~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/ext4/balloc.c

Version: ~ [ linux-5.7 ] ~ [ linux-5.6.15 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.43 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.125 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.182 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.225 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.225 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.84 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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 ext4_group_t ext4_get_group_number(struct super_block *sb,
 36                                    ext4_fsblk_t block)
 37 {
 38         ext4_group_t group;
 39 
 40         if (test_opt2(sb, STD_GROUP_SIZE))
 41                 group = (block -
 42                          le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) >>
 43                         (EXT4_BLOCK_SIZE_BITS(sb) + EXT4_CLUSTER_BITS(sb) + 3);
 44         else
 45                 ext4_get_group_no_and_offset(sb, block, &group, NULL);
 46         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 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         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 57         ext4_grpblk_t offset;
 58 
 59         blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
 60         offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) >>
 61                 EXT4_SB(sb)->s_cluster_bits;
 62         if (offsetp)
 63                 *offsetp = offset;
 64         if (blockgrpp)
 65                 *blockgrpp = blocknr;
 66 
 67 }
 68 
 69 /*
 70  * Check whether the 'block' lives within the 'block_group'. Returns 1 if so
 71  * and 0 otherwise.
 72  */
 73 static inline int ext4_block_in_group(struct super_block *sb,
 74                                       ext4_fsblk_t block,
 75                                       ext4_group_t block_group)
 76 {
 77         ext4_group_t actual_group;
 78 
 79         actual_group = ext4_get_group_number(sb, block);
 80         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 static unsigned ext4_num_overhead_clusters(struct super_block *sb,
 87                                            ext4_group_t block_group,
 88                                            struct ext4_group_desc *gdp)
 89 {
 90         unsigned num_clusters;
 91         int block_cluster = -1, inode_cluster = -1, itbl_cluster = -1, i, c;
 92         ext4_fsblk_t start = ext4_group_first_block_no(sb, block_group);
 93         ext4_fsblk_t itbl_blk;
 94         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         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         if (ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), block_group)) {
114                 block_cluster = EXT4_B2C(sbi,
115                                          ext4_block_bitmap(sb, gdp) - start);
116                 if (block_cluster < num_clusters)
117                         block_cluster = -1;
118                 else if (block_cluster == num_clusters) {
119                         num_clusters++;
120                         block_cluster = -1;
121                 }
122         }
123 
124         if (ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), block_group)) {
125                 inode_cluster = EXT4_B2C(sbi,
126                                          ext4_inode_bitmap(sb, gdp) - start);
127                 if (inode_cluster < num_clusters)
128                         inode_cluster = -1;
129                 else if (inode_cluster == num_clusters) {
130                         num_clusters++;
131                         inode_cluster = -1;
132                 }
133         }
134 
135         itbl_blk = ext4_inode_table(sb, gdp);
136         for (i = 0; i < sbi->s_itb_per_group; i++) {
137                 if (ext4_block_in_group(sb, itbl_blk + i, block_group)) {
138                         c = EXT4_B2C(sbi, itbl_blk + i - start);
139                         if ((c < num_clusters) || (c == inode_cluster) ||
140                             (c == block_cluster) || (c == itbl_cluster))
141                                 continue;
142                         if (c == num_clusters) {
143                                 num_clusters++;
144                                 continue;
145                         }
146                         num_clusters++;
147                         itbl_cluster = c;
148                 }
149         }
150 
151         if (block_cluster != -1)
152                 num_clusters++;
153         if (inode_cluster != -1)
154                 num_clusters++;
155 
156         return num_clusters;
157 }
158 
159 static unsigned int num_clusters_in_group(struct super_block *sb,
160                                           ext4_group_t block_group)
161 {
162         unsigned int blocks;
163 
164         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                 blocks = ext4_blocks_count(EXT4_SB(sb)->s_es) -
172                         ext4_group_first_block_no(sb, block_group);
173         } else
174                 blocks = EXT4_BLOCKS_PER_GROUP(sb);
175         return EXT4_NUM_B2C(EXT4_SB(sb), blocks);
176 }
177 
178 /* Initializes an uninitialized block bitmap */
179 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         unsigned int bit, bit_max;
185         struct ext4_sb_info *sbi = EXT4_SB(sb);
186         ext4_fsblk_t start, tmp;
187 
188         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         if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) {
193                 ext4_mark_group_bitmap_corrupted(sb, block_group,
194                                         EXT4_GROUP_INFO_BBITMAP_CORRUPT |
195                                         EXT4_GROUP_INFO_IBITMAP_CORRUPT);
196                 return -EFSBADCRC;
197         }
198         memset(bh->b_data, 0, sb->s_blocksize);
199 
200         bit_max = ext4_num_base_meta_clusters(sb, block_group);
201         if ((bit_max >> 3) >= bh->b_size)
202                 return -EFSCORRUPTED;
203 
204         for (bit = 0; bit < bit_max; bit++)
205                 ext4_set_bit(bit, bh->b_data);
206 
207         start = ext4_group_first_block_no(sb, block_group);
208 
209         /* Set bits for block and inode bitmaps, and inode table */
210         tmp = ext4_block_bitmap(sb, gdp);
211         if (ext4_block_in_group(sb, tmp, block_group))
212                 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
213 
214         tmp = ext4_inode_bitmap(sb, gdp);
215         if (ext4_block_in_group(sb, tmp, block_group))
216                 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
217 
218         tmp = ext4_inode_table(sb, gdp);
219         for (; tmp < ext4_inode_table(sb, gdp) +
220                      sbi->s_itb_per_group; tmp++) {
221                 if (ext4_block_in_group(sb, tmp, block_group))
222                         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         ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group),
231                              sb->s_blocksize * 8, bh->b_data);
232         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 unsigned ext4_free_clusters_after_init(struct super_block *sb,
239                                        ext4_group_t block_group,
240                                        struct ext4_group_desc *gdp)
241 {
242         return num_clusters_in_group(sb, block_group) - 
243                 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 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         unsigned int group_desc;
269         unsigned int offset;
270         ext4_group_t ngroups = ext4_get_groups_count(sb);
271         struct ext4_group_desc *desc;
272         struct ext4_sb_info *sbi = EXT4_SB(sb);
273 
274         if (block_group >= ngroups) {
275                 ext4_error(sb, "block_group >= groups_count - block_group = %u,"
276                            " groups_count = %u", block_group, ngroups);
277 
278                 return NULL;
279         }
280 
281         group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
282         offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
283         if (!sbi->s_group_desc[group_desc]) {
284                 ext4_error(sb, "Group descriptor not loaded - "
285                            "block_group = %u, group_desc = %u, desc = %u",
286                            block_group, group_desc, offset);
287                 return NULL;
288         }
289 
290         desc = (struct ext4_group_desc *)(
291                 (__u8 *)sbi->s_group_desc[group_desc]->b_data +
292                 offset * EXT4_DESC_SIZE(sb));
293         if (bh)
294                 *bh = sbi->s_group_desc[group_desc];
295         return desc;
296 }
297 
298 /*
299  * Return the block number which was discovered to be invalid, or 0 if
300  * the block bitmap is valid.
301  */
302 static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb,
303                                             struct ext4_group_desc *desc,
304                                             ext4_group_t block_group,
305                                             struct buffer_head *bh)
306 {
307         struct ext4_sb_info *sbi = EXT4_SB(sb);
308         ext4_grpblk_t offset;
309         ext4_grpblk_t next_zero_bit;
310         ext4_grpblk_t max_bit = EXT4_CLUSTERS_PER_GROUP(sb);
311         ext4_fsblk_t blk;
312         ext4_fsblk_t group_first_block;
313 
314         if (ext4_has_feature_flex_bg(sb)) {
315                 /* with FLEX_BG, the inode/block bitmaps and itable
316                  * blocks may not be in the group at all
317                  * so the bitmap validation will be skipped for those groups
318                  * or it has to also read the block group where the bitmaps
319                  * are located to verify they are set.
320                  */
321                 return 0;
322         }
323         group_first_block = ext4_group_first_block_no(sb, block_group);
324 
325         /* check whether block bitmap block number is set */
326         blk = ext4_block_bitmap(sb, desc);
327         offset = blk - group_first_block;
328         if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
329             !ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
330                 /* bad block bitmap */
331                 return blk;
332 
333         /* check whether the inode bitmap block number is set */
334         blk = ext4_inode_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 table block number is set */
342         blk = ext4_inode_table(sb, desc);
343         offset = blk - group_first_block;
344         if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
345             EXT4_B2C(sbi, offset + sbi->s_itb_per_group) >= max_bit)
346                 return blk;
347         next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
348                         EXT4_B2C(sbi, offset + sbi->s_itb_per_group),
349                         EXT4_B2C(sbi, offset));
350         if (next_zero_bit <
351             EXT4_B2C(sbi, offset + sbi->s_itb_per_group))
352                 /* bad bitmap for inode tables */
353                 return blk;
354         return 0;
355 }
356 
357 static int ext4_validate_block_bitmap(struct super_block *sb,
358                                       struct ext4_group_desc *desc,
359                                       ext4_group_t block_group,
360                                       struct buffer_head *bh)
361 {
362         ext4_fsblk_t    blk;
363         struct ext4_group_info *grp = ext4_get_group_info(sb, block_group);
364 
365         if (buffer_verified(bh))
366                 return 0;
367         if (EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
368                 return -EFSCORRUPTED;
369 
370         ext4_lock_group(sb, block_group);
371         if (buffer_verified(bh))
372                 goto verified;
373         if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group,
374                         desc, bh))) {
375                 ext4_unlock_group(sb, block_group);
376                 ext4_error(sb, "bg %u: bad block bitmap checksum", block_group);
377                 ext4_mark_group_bitmap_corrupted(sb, block_group,
378                                         EXT4_GROUP_INFO_BBITMAP_CORRUPT);
379                 return -EFSBADCRC;
380         }
381         blk = ext4_valid_block_bitmap(sb, desc, block_group, bh);
382         if (unlikely(blk != 0)) {
383                 ext4_unlock_group(sb, block_group);
384                 ext4_error(sb, "bg %u: block %llu: invalid block bitmap",
385                            block_group, blk);
386                 ext4_mark_group_bitmap_corrupted(sb, block_group,
387                                         EXT4_GROUP_INFO_BBITMAP_CORRUPT);
388                 return -EFSCORRUPTED;
389         }
390         set_buffer_verified(bh);
391 verified:
392         ext4_unlock_group(sb, block_group);
393         return 0;
394 }
395 
396 /**
397  * ext4_read_block_bitmap_nowait()
398  * @sb:                 super block
399  * @block_group:        given block group
400  *
401  * Read the bitmap for a given block_group,and validate the
402  * bits for block/inode/inode tables are set in the bitmaps
403  *
404  * Return buffer_head on success or NULL in case of failure.
405  */
406 struct buffer_head *
407 ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group)
408 {
409         struct ext4_group_desc *desc;
410         struct ext4_sb_info *sbi = EXT4_SB(sb);
411         struct buffer_head *bh;
412         ext4_fsblk_t bitmap_blk;
413         int err;
414 
415         desc = ext4_get_group_desc(sb, block_group, NULL);
416         if (!desc)
417                 return ERR_PTR(-EFSCORRUPTED);
418         bitmap_blk = ext4_block_bitmap(sb, desc);
419         if ((bitmap_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) ||
420             (bitmap_blk >= ext4_blocks_count(sbi->s_es))) {
421                 ext4_error(sb, "Invalid block bitmap block %llu in "
422                            "block_group %u", bitmap_blk, block_group);
423                 ext4_mark_group_bitmap_corrupted(sb, block_group,
424                                         EXT4_GROUP_INFO_BBITMAP_CORRUPT);
425                 return ERR_PTR(-EFSCORRUPTED);
426         }
427         bh = sb_getblk(sb, bitmap_blk);
428         if (unlikely(!bh)) {
429                 ext4_warning(sb, "Cannot get buffer for block bitmap - "
430                              "block_group = %u, block_bitmap = %llu",
431                              block_group, bitmap_blk);
432                 return ERR_PTR(-ENOMEM);
433         }
434 
435         if (bitmap_uptodate(bh))
436                 goto verify;
437 
438         lock_buffer(bh);
439         if (bitmap_uptodate(bh)) {
440                 unlock_buffer(bh);
441                 goto verify;
442         }
443         ext4_lock_group(sb, block_group);
444         if (ext4_has_group_desc_csum(sb) &&
445             (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))) {
446                 if (block_group == 0) {
447                         ext4_unlock_group(sb, block_group);
448                         unlock_buffer(bh);
449                         ext4_error(sb, "Block bitmap for bg 0 marked "
450                                    "uninitialized");
451                         err = -EFSCORRUPTED;
452                         goto out;
453                 }
454                 err = ext4_init_block_bitmap(sb, bh, block_group, desc);
455                 set_bitmap_uptodate(bh);
456                 set_buffer_uptodate(bh);
457                 set_buffer_verified(bh);
458                 ext4_unlock_group(sb, block_group);
459                 unlock_buffer(bh);
460                 if (err) {
461                         ext4_error(sb, "Failed to init block bitmap for group "
462                                    "%u: %d", block_group, err);
463                         goto out;
464                 }
465                 goto verify;
466         }
467         ext4_unlock_group(sb, block_group);
468         if (buffer_uptodate(bh)) {
469                 /*
470                  * if not uninit if bh is uptodate,
471                  * bitmap is also uptodate
472                  */
473                 set_bitmap_uptodate(bh);
474                 unlock_buffer(bh);
475                 goto verify;
476         }
477         /*
478          * submit the buffer_head for reading
479          */
480         set_buffer_new(bh);
481         trace_ext4_read_block_bitmap_load(sb, block_group);
482         bh->b_end_io = ext4_end_bitmap_read;
483         get_bh(bh);
484         submit_bh(REQ_OP_READ, REQ_META | REQ_PRIO, bh);
485         return bh;
486 verify:
487         err = ext4_validate_block_bitmap(sb, desc, block_group, bh);
488         if (err)
489                 goto out;
490         return bh;
491 out:
492         put_bh(bh);
493         return ERR_PTR(err);
494 }
495 
496 /* Returns 0 on success, 1 on error */
497 int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group,
498                            struct buffer_head *bh)
499 {
500         struct ext4_group_desc *desc;
501 
502         if (!buffer_new(bh))
503                 return 0;
504         desc = ext4_get_group_desc(sb, block_group, NULL);
505         if (!desc)
506                 return -EFSCORRUPTED;
507         wait_on_buffer(bh);
508         if (!buffer_uptodate(bh)) {
509                 ext4_error(sb, "Cannot read block bitmap - "
510                            "block_group = %u, block_bitmap = %llu",
511                            block_group, (unsigned long long) bh->b_blocknr);
512                 ext4_mark_group_bitmap_corrupted(sb, block_group,
513                                         EXT4_GROUP_INFO_BBITMAP_CORRUPT);
514                 return -EIO;
515         }
516         clear_buffer_new(bh);
517         /* Panic or remount fs read-only if block bitmap is invalid */
518         return ext4_validate_block_bitmap(sb, desc, block_group, bh);
519 }
520 
521 struct buffer_head *
522 ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
523 {
524         struct buffer_head *bh;
525         int err;
526 
527         bh = ext4_read_block_bitmap_nowait(sb, block_group);
528         if (IS_ERR(bh))
529                 return bh;
530         err = ext4_wait_block_bitmap(sb, block_group, bh);
531         if (err) {
532                 put_bh(bh);
533                 return ERR_PTR(err);
534         }
535         return bh;
536 }
537 
538 /**
539  * ext4_has_free_clusters()
540  * @sbi:        in-core super block structure.
541  * @nclusters:  number of needed blocks
542  * @flags:      flags from ext4_mb_new_blocks()
543  *
544  * Check if filesystem has nclusters free & available for allocation.
545  * On success return 1, return 0 on failure.
546  */
547 static int ext4_has_free_clusters(struct ext4_sb_info *sbi,
548                                   s64 nclusters, unsigned int flags)
549 {
550         s64 free_clusters, dirty_clusters, rsv, resv_clusters;
551         struct percpu_counter *fcc = &sbi->s_freeclusters_counter;
552         struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter;
553 
554         free_clusters  = percpu_counter_read_positive(fcc);
555         dirty_clusters = percpu_counter_read_positive(dcc);
556         resv_clusters = atomic64_read(&sbi->s_resv_clusters);
557 
558         /*
559          * r_blocks_count should always be multiple of the cluster ratio so
560          * we are safe to do a plane bit shift only.
561          */
562         rsv = (ext4_r_blocks_count(sbi->s_es) >> sbi->s_cluster_bits) +
563               resv_clusters;
564 
565         if (free_clusters - (nclusters + rsv + dirty_clusters) <
566                                         EXT4_FREECLUSTERS_WATERMARK) {
567                 free_clusters  = percpu_counter_sum_positive(fcc);
568                 dirty_clusters = percpu_counter_sum_positive(dcc);
569         }
570         /* Check whether we have space after accounting for current
571          * dirty clusters & root reserved clusters.
572          */
573         if (free_clusters >= (rsv + nclusters + dirty_clusters))
574                 return 1;
575 
576         /* Hm, nope.  Are (enough) root reserved clusters available? */
577         if (uid_eq(sbi->s_resuid, current_fsuid()) ||
578             (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) ||
579             capable(CAP_SYS_RESOURCE) ||
580             (flags & EXT4_MB_USE_ROOT_BLOCKS)) {
581 
582                 if (free_clusters >= (nclusters + dirty_clusters +
583                                       resv_clusters))
584                         return 1;
585         }
586         /* No free blocks. Let's see if we can dip into reserved pool */
587         if (flags & EXT4_MB_USE_RESERVED) {
588                 if (free_clusters >= (nclusters + dirty_clusters))
589                         return 1;
590         }
591 
592         return 0;
593 }
594 
595 int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
596                              s64 nclusters, unsigned int flags)
597 {
598         if (ext4_has_free_clusters(sbi, nclusters, flags)) {
599                 percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters);
600                 return 0;
601         } else
602                 return -ENOSPC;
603 }
604 
605 /**
606  * ext4_should_retry_alloc()
607  * @sb:                 super block
608  * @retries             number of attemps has been made
609  *
610  * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
611  * it is profitable to retry the operation, this function will wait
612  * for the current or committing transaction to complete, and then
613  * return TRUE.  We will only retry once.
614  */
615 int ext4_should_retry_alloc(struct super_block *sb, int *retries)
616 {
617         if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) ||
618             (*retries)++ > 1 ||
619             !EXT4_SB(sb)->s_journal)
620                 return 0;
621 
622         smp_mb();
623         if (EXT4_SB(sb)->s_mb_free_pending == 0)
624                 return 0;
625 
626         jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
627         jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
628         return 1;
629 }
630 
631 /*
632  * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
633  *
634  * @handle:             handle to this transaction
635  * @inode:              file inode
636  * @goal:               given target block(filesystem wide)
637  * @count:              pointer to total number of clusters needed
638  * @errp:               error code
639  *
640  * Return 1st allocated block number on success, *count stores total account
641  * error stores in errp pointer
642  */
643 ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
644                                   ext4_fsblk_t goal, unsigned int flags,
645                                   unsigned long *count, int *errp)
646 {
647         struct ext4_allocation_request ar;
648         ext4_fsblk_t ret;
649 
650         memset(&ar, 0, sizeof(ar));
651         /* Fill with neighbour allocated blocks */
652         ar.inode = inode;
653         ar.goal = goal;
654         ar.len = count ? *count : 1;
655         ar.flags = flags;
656 
657         ret = ext4_mb_new_blocks(handle, &ar, errp);
658         if (count)
659                 *count = ar.len;
660         /*
661          * Account for the allocated meta blocks.  We will never
662          * fail EDQUOT for metdata, but we do account for it.
663          */
664         if (!(*errp) && (flags & EXT4_MB_DELALLOC_RESERVED)) {
665                 dquot_alloc_block_nofail(inode,
666                                 EXT4_C2B(EXT4_SB(inode->i_sb), ar.len));
667         }
668         return ret;
669 }
670 
671 /**
672  * ext4_count_free_clusters() -- count filesystem free clusters
673  * @sb:         superblock
674  *
675  * Adds up the number of free clusters from each block group.
676  */
677 ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb)
678 {
679         ext4_fsblk_t desc_count;
680         struct ext4_group_desc *gdp;
681         ext4_group_t i;
682         ext4_group_t ngroups = ext4_get_groups_count(sb);
683         struct ext4_group_info *grp;
684 #ifdef EXT4FS_DEBUG
685         struct ext4_super_block *es;
686         ext4_fsblk_t bitmap_count;
687         unsigned int x;
688         struct buffer_head *bitmap_bh = NULL;
689 
690         es = EXT4_SB(sb)->s_es;
691         desc_count = 0;
692         bitmap_count = 0;
693         gdp = NULL;
694 
695         for (i = 0; i < ngroups; i++) {
696                 gdp = ext4_get_group_desc(sb, i, NULL);
697                 if (!gdp)
698                         continue;
699                 grp = NULL;
700                 if (EXT4_SB(sb)->s_group_info)
701                         grp = ext4_get_group_info(sb, i);
702                 if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
703                         desc_count += ext4_free_group_clusters(sb, gdp);
704                 brelse(bitmap_bh);
705                 bitmap_bh = ext4_read_block_bitmap(sb, i);
706                 if (IS_ERR(bitmap_bh)) {
707                         bitmap_bh = NULL;
708                         continue;
709                 }
710 
711                 x = ext4_count_free(bitmap_bh->b_data,
712                                     EXT4_CLUSTERS_PER_GROUP(sb) / 8);
713                 printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
714                         i, ext4_free_group_clusters(sb, gdp), x);
715                 bitmap_count += x;
716         }
717         brelse(bitmap_bh);
718         printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu"
719                ", computed = %llu, %llu\n",
720                EXT4_NUM_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)),
721                desc_count, bitmap_count);
722         return bitmap_count;
723 #else
724         desc_count = 0;
725         for (i = 0; i < ngroups; i++) {
726                 gdp = ext4_get_group_desc(sb, i, NULL);
727                 if (!gdp)
728                         continue;
729                 grp = NULL;
730                 if (EXT4_SB(sb)->s_group_info)
731                         grp = ext4_get_group_info(sb, i);
732                 if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
733                         desc_count += ext4_free_group_clusters(sb, gdp);
734         }
735 
736         return desc_count;
737 #endif
738 }
739 
740 static inline int test_root(ext4_group_t a, int b)
741 {
742         while (1) {
743                 if (a < b)
744                         return 0;
745                 if (a == b)
746                         return 1;
747                 if ((a % b) != 0)
748                         return 0;
749                 a = a / b;
750         }
751 }
752 
753 /**
754  *      ext4_bg_has_super - number of blocks used by the superblock in group
755  *      @sb: superblock for filesystem
756  *      @group: group number to check
757  *
758  *      Return the number of blocks used by the superblock (primary or backup)
759  *      in this group.  Currently this will be only 0 or 1.
760  */
761 int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
762 {
763         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
764 
765         if (group == 0)
766                 return 1;
767         if (ext4_has_feature_sparse_super2(sb)) {
768                 if (group == le32_to_cpu(es->s_backup_bgs[0]) ||
769                     group == le32_to_cpu(es->s_backup_bgs[1]))
770                         return 1;
771                 return 0;
772         }
773         if ((group <= 1) || !ext4_has_feature_sparse_super(sb))
774                 return 1;
775         if (!(group & 1))
776                 return 0;
777         if (test_root(group, 3) || (test_root(group, 5)) ||
778             test_root(group, 7))
779                 return 1;
780 
781         return 0;
782 }
783 
784 static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
785                                         ext4_group_t group)
786 {
787         unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
788         ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
789         ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
790 
791         if (group == first || group == first + 1 || group == last)
792                 return 1;
793         return 0;
794 }
795 
796 static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
797                                         ext4_group_t group)
798 {
799         if (!ext4_bg_has_super(sb, group))
800                 return 0;
801 
802         if (ext4_has_feature_meta_bg(sb))
803                 return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
804         else
805                 return EXT4_SB(sb)->s_gdb_count;
806 }
807 
808 /**
809  *      ext4_bg_num_gdb - number of blocks used by the group table in group
810  *      @sb: superblock for filesystem
811  *      @group: group number to check
812  *
813  *      Return the number of blocks used by the group descriptor table
814  *      (primary or backup) in this group.  In the future there may be a
815  *      different number of descriptor blocks in each group.
816  */
817 unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
818 {
819         unsigned long first_meta_bg =
820                         le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
821         unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
822 
823         if (!ext4_has_feature_meta_bg(sb) || metagroup < first_meta_bg)
824                 return ext4_bg_num_gdb_nometa(sb, group);
825 
826         return ext4_bg_num_gdb_meta(sb,group);
827 
828 }
829 
830 /*
831  * This function returns the number of file system metadata clusters at
832  * the beginning of a block group, including the reserved gdt blocks.
833  */
834 static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
835                                      ext4_group_t block_group)
836 {
837         struct ext4_sb_info *sbi = EXT4_SB(sb);
838         unsigned num;
839 
840         /* Check for superblock and gdt backups in this group */
841         num = ext4_bg_has_super(sb, block_group);
842 
843         if (!ext4_has_feature_meta_bg(sb) ||
844             block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
845                           sbi->s_desc_per_block) {
846                 if (num) {
847                         num += ext4_bg_num_gdb(sb, block_group);
848                         num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
849                 }
850         } else { /* For META_BG_BLOCK_GROUPS */
851                 num += ext4_bg_num_gdb(sb, block_group);
852         }
853         return EXT4_NUM_B2C(sbi, num);
854 }
855 /**
856  *      ext4_inode_to_goal_block - return a hint for block allocation
857  *      @inode: inode for block allocation
858  *
859  *      Return the ideal location to start allocating blocks for a
860  *      newly created inode.
861  */
862 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode)
863 {
864         struct ext4_inode_info *ei = EXT4_I(inode);
865         ext4_group_t block_group;
866         ext4_grpblk_t colour;
867         int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
868         ext4_fsblk_t bg_start;
869         ext4_fsblk_t last_block;
870 
871         block_group = ei->i_block_group;
872         if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
873                 /*
874                  * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
875                  * block groups per flexgroup, reserve the first block
876                  * group for directories and special files.  Regular
877                  * files will start at the second block group.  This
878                  * tends to speed up directory access and improves
879                  * fsck times.
880                  */
881                 block_group &= ~(flex_size-1);
882                 if (S_ISREG(inode->i_mode))
883                         block_group++;
884         }
885         bg_start = ext4_group_first_block_no(inode->i_sb, block_group);
886         last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;
887 
888         /*
889          * If we are doing delayed allocation, we don't need take
890          * colour into account.
891          */
892         if (test_opt(inode->i_sb, DELALLOC))
893                 return bg_start;
894 
895         if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
896                 colour = (current->pid % 16) *
897                         (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
898         else
899                 colour = (current->pid % 16) * ((last_block - bg_start) / 16);
900         return bg_start + colour;
901 }
902 
903 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp