blkbits; |
a2df2a63 | 4385 | |
fd28784a | 4386 | ext4_falloc_update_inode(inode, mode, new_size, |
2ed88685 | 4387 | (map.m_flags & EXT4_MAP_NEW)); |
a2df2a63 AA |
4388 | ext4_mark_inode_dirty(handle, inode); |
4389 | ret2 = ext4_journal_stop(handle); | |
4390 | if (ret2) | |
4391 | break; | |
4392 | } | |
fd28784a AK |
4393 | if (ret == -ENOSPC && |
4394 | ext4_should_retry_alloc(inode->i_sb, &retries)) { | |
4395 | ret = 0; | |
a2df2a63 | 4396 | goto retry; |
a2df2a63 | 4397 | } |
55bd725a | 4398 | mutex_unlock(&inode->i_mutex); |
0562e0ba JZ |
4399 | trace_ext4_fallocate_exit(inode, offset, max_blocks, |
4400 | ret > 0 ? ret2 : ret); | |
a2df2a63 AA |
4401 | return ret > 0 ? ret2 : ret; |
4402 | } | |
6873fa0d | 4403 | |
0031462b MC |
4404 | /* |
4405 | * This function convert a range of blocks to written extents | |
4406 | * The caller of this function will pass the start offset and the size. | |
4407 | * all unwritten extents within this range will be converted to | |
4408 | * written extents. | |
4409 | * | |
4410 | * This function is called from the direct IO end io call back | |
4411 | * function, to convert the fallocated extents after IO is completed. | |
109f5565 | 4412 | * Returns 0 on success. |
0031462b MC |
4413 | */ |
4414 | int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset, | |
a1de02dc | 4415 | ssize_t len) |
0031462b MC |
4416 | { |
4417 | handle_t *handle; | |
0031462b MC |
4418 | unsigned int max_blocks; |
4419 | int ret = 0; | |
4420 | int ret2 = 0; | |
2ed88685 | 4421 | struct ext4_map_blocks map; |
0031462b MC |
4422 | unsigned int credits, blkbits = inode->i_blkbits; |
4423 | ||
2ed88685 | 4424 | map.m_lblk = offset >> blkbits; |
0031462b MC |
4425 | /* |
4426 | * We can't just convert len to max_blocks because | |
4427 | * If blocksize = 4096 offset = 3072 and len = 2048 | |
4428 | */ | |
2ed88685 TT |
4429 | max_blocks = ((EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) - |
4430 | map.m_lblk); | |
0031462b MC |
4431 | /* |
4432 | * credits to insert 1 extent into extent tree | |
4433 | */ | |
4434 | credits = ext4_chunk_trans_blocks(inode, max_blocks); | |
4435 | while (ret >= 0 && ret < max_blocks) { | |
2ed88685 TT |
4436 | map.m_lblk += ret; |
4437 | map.m_len = (max_blocks -= ret); | |
0031462b MC |
4438 | handle = ext4_journal_start(inode, credits); |
4439 | if (IS_ERR(handle)) { | |
4440 | ret = PTR_ERR(handle); | |
4441 | break; | |
4442 | } | |
2ed88685 | 4443 | ret = ext4_map_blocks(handle, inode, &map, |
c7064ef1 | 4444 | EXT4_GET_BLOCKS_IO_CONVERT_EXT); |
0031462b MC |
4445 | if (ret <= 0) { |
4446 | WARN_ON(ret <= 0); | |
e35fd660 | 4447 | printk(KERN_ERR "%s: ext4_ext_map_blocks " |
0031462b MC |
4448 | "returned error inode#%lu, block=%u, " |
4449 | "max_blocks=%u", __func__, | |
2ed88685 | 4450 | inode->i_ino, map.m_lblk, map.m_len); |
0031462b MC |
4451 | } |
4452 | ext4_mark_inode_dirty(handle, inode); | |
4453 | ret2 = ext4_journal_stop(handle); | |
4454 | if (ret <= 0 || ret2 ) | |
4455 | break; | |
4456 | } | |
4457 | return ret > 0 ? ret2 : ret; | |
4458 | } | |
6d9c85eb | 4459 | |
6873fa0d ES |
4460 | /* |
4461 | * Callback function called for each extent to gather FIEMAP information. | |
4462 | */ | |
c03f8aa9 | 4463 | static int ext4_ext_fiemap_cb(struct inode *inode, ext4_lblk_t next, |
6873fa0d ES |
4464 | struct ext4_ext_cache *newex, struct ext4_extent *ex, |
4465 | void *data) | |
4466 | { | |
6873fa0d ES |
4467 | __u64 logical; |
4468 | __u64 physical; | |
4469 | __u64 length; | |
4470 | __u32 flags = 0; | |
6d9c85eb YY |
4471 | int ret = 0; |
4472 | struct fiemap_extent_info *fieinfo = data; | |
4473 | unsigned char blksize_bits; | |
6873fa0d | 4474 | |
6d9c85eb YY |
4475 | blksize_bits = inode->i_sb->s_blocksize_bits; |
4476 | logical = (__u64)newex->ec_block << blksize_bits; | |
6873fa0d | 4477 | |
b05e6ae5 | 4478 | if (newex->ec_start == 0) { |
6d9c85eb YY |
4479 | /* |
4480 | * No extent in extent-tree contains block @newex->ec_start, | |
4481 | * then the block may stay in 1)a hole or 2)delayed-extent. | |
4482 | * | |
4483 | * Holes or delayed-extents are processed as follows. | |
4484 | * 1. lookup dirty pages with specified range in pagecache. | |
4485 | * If no page is got, then there is no delayed-extent and | |
4486 | * return with EXT_CONTINUE. | |
4487 | * 2. find the 1st mapped buffer, | |
4488 | * 3. check if the mapped buffer is both in the request range | |
4489 | * and a delayed buffer. If not, there is no delayed-extent, | |
4490 | * then return. | |
4491 | * 4. a delayed-extent is found, the extent will be collected. | |
4492 | */ | |
4493 | ext4_lblk_t end = 0; | |
4494 | pgoff_t last_offset; | |
4495 | pgoff_t offset; | |
4496 | pgoff_t index; | |
b221349f | 4497 | pgoff_t start_index = 0; |
6d9c85eb | 4498 | struct page **pages = NULL; |
6873fa0d | 4499 | struct buffer_head *bh = NULL; |
6d9c85eb YY |
4500 | struct buffer_head *head = NULL; |
4501 | unsigned int nr_pages = PAGE_SIZE / sizeof(struct page *); | |
4502 | ||
4503 | pages = kmalloc(PAGE_SIZE, GFP_KERNEL); | |
4504 | if (pages == NULL) | |
4505 | return -ENOMEM; | |
6873fa0d ES |
4506 | |
4507 | offset = logical >> PAGE_SHIFT; | |
6d9c85eb YY |
4508 | repeat: |
4509 | last_offset = offset; | |
4510 | head = NULL; | |
4511 | ret = find_get_pages_tag(inode->i_mapping, &offset, | |
4512 | PAGECACHE_TAG_DIRTY, nr_pages, pages); | |
4513 | ||
4514 | if (!(flags & FIEMAP_EXTENT_DELALLOC)) { | |
4515 | /* First time, try to find a mapped buffer. */ | |
4516 | if (ret == 0) { | |
4517 | out: | |
4518 | for (index = 0; index < ret; index++) | |
4519 | page_cache_release(pages[index]); | |
4520 | /* just a hole. */ | |
4521 | kfree(pages); | |
4522 | return EXT_CONTINUE; | |
4523 | } | |
b221349f | 4524 | index = 0; |
6873fa0d | 4525 | |
b221349f | 4526 | next_page: |
6d9c85eb | 4527 | /* Try to find the 1st mapped buffer. */ |
b221349f | 4528 | end = ((__u64)pages[index]->index << PAGE_SHIFT) >> |
6d9c85eb | 4529 | blksize_bits; |
b221349f | 4530 | if (!page_has_buffers(pages[index])) |
6d9c85eb | 4531 | goto out; |
b221349f | 4532 | head = page_buffers(pages[index]); |
6d9c85eb YY |
4533 | if (!head) |
4534 | goto out; | |
6873fa0d | 4535 | |
b221349f | 4536 | index++; |
6d9c85eb YY |
4537 | bh = head; |
4538 | do { | |
b221349f YY |
4539 | if (end >= newex->ec_block + |
4540 | newex->ec_len) | |
4541 | /* The buffer is out of | |
4542 | * the request range. | |
4543 | */ | |
4544 | goto out; | |
4545 | ||
4546 | if (buffer_mapped(bh) && | |
4547 | end >= newex->ec_block) { | |
4548 | start_index = index - 1; | |
6d9c85eb | 4549 | /* get the 1st mapped buffer. */ |
6d9c85eb YY |
4550 | goto found_mapped_buffer; |
4551 | } | |
b221349f | 4552 | |
6d9c85eb YY |
4553 | bh = bh->b_this_page; |
4554 | end++; | |
4555 | } while (bh != head); | |
6873fa0d | 4556 | |
b221349f YY |
4557 | /* No mapped buffer in the range found in this page, |
4558 | * We need to look up next page. | |
4559 | */ | |
4560 | if (index >= ret) { | |
4561 | /* There is no page left, but we need to limit | |
4562 | * newex->ec_len. | |
4563 | */ | |
4564 | newex->ec_len = end - newex->ec_block; | |
4565 | goto out; | |
4566 | } | |
4567 | goto next_page; | |
6873fa0d | 4568 | } else { |
6d9c85eb YY |
4569 | /*Find contiguous delayed buffers. */ |
4570 | if (ret > 0 && pages[0]->index == last_offset) | |
4571 | head = page_buffers(pages[0]); | |
4572 | bh = head; | |
b221349f YY |
4573 | index = 1; |
4574 | start_index = 0; | |
6873fa0d | 4575 | } |
6d9c85eb YY |
4576 | |
4577 | found_mapped_buffer: | |
4578 | if (bh != NULL && buffer_delay(bh)) { | |
4579 | /* 1st or contiguous delayed buffer found. */ | |
4580 | if (!(flags & FIEMAP_EXTENT_DELALLOC)) { | |
4581 | /* | |
4582 | * 1st delayed buffer found, record | |
4583 | * the start of extent. | |
4584 | */ | |
4585 | flags |= FIEMAP_EXTENT_DELALLOC; | |
4586 | newex->ec_block = end; | |
4587 | logical = (__u64)end << blksize_bits; | |
4588 | } | |
4589 | /* Find contiguous delayed buffers. */ | |
4590 | do { | |
4591 | if (!buffer_delay(bh)) | |
4592 | goto found_delayed_extent; | |
4593 | bh = bh->b_this_page; | |
4594 | end++; | |
4595 | } while (bh != head); | |
4596 | ||
b221349f | 4597 | for (; index < ret; index++) { |
6d9c85eb YY |
4598 | if (!page_has_buffers(pages[index])) { |
4599 | bh = NULL; | |
4600 | break; | |
4601 | } | |
4602 | head = page_buffers(pages[index]); | |
4603 | if (!head) { | |
4604 | bh = NULL; | |
4605 | break; | |
4606 | } | |
b221349f | 4607 | |
6d9c85eb | 4608 | if (pages[index]->index != |
b221349f YY |
4609 | pages[start_index]->index + index |
4610 | - start_index) { | |
6d9c85eb YY |
4611 | /* Blocks are not contiguous. */ |
4612 | bh = NULL; | |
4613 | break; | |
4614 | } | |
4615 | bh = head; | |
4616 | do { | |
4617 | if (!buffer_delay(bh)) | |
4618 | /* Delayed-extent ends. */ | |
4619 | goto found_delayed_extent; | |
4620 | bh = bh->b_this_page; | |
4621 | end++; | |
4622 | } while (bh != head); | |
4623 | } | |
4624 | } else if (!(flags & FIEMAP_EXTENT_DELALLOC)) | |
4625 | /* a hole found. */ | |
4626 | goto out; | |
4627 | ||
4628 | found_delayed_extent: | |
4629 | newex->ec_len = min(end - newex->ec_block, | |
4630 | (ext4_lblk_t)EXT_INIT_MAX_LEN); | |
4631 | if (ret == nr_pages && bh != NULL && | |
4632 | newex->ec_len < EXT_INIT_MAX_LEN && | |
4633 | buffer_delay(bh)) { | |
4634 | /* Have not collected an extent and continue. */ | |
4635 | for (index = 0; index < ret; index++) | |
4636 | page_cache_release(pages[index]); | |
4637 | goto repeat; | |
6873fa0d | 4638 | } |
6d9c85eb YY |
4639 | |
4640 | for (index = 0; index < ret; index++) | |
4641 | page_cache_release(pages[index]); | |
4642 | kfree(pages); | |
6873fa0d ES |
4643 | } |
4644 | ||
4645 | physical = (__u64)newex->ec_start << blksize_bits; | |
4646 | length = (__u64)newex->ec_len << blksize_bits; | |
4647 | ||
4648 | if (ex && ext4_ext_is_uninitialized(ex)) | |
4649 | flags |= FIEMAP_EXTENT_UNWRITTEN; | |
4650 | ||
c03f8aa9 | 4651 | if (next == EXT_MAX_BLOCKS) |
6873fa0d ES |
4652 | flags |= FIEMAP_EXTENT_LAST; |
4653 | ||
6d9c85eb | 4654 | ret = fiemap_fill_next_extent(fieinfo, logical, physical, |
6873fa0d | 4655 | length, flags); |
6d9c85eb YY |
4656 | if (ret < 0) |
4657 | return ret; | |
4658 | if (ret == 1) | |
6873fa0d | 4659 | return EXT_BREAK; |
6873fa0d ES |
4660 | return EXT_CONTINUE; |
4661 | } | |
6873fa0d ES |
4662 | /* fiemap flags we can handle specified here */ |
4663 | #define EXT4_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR) | |
4664 | ||
3a06d778 AK |
4665 | static int ext4_xattr_fiemap(struct inode *inode, |
4666 | struct fiemap_extent_info *fieinfo) | |
6873fa0d ES |
4667 | { |
4668 | __u64 physical = 0; | |
4669 | __u64 length; | |
4670 | __u32 flags = FIEMAP_EXTENT_LAST; | |
4671 | int blockbits = inode->i_sb->s_blocksize_bits; | |
4672 | int error = 0; | |
4673 | ||
4674 | /* in-inode? */ | |
19f5fb7a | 4675 | if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) { |
6873fa0d ES |
4676 | struct ext4_iloc iloc; |
4677 | int offset; /* offset of xattr in inode */ | |
4678 | ||
4679 | error = ext4_get_inode_loc(inode, &iloc); | |
4680 | if (error) | |
4681 | return error; | |
4682 | physical = iloc.bh->b_blocknr << blockbits; | |
4683 | offset = EXT4_GOOD_OLD_INODE_SIZE + | |
4684 | EXT4_I(inode)->i_extra_isize; | |
4685 | physical += offset; | |
4686 | length = EXT4_SB(inode->i_sb)->s_inode_size - offset; | |
4687 | flags |= FIEMAP_EXTENT_DATA_INLINE; | |
fd2dd9fb | 4688 | brelse(iloc.bh); |
6873fa0d ES |
4689 | } else { /* external block */ |
4690 | physical = EXT4_I(inode)->i_file_acl << blockbits; | |
4691 | length = inode->i_sb->s_blocksize; | |
4692 | } | |
4693 | ||
4694 | if (physical) | |
4695 | error = fiemap_fill_next_extent(fieinfo, 0, physical, | |
4696 | length, flags); | |
4697 | return (error < 0 ? error : 0); | |
4698 | } | |
4699 | ||
a4bb6b64 AH |
4700 | /* |
4701 | * ext4_ext_punch_hole | |
4702 | * | |
4703 | * Punches a hole of "length" bytes in a file starting | |
4704 | * at byte "offset" | |
4705 | * | |
4706 | * @inode: The inode of the file to punch a hole in | |
4707 | * @offset: The starting byte offset of the hole | |
4708 | * @length: The length of the hole | |
4709 | * | |
4710 | * Returns the number of blocks removed or negative on err | |
4711 | */ | |
4712 | int ext4_ext_punch_hole(struct file *file, loff_t offset, loff_t length) | |
4713 | { | |
4714 | struct inode *inode = file->f_path.dentry->d_inode; | |
4715 | struct super_block *sb = inode->i_sb; | |
4716 | struct ext4_ext_cache cache_ex; | |
4717 | ext4_lblk_t first_block, last_block, num_blocks, iblock, max_blocks; | |
4718 | struct address_space *mapping = inode->i_mapping; | |
4719 | struct ext4_map_blocks map; | |
4720 | handle_t *handle; | |
ba06208a AH |
4721 | loff_t first_page, last_page, page_len; |
4722 | loff_t first_page_offset, last_page_offset; | |
a4bb6b64 AH |
4723 | int ret, credits, blocks_released, err = 0; |
4724 | ||
2be4751b AH |
4725 | /* No need to punch hole beyond i_size */ |
4726 | if (offset >= inode->i_size) | |
4727 | return 0; | |
4728 | ||
4729 | /* | |
4730 | * If the hole extends beyond i_size, set the hole | |
4731 | * to end after the page that contains i_size | |
4732 | */ | |
4733 | if (offset + length > inode->i_size) { | |
4734 | length = inode->i_size + | |
4735 | PAGE_CACHE_SIZE - (inode->i_size & (PAGE_CACHE_SIZE - 1)) - | |
4736 | offset; | |
4737 | } | |
4738 | ||
a4bb6b64 AH |
4739 | first_block = (offset + sb->s_blocksize - 1) >> |
4740 | EXT4_BLOCK_SIZE_BITS(sb); | |
4741 | last_block = (offset + length) >> EXT4_BLOCK_SIZE_BITS(sb); | |
4742 | ||
a4bb6b64 AH |
4743 | first_page = (offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; |
4744 | last_page = (offset + length) >> PAGE_CACHE_SHIFT; | |
4745 | ||
4746 | first_page_offset = first_page << PAGE_CACHE_SHIFT; | |
4747 | last_page_offset = last_page << PAGE_CACHE_SHIFT; | |
4748 | ||
4749 | /* | |
4750 | * Write out all dirty pages to avoid race conditions | |
4751 | * Then release them. | |
4752 | */ | |
4753 | if (mapping->nrpages && mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) { | |
4754 | err = filemap_write_and_wait_range(mapping, | |
2be4751b | 4755 | offset, offset + length - 1); |
a4bb6b64 | 4756 | |
2be4751b AH |
4757 | if (err) |
4758 | return err; | |
a4bb6b64 AH |
4759 | } |
4760 | ||
4761 | /* Now release the pages */ | |
4762 | if (last_page_offset > first_page_offset) { | |
4763 | truncate_inode_pages_range(mapping, first_page_offset, | |
4764 | last_page_offset-1); | |
4765 | } | |
4766 | ||
4767 | /* finish any pending end_io work */ | |
4768 | ext4_flush_completed_IO(inode); | |
4769 | ||
4770 | credits = ext4_writepage_trans_blocks(inode); | |
4771 | handle = ext4_journal_start(inode, credits); | |
4772 | if (IS_ERR(handle)) | |
4773 | return PTR_ERR(handle); | |
4774 | ||
4775 | err = ext4_orphan_add(handle, inode); | |
4776 | if (err) | |
4777 | goto out; | |
4778 | ||
4779 | /* | |
ba06208a AH |
4780 | * Now we need to zero out the non-page-aligned data in the |
4781 | * pages at the start and tail of the hole, and unmap the buffer | |
4782 | * heads for the block aligned regions of the page that were | |
4783 | * completely zeroed. | |
a4bb6b64 | 4784 | */ |
ba06208a AH |
4785 | if (first_page > last_page) { |
4786 | /* | |
4787 | * If the file space being truncated is contained within a page | |
4788 | * just zero out and unmap the middle of that page | |
4789 | */ | |
4790 | err = ext4_discard_partial_page_buffers(handle, | |
4791 | mapping, offset, length, 0); | |
4792 | ||
4793 | if (err) | |
4794 | goto out; | |
4795 | } else { | |
4796 | /* | |
4797 | * zero out and unmap the partial page that contains | |
4798 | * the start of the hole | |
4799 | */ | |
4800 | page_len = first_page_offset - offset; | |
4801 | if (page_len > 0) { | |
4802 | err = ext4_discard_partial_page_buffers(handle, mapping, | |
4803 | offset, page_len, 0); | |
4804 | if (err) | |
4805 | goto out; | |
4806 | } | |
4807 | ||
4808 | /* | |
4809 | * zero out and unmap the partial page that contains | |
4810 | * the end of the hole | |
4811 | */ | |
4812 | page_len = offset + length - last_page_offset; | |
4813 | if (page_len > 0) { | |
4814 | err = ext4_discard_partial_page_buffers(handle, mapping, | |
4815 | last_page_offset, page_len, 0); | |
4816 | if (err) | |
4817 | goto out; | |
a4bb6b64 AH |
4818 | } |
4819 | } | |
4820 | ||
2be4751b AH |
4821 | |
4822 | /* | |
4823 | * If i_size is contained in the last page, we need to | |
4824 | * unmap and zero the partial page after i_size | |
4825 | */ | |
4826 | if (inode->i_size >> PAGE_CACHE_SHIFT == last_page && | |
4827 | inode->i_size % PAGE_CACHE_SIZE != 0) { | |
4828 | ||
4829 | page_len = PAGE_CACHE_SIZE - | |
4830 | (inode->i_size & (PAGE_CACHE_SIZE - 1)); | |
4831 | ||
4832 | if (page_len > 0) { | |
4833 | err = ext4_discard_partial_page_buffers(handle, | |
4834 | mapping, inode->i_size, page_len, 0); | |
4835 | ||
4836 | if (err) | |
4837 | goto out; | |
4838 | } | |
4839 | } | |
4840 | ||
a4bb6b64 AH |
4841 | /* If there are no blocks to remove, return now */ |
4842 | if (first_block >= last_block) | |
4843 | goto out; | |
4844 | ||
4845 | down_write(&EXT4_I(inode)->i_data_sem); | |
4846 | ext4_ext_invalidate_cache(inode); | |
4847 | ext4_discard_preallocations(inode); | |
4848 | ||
4849 | /* | |
4850 | * Loop over all the blocks and identify blocks | |
4851 | * that need to be punched out | |
4852 | */ | |
4853 | iblock = first_block; | |
4854 | blocks_released = 0; | |
4855 | while (iblock < last_block) { | |
4856 | max_blocks = last_block - iblock; | |
4857 | num_blocks = 1; | |
4858 | memset(&map, 0, sizeof(map)); | |
4859 | map.m_lblk = iblock; | |
4860 | map.m_len = max_blocks; | |
4861 | ret = ext4_ext_map_blocks(handle, inode, &map, | |
4862 | EXT4_GET_BLOCKS_PUNCH_OUT_EXT); | |
4863 | ||
4864 | if (ret > 0) { | |
4865 | blocks_released += ret; | |
4866 | num_blocks = ret; | |
4867 | } else if (ret == 0) { | |
4868 | /* | |
4869 | * If map blocks could not find the block, | |
4870 | * then it is in a hole. If the hole was | |
4871 | * not already cached, then map blocks should | |
4872 | * put it in the cache. So we can get the hole | |
4873 | * out of the cache | |
4874 | */ | |
4875 | memset(&cache_ex, 0, sizeof(cache_ex)); | |
4876 | if ((ext4_ext_check_cache(inode, iblock, &cache_ex)) && | |
4877 | !cache_ex.ec_start) { | |
4878 | ||
4879 | /* The hole is cached */ | |
4880 | num_blocks = cache_ex.ec_block + | |
4881 | cache_ex.ec_len - iblock; | |
4882 | ||
4883 | } else { | |
4884 | /* The block could not be identified */ | |
4885 | err = -EIO; | |
4886 | break; | |
4887 | } | |
4888 | } else { | |
4889 | /* Map blocks error */ | |
4890 | err = ret; | |
4891 | break; | |
4892 | } | |
4893 | ||
4894 | if (num_blocks == 0) { | |
4895 | /* This condition should never happen */ | |
4896 | ext_debug("Block lookup failed"); | |
4897 | err = -EIO; | |
4898 | break; | |
4899 | } | |
4900 | ||
4901 | iblock += num_blocks; | |
4902 | } | |
4903 | ||
4904 | if (blocks_released > 0) { | |
4905 | ext4_ext_invalidate_cache(inode); | |
4906 | ext4_discard_preallocations(inode); | |
4907 | } | |
4908 | ||
4909 | if (IS_SYNC(inode)) | |
4910 | ext4_handle_sync(handle); | |
4911 | ||
4912 | up_write(&EXT4_I(inode)->i_data_sem); | |
4913 | ||
4914 | out: | |
4915 | ext4_orphan_del(handle, inode); | |
4916 | inode->i_mtime = inode->i_ctime = ext4_current_time(inode); | |
4917 | ext4_mark_inode_dirty(handle, inode); | |
4918 | ext4_journal_stop(handle); | |
4919 | return err; | |
4920 | } | |
6873fa0d ES |
4921 | int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, |
4922 | __u64 start, __u64 len) | |
4923 | { | |
4924 | ext4_lblk_t start_blk; | |
6873fa0d ES |
4925 | int error = 0; |
4926 | ||
4927 | /* fallback to generic here if not in extents fmt */ | |
12e9b892 | 4928 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) |
6873fa0d ES |
4929 | return generic_block_fiemap(inode, fieinfo, start, len, |
4930 | ext4_get_block); | |
4931 | ||
4932 | if (fiemap_check_flags(fieinfo, EXT4_FIEMAP_FLAGS)) | |
4933 | return -EBADR; | |
4934 | ||
4935 | if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) { | |
4936 | error = ext4_xattr_fiemap(inode, fieinfo); | |
4937 | } else { | |
aca92ff6 LM |
4938 | ext4_lblk_t len_blks; |
4939 | __u64 last_blk; | |
4940 | ||
6873fa0d | 4941 | start_blk = start >> inode->i_sb->s_blocksize_bits; |
aca92ff6 | 4942 | last_blk = (start + len - 1) >> inode->i_sb->s_blocksize_bits; |
f17722f9 LC |
4943 | if (last_blk >= EXT_MAX_BLOCKS) |
4944 | last_blk = EXT_MAX_BLOCKS-1; | |
aca92ff6 | 4945 | len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1; |
6873fa0d ES |
4946 | |
4947 | /* | |
4948 | * Walk the extent tree gathering extent information. | |
4949 | * ext4_ext_fiemap_cb will push extents back to user. | |
4950 | */ | |
6873fa0d ES |
4951 | error = ext4_ext_walk_space(inode, start_blk, len_blks, |
4952 | ext4_ext_fiemap_cb, fieinfo); | |
6873fa0d ES |
4953 | } |
4954 | ||
4955 | return error; | |
4956 | } |