balloc.c 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746
  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. #include <trace/events/ext4.h>
  23. /*
  24. * balloc.c contains the blocks allocation and deallocation routines
  25. */
  26. /*
  27. * Calculate the block group number and offset, given a block number
  28. */
  29. void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
  30. ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
  31. {
  32. struct ext4_super_block *es = EXT4_SB(sb)->s_es;
  33. ext4_grpblk_t offset;
  34. blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
  35. offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb));
  36. if (offsetp)
  37. *offsetp = offset;
  38. if (blockgrpp)
  39. *blockgrpp = blocknr;
  40. }
  41. static int ext4_block_in_group(struct super_block *sb, ext4_fsblk_t block,
  42. ext4_group_t block_group)
  43. {
  44. ext4_group_t actual_group;
  45. ext4_get_group_no_and_offset(sb, block, &actual_group, NULL);
  46. if (actual_group == block_group)
  47. return 1;
  48. return 0;
  49. }
  50. static int ext4_group_used_meta_blocks(struct super_block *sb,
  51. ext4_group_t block_group,
  52. struct ext4_group_desc *gdp)
  53. {
  54. ext4_fsblk_t tmp;
  55. struct ext4_sb_info *sbi = EXT4_SB(sb);
  56. /* block bitmap, inode bitmap, and inode table blocks */
  57. int used_blocks = sbi->s_itb_per_group + 2;
  58. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
  59. if (!ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp),
  60. block_group))
  61. used_blocks--;
  62. if (!ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp),
  63. block_group))
  64. used_blocks--;
  65. tmp = ext4_inode_table(sb, gdp);
  66. for (; tmp < ext4_inode_table(sb, gdp) +
  67. sbi->s_itb_per_group; tmp++) {
  68. if (!ext4_block_in_group(sb, tmp, block_group))
  69. used_blocks -= 1;
  70. }
  71. }
  72. return used_blocks;
  73. }
  74. /* Initializes an uninitialized block bitmap if given, and returns the
  75. * number of blocks free in the group. */
  76. unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
  77. ext4_group_t block_group, struct ext4_group_desc *gdp)
  78. {
  79. int bit, bit_max;
  80. ext4_group_t ngroups = ext4_get_groups_count(sb);
  81. unsigned free_blocks, group_blocks;
  82. struct ext4_sb_info *sbi = EXT4_SB(sb);
  83. if (bh) {
  84. J_ASSERT_BH(bh, buffer_locked(bh));
  85. /* If checksum is bad mark all blocks used to prevent allocation
  86. * essentially implementing a per-group read-only flag. */
  87. if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) {
  88. ext4_error(sb, "Checksum bad for group %u",
  89. block_group);
  90. ext4_free_blks_set(sb, gdp, 0);
  91. ext4_free_inodes_set(sb, gdp, 0);
  92. ext4_itable_unused_set(sb, gdp, 0);
  93. memset(bh->b_data, 0xff, sb->s_blocksize);
  94. return 0;
  95. }
  96. memset(bh->b_data, 0, sb->s_blocksize);
  97. }
  98. /* Check for superblock and gdt backups in this group */
  99. bit_max = ext4_bg_has_super(sb, block_group);
  100. if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
  101. block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
  102. sbi->s_desc_per_block) {
  103. if (bit_max) {
  104. bit_max += ext4_bg_num_gdb(sb, block_group);
  105. bit_max +=
  106. le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
  107. }
  108. } else { /* For META_BG_BLOCK_GROUPS */
  109. bit_max += ext4_bg_num_gdb(sb, block_group);
  110. }
  111. if (block_group == ngroups - 1) {
  112. /*
  113. * Even though mke2fs always initialize first and last group
  114. * if some other tool enabled the EXT4_BG_BLOCK_UNINIT we need
  115. * to make sure we calculate the right free blocks
  116. */
  117. group_blocks = ext4_blocks_count(sbi->s_es) -
  118. ext4_group_first_block_no(sb, ngroups - 1);
  119. } else {
  120. group_blocks = EXT4_BLOCKS_PER_GROUP(sb);
  121. }
  122. free_blocks = group_blocks - bit_max;
  123. if (bh) {
  124. ext4_fsblk_t start, tmp;
  125. int flex_bg = 0;
  126. for (bit = 0; bit < bit_max; bit++)
  127. ext4_set_bit(bit, bh->b_data);
  128. start = ext4_group_first_block_no(sb, block_group);
  129. if (EXT4_HAS_INCOMPAT_FEATURE(sb,
  130. EXT4_FEATURE_INCOMPAT_FLEX_BG))
  131. flex_bg = 1;
  132. /* Set bits for block and inode bitmaps, and inode table */
  133. tmp = ext4_block_bitmap(sb, gdp);
  134. if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
  135. ext4_set_bit(tmp - start, bh->b_data);
  136. tmp = ext4_inode_bitmap(sb, gdp);
  137. if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
  138. ext4_set_bit(tmp - start, bh->b_data);
  139. tmp = ext4_inode_table(sb, gdp);
  140. for (; tmp < ext4_inode_table(sb, gdp) +
  141. sbi->s_itb_per_group; tmp++) {
  142. if (!flex_bg ||
  143. ext4_block_in_group(sb, tmp, block_group))
  144. ext4_set_bit(tmp - start, bh->b_data);
  145. }
  146. /*
  147. * Also if the number of blocks within the group is
  148. * less than the blocksize * 8 ( which is the size
  149. * of bitmap ), set rest of the block bitmap to 1
  150. */
  151. ext4_mark_bitmap_end(group_blocks, sb->s_blocksize * 8,
  152. bh->b_data);
  153. }
  154. return free_blocks - ext4_group_used_meta_blocks(sb, block_group, gdp);
  155. }
  156. /*
  157. * The free blocks are managed by bitmaps. A file system contains several
  158. * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
  159. * block for inodes, N blocks for the inode table and data blocks.
  160. *
  161. * The file system contains group descriptors which are located after the
  162. * super block. Each descriptor contains the number of the bitmap block and
  163. * the free blocks count in the block. The descriptors are loaded in memory
  164. * when a file system is mounted (see ext4_fill_super).
  165. */
  166. /**
  167. * ext4_get_group_desc() -- load group descriptor from disk
  168. * @sb: super block
  169. * @block_group: given block group
  170. * @bh: pointer to the buffer head to store the block
  171. * group descriptor
  172. */
  173. struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
  174. ext4_group_t block_group,
  175. struct buffer_head **bh)
  176. {
  177. unsigned int group_desc;
  178. unsigned int offset;
  179. ext4_group_t ngroups = ext4_get_groups_count(sb);
  180. struct ext4_group_desc *desc;
  181. struct ext4_sb_info *sbi = EXT4_SB(sb);
  182. if (block_group >= ngroups) {
  183. ext4_error(sb, "block_group >= groups_count - block_group = %u,"
  184. " groups_count = %u", block_group, ngroups);
  185. return NULL;
  186. }
  187. group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
  188. offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
  189. if (!sbi->s_group_desc[group_desc]) {
  190. ext4_error(sb, "Group descriptor not loaded - "
  191. "block_group = %u, group_desc = %u, desc = %u",
  192. block_group, group_desc, offset);
  193. return NULL;
  194. }
  195. desc = (struct ext4_group_desc *)(
  196. (__u8 *)sbi->s_group_desc[group_desc]->b_data +
  197. offset * EXT4_DESC_SIZE(sb));
  198. if (bh)
  199. *bh = sbi->s_group_desc[group_desc];
  200. return desc;
  201. }
  202. static int ext4_valid_block_bitmap(struct super_block *sb,
  203. struct ext4_group_desc *desc,
  204. unsigned int block_group,
  205. struct buffer_head *bh)
  206. {
  207. ext4_grpblk_t offset;
  208. ext4_grpblk_t next_zero_bit;
  209. ext4_fsblk_t bitmap_blk;
  210. ext4_fsblk_t group_first_block;
  211. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
  212. /* with FLEX_BG, the inode/block bitmaps and itable
  213. * blocks may not be in the group at all
  214. * so the bitmap validation will be skipped for those groups
  215. * or it has to also read the block group where the bitmaps
  216. * are located to verify they are set.
  217. */
  218. return 1;
  219. }
  220. group_first_block = ext4_group_first_block_no(sb, block_group);
  221. /* check whether block bitmap block number is set */
  222. bitmap_blk = ext4_block_bitmap(sb, desc);
  223. offset = bitmap_blk - group_first_block;
  224. if (!ext4_test_bit(offset, bh->b_data))
  225. /* bad block bitmap */
  226. goto err_out;
  227. /* check whether the inode bitmap block number is set */
  228. bitmap_blk = ext4_inode_bitmap(sb, desc);
  229. offset = bitmap_blk - group_first_block;
  230. if (!ext4_test_bit(offset, bh->b_data))
  231. /* bad block bitmap */
  232. goto err_out;
  233. /* check whether the inode table block number is set */
  234. bitmap_blk = ext4_inode_table(sb, desc);
  235. offset = bitmap_blk - group_first_block;
  236. next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
  237. offset + EXT4_SB(sb)->s_itb_per_group,
  238. offset);
  239. if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group)
  240. /* good bitmap for inode tables */
  241. return 1;
  242. err_out:
  243. ext4_error(sb, "Invalid block bitmap - block_group = %d, block = %llu",
  244. block_group, bitmap_blk);
  245. return 0;
  246. }
  247. /**
  248. * ext4_read_block_bitmap()
  249. * @sb: super block
  250. * @block_group: given block group
  251. *
  252. * Read the bitmap for a given block_group,and validate the
  253. * bits for block/inode/inode tables are set in the bitmaps
  254. *
  255. * Return buffer_head on success or NULL in case of failure.
  256. */
  257. struct buffer_head *
  258. ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
  259. {
  260. struct ext4_group_desc *desc;
  261. struct buffer_head *bh = NULL;
  262. ext4_fsblk_t bitmap_blk;
  263. desc = ext4_get_group_desc(sb, block_group, NULL);
  264. if (!desc)
  265. return NULL;
  266. bitmap_blk = ext4_block_bitmap(sb, desc);
  267. bh = sb_getblk(sb, bitmap_blk);
  268. if (unlikely(!bh)) {
  269. ext4_error(sb, "Cannot read block bitmap - "
  270. "block_group = %u, block_bitmap = %llu",
  271. block_group, bitmap_blk);
  272. return NULL;
  273. }
  274. if (bitmap_uptodate(bh))
  275. return bh;
  276. lock_buffer(bh);
  277. if (bitmap_uptodate(bh)) {
  278. unlock_buffer(bh);
  279. return bh;
  280. }
  281. ext4_lock_group(sb, block_group);
  282. if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
  283. ext4_init_block_bitmap(sb, bh, block_group, desc);
  284. set_bitmap_uptodate(bh);
  285. set_buffer_uptodate(bh);
  286. ext4_unlock_group(sb, block_group);
  287. unlock_buffer(bh);
  288. return bh;
  289. }
  290. ext4_unlock_group(sb, block_group);
  291. if (buffer_uptodate(bh)) {
  292. /*
  293. * if not uninit if bh is uptodate,
  294. * bitmap is also uptodate
  295. */
  296. set_bitmap_uptodate(bh);
  297. unlock_buffer(bh);
  298. return bh;
  299. }
  300. /*
  301. * submit the buffer_head for read. We can
  302. * safely mark the bitmap as uptodate now.
  303. * We do it here so the bitmap uptodate bit
  304. * get set with buffer lock held.
  305. */
  306. trace_ext4_read_block_bitmap_load(sb, block_group);
  307. set_bitmap_uptodate(bh);
  308. if (bh_submit_read(bh) < 0) {
  309. put_bh(bh);
  310. ext4_error(sb, "Cannot read block bitmap - "
  311. "block_group = %u, block_bitmap = %llu",
  312. block_group, bitmap_blk);
  313. return NULL;
  314. }
  315. ext4_valid_block_bitmap(sb, desc, block_group, bh);
  316. /*
  317. * file system mounted not to panic on error,
  318. * continue with corrupt bitmap
  319. */
  320. return bh;
  321. }
  322. /**
  323. * ext4_add_groupblocks() -- Add given blocks to an existing group
  324. * @handle: handle to this transaction
  325. * @sb: super block
  326. * @block: start physcial block to add to the block group
  327. * @count: number of blocks to free
  328. *
  329. * This marks the blocks as free in the bitmap. We ask the
  330. * mballoc to reload the buddy after this by setting group
  331. * EXT4_GROUP_INFO_NEED_INIT_BIT flag
  332. */
  333. void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
  334. ext4_fsblk_t block, unsigned long count)
  335. {
  336. struct buffer_head *bitmap_bh = NULL;
  337. struct buffer_head *gd_bh;
  338. ext4_group_t block_group;
  339. ext4_grpblk_t bit;
  340. unsigned int i;
  341. struct ext4_group_desc *desc;
  342. struct ext4_sb_info *sbi = EXT4_SB(sb);
  343. int err = 0, ret, blk_free_count;
  344. ext4_grpblk_t blocks_freed;
  345. struct ext4_group_info *grp;
  346. ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1);
  347. ext4_get_group_no_and_offset(sb, block, &block_group, &bit);
  348. grp = ext4_get_group_info(sb, block_group);
  349. /*
  350. * Check to see if we are freeing blocks across a group
  351. * boundary.
  352. */
  353. if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) {
  354. goto error_return;
  355. }
  356. bitmap_bh = ext4_read_block_bitmap(sb, block_group);
  357. if (!bitmap_bh)
  358. goto error_return;
  359. desc = ext4_get_group_desc(sb, block_group, &gd_bh);
  360. if (!desc)
  361. goto error_return;
  362. if (in_range(ext4_block_bitmap(sb, desc), block, count) ||
  363. in_range(ext4_inode_bitmap(sb, desc), block, count) ||
  364. in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) ||
  365. in_range(block + count - 1, ext4_inode_table(sb, desc),
  366. sbi->s_itb_per_group)) {
  367. ext4_error(sb, "Adding blocks in system zones - "
  368. "Block = %llu, count = %lu",
  369. block, count);
  370. goto error_return;
  371. }
  372. /*
  373. * We are about to add blocks to the bitmap,
  374. * so we need undo access.
  375. */
  376. BUFFER_TRACE(bitmap_bh, "getting undo access");
  377. err = ext4_journal_get_undo_access(handle, bitmap_bh);
  378. if (err)
  379. goto error_return;
  380. /*
  381. * We are about to modify some metadata. Call the journal APIs
  382. * to unshare ->b_data if a currently-committing transaction is
  383. * using it
  384. */
  385. BUFFER_TRACE(gd_bh, "get_write_access");
  386. err = ext4_journal_get_write_access(handle, gd_bh);
  387. if (err)
  388. goto error_return;
  389. /*
  390. * make sure we don't allow a parallel init on other groups in the
  391. * same buddy cache
  392. */
  393. down_write(&grp->alloc_sem);
  394. for (i = 0, blocks_freed = 0; i < count; i++) {
  395. BUFFER_TRACE(bitmap_bh, "clear bit");
  396. if (!ext4_clear_bit_atomic(ext4_group_lock_ptr(sb, block_group),
  397. bit + i, bitmap_bh->b_data)) {
  398. ext4_error(sb, "bit already cleared for block %llu",
  399. (ext4_fsblk_t)(block + i));
  400. BUFFER_TRACE(bitmap_bh, "bit already cleared");
  401. } else {
  402. blocks_freed++;
  403. }
  404. }
  405. ext4_lock_group(sb, block_group);
  406. blk_free_count = blocks_freed + ext4_free_blks_count(sb, desc);
  407. ext4_free_blks_set(sb, desc, blk_free_count);
  408. desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc);
  409. ext4_unlock_group(sb, block_group);
  410. percpu_counter_add(&sbi->s_freeblocks_counter, blocks_freed);
  411. if (sbi->s_log_groups_per_flex) {
  412. ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
  413. atomic_add(blocks_freed,
  414. &sbi->s_flex_groups[flex_group].free_blocks);
  415. }
  416. /*
  417. * request to reload the buddy with the
  418. * new bitmap information
  419. */
  420. set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state));
  421. grp->bb_free += blocks_freed;
  422. up_write(&grp->alloc_sem);
  423. /* We dirtied the bitmap block */
  424. BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
  425. err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
  426. /* And the group descriptor block */
  427. BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
  428. ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh);
  429. if (!err)
  430. err = ret;
  431. error_return:
  432. brelse(bitmap_bh);
  433. ext4_std_error(sb, err);
  434. return;
  435. }
  436. /**
  437. * ext4_has_free_blocks()
  438. * @sbi: in-core super block structure.
  439. * @nblocks: number of needed blocks
  440. *
  441. * Check if filesystem has nblocks free & available for allocation.
  442. * On success return 1, return 0 on failure.
  443. */
  444. static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
  445. {
  446. s64 free_blocks, dirty_blocks, root_blocks;
  447. struct percpu_counter *fbc = &sbi->s_freeblocks_counter;
  448. struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter;
  449. free_blocks = percpu_counter_read_positive(fbc);
  450. dirty_blocks = percpu_counter_read_positive(dbc);
  451. root_blocks = ext4_r_blocks_count(sbi->s_es);
  452. if (free_blocks - (nblocks + root_blocks + dirty_blocks) <
  453. EXT4_FREEBLOCKS_WATERMARK) {
  454. free_blocks = percpu_counter_sum_positive(fbc);
  455. dirty_blocks = percpu_counter_sum_positive(dbc);
  456. if (dirty_blocks < 0) {
  457. printk(KERN_CRIT "Dirty block accounting "
  458. "went wrong %lld\n",
  459. (long long)dirty_blocks);
  460. }
  461. }
  462. /* Check whether we have space after
  463. * accounting for current dirty blocks & root reserved blocks.
  464. */
  465. if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks))
  466. return 1;
  467. /* Hm, nope. Are (enough) root reserved blocks available? */
  468. if (sbi->s_resuid == current_fsuid() ||
  469. ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) ||
  470. capable(CAP_SYS_RESOURCE)) {
  471. if (free_blocks >= (nblocks + dirty_blocks))
  472. return 1;
  473. }
  474. return 0;
  475. }
  476. int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
  477. s64 nblocks)
  478. {
  479. if (ext4_has_free_blocks(sbi, nblocks)) {
  480. percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks);
  481. return 0;
  482. } else
  483. return -ENOSPC;
  484. }
  485. /**
  486. * ext4_should_retry_alloc()
  487. * @sb: super block
  488. * @retries number of attemps has been made
  489. *
  490. * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
  491. * it is profitable to retry the operation, this function will wait
  492. * for the current or commiting transaction to complete, and then
  493. * return TRUE.
  494. *
  495. * if the total number of retries exceed three times, return FALSE.
  496. */
  497. int ext4_should_retry_alloc(struct super_block *sb, int *retries)
  498. {
  499. if (!ext4_has_free_blocks(EXT4_SB(sb), 1) ||
  500. (*retries)++ > 3 ||
  501. !EXT4_SB(sb)->s_journal)
  502. return 0;
  503. jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
  504. return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
  505. }
  506. /*
  507. * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
  508. *
  509. * @handle: handle to this transaction
  510. * @inode: file inode
  511. * @goal: given target block(filesystem wide)
  512. * @count: pointer to total number of blocks needed
  513. * @errp: error code
  514. *
  515. * Return 1st allocated block number on success, *count stores total account
  516. * error stores in errp pointer
  517. */
  518. ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
  519. ext4_fsblk_t goal, unsigned long *count, int *errp)
  520. {
  521. struct ext4_allocation_request ar;
  522. ext4_fsblk_t ret;
  523. memset(&ar, 0, sizeof(ar));
  524. /* Fill with neighbour allocated blocks */
  525. ar.inode = inode;
  526. ar.goal = goal;
  527. ar.len = count ? *count : 1;
  528. ret = ext4_mb_new_blocks(handle, &ar, errp);
  529. if (count)
  530. *count = ar.len;
  531. /*
  532. * Account for the allocated meta blocks. We will never
  533. * fail EDQUOT for metdata, but we do account for it.
  534. */
  535. if (!(*errp) &&
  536. ext4_test_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED)) {
  537. spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
  538. EXT4_I(inode)->i_allocated_meta_blocks += ar.len;
  539. spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
  540. dquot_alloc_block_nofail(inode, ar.len);
  541. }
  542. return ret;
  543. }
  544. /**
  545. * ext4_count_free_blocks() -- count filesystem free blocks
  546. * @sb: superblock
  547. *
  548. * Adds up the number of free blocks from each block group.
  549. */
  550. ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb)
  551. {
  552. ext4_fsblk_t desc_count;
  553. struct ext4_group_desc *gdp;
  554. ext4_group_t i;
  555. ext4_group_t ngroups = ext4_get_groups_count(sb);
  556. #ifdef EXT4FS_DEBUG
  557. struct ext4_super_block *es;
  558. ext4_fsblk_t bitmap_count;
  559. unsigned int x;
  560. struct buffer_head *bitmap_bh = NULL;
  561. es = EXT4_SB(sb)->s_es;
  562. desc_count = 0;
  563. bitmap_count = 0;
  564. gdp = NULL;
  565. for (i = 0; i < ngroups; i++) {
  566. gdp = ext4_get_group_desc(sb, i, NULL);
  567. if (!gdp)
  568. continue;
  569. desc_count += ext4_free_blks_count(sb, gdp);
  570. brelse(bitmap_bh);
  571. bitmap_bh = ext4_read_block_bitmap(sb, i);
  572. if (bitmap_bh == NULL)
  573. continue;
  574. x = ext4_count_free(bitmap_bh, sb->s_blocksize);
  575. printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
  576. i, ext4_free_blks_count(sb, gdp), x);
  577. bitmap_count += x;
  578. }
  579. brelse(bitmap_bh);
  580. printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu"
  581. ", computed = %llu, %llu\n", ext4_free_blocks_count(es),
  582. desc_count, bitmap_count);
  583. return bitmap_count;
  584. #else
  585. desc_count = 0;
  586. for (i = 0; i < ngroups; i++) {
  587. gdp = ext4_get_group_desc(sb, i, NULL);
  588. if (!gdp)
  589. continue;
  590. desc_count += ext4_free_blks_count(sb, gdp);
  591. }
  592. return desc_count;
  593. #endif
  594. }
  595. static inline int test_root(ext4_group_t a, int b)
  596. {
  597. int num = b;
  598. while (a > num)
  599. num *= b;
  600. return num == a;
  601. }
  602. static int ext4_group_sparse(ext4_group_t group)
  603. {
  604. if (group <= 1)
  605. return 1;
  606. if (!(group & 1))
  607. return 0;
  608. return (test_root(group, 7) || test_root(group, 5) ||
  609. test_root(group, 3));
  610. }
  611. /**
  612. * ext4_bg_has_super - number of blocks used by the superblock in group
  613. * @sb: superblock for filesystem
  614. * @group: group number to check
  615. *
  616. * Return the number of blocks used by the superblock (primary or backup)
  617. * in this group. Currently this will be only 0 or 1.
  618. */
  619. int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
  620. {
  621. if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
  622. EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
  623. !ext4_group_sparse(group))
  624. return 0;
  625. return 1;
  626. }
  627. static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
  628. ext4_group_t group)
  629. {
  630. unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
  631. ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
  632. ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
  633. if (group == first || group == first + 1 || group == last)
  634. return 1;
  635. return 0;
  636. }
  637. static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
  638. ext4_group_t group)
  639. {
  640. if (!ext4_bg_has_super(sb, group))
  641. return 0;
  642. if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG))
  643. return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
  644. else
  645. return EXT4_SB(sb)->s_gdb_count;
  646. }
  647. /**
  648. * ext4_bg_num_gdb - number of blocks used by the group table in group
  649. * @sb: superblock for filesystem
  650. * @group: group number to check
  651. *
  652. * Return the number of blocks used by the group descriptor table
  653. * (primary or backup) in this group. In the future there may be a
  654. * different number of descriptor blocks in each group.
  655. */
  656. unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
  657. {
  658. unsigned long first_meta_bg =
  659. le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
  660. unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
  661. if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) ||
  662. metagroup < first_meta_bg)
  663. return ext4_bg_num_gdb_nometa(sb, group);
  664. return ext4_bg_num_gdb_meta(sb,group);
  665. }