balloc.c 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743
  1. /*
  2. * linux/fs/ext4/balloc.c
  3. *
  4. * Copyright (C) 1992, 1993, 1994, 1995
  5. * Remy Card (card@masi.ibp.fr)
  6. * Laboratoire MASI - Institut Blaise Pascal
  7. * Universite Pierre et Marie Curie (Paris VI)
  8. *
  9. * Enhanced block allocation by Stephen Tweedie (sct@redhat.com), 1993
  10. * Big-endian to little-endian byte-swapping/bitmaps by
  11. * David S. Miller (davem@caip.rutgers.edu), 1995
  12. */
  13. #include <linux/time.h>
  14. #include <linux/capability.h>
  15. #include <linux/fs.h>
  16. #include <linux/jbd2.h>
  17. #include <linux/quotaops.h>
  18. #include <linux/buffer_head.h>
  19. #include "ext4.h"
  20. #include "ext4_jbd2.h"
  21. #include "mballoc.h"
  22. /*
  23. * balloc.c contains the blocks allocation and deallocation routines
  24. */
  25. /*
  26. * Calculate the block group number and offset, given a block number
  27. */
  28. void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
  29. ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
  30. {
  31. struct ext4_super_block *es = EXT4_SB(sb)->s_es;
  32. ext4_grpblk_t offset;
  33. blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
  34. offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb));
  35. if (offsetp)
  36. *offsetp = offset;
  37. if (blockgrpp)
  38. *blockgrpp = blocknr;
  39. }
  40. static int ext4_block_in_group(struct super_block *sb, ext4_fsblk_t block,
  41. ext4_group_t block_group)
  42. {
  43. ext4_group_t actual_group;
  44. ext4_get_group_no_and_offset(sb, block, &actual_group, NULL);
  45. if (actual_group == block_group)
  46. return 1;
  47. return 0;
  48. }
  49. static int ext4_group_used_meta_blocks(struct super_block *sb,
  50. ext4_group_t block_group,
  51. struct ext4_group_desc *gdp)
  52. {
  53. ext4_fsblk_t tmp;
  54. struct ext4_sb_info *sbi = EXT4_SB(sb);
  55. /* block bitmap, inode bitmap, and inode table blocks */
  56. int used_blocks = sbi->s_itb_per_group + 2;
  57. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
  58. if (!ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp),
  59. block_group))
  60. used_blocks--;
  61. if (!ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp),
  62. block_group))
  63. used_blocks--;
  64. tmp = ext4_inode_table(sb, gdp);
  65. for (; tmp < ext4_inode_table(sb, gdp) +
  66. sbi->s_itb_per_group; tmp++) {
  67. if (!ext4_block_in_group(sb, tmp, block_group))
  68. used_blocks -= 1;
  69. }
  70. }
  71. return used_blocks;
  72. }
  73. /* Initializes an uninitialized block bitmap if given, and returns the
  74. * number of blocks free in the group. */
  75. unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
  76. ext4_group_t block_group, struct ext4_group_desc *gdp)
  77. {
  78. int bit, bit_max;
  79. ext4_group_t ngroups = ext4_get_groups_count(sb);
  80. unsigned free_blocks, group_blocks;
  81. struct ext4_sb_info *sbi = EXT4_SB(sb);
  82. if (bh) {
  83. J_ASSERT_BH(bh, buffer_locked(bh));
  84. /* If checksum is bad mark all blocks used to prevent allocation
  85. * essentially implementing a per-group read-only flag. */
  86. if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) {
  87. ext4_error(sb, "Checksum bad for group %u",
  88. block_group);
  89. ext4_free_blks_set(sb, gdp, 0);
  90. ext4_free_inodes_set(sb, gdp, 0);
  91. ext4_itable_unused_set(sb, gdp, 0);
  92. memset(bh->b_data, 0xff, sb->s_blocksize);
  93. return 0;
  94. }
  95. memset(bh->b_data, 0, sb->s_blocksize);
  96. }
  97. /* Check for superblock and gdt backups in this group */
  98. bit_max = ext4_bg_has_super(sb, block_group);
  99. if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
  100. block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
  101. sbi->s_desc_per_block) {
  102. if (bit_max) {
  103. bit_max += ext4_bg_num_gdb(sb, block_group);
  104. bit_max +=
  105. le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
  106. }
  107. } else { /* For META_BG_BLOCK_GROUPS */
  108. bit_max += ext4_bg_num_gdb(sb, block_group);
  109. }
  110. if (block_group == ngroups - 1) {
  111. /*
  112. * Even though mke2fs always initialize first and last group
  113. * if some other tool enabled the EXT4_BG_BLOCK_UNINIT we need
  114. * to make sure we calculate the right free blocks
  115. */
  116. group_blocks = ext4_blocks_count(sbi->s_es) -
  117. ext4_group_first_block_no(sb, ngroups - 1);
  118. } else {
  119. group_blocks = EXT4_BLOCKS_PER_GROUP(sb);
  120. }
  121. free_blocks = group_blocks - bit_max;
  122. if (bh) {
  123. ext4_fsblk_t start, tmp;
  124. int flex_bg = 0;
  125. for (bit = 0; bit < bit_max; bit++)
  126. ext4_set_bit(bit, bh->b_data);
  127. start = ext4_group_first_block_no(sb, block_group);
  128. if (EXT4_HAS_INCOMPAT_FEATURE(sb,
  129. EXT4_FEATURE_INCOMPAT_FLEX_BG))
  130. flex_bg = 1;
  131. /* Set bits for block and inode bitmaps, and inode table */
  132. tmp = ext4_block_bitmap(sb, gdp);
  133. if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
  134. ext4_set_bit(tmp - start, bh->b_data);
  135. tmp = ext4_inode_bitmap(sb, gdp);
  136. if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
  137. ext4_set_bit(tmp - start, bh->b_data);
  138. tmp = ext4_inode_table(sb, gdp);
  139. for (; tmp < ext4_inode_table(sb, gdp) +
  140. sbi->s_itb_per_group; tmp++) {
  141. if (!flex_bg ||
  142. ext4_block_in_group(sb, tmp, block_group))
  143. ext4_set_bit(tmp - start, bh->b_data);
  144. }
  145. /*
  146. * Also if the number of blocks within the group is
  147. * less than the blocksize * 8 ( which is the size
  148. * of bitmap ), set rest of the block bitmap to 1
  149. */
  150. ext4_mark_bitmap_end(group_blocks, sb->s_blocksize * 8,
  151. bh->b_data);
  152. }
  153. return free_blocks - ext4_group_used_meta_blocks(sb, block_group, gdp);
  154. }
  155. /*
  156. * The free blocks are managed by bitmaps. A file system contains several
  157. * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
  158. * block for inodes, N blocks for the inode table and data blocks.
  159. *
  160. * The file system contains group descriptors which are located after the
  161. * super block. Each descriptor contains the number of the bitmap block and
  162. * the free blocks count in the block. The descriptors are loaded in memory
  163. * when a file system is mounted (see ext4_fill_super).
  164. */
  165. /**
  166. * ext4_get_group_desc() -- load group descriptor from disk
  167. * @sb: super block
  168. * @block_group: given block group
  169. * @bh: pointer to the buffer head to store the block
  170. * group descriptor
  171. */
  172. struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
  173. ext4_group_t block_group,
  174. struct buffer_head **bh)
  175. {
  176. unsigned int group_desc;
  177. unsigned int offset;
  178. ext4_group_t ngroups = ext4_get_groups_count(sb);
  179. struct ext4_group_desc *desc;
  180. struct ext4_sb_info *sbi = EXT4_SB(sb);
  181. if (block_group >= ngroups) {
  182. ext4_error(sb, "block_group >= groups_count - block_group = %u,"
  183. " groups_count = %u", block_group, ngroups);
  184. return NULL;
  185. }
  186. group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
  187. offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
  188. if (!sbi->s_group_desc[group_desc]) {
  189. ext4_error(sb, "Group descriptor not loaded - "
  190. "block_group = %u, group_desc = %u, desc = %u",
  191. block_group, group_desc, offset);
  192. return NULL;
  193. }
  194. desc = (struct ext4_group_desc *)(
  195. (__u8 *)sbi->s_group_desc[group_desc]->b_data +
  196. offset * EXT4_DESC_SIZE(sb));
  197. if (bh)
  198. *bh = sbi->s_group_desc[group_desc];
  199. return desc;
  200. }
  201. static int ext4_valid_block_bitmap(struct super_block *sb,
  202. struct ext4_group_desc *desc,
  203. unsigned int block_group,
  204. struct buffer_head *bh)
  205. {
  206. ext4_grpblk_t offset;
  207. ext4_grpblk_t next_zero_bit;
  208. ext4_fsblk_t bitmap_blk;
  209. ext4_fsblk_t group_first_block;
  210. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
  211. /* with FLEX_BG, the inode/block bitmaps and itable
  212. * blocks may not be in the group at all
  213. * so the bitmap validation will be skipped for those groups
  214. * or it has to also read the block group where the bitmaps
  215. * are located to verify they are set.
  216. */
  217. return 1;
  218. }
  219. group_first_block = ext4_group_first_block_no(sb, block_group);
  220. /* check whether block bitmap block number is set */
  221. bitmap_blk = ext4_block_bitmap(sb, desc);
  222. offset = bitmap_blk - group_first_block;
  223. if (!ext4_test_bit(offset, bh->b_data))
  224. /* bad block bitmap */
  225. goto err_out;
  226. /* check whether the inode bitmap block number is set */
  227. bitmap_blk = ext4_inode_bitmap(sb, desc);
  228. offset = bitmap_blk - group_first_block;
  229. if (!ext4_test_bit(offset, bh->b_data))
  230. /* bad block bitmap */
  231. goto err_out;
  232. /* check whether the inode table block number is set */
  233. bitmap_blk = ext4_inode_table(sb, desc);
  234. offset = bitmap_blk - group_first_block;
  235. next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
  236. offset + EXT4_SB(sb)->s_itb_per_group,
  237. offset);
  238. if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group)
  239. /* good bitmap for inode tables */
  240. return 1;
  241. err_out:
  242. ext4_error(sb, "Invalid block bitmap - block_group = %d, block = %llu",
  243. block_group, bitmap_blk);
  244. return 0;
  245. }
  246. /**
  247. * ext4_read_block_bitmap()
  248. * @sb: super block
  249. * @block_group: given block group
  250. *
  251. * Read the bitmap for a given block_group,and validate the
  252. * bits for block/inode/inode tables are set in the bitmaps
  253. *
  254. * Return buffer_head on success or NULL in case of failure.
  255. */
  256. struct buffer_head *
  257. ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
  258. {
  259. struct ext4_group_desc *desc;
  260. struct buffer_head *bh = NULL;
  261. ext4_fsblk_t bitmap_blk;
  262. desc = ext4_get_group_desc(sb, block_group, NULL);
  263. if (!desc)
  264. return NULL;
  265. bitmap_blk = ext4_block_bitmap(sb, desc);
  266. bh = sb_getblk(sb, bitmap_blk);
  267. if (unlikely(!bh)) {
  268. ext4_error(sb, "Cannot read block bitmap - "
  269. "block_group = %u, block_bitmap = %llu",
  270. block_group, bitmap_blk);
  271. return NULL;
  272. }
  273. if (bitmap_uptodate(bh))
  274. return bh;
  275. lock_buffer(bh);
  276. if (bitmap_uptodate(bh)) {
  277. unlock_buffer(bh);
  278. return bh;
  279. }
  280. ext4_lock_group(sb, block_group);
  281. if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
  282. ext4_init_block_bitmap(sb, bh, block_group, desc);
  283. set_bitmap_uptodate(bh);
  284. set_buffer_uptodate(bh);
  285. ext4_unlock_group(sb, block_group);
  286. unlock_buffer(bh);
  287. return bh;
  288. }
  289. ext4_unlock_group(sb, block_group);
  290. if (buffer_uptodate(bh)) {
  291. /*
  292. * if not uninit if bh is uptodate,
  293. * bitmap is also uptodate
  294. */
  295. set_bitmap_uptodate(bh);
  296. unlock_buffer(bh);
  297. return bh;
  298. }
  299. /*
  300. * submit the buffer_head for read. We can
  301. * safely mark the bitmap as uptodate now.
  302. * We do it here so the bitmap uptodate bit
  303. * get set with buffer lock held.
  304. */
  305. set_bitmap_uptodate(bh);
  306. if (bh_submit_read(bh) < 0) {
  307. put_bh(bh);
  308. ext4_error(sb, "Cannot read block bitmap - "
  309. "block_group = %u, block_bitmap = %llu",
  310. block_group, bitmap_blk);
  311. return NULL;
  312. }
  313. ext4_valid_block_bitmap(sb, desc, block_group, bh);
  314. /*
  315. * file system mounted not to panic on error,
  316. * continue with corrupt bitmap
  317. */
  318. return bh;
  319. }
  320. /**
  321. * ext4_add_groupblocks() -- Add given blocks to an existing group
  322. * @handle: handle to this transaction
  323. * @sb: super block
  324. * @block: start physcial block to add to the block group
  325. * @count: number of blocks to free
  326. *
  327. * This marks the blocks as free in the bitmap. We ask the
  328. * mballoc to reload the buddy after this by setting group
  329. * EXT4_GROUP_INFO_NEED_INIT_BIT flag
  330. */
  331. void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
  332. ext4_fsblk_t block, unsigned long count)
  333. {
  334. struct buffer_head *bitmap_bh = NULL;
  335. struct buffer_head *gd_bh;
  336. ext4_group_t block_group;
  337. ext4_grpblk_t bit;
  338. unsigned int i;
  339. struct ext4_group_desc *desc;
  340. struct ext4_sb_info *sbi = EXT4_SB(sb);
  341. int err = 0, ret, blk_free_count;
  342. ext4_grpblk_t blocks_freed;
  343. struct ext4_group_info *grp;
  344. ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1);
  345. ext4_get_group_no_and_offset(sb, block, &block_group, &bit);
  346. grp = ext4_get_group_info(sb, block_group);
  347. /*
  348. * Check to see if we are freeing blocks across a group
  349. * boundary.
  350. */
  351. if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) {
  352. goto error_return;
  353. }
  354. bitmap_bh = ext4_read_block_bitmap(sb, block_group);
  355. if (!bitmap_bh)
  356. goto error_return;
  357. desc = ext4_get_group_desc(sb, block_group, &gd_bh);
  358. if (!desc)
  359. goto error_return;
  360. if (in_range(ext4_block_bitmap(sb, desc), block, count) ||
  361. in_range(ext4_inode_bitmap(sb, desc), block, count) ||
  362. in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) ||
  363. in_range(block + count - 1, ext4_inode_table(sb, desc),
  364. sbi->s_itb_per_group)) {
  365. ext4_error(sb, "Adding blocks in system zones - "
  366. "Block = %llu, count = %lu",
  367. block, count);
  368. goto error_return;
  369. }
  370. /*
  371. * We are about to add blocks to the bitmap,
  372. * so we need undo access.
  373. */
  374. BUFFER_TRACE(bitmap_bh, "getting undo access");
  375. err = ext4_journal_get_undo_access(handle, bitmap_bh);
  376. if (err)
  377. goto error_return;
  378. /*
  379. * We are about to modify some metadata. Call the journal APIs
  380. * to unshare ->b_data if a currently-committing transaction is
  381. * using it
  382. */
  383. BUFFER_TRACE(gd_bh, "get_write_access");
  384. err = ext4_journal_get_write_access(handle, gd_bh);
  385. if (err)
  386. goto error_return;
  387. /*
  388. * make sure we don't allow a parallel init on other groups in the
  389. * same buddy cache
  390. */
  391. down_write(&grp->alloc_sem);
  392. for (i = 0, blocks_freed = 0; i < count; i++) {
  393. BUFFER_TRACE(bitmap_bh, "clear bit");
  394. if (!ext4_clear_bit_atomic(ext4_group_lock_ptr(sb, block_group),
  395. bit + i, bitmap_bh->b_data)) {
  396. ext4_error(sb, "bit already cleared for block %llu",
  397. (ext4_fsblk_t)(block + i));
  398. BUFFER_TRACE(bitmap_bh, "bit already cleared");
  399. } else {
  400. blocks_freed++;
  401. }
  402. }
  403. ext4_lock_group(sb, block_group);
  404. blk_free_count = blocks_freed + ext4_free_blks_count(sb, desc);
  405. ext4_free_blks_set(sb, desc, blk_free_count);
  406. desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc);
  407. ext4_unlock_group(sb, block_group);
  408. percpu_counter_add(&sbi->s_freeblocks_counter, blocks_freed);
  409. if (sbi->s_log_groups_per_flex) {
  410. ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
  411. atomic_add(blocks_freed,
  412. &sbi->s_flex_groups[flex_group].free_blocks);
  413. }
  414. /*
  415. * request to reload the buddy with the
  416. * new bitmap information
  417. */
  418. set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state));
  419. grp->bb_free += blocks_freed;
  420. up_write(&grp->alloc_sem);
  421. /* We dirtied the bitmap block */
  422. BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
  423. err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
  424. /* And the group descriptor block */
  425. BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
  426. ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh);
  427. if (!err)
  428. err = ret;
  429. error_return:
  430. brelse(bitmap_bh);
  431. ext4_std_error(sb, err);
  432. return;
  433. }
  434. /**
  435. * ext4_has_free_blocks()
  436. * @sbi: in-core super block structure.
  437. * @nblocks: number of needed blocks
  438. *
  439. * Check if filesystem has nblocks free & available for allocation.
  440. * On success return 1, return 0 on failure.
  441. */
  442. static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
  443. {
  444. s64 free_blocks, dirty_blocks, root_blocks;
  445. struct percpu_counter *fbc = &sbi->s_freeblocks_counter;
  446. struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter;
  447. free_blocks = percpu_counter_read_positive(fbc);
  448. dirty_blocks = percpu_counter_read_positive(dbc);
  449. root_blocks = ext4_r_blocks_count(sbi->s_es);
  450. if (free_blocks - (nblocks + root_blocks + dirty_blocks) <
  451. EXT4_FREEBLOCKS_WATERMARK) {
  452. free_blocks = percpu_counter_sum_positive(fbc);
  453. dirty_blocks = percpu_counter_sum_positive(dbc);
  454. if (dirty_blocks < 0) {
  455. printk(KERN_CRIT "Dirty block accounting "
  456. "went wrong %lld\n",
  457. (long long)dirty_blocks);
  458. }
  459. }
  460. /* Check whether we have space after
  461. * accounting for current dirty blocks & root reserved blocks.
  462. */
  463. if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks))
  464. return 1;
  465. /* Hm, nope. Are (enough) root reserved blocks available? */
  466. if (sbi->s_resuid == current_fsuid() ||
  467. ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) ||
  468. capable(CAP_SYS_RESOURCE)) {
  469. if (free_blocks >= (nblocks + dirty_blocks))
  470. return 1;
  471. }
  472. return 0;
  473. }
  474. int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
  475. s64 nblocks)
  476. {
  477. if (ext4_has_free_blocks(sbi, nblocks)) {
  478. percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks);
  479. return 0;
  480. } else
  481. return -ENOSPC;
  482. }
  483. /**
  484. * ext4_should_retry_alloc()
  485. * @sb: super block
  486. * @retries number of attemps has been made
  487. *
  488. * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
  489. * it is profitable to retry the operation, this function will wait
  490. * for the current or commiting transaction to complete, and then
  491. * return TRUE.
  492. *
  493. * if the total number of retries exceed three times, return FALSE.
  494. */
  495. int ext4_should_retry_alloc(struct super_block *sb, int *retries)
  496. {
  497. if (!ext4_has_free_blocks(EXT4_SB(sb), 1) ||
  498. (*retries)++ > 3 ||
  499. !EXT4_SB(sb)->s_journal)
  500. return 0;
  501. jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
  502. return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
  503. }
  504. /*
  505. * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
  506. *
  507. * @handle: handle to this transaction
  508. * @inode: file inode
  509. * @goal: given target block(filesystem wide)
  510. * @count: pointer to total number of blocks needed
  511. * @errp: error code
  512. *
  513. * Return 1st allocated block number on success, *count stores total account
  514. * error stores in errp pointer
  515. */
  516. ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
  517. ext4_fsblk_t goal, unsigned long *count, int *errp)
  518. {
  519. struct ext4_allocation_request ar;
  520. ext4_fsblk_t ret;
  521. memset(&ar, 0, sizeof(ar));
  522. /* Fill with neighbour allocated blocks */
  523. ar.inode = inode;
  524. ar.goal = goal;
  525. ar.len = count ? *count : 1;
  526. ret = ext4_mb_new_blocks(handle, &ar, errp);
  527. if (count)
  528. *count = ar.len;
  529. /*
  530. * Account for the allocated meta blocks. We will never
  531. * fail EDQUOT for metdata, but we do account for it.
  532. */
  533. if (!(*errp) &&
  534. ext4_test_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED)) {
  535. spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
  536. EXT4_I(inode)->i_allocated_meta_blocks += ar.len;
  537. spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
  538. dquot_alloc_block_nofail(inode, ar.len);
  539. }
  540. return ret;
  541. }
  542. /**
  543. * ext4_count_free_blocks() -- count filesystem free blocks
  544. * @sb: superblock
  545. *
  546. * Adds up the number of free blocks from each block group.
  547. */
  548. ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb)
  549. {
  550. ext4_fsblk_t desc_count;
  551. struct ext4_group_desc *gdp;
  552. ext4_group_t i;
  553. ext4_group_t ngroups = ext4_get_groups_count(sb);
  554. #ifdef EXT4FS_DEBUG
  555. struct ext4_super_block *es;
  556. ext4_fsblk_t bitmap_count;
  557. unsigned int x;
  558. struct buffer_head *bitmap_bh = NULL;
  559. es = EXT4_SB(sb)->s_es;
  560. desc_count = 0;
  561. bitmap_count = 0;
  562. gdp = NULL;
  563. for (i = 0; i < ngroups; i++) {
  564. gdp = ext4_get_group_desc(sb, i, NULL);
  565. if (!gdp)
  566. continue;
  567. desc_count += ext4_free_blks_count(sb, gdp);
  568. brelse(bitmap_bh);
  569. bitmap_bh = ext4_read_block_bitmap(sb, i);
  570. if (bitmap_bh == NULL)
  571. continue;
  572. x = ext4_count_free(bitmap_bh, sb->s_blocksize);
  573. printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
  574. i, ext4_free_blks_count(sb, gdp), x);
  575. bitmap_count += x;
  576. }
  577. brelse(bitmap_bh);
  578. printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu"
  579. ", computed = %llu, %llu\n", ext4_free_blocks_count(es),
  580. desc_count, bitmap_count);
  581. return bitmap_count;
  582. #else
  583. desc_count = 0;
  584. for (i = 0; i < ngroups; i++) {
  585. gdp = ext4_get_group_desc(sb, i, NULL);
  586. if (!gdp)
  587. continue;
  588. desc_count += ext4_free_blks_count(sb, gdp);
  589. }
  590. return desc_count;
  591. #endif
  592. }
  593. static inline int test_root(ext4_group_t a, int b)
  594. {
  595. int num = b;
  596. while (a > num)
  597. num *= b;
  598. return num == a;
  599. }
  600. static int ext4_group_sparse(ext4_group_t group)
  601. {
  602. if (group <= 1)
  603. return 1;
  604. if (!(group & 1))
  605. return 0;
  606. return (test_root(group, 7) || test_root(group, 5) ||
  607. test_root(group, 3));
  608. }
  609. /**
  610. * ext4_bg_has_super - number of blocks used by the superblock in group
  611. * @sb: superblock for filesystem
  612. * @group: group number to check
  613. *
  614. * Return the number of blocks used by the superblock (primary or backup)
  615. * in this group. Currently this will be only 0 or 1.
  616. */
  617. int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
  618. {
  619. if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
  620. EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
  621. !ext4_group_sparse(group))
  622. return 0;
  623. return 1;
  624. }
  625. static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
  626. ext4_group_t group)
  627. {
  628. unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
  629. ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
  630. ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
  631. if (group == first || group == first + 1 || group == last)
  632. return 1;
  633. return 0;
  634. }
  635. static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
  636. ext4_group_t group)
  637. {
  638. if (!ext4_bg_has_super(sb, group))
  639. return 0;
  640. if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG))
  641. return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
  642. else
  643. return EXT4_SB(sb)->s_gdb_count;
  644. }
  645. /**
  646. * ext4_bg_num_gdb - number of blocks used by the group table in group
  647. * @sb: superblock for filesystem
  648. * @group: group number to check
  649. *
  650. * Return the number of blocks used by the group descriptor table
  651. * (primary or backup) in this group. In the future there may be a
  652. * different number of descriptor blocks in each group.
  653. */
  654. unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
  655. {
  656. unsigned long first_meta_bg =
  657. le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
  658. unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
  659. if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) ||
  660. metagroup < first_meta_bg)
  661. return ext4_bg_num_gdb_nometa(sb, group);
  662. return ext4_bg_num_gdb_meta(sb,group);
  663. }