Branch data Line data Source code
1 : : // SPDX-License-Identifier: GPL-2.0
2 : : /*
3 : : * linux/fs/ext4/balloc.c
4 : : *
5 : : * Copyright (C) 1992, 1993, 1994, 1995
6 : : * Remy Card (card@masi.ibp.fr)
7 : : * Laboratoire MASI - Institut Blaise Pascal
8 : : * Universite Pierre et Marie Curie (Paris VI)
9 : : *
10 : : * Enhanced block allocation by Stephen Tweedie (sct@redhat.com), 1993
11 : : * Big-endian to little-endian byte-swapping/bitmaps by
12 : : * David S. Miller (davem@caip.rutgers.edu), 1995
13 : : */
14 : :
15 : : #include <linux/time.h>
16 : : #include <linux/capability.h>
17 : : #include <linux/fs.h>
18 : : #include <linux/quotaops.h>
19 : : #include <linux/buffer_head.h>
20 : : #include "ext4.h"
21 : : #include "ext4_jbd2.h"
22 : : #include "mballoc.h"
23 : :
24 : : #include <trace/events/ext4.h>
25 : :
26 : : static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
27 : : ext4_group_t block_group);
28 : : /*
29 : : * balloc.c contains the blocks allocation and deallocation routines
30 : : */
31 : :
32 : : /*
33 : : * Calculate block group number for a given block number
34 : : */
35 : 418396 : ext4_group_t ext4_get_group_number(struct super_block *sb,
36 : : ext4_fsblk_t block)
37 : : {
38 : 418396 : ext4_group_t group;
39 : :
40 [ # # ]: 0 : if (test_opt2(sb, STD_GROUP_SIZE))
41 : 418396 : group = (block -
42 : 418396 : le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) >>
43 : 418396 : (EXT4_BLOCK_SIZE_BITS(sb) + EXT4_CLUSTER_BITS(sb) + 3);
44 : : else
45 : 0 : ext4_get_group_no_and_offset(sb, block, &group, NULL);
46 : 418396 : return group;
47 : : }
48 : :
49 : : /*
50 : : * Calculate the block group number and offset into the block/cluster
51 : : * allocation bitmap, given a block number
52 : : */
53 : 3025 : void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
54 : : ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
55 : : {
56 [ + - ]: 3025 : struct ext4_super_block *es = EXT4_SB(sb)->s_es;
57 : 3025 : ext4_grpblk_t offset;
58 : :
59 : 3025 : blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
60 : 3025 : offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) >>
61 : 3025 : EXT4_SB(sb)->s_cluster_bits;
62 [ + - ]: 3025 : if (offsetp)
63 : 3025 : *offsetp = offset;
64 [ + - ]: 3025 : if (blockgrpp)
65 : 3025 : *blockgrpp = blocknr;
66 : :
67 : 3025 : }
68 : :
69 : : /*
70 : : * Check whether the 'block' lives within the 'block_group'. Returns 1 if so
71 : : * and 0 otherwise.
72 : : */
73 : 418396 : static inline int ext4_block_in_group(struct super_block *sb,
74 : : ext4_fsblk_t block,
75 : : ext4_group_t block_group)
76 : : {
77 : 418396 : ext4_group_t actual_group;
78 : :
79 [ + - + - ]: 1628 : actual_group = ext4_get_group_number(sb, block);
80 : 418396 : return (actual_group == block_group) ? 1 : 0;
81 : : }
82 : :
83 : : /* Return the number of clusters used for file system metadata; this
84 : : * represents the overhead needed by the file system.
85 : : */
86 : 814 : static unsigned ext4_num_overhead_clusters(struct super_block *sb,
87 : : ext4_group_t block_group,
88 : : struct ext4_group_desc *gdp)
89 : : {
90 : 814 : unsigned num_clusters;
91 : 814 : int block_cluster = -1, inode_cluster = -1, itbl_cluster = -1, i, c;
92 : 814 : ext4_fsblk_t start = ext4_group_first_block_no(sb, block_group);
93 : 814 : ext4_fsblk_t itbl_blk;
94 : 814 : struct ext4_sb_info *sbi = EXT4_SB(sb);
95 : :
96 : : /* This is the number of clusters used by the superblock,
97 : : * block group descriptors, and reserved block group
98 : : * descriptor blocks */
99 : 814 : num_clusters = ext4_num_base_meta_clusters(sb, block_group);
100 : :
101 : : /*
102 : : * For the allocation bitmaps and inode table, we first need
103 : : * to check to see if the block is in the block group. If it
104 : : * is, then check to see if the cluster is already accounted
105 : : * for in the clusters used for the base metadata cluster, or
106 : : * if we can increment the base metadata cluster to include
107 : : * that block. Otherwise, we will have to track the cluster
108 : : * used for the allocation bitmap or inode table explicitly.
109 : : * Normally all of these blocks are contiguous, so the special
110 : : * case handling shouldn't be necessary except for *very*
111 : : * unusual file system layouts.
112 : : */
113 [ - + ]: 1628 : if (ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), block_group)) {
114 : 0 : block_cluster = EXT4_B2C(sbi,
115 : : ext4_block_bitmap(sb, gdp) - start);
116 [ # # ]: 0 : if (block_cluster < num_clusters)
117 : : block_cluster = -1;
118 [ # # ]: 0 : else if (block_cluster == num_clusters) {
119 : 0 : num_clusters++;
120 : 0 : block_cluster = -1;
121 : : }
122 : : }
123 : :
124 [ - + ]: 1628 : if (ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), block_group)) {
125 : 0 : inode_cluster = EXT4_B2C(sbi,
126 : : ext4_inode_bitmap(sb, gdp) - start);
127 [ # # ]: 0 : if (inode_cluster < num_clusters)
128 : : inode_cluster = -1;
129 [ # # ]: 0 : else if (inode_cluster == num_clusters) {
130 : 0 : num_clusters++;
131 : 0 : inode_cluster = -1;
132 : : }
133 : : }
134 : :
135 : 814 : itbl_blk = ext4_inode_table(sb, gdp);
136 [ + + ]: 418396 : for (i = 0; i < sbi->s_itb_per_group; i++) {
137 [ + - - + ]: 833536 : if (ext4_block_in_group(sb, itbl_blk + i, block_group)) {
138 : 0 : c = EXT4_B2C(sbi, itbl_blk + i - start);
139 [ # # ]: 0 : if ((c < num_clusters) || (c == inode_cluster) ||
140 [ # # ]: 0 : (c == block_cluster) || (c == itbl_cluster))
141 : 0 : continue;
142 [ # # ]: 0 : if (c == num_clusters) {
143 : 0 : num_clusters++;
144 : 0 : continue;
145 : : }
146 : 0 : num_clusters++;
147 : 0 : itbl_cluster = c;
148 : : }
149 : : }
150 : :
151 [ - + ]: 814 : if (block_cluster != -1)
152 : 0 : num_clusters++;
153 [ - + ]: 814 : if (inode_cluster != -1)
154 : 0 : num_clusters++;
155 : :
156 : 814 : return num_clusters;
157 : : }
158 : :
159 : 814 : static unsigned int num_clusters_in_group(struct super_block *sb,
160 : : ext4_group_t block_group)
161 : : {
162 : 814 : unsigned int blocks;
163 : :
164 [ - + ]: 814 : if (block_group == ext4_get_groups_count(sb) - 1) {
165 : : /*
166 : : * Even though mke2fs always initializes the first and
167 : : * last group, just in case some other tool was used,
168 : : * we need to make sure we calculate the right free
169 : : * blocks.
170 : : */
171 : 0 : blocks = ext4_blocks_count(EXT4_SB(sb)->s_es) -
172 : : ext4_group_first_block_no(sb, block_group);
173 : : } else
174 : 814 : blocks = EXT4_BLOCKS_PER_GROUP(sb);
175 : 814 : return EXT4_NUM_B2C(EXT4_SB(sb), blocks);
176 : : }
177 : :
178 : : /* Initializes an uninitialized block bitmap */
179 : 0 : static int ext4_init_block_bitmap(struct super_block *sb,
180 : : struct buffer_head *bh,
181 : : ext4_group_t block_group,
182 : : struct ext4_group_desc *gdp)
183 : : {
184 : 0 : unsigned int bit, bit_max;
185 : 0 : struct ext4_sb_info *sbi = EXT4_SB(sb);
186 : 0 : ext4_fsblk_t start, tmp;
187 : :
188 [ # # ]: 0 : J_ASSERT_BH(bh, buffer_locked(bh));
189 : :
190 : : /* If checksum is bad mark all blocks used to prevent allocation
191 : : * essentially implementing a per-group read-only flag. */
192 [ # # ]: 0 : if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) {
193 : 0 : ext4_mark_group_bitmap_corrupted(sb, block_group,
194 : : EXT4_GROUP_INFO_BBITMAP_CORRUPT |
195 : : EXT4_GROUP_INFO_IBITMAP_CORRUPT);
196 : 0 : return -EFSBADCRC;
197 : : }
198 : 0 : memset(bh->b_data, 0, sb->s_blocksize);
199 : :
200 : 0 : bit_max = ext4_num_base_meta_clusters(sb, block_group);
201 [ # # ]: 0 : if ((bit_max >> 3) >= bh->b_size)
202 : : return -EFSCORRUPTED;
203 : :
204 [ # # ]: 0 : for (bit = 0; bit < bit_max; bit++)
205 : 0 : ext4_set_bit(bit, bh->b_data);
206 : :
207 : 0 : start = ext4_group_first_block_no(sb, block_group);
208 : :
209 : : /* Set bits for block and inode bitmaps, and inode table */
210 : 0 : tmp = ext4_block_bitmap(sb, gdp);
211 [ # # # # ]: 0 : if (ext4_block_in_group(sb, tmp, block_group))
212 : 0 : ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
213 : :
214 : 0 : tmp = ext4_inode_bitmap(sb, gdp);
215 [ # # # # ]: 0 : if (ext4_block_in_group(sb, tmp, block_group))
216 : 0 : ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
217 : :
218 : 0 : tmp = ext4_inode_table(sb, gdp);
219 : 0 : for (; tmp < ext4_inode_table(sb, gdp) +
220 [ # # ]: 0 : sbi->s_itb_per_group; tmp++) {
221 [ # # # # ]: 0 : if (ext4_block_in_group(sb, tmp, block_group))
222 : 0 : ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
223 : : }
224 : :
225 : : /*
226 : : * Also if the number of blocks within the group is less than
227 : : * the blocksize * 8 ( which is the size of bitmap ), set rest
228 : : * of the block bitmap to 1
229 : : */
230 : 0 : ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group),
231 : 0 : sb->s_blocksize * 8, bh->b_data);
232 : 0 : return 0;
233 : : }
234 : :
235 : : /* Return the number of free blocks in a block group. It is used when
236 : : * the block bitmap is uninitialized, so we can't just count the bits
237 : : * in the bitmap. */
238 : 814 : unsigned ext4_free_clusters_after_init(struct super_block *sb,
239 : : ext4_group_t block_group,
240 : : struct ext4_group_desc *gdp)
241 : : {
242 : 814 : return num_clusters_in_group(sb, block_group) -
243 : 814 : ext4_num_overhead_clusters(sb, block_group, gdp);
244 : : }
245 : :
246 : : /*
247 : : * The free blocks are managed by bitmaps. A file system contains several
248 : : * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
249 : : * block for inodes, N blocks for the inode table and data blocks.
250 : : *
251 : : * The file system contains group descriptors which are located after the
252 : : * super block. Each descriptor contains the number of the bitmap block and
253 : : * the free blocks count in the block. The descriptors are loaded in memory
254 : : * when a file system is mounted (see ext4_fill_super).
255 : : */
256 : :
257 : : /**
258 : : * ext4_get_group_desc() -- load group descriptor from disk
259 : : * @sb: super block
260 : : * @block_group: given block group
261 : : * @bh: pointer to the buffer head to store the block
262 : : * group descriptor
263 : : */
264 : 104309 : struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
265 : : ext4_group_t block_group,
266 : : struct buffer_head **bh)
267 : : {
268 : 104309 : unsigned int group_desc;
269 : 104309 : unsigned int offset;
270 : 104309 : ext4_group_t ngroups = ext4_get_groups_count(sb);
271 : 104309 : struct ext4_group_desc *desc;
272 [ - + ]: 104309 : struct ext4_sb_info *sbi = EXT4_SB(sb);
273 : 104309 : struct buffer_head *bh_p;
274 : :
275 [ - + ]: 104309 : if (block_group >= ngroups) {
276 : 0 : ext4_error(sb, "block_group >= groups_count - block_group = %u,"
277 : : " groups_count = %u", block_group, ngroups);
278 : :
279 : 0 : return NULL;
280 : : }
281 : :
282 : 104309 : group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
283 : 104309 : offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
284 : 104309 : bh_p = sbi_array_rcu_deref(sbi, s_group_desc, group_desc);
285 : : /*
286 : : * sbi_array_rcu_deref returns with rcu unlocked, this is ok since
287 : : * the pointer being dereferenced won't be dereferenced again. By
288 : : * looking at the usage in add_new_gdb() the value isn't modified,
289 : : * just the pointer, and so it remains valid.
290 : : */
291 [ - + ]: 104309 : if (!bh_p) {
292 : 0 : ext4_error(sb, "Group descriptor not loaded - "
293 : : "block_group = %u, group_desc = %u, desc = %u",
294 : : block_group, group_desc, offset);
295 : 0 : return NULL;
296 : : }
297 : :
298 : 104309 : desc = (struct ext4_group_desc *)(
299 : 104309 : (__u8 *)bh_p->b_data +
300 [ + + ]: 104309 : offset * EXT4_DESC_SIZE(sb));
301 [ + + ]: 104309 : if (bh)
302 : 10494 : *bh = bh_p;
303 : : return desc;
304 : : }
305 : :
306 : : /*
307 : : * Return the block number which was discovered to be invalid, or 0 if
308 : : * the block bitmap is valid.
309 : : */
310 : : static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb,
311 : : struct ext4_group_desc *desc,
312 : : ext4_group_t block_group,
313 : : struct buffer_head *bh)
314 : : {
315 : : struct ext4_sb_info *sbi = EXT4_SB(sb);
316 : : ext4_grpblk_t offset;
317 : : ext4_grpblk_t next_zero_bit;
318 : : ext4_grpblk_t max_bit = EXT4_CLUSTERS_PER_GROUP(sb);
319 : : ext4_fsblk_t blk;
320 : : ext4_fsblk_t group_first_block;
321 : :
322 : : if (ext4_has_feature_flex_bg(sb)) {
323 : : /* with FLEX_BG, the inode/block bitmaps and itable
324 : : * blocks may not be in the group at all
325 : : * so the bitmap validation will be skipped for those groups
326 : : * or it has to also read the block group where the bitmaps
327 : : * are located to verify they are set.
328 : : */
329 : : return 0;
330 : : }
331 : : group_first_block = ext4_group_first_block_no(sb, block_group);
332 : :
333 : : /* check whether block bitmap block number is set */
334 : : blk = ext4_block_bitmap(sb, desc);
335 : : offset = blk - group_first_block;
336 : : if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
337 : : !ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
338 : : /* bad block bitmap */
339 : : return blk;
340 : :
341 : : /* check whether the inode bitmap block number is set */
342 : : blk = ext4_inode_bitmap(sb, desc);
343 : : offset = blk - group_first_block;
344 : : if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
345 : : !ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
346 : : /* bad block bitmap */
347 : : return blk;
348 : :
349 : : /* check whether the inode table block number is set */
350 : : blk = ext4_inode_table(sb, desc);
351 : : offset = blk - group_first_block;
352 : : if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
353 : : EXT4_B2C(sbi, offset + sbi->s_itb_per_group) >= max_bit)
354 : : return blk;
355 : : next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
356 : : EXT4_B2C(sbi, offset + sbi->s_itb_per_group),
357 : : EXT4_B2C(sbi, offset));
358 : : if (next_zero_bit <
359 : : EXT4_B2C(sbi, offset + sbi->s_itb_per_group))
360 : : /* bad bitmap for inode tables */
361 : : return blk;
362 : : return 0;
363 : : }
364 : :
365 : 3003 : static int ext4_validate_block_bitmap(struct super_block *sb,
366 : : struct ext4_group_desc *desc,
367 : : ext4_group_t block_group,
368 : : struct buffer_head *bh)
369 : : {
370 : 3003 : ext4_fsblk_t blk;
371 : 3003 : struct ext4_group_info *grp = ext4_get_group_info(sb, block_group);
372 : :
373 [ + + ]: 3003 : if (buffer_verified(bh))
374 : : return 0;
375 [ + - ]: 33 : if (EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
376 : : return -EFSCORRUPTED;
377 : :
378 : 33 : ext4_lock_group(sb, block_group);
379 [ - + ]: 33 : if (buffer_verified(bh))
380 : 0 : goto verified;
381 [ - + ]: 33 : if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group,
382 : : desc, bh) ||
383 : : ext4_simulate_fail(sb, EXT4_SIM_BBITMAP_CRC))) {
384 : 0 : ext4_unlock_group(sb, block_group);
385 : 0 : ext4_error(sb, "bg %u: bad block bitmap checksum", block_group);
386 : 0 : ext4_mark_group_bitmap_corrupted(sb, block_group,
387 : : EXT4_GROUP_INFO_BBITMAP_CORRUPT);
388 : 0 : return -EFSBADCRC;
389 : : }
390 : 33 : blk = ext4_valid_block_bitmap(sb, desc, block_group, bh);
391 [ - + ]: 33 : if (unlikely(blk != 0)) {
392 : 0 : ext4_unlock_group(sb, block_group);
393 : 0 : ext4_error(sb, "bg %u: block %llu: invalid block bitmap",
394 : : block_group, blk);
395 : 0 : ext4_mark_group_bitmap_corrupted(sb, block_group,
396 : : EXT4_GROUP_INFO_BBITMAP_CORRUPT);
397 : 0 : return -EFSCORRUPTED;
398 : : }
399 : 33 : set_buffer_verified(bh);
400 : 33 : verified:
401 : 33 : ext4_unlock_group(sb, block_group);
402 : 33 : return 0;
403 : : }
404 : :
405 : : /**
406 : : * ext4_read_block_bitmap_nowait()
407 : : * @sb: super block
408 : : * @block_group: given block group
409 : : *
410 : : * Read the bitmap for a given block_group,and validate the
411 : : * bits for block/inode/inode tables are set in the bitmaps
412 : : *
413 : : * Return buffer_head on success or NULL in case of failure.
414 : : */
415 : : struct buffer_head *
416 : 3003 : ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group)
417 : : {
418 : 3003 : struct ext4_group_desc *desc;
419 : 3003 : struct ext4_sb_info *sbi = EXT4_SB(sb);
420 : 3003 : struct buffer_head *bh;
421 : 3003 : ext4_fsblk_t bitmap_blk;
422 : 3003 : int err;
423 : :
424 : 3003 : desc = ext4_get_group_desc(sb, block_group, NULL);
425 [ + - ]: 3003 : if (!desc)
426 : : return ERR_PTR(-EFSCORRUPTED);
427 : 3003 : bitmap_blk = ext4_block_bitmap(sb, desc);
428 [ + - - + ]: 3003 : if ((bitmap_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) ||
429 [ - + ]: 3003 : (bitmap_blk >= ext4_blocks_count(sbi->s_es))) {
430 : 0 : ext4_error(sb, "Invalid block bitmap block %llu in "
431 : : "block_group %u", bitmap_blk, block_group);
432 : 0 : ext4_mark_group_bitmap_corrupted(sb, block_group,
433 : : EXT4_GROUP_INFO_BBITMAP_CORRUPT);
434 : 0 : return ERR_PTR(-EFSCORRUPTED);
435 : : }
436 : 3003 : bh = sb_getblk(sb, bitmap_blk);
437 [ - + ]: 3003 : if (unlikely(!bh)) {
438 : 0 : ext4_warning(sb, "Cannot get buffer for block bitmap - "
439 : : "block_group = %u, block_bitmap = %llu",
440 : : block_group, bitmap_blk);
441 : 0 : return ERR_PTR(-ENOMEM);
442 : : }
443 : :
444 [ + + ]: 3003 : if (bitmap_uptodate(bh))
445 : 2970 : goto verify;
446 : :
447 : 33 : lock_buffer(bh);
448 [ - + ]: 33 : if (bitmap_uptodate(bh)) {
449 : 0 : unlock_buffer(bh);
450 : 0 : goto verify;
451 : : }
452 : 33 : ext4_lock_group(sb, block_group);
453 [ + - ]: 33 : if (ext4_has_group_desc_csum(sb) &&
454 [ - + ]: 33 : (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))) {
455 [ # # ]: 0 : if (block_group == 0) {
456 : 0 : ext4_unlock_group(sb, block_group);
457 : 0 : unlock_buffer(bh);
458 : 0 : ext4_error(sb, "Block bitmap for bg 0 marked "
459 : : "uninitialized");
460 : 0 : err = -EFSCORRUPTED;
461 : 0 : goto out;
462 : : }
463 : 0 : err = ext4_init_block_bitmap(sb, bh, block_group, desc);
464 : 0 : set_bitmap_uptodate(bh);
465 : 0 : set_buffer_uptodate(bh);
466 : 0 : set_buffer_verified(bh);
467 : 0 : ext4_unlock_group(sb, block_group);
468 : 0 : unlock_buffer(bh);
469 [ # # ]: 0 : if (err) {
470 : 0 : ext4_error(sb, "Failed to init block bitmap for group "
471 : : "%u: %d", block_group, err);
472 : 0 : goto out;
473 : : }
474 : 0 : goto verify;
475 : : }
476 : 33 : ext4_unlock_group(sb, block_group);
477 [ - + ]: 33 : if (buffer_uptodate(bh)) {
478 : : /*
479 : : * if not uninit if bh is uptodate,
480 : : * bitmap is also uptodate
481 : : */
482 : 0 : set_bitmap_uptodate(bh);
483 : 0 : unlock_buffer(bh);
484 : 0 : goto verify;
485 : : }
486 : : /*
487 : : * submit the buffer_head for reading
488 : : */
489 : 33 : set_buffer_new(bh);
490 : 33 : trace_ext4_read_block_bitmap_load(sb, block_group);
491 : 33 : bh->b_end_io = ext4_end_bitmap_read;
492 : 33 : get_bh(bh);
493 : 33 : submit_bh(REQ_OP_READ, REQ_META | REQ_PRIO, bh);
494 : 33 : return bh;
495 : 2970 : verify:
496 : 2970 : err = ext4_validate_block_bitmap(sb, desc, block_group, bh);
497 [ - + ]: 2970 : if (err)
498 : 0 : goto out;
499 : : return bh;
500 : 0 : out:
501 : 0 : put_bh(bh);
502 : 0 : return ERR_PTR(err);
503 : : }
504 : :
505 : : /* Returns 0 on success, 1 on error */
506 : 3003 : int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group,
507 : : struct buffer_head *bh)
508 : : {
509 : 3003 : struct ext4_group_desc *desc;
510 : :
511 [ + + ]: 3003 : if (!buffer_new(bh))
512 : : return 0;
513 : 33 : desc = ext4_get_group_desc(sb, block_group, NULL);
514 [ + - ]: 33 : if (!desc)
515 : : return -EFSCORRUPTED;
516 : 33 : wait_on_buffer(bh);
517 : 33 : ext4_simulate_fail_bh(sb, bh, EXT4_SIM_BBITMAP_EIO);
518 [ - + ]: 33 : if (!buffer_uptodate(bh)) {
519 : 0 : ext4_set_errno(sb, EIO);
520 : 0 : ext4_error(sb, "Cannot read block bitmap - "
521 : : "block_group = %u, block_bitmap = %llu",
522 : : block_group, (unsigned long long) bh->b_blocknr);
523 : 0 : ext4_mark_group_bitmap_corrupted(sb, block_group,
524 : : EXT4_GROUP_INFO_BBITMAP_CORRUPT);
525 : 0 : return -EIO;
526 : : }
527 : 33 : clear_buffer_new(bh);
528 : : /* Panic or remount fs read-only if block bitmap is invalid */
529 : 33 : return ext4_validate_block_bitmap(sb, desc, block_group, bh);
530 : : }
531 : :
532 : : struct buffer_head *
533 : 2937 : ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
534 : : {
535 : 2937 : struct buffer_head *bh;
536 : 2937 : int err;
537 : :
538 : 2937 : bh = ext4_read_block_bitmap_nowait(sb, block_group);
539 [ + - ]: 2937 : if (IS_ERR(bh))
540 : : return bh;
541 : 2937 : err = ext4_wait_block_bitmap(sb, block_group, bh);
542 [ - + ]: 2937 : if (err) {
543 : 0 : put_bh(bh);
544 : 0 : return ERR_PTR(err);
545 : : }
546 : : return bh;
547 : : }
548 : :
549 : : /**
550 : : * ext4_has_free_clusters()
551 : : * @sbi: in-core super block structure.
552 : : * @nclusters: number of needed blocks
553 : : * @flags: flags from ext4_mb_new_blocks()
554 : : *
555 : : * Check if filesystem has nclusters free & available for allocation.
556 : : * On success return 1, return 0 on failure.
557 : : */
558 : 10451 : static int ext4_has_free_clusters(struct ext4_sb_info *sbi,
559 : : s64 nclusters, unsigned int flags)
560 : : {
561 : 10451 : s64 free_clusters, dirty_clusters, rsv, resv_clusters;
562 : 10451 : struct percpu_counter *fcc = &sbi->s_freeclusters_counter;
563 : 10451 : struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter;
564 : :
565 : 10451 : free_clusters = percpu_counter_read_positive(fcc);
566 : 10451 : dirty_clusters = percpu_counter_read_positive(dcc);
567 : 10451 : resv_clusters = atomic64_read(&sbi->s_resv_clusters);
568 : :
569 : : /*
570 : : * r_blocks_count should always be multiple of the cluster ratio so
571 : : * we are safe to do a plane bit shift only.
572 : : */
573 [ - + ]: 10451 : rsv = (ext4_r_blocks_count(sbi->s_es) >> sbi->s_cluster_bits) +
574 : : resv_clusters;
575 : :
576 : 10451 : if (free_clusters - (nclusters + rsv + dirty_clusters) <
577 [ - + ]: 10451 : EXT4_FREECLUSTERS_WATERMARK) {
578 : 0 : free_clusters = percpu_counter_sum_positive(fcc);
579 : 0 : dirty_clusters = percpu_counter_sum_positive(dcc);
580 : : }
581 : : /* Check whether we have space after accounting for current
582 : : * dirty clusters & root reserved clusters.
583 : : */
584 [ - + ]: 10451 : if (free_clusters >= (rsv + nclusters + dirty_clusters))
585 : : return 1;
586 : :
587 : : /* Hm, nope. Are (enough) root reserved clusters available? */
588 [ # # # # ]: 0 : if (uid_eq(sbi->s_resuid, current_fsuid()) ||
589 [ # # # # ]: 0 : (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) ||
590 : 0 : capable(CAP_SYS_RESOURCE) ||
591 [ # # ]: 0 : (flags & EXT4_MB_USE_ROOT_BLOCKS)) {
592 : :
593 [ # # ]: 0 : if (free_clusters >= (nclusters + dirty_clusters +
594 : : resv_clusters))
595 : : return 1;
596 : : }
597 : : /* No free blocks. Let's see if we can dip into reserved pool */
598 [ # # ]: 0 : if (flags & EXT4_MB_USE_RESERVED) {
599 [ # # ]: 0 : if (free_clusters >= (nclusters + dirty_clusters))
600 : 0 : return 1;
601 : : }
602 : :
603 : : return 0;
604 : : }
605 : :
606 : 10451 : int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
607 : : s64 nclusters, unsigned int flags)
608 : : {
609 [ + - ]: 10451 : if (ext4_has_free_clusters(sbi, nclusters, flags)) {
610 : 10451 : percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters);
611 : 10451 : return 0;
612 : : } else
613 : : return -ENOSPC;
614 : : }
615 : :
616 : : /**
617 : : * ext4_should_retry_alloc() - check if a block allocation should be retried
618 : : * @sb: super block
619 : : * @retries: number of attemps has been made
620 : : *
621 : : * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
622 : : * it is profitable to retry the operation, this function will wait
623 : : * for the current or committing transaction to complete, and then
624 : : * return TRUE. We will only retry once.
625 : : */
626 : 0 : int ext4_should_retry_alloc(struct super_block *sb, int *retries)
627 : : {
628 [ # # ]: 0 : if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) ||
629 [ # # # # ]: 0 : (*retries)++ > 1 ||
630 [ # # ]: 0 : !EXT4_SB(sb)->s_journal)
631 : 0 : return 0;
632 : :
633 : 0 : smp_mb();
634 [ # # ]: 0 : if (EXT4_SB(sb)->s_mb_free_pending == 0)
635 : : return 0;
636 : :
637 : 0 : jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
638 : 0 : jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
639 : 0 : return 1;
640 : : }
641 : :
642 : : /*
643 : : * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
644 : : *
645 : : * @handle: handle to this transaction
646 : : * @inode: file inode
647 : : * @goal: given target block(filesystem wide)
648 : : * @count: pointer to total number of clusters needed
649 : : * @errp: error code
650 : : *
651 : : * Return 1st allocated block number on success, *count stores total account
652 : : * error stores in errp pointer
653 : : */
654 : 0 : ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
655 : : ext4_fsblk_t goal, unsigned int flags,
656 : : unsigned long *count, int *errp)
657 : : {
658 : 0 : struct ext4_allocation_request ar;
659 : 0 : ext4_fsblk_t ret;
660 : :
661 : 0 : memset(&ar, 0, sizeof(ar));
662 : : /* Fill with neighbour allocated blocks */
663 : 0 : ar.inode = inode;
664 : 0 : ar.goal = goal;
665 [ # # ]: 0 : ar.len = count ? *count : 1;
666 : 0 : ar.flags = flags;
667 : :
668 : 0 : ret = ext4_mb_new_blocks(handle, &ar, errp);
669 [ # # ]: 0 : if (count)
670 : 0 : *count = ar.len;
671 : : /*
672 : : * Account for the allocated meta blocks. We will never
673 : : * fail EDQUOT for metdata, but we do account for it.
674 : : */
675 [ # # # # ]: 0 : if (!(*errp) && (flags & EXT4_MB_DELALLOC_RESERVED)) {
676 : 0 : dquot_alloc_block_nofail(inode,
677 : 0 : EXT4_C2B(EXT4_SB(inode->i_sb), ar.len));
678 : : }
679 : 0 : return ret;
680 : : }
681 : :
682 : : /**
683 : : * ext4_count_free_clusters() -- count filesystem free clusters
684 : : * @sb: superblock
685 : : *
686 : : * Adds up the number of free clusters from each block group.
687 : : */
688 : 22 : ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb)
689 : : {
690 : 22 : ext4_fsblk_t desc_count;
691 : 22 : struct ext4_group_desc *gdp;
692 : 22 : ext4_group_t i;
693 : 22 : ext4_group_t ngroups = ext4_get_groups_count(sb);
694 : 22 : struct ext4_group_info *grp;
695 : : #ifdef EXT4FS_DEBUG
696 : : struct ext4_super_block *es;
697 : : ext4_fsblk_t bitmap_count;
698 : : unsigned int x;
699 : : struct buffer_head *bitmap_bh = NULL;
700 : :
701 : : es = EXT4_SB(sb)->s_es;
702 : : desc_count = 0;
703 : : bitmap_count = 0;
704 : : gdp = NULL;
705 : :
706 : : for (i = 0; i < ngroups; i++) {
707 : : gdp = ext4_get_group_desc(sb, i, NULL);
708 : : if (!gdp)
709 : : continue;
710 : : grp = NULL;
711 : : if (EXT4_SB(sb)->s_group_info)
712 : : grp = ext4_get_group_info(sb, i);
713 : : if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
714 : : desc_count += ext4_free_group_clusters(sb, gdp);
715 : : brelse(bitmap_bh);
716 : : bitmap_bh = ext4_read_block_bitmap(sb, i);
717 : : if (IS_ERR(bitmap_bh)) {
718 : : bitmap_bh = NULL;
719 : : continue;
720 : : }
721 : :
722 : : x = ext4_count_free(bitmap_bh->b_data,
723 : : EXT4_CLUSTERS_PER_GROUP(sb) / 8);
724 : : printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
725 : : i, ext4_free_group_clusters(sb, gdp), x);
726 : : bitmap_count += x;
727 : : }
728 : : brelse(bitmap_bh);
729 : : printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu"
730 : : ", computed = %llu, %llu\n",
731 : : EXT4_NUM_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)),
732 : : desc_count, bitmap_count);
733 : : return bitmap_count;
734 : : #else
735 : 22 : desc_count = 0;
736 [ + + ]: 1100 : for (i = 0; i < ngroups; i++) {
737 : 1056 : gdp = ext4_get_group_desc(sb, i, NULL);
738 [ - + ]: 1056 : if (!gdp)
739 : 0 : continue;
740 : 1056 : grp = NULL;
741 [ + - ]: 1056 : if (EXT4_SB(sb)->s_group_info)
742 : 1056 : grp = ext4_get_group_info(sb, i);
743 [ + - + - ]: 1056 : if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
744 : 1056 : desc_count += ext4_free_group_clusters(sb, gdp);
745 : : }
746 : :
747 : 22 : return desc_count;
748 : : #endif
749 : : }
750 : :
751 : : static inline int test_root(ext4_group_t a, int b)
752 : : {
753 : 8536 : while (1) {
754 [ + - + + : 7095 : if (a < b)
+ + ]
755 : : return 0;
756 [ + + + + : 6886 : if (a == b)
+ + ]
757 : : return 1;
758 [ + + + + : 6248 : if ((a % b) != 0)
+ + ]
759 : : return 0;
760 : 1441 : a = a / b;
761 : : }
762 : : }
763 : :
764 : : /**
765 : : * ext4_bg_has_super - number of blocks used by the superblock in group
766 : : * @sb: superblock for filesystem
767 : : * @group: group number to check
768 : : *
769 : : * Return the number of blocks used by the superblock (primary or backup)
770 : : * in this group. Currently this will be only 0 or 1.
771 : : */
772 : 4246 : int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
773 : : {
774 [ + + ]: 4246 : struct ext4_super_block *es = EXT4_SB(sb)->s_es;
775 : :
776 [ + + ]: 4246 : if (group == 0)
777 : : return 1;
778 [ - + ]: 4136 : if (ext4_has_feature_sparse_super2(sb)) {
779 [ # # ]: 0 : if (group == le32_to_cpu(es->s_backup_bgs[0]) ||
780 [ # # ]: 0 : group == le32_to_cpu(es->s_backup_bgs[1]))
781 : : return 1;
782 : 0 : return 0;
783 : : }
784 [ + + + - ]: 4136 : if ((group <= 1) || !ext4_has_feature_sparse_super(sb))
785 : : return 1;
786 [ + + ]: 4048 : if (!(group & 1))
787 : : return 0;
788 [ + + + + : 5654 : if (test_root(group, 3) || (test_root(group, 5)) ||
+ + ]
789 : : test_root(group, 7))
790 : 638 : return 1;
791 : :
792 : : return 0;
793 : : }
794 : :
795 : 0 : static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
796 : : ext4_group_t group)
797 : : {
798 : 0 : unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
799 : 0 : ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
800 : 0 : ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
801 : :
802 [ # # # # : 0 : if (group == first || group == first + 1 || group == last)
# # ]
803 : 0 : return 1;
804 : : return 0;
805 : : }
806 : :
807 : 1320 : static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
808 : : ext4_group_t group)
809 : : {
810 [ + + ]: 1320 : if (!ext4_bg_has_super(sb, group))
811 : : return 0;
812 : :
813 [ - + ]: 429 : if (ext4_has_feature_meta_bg(sb))
814 : 0 : return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
815 : : else
816 : 429 : return EXT4_SB(sb)->s_gdb_count;
817 : : }
818 : :
819 : : /**
820 : : * ext4_bg_num_gdb - number of blocks used by the group table in group
821 : : * @sb: superblock for filesystem
822 : : * @group: group number to check
823 : : *
824 : : * Return the number of blocks used by the group descriptor table
825 : : * (primary or backup) in this group. In the future there may be a
826 : : * different number of descriptor blocks in each group.
827 : : */
828 : 1320 : unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
829 : : {
830 : 1320 : unsigned long first_meta_bg =
831 [ - + ]: 1320 : le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
832 : 1320 : unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
833 : :
834 [ - + - - ]: 1320 : if (!ext4_has_feature_meta_bg(sb) || metagroup < first_meta_bg)
835 : 1320 : return ext4_bg_num_gdb_nometa(sb, group);
836 : :
837 [ # # ]: 0 : return ext4_bg_num_gdb_meta(sb,group);
838 : :
839 : : }
840 : :
841 : : /*
842 : : * This function returns the number of file system metadata clusters at
843 : : * the beginning of a block group, including the reserved gdt blocks.
844 : : */
845 : 814 : static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
846 : : ext4_group_t block_group)
847 : : {
848 : 814 : struct ext4_sb_info *sbi = EXT4_SB(sb);
849 : 814 : unsigned num;
850 : :
851 : : /* Check for superblock and gdt backups in this group */
852 : 814 : num = ext4_bg_has_super(sb, block_group);
853 : :
854 [ - + ]: 814 : if (!ext4_has_feature_meta_bg(sb) ||
855 : 0 : block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
856 [ # # ]: 0 : sbi->s_desc_per_block) {
857 [ + + ]: 814 : if (num) {
858 : 77 : num += ext4_bg_num_gdb(sb, block_group);
859 : 77 : num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
860 : : }
861 : : } else { /* For META_BG_BLOCK_GROUPS */
862 : 0 : num += ext4_bg_num_gdb(sb, block_group);
863 : : }
864 : 814 : return EXT4_NUM_B2C(sbi, num);
865 : : }
866 : : /**
867 : : * ext4_inode_to_goal_block - return a hint for block allocation
868 : : * @inode: inode for block allocation
869 : : *
870 : : * Return the ideal location to start allocating blocks for a
871 : : * newly created inode.
872 : : */
873 : 2827 : ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode)
874 : : {
875 : 2827 : struct ext4_inode_info *ei = EXT4_I(inode);
876 : 2827 : ext4_group_t block_group;
877 : 2827 : ext4_grpblk_t colour;
878 [ + - ]: 2827 : int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
879 : 2827 : ext4_fsblk_t bg_start;
880 : 2827 : ext4_fsblk_t last_block;
881 : :
882 : 2827 : block_group = ei->i_block_group;
883 [ + - ]: 2827 : if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
884 : : /*
885 : : * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
886 : : * block groups per flexgroup, reserve the first block
887 : : * group for directories and special files. Regular
888 : : * files will start at the second block group. This
889 : : * tends to speed up directory access and improves
890 : : * fsck times.
891 : : */
892 : 2827 : block_group &= ~(flex_size-1);
893 [ + + ]: 2827 : if (S_ISREG(inode->i_mode))
894 : 88 : block_group++;
895 : : }
896 [ - + ]: 2827 : bg_start = ext4_group_first_block_no(inode->i_sb, block_group);
897 [ - + ]: 2827 : last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;
898 : :
899 : : /*
900 : : * If we are doing delayed allocation, we don't need take
901 : : * colour into account.
902 : : */
903 [ - + ]: 2827 : if (test_opt(inode->i_sb, DELALLOC))
904 : : return bg_start;
905 : :
906 [ # # ]: 0 : if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
907 : 0 : colour = (current->pid % 16) *
908 : 0 : (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
909 : : else
910 : 0 : colour = (current->pid % 16) * ((last_block - bg_start) / 16);
911 : 0 : return bg_start + colour;
912 : : }
913 : :
|
