balloc.c 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789
  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, __func__,
  88. "Checksum bad for group %u", 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. le32_to_cpu(sbi->s_es->s_first_data_block) -
  118. (EXT4_BLOCKS_PER_GROUP(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. mark_bitmap_end(group_blocks, sb->s_blocksize * 8, 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. #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
  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, "ext4_get_group_desc",
  184. "block_group >= groups_count - "
  185. "block_group = %u, groups_count = %u",
  186. block_group, ngroups);
  187. return NULL;
  188. }
  189. group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
  190. offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
  191. if (!sbi->s_group_desc[group_desc]) {
  192. ext4_error(sb, "ext4_get_group_desc",
  193. "Group descriptor not loaded - "
  194. "block_group = %u, group_desc = %u, desc = %u",
  195. block_group, group_desc, offset);
  196. return NULL;
  197. }
  198. desc = (struct ext4_group_desc *)(
  199. (__u8 *)sbi->s_group_desc[group_desc]->b_data +
  200. offset * EXT4_DESC_SIZE(sb));
  201. if (bh)
  202. *bh = sbi->s_group_desc[group_desc];
  203. return desc;
  204. }
  205. static int ext4_valid_block_bitmap(struct super_block *sb,
  206. struct ext4_group_desc *desc,
  207. unsigned int block_group,
  208. struct buffer_head *bh)
  209. {
  210. ext4_grpblk_t offset;
  211. ext4_grpblk_t next_zero_bit;
  212. ext4_fsblk_t bitmap_blk;
  213. ext4_fsblk_t group_first_block;
  214. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
  215. /* with FLEX_BG, the inode/block bitmaps and itable
  216. * blocks may not be in the group at all
  217. * so the bitmap validation will be skipped for those groups
  218. * or it has to also read the block group where the bitmaps
  219. * are located to verify they are set.
  220. */
  221. return 1;
  222. }
  223. group_first_block = ext4_group_first_block_no(sb, block_group);
  224. /* check whether block bitmap block number is set */
  225. bitmap_blk = ext4_block_bitmap(sb, desc);
  226. offset = bitmap_blk - group_first_block;
  227. if (!ext4_test_bit(offset, bh->b_data))
  228. /* bad block bitmap */
  229. goto err_out;
  230. /* check whether the inode bitmap block number is set */
  231. bitmap_blk = ext4_inode_bitmap(sb, desc);
  232. offset = bitmap_blk - group_first_block;
  233. if (!ext4_test_bit(offset, bh->b_data))
  234. /* bad block bitmap */
  235. goto err_out;
  236. /* check whether the inode table block number is set */
  237. bitmap_blk = ext4_inode_table(sb, desc);
  238. offset = bitmap_blk - group_first_block;
  239. next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
  240. offset + EXT4_SB(sb)->s_itb_per_group,
  241. offset);
  242. if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group)
  243. /* good bitmap for inode tables */
  244. return 1;
  245. err_out:
  246. ext4_error(sb, __func__,
  247. "Invalid block bitmap - "
  248. "block_group = %d, block = %llu",
  249. block_group, bitmap_blk);
  250. return 0;
  251. }
  252. /**
  253. * ext4_read_block_bitmap()
  254. * @sb: super block
  255. * @block_group: given block group
  256. *
  257. * Read the bitmap for a given block_group,and validate the
  258. * bits for block/inode/inode tables are set in the bitmaps
  259. *
  260. * Return buffer_head on success or NULL in case of failure.
  261. */
  262. struct buffer_head *
  263. ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
  264. {
  265. struct ext4_group_desc *desc;
  266. struct buffer_head *bh = NULL;
  267. ext4_fsblk_t bitmap_blk;
  268. desc = ext4_get_group_desc(sb, block_group, NULL);
  269. if (!desc)
  270. return NULL;
  271. bitmap_blk = ext4_block_bitmap(sb, desc);
  272. bh = sb_getblk(sb, bitmap_blk);
  273. if (unlikely(!bh)) {
  274. ext4_error(sb, __func__,
  275. "Cannot read block bitmap - "
  276. "block_group = %u, block_bitmap = %llu",
  277. block_group, bitmap_blk);
  278. return NULL;
  279. }
  280. if (bitmap_uptodate(bh))
  281. return bh;
  282. lock_buffer(bh);
  283. if (bitmap_uptodate(bh)) {
  284. unlock_buffer(bh);
  285. return bh;
  286. }
  287. ext4_lock_group(sb, block_group);
  288. if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
  289. ext4_init_block_bitmap(sb, bh, block_group, desc);
  290. set_bitmap_uptodate(bh);
  291. set_buffer_uptodate(bh);
  292. ext4_unlock_group(sb, block_group);
  293. unlock_buffer(bh);
  294. return bh;
  295. }
  296. ext4_unlock_group(sb, block_group);
  297. if (buffer_uptodate(bh)) {
  298. /*
  299. * if not uninit if bh is uptodate,
  300. * bitmap is also uptodate
  301. */
  302. set_bitmap_uptodate(bh);
  303. unlock_buffer(bh);
  304. return bh;
  305. }
  306. /*
  307. * submit the buffer_head for read. We can
  308. * safely mark the bitmap as uptodate now.
  309. * We do it here so the bitmap uptodate bit
  310. * get set with buffer lock held.
  311. */
  312. set_bitmap_uptodate(bh);
  313. if (bh_submit_read(bh) < 0) {
  314. put_bh(bh);
  315. ext4_error(sb, __func__,
  316. "Cannot read block bitmap - "
  317. "block_group = %u, block_bitmap = %llu",
  318. block_group, bitmap_blk);
  319. return NULL;
  320. }
  321. ext4_valid_block_bitmap(sb, desc, block_group, bh);
  322. /*
  323. * file system mounted not to panic on error,
  324. * continue with corrupt bitmap
  325. */
  326. return bh;
  327. }
  328. /**
  329. * ext4_add_groupblocks() -- Add given blocks to an existing group
  330. * @handle: handle to this transaction
  331. * @sb: super block
  332. * @block: start physcial block to add to the block group
  333. * @count: number of blocks to free
  334. *
  335. * This marks the blocks as free in the bitmap. We ask the
  336. * mballoc to reload the buddy after this by setting group
  337. * EXT4_GROUP_INFO_NEED_INIT_BIT flag
  338. */
  339. void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
  340. ext4_fsblk_t block, unsigned long count)
  341. {
  342. struct buffer_head *bitmap_bh = NULL;
  343. struct buffer_head *gd_bh;
  344. ext4_group_t block_group;
  345. ext4_grpblk_t bit;
  346. unsigned int i;
  347. struct ext4_group_desc *desc;
  348. struct ext4_super_block *es;
  349. struct ext4_sb_info *sbi;
  350. int err = 0, ret, blk_free_count;
  351. ext4_grpblk_t blocks_freed;
  352. struct ext4_group_info *grp;
  353. sbi = EXT4_SB(sb);
  354. es = sbi->s_es;
  355. ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1);
  356. ext4_get_group_no_and_offset(sb, block, &block_group, &bit);
  357. grp = ext4_get_group_info(sb, block_group);
  358. /*
  359. * Check to see if we are freeing blocks across a group
  360. * boundary.
  361. */
  362. if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) {
  363. goto error_return;
  364. }
  365. bitmap_bh = ext4_read_block_bitmap(sb, block_group);
  366. if (!bitmap_bh)
  367. goto error_return;
  368. desc = ext4_get_group_desc(sb, block_group, &gd_bh);
  369. if (!desc)
  370. goto error_return;
  371. if (in_range(ext4_block_bitmap(sb, desc), block, count) ||
  372. in_range(ext4_inode_bitmap(sb, desc), block, count) ||
  373. in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) ||
  374. in_range(block + count - 1, ext4_inode_table(sb, desc),
  375. sbi->s_itb_per_group)) {
  376. ext4_error(sb, __func__,
  377. "Adding blocks in system zones - "
  378. "Block = %llu, count = %lu",
  379. block, count);
  380. goto error_return;
  381. }
  382. /*
  383. * We are about to add blocks to the bitmap,
  384. * so we need undo access.
  385. */
  386. BUFFER_TRACE(bitmap_bh, "getting undo access");
  387. err = ext4_journal_get_undo_access(handle, bitmap_bh);
  388. if (err)
  389. goto error_return;
  390. /*
  391. * We are about to modify some metadata. Call the journal APIs
  392. * to unshare ->b_data if a currently-committing transaction is
  393. * using it
  394. */
  395. BUFFER_TRACE(gd_bh, "get_write_access");
  396. err = ext4_journal_get_write_access(handle, gd_bh);
  397. if (err)
  398. goto error_return;
  399. /*
  400. * make sure we don't allow a parallel init on other groups in the
  401. * same buddy cache
  402. */
  403. down_write(&grp->alloc_sem);
  404. for (i = 0, blocks_freed = 0; i < count; i++) {
  405. BUFFER_TRACE(bitmap_bh, "clear bit");
  406. if (!ext4_clear_bit_atomic(ext4_group_lock_ptr(sb, block_group),
  407. bit + i, bitmap_bh->b_data)) {
  408. ext4_error(sb, __func__,
  409. "bit already cleared for block %llu",
  410. (ext4_fsblk_t)(block + i));
  411. BUFFER_TRACE(bitmap_bh, "bit already cleared");
  412. } else {
  413. blocks_freed++;
  414. }
  415. }
  416. ext4_lock_group(sb, block_group);
  417. blk_free_count = blocks_freed + ext4_free_blks_count(sb, desc);
  418. ext4_free_blks_set(sb, desc, blk_free_count);
  419. desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc);
  420. ext4_unlock_group(sb, block_group);
  421. percpu_counter_add(&sbi->s_freeblocks_counter, blocks_freed);
  422. if (sbi->s_log_groups_per_flex) {
  423. ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
  424. atomic_add(blocks_freed,
  425. &sbi->s_flex_groups[flex_group].free_blocks);
  426. }
  427. /*
  428. * request to reload the buddy with the
  429. * new bitmap information
  430. */
  431. set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state));
  432. ext4_mb_update_group_info(grp, blocks_freed);
  433. up_write(&grp->alloc_sem);
  434. /* We dirtied the bitmap block */
  435. BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
  436. err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
  437. /* And the group descriptor block */
  438. BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
  439. ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh);
  440. if (!err)
  441. err = ret;
  442. sb->s_dirt = 1;
  443. error_return:
  444. brelse(bitmap_bh);
  445. ext4_std_error(sb, err);
  446. return;
  447. }
  448. /**
  449. * ext4_free_blocks() -- Free given blocks and update quota
  450. * @handle: handle for this transaction
  451. * @inode: inode
  452. * @block: start physical block to free
  453. * @count: number of blocks to count
  454. * @metadata: Are these metadata blocks
  455. */
  456. void ext4_free_blocks(handle_t *handle, struct inode *inode,
  457. ext4_fsblk_t block, unsigned long count,
  458. int metadata)
  459. {
  460. struct super_block *sb;
  461. unsigned long dquot_freed_blocks;
  462. /* this isn't the right place to decide whether block is metadata
  463. * inode.c/extents.c knows better, but for safety ... */
  464. if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
  465. metadata = 1;
  466. /* We need to make sure we don't reuse
  467. * block released untill the transaction commit.
  468. * writeback mode have weak data consistency so
  469. * don't force data as metadata when freeing block
  470. * for writeback mode.
  471. */
  472. if (metadata == 0 && !ext4_should_writeback_data(inode))
  473. metadata = 1;
  474. sb = inode->i_sb;
  475. ext4_mb_free_blocks(handle, inode, block, count,
  476. metadata, &dquot_freed_blocks);
  477. if (dquot_freed_blocks)
  478. vfs_dq_free_block(inode, dquot_freed_blocks);
  479. return;
  480. }
  481. /**
  482. * ext4_has_free_blocks()
  483. * @sbi: in-core super block structure.
  484. * @nblocks: number of needed blocks
  485. *
  486. * Check if filesystem has nblocks free & available for allocation.
  487. * On success return 1, return 0 on failure.
  488. */
  489. int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
  490. {
  491. s64 free_blocks, dirty_blocks, root_blocks;
  492. struct percpu_counter *fbc = &sbi->s_freeblocks_counter;
  493. struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter;
  494. free_blocks = percpu_counter_read_positive(fbc);
  495. dirty_blocks = percpu_counter_read_positive(dbc);
  496. root_blocks = ext4_r_blocks_count(sbi->s_es);
  497. if (free_blocks - (nblocks + root_blocks + dirty_blocks) <
  498. EXT4_FREEBLOCKS_WATERMARK) {
  499. free_blocks = percpu_counter_sum_positive(fbc);
  500. dirty_blocks = percpu_counter_sum_positive(dbc);
  501. if (dirty_blocks < 0) {
  502. printk(KERN_CRIT "Dirty block accounting "
  503. "went wrong %lld\n",
  504. (long long)dirty_blocks);
  505. }
  506. }
  507. /* Check whether we have space after
  508. * accounting for current dirty blocks & root reserved blocks.
  509. */
  510. if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks))
  511. return 1;
  512. /* Hm, nope. Are (enough) root reserved blocks available? */
  513. if (sbi->s_resuid == current_fsuid() ||
  514. ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) ||
  515. capable(CAP_SYS_RESOURCE)) {
  516. if (free_blocks >= (nblocks + dirty_blocks))
  517. return 1;
  518. }
  519. return 0;
  520. }
  521. int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
  522. s64 nblocks)
  523. {
  524. if (ext4_has_free_blocks(sbi, nblocks)) {
  525. percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks);
  526. return 0;
  527. } else
  528. return -ENOSPC;
  529. }
  530. /**
  531. * ext4_should_retry_alloc()
  532. * @sb: super block
  533. * @retries number of attemps has been made
  534. *
  535. * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
  536. * it is profitable to retry the operation, this function will wait
  537. * for the current or commiting transaction to complete, and then
  538. * return TRUE.
  539. *
  540. * if the total number of retries exceed three times, return FALSE.
  541. */
  542. int ext4_should_retry_alloc(struct super_block *sb, int *retries)
  543. {
  544. if (!ext4_has_free_blocks(EXT4_SB(sb), 1) ||
  545. (*retries)++ > 3 ||
  546. !EXT4_SB(sb)->s_journal)
  547. return 0;
  548. jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
  549. return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
  550. }
  551. /*
  552. * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
  553. *
  554. * @handle: handle to this transaction
  555. * @inode: file inode
  556. * @goal: given target block(filesystem wide)
  557. * @count: pointer to total number of blocks needed
  558. * @errp: error code
  559. *
  560. * Return 1st allocated block number on success, *count stores total account
  561. * error stores in errp pointer
  562. */
  563. ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
  564. ext4_fsblk_t goal, unsigned long *count, int *errp)
  565. {
  566. struct ext4_allocation_request ar;
  567. ext4_fsblk_t ret;
  568. memset(&ar, 0, sizeof(ar));
  569. /* Fill with neighbour allocated blocks */
  570. ar.inode = inode;
  571. ar.goal = goal;
  572. ar.len = count ? *count : 1;
  573. ret = ext4_mb_new_blocks(handle, &ar, errp);
  574. if (count)
  575. *count = ar.len;
  576. /*
  577. * Account for the allocated meta blocks
  578. */
  579. if (!(*errp) && EXT4_I(inode)->i_delalloc_reserved_flag) {
  580. spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
  581. EXT4_I(inode)->i_allocated_meta_blocks += ar.len;
  582. spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
  583. }
  584. return ret;
  585. }
  586. /**
  587. * ext4_count_free_blocks() -- count filesystem free blocks
  588. * @sb: superblock
  589. *
  590. * Adds up the number of free blocks from each block group.
  591. */
  592. ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb)
  593. {
  594. ext4_fsblk_t desc_count;
  595. struct ext4_group_desc *gdp;
  596. ext4_group_t i;
  597. ext4_group_t ngroups = ext4_get_groups_count(sb);
  598. #ifdef EXT4FS_DEBUG
  599. struct ext4_super_block *es;
  600. ext4_fsblk_t bitmap_count;
  601. unsigned int x;
  602. struct buffer_head *bitmap_bh = NULL;
  603. es = EXT4_SB(sb)->s_es;
  604. desc_count = 0;
  605. bitmap_count = 0;
  606. gdp = NULL;
  607. for (i = 0; i < ngroups; i++) {
  608. gdp = ext4_get_group_desc(sb, i, NULL);
  609. if (!gdp)
  610. continue;
  611. desc_count += ext4_free_blks_count(sb, gdp);
  612. brelse(bitmap_bh);
  613. bitmap_bh = ext4_read_block_bitmap(sb, i);
  614. if (bitmap_bh == NULL)
  615. continue;
  616. x = ext4_count_free(bitmap_bh, sb->s_blocksize);
  617. printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
  618. i, ext4_free_blks_count(sb, gdp), x);
  619. bitmap_count += x;
  620. }
  621. brelse(bitmap_bh);
  622. printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu"
  623. ", computed = %llu, %llu\n", ext4_free_blocks_count(es),
  624. desc_count, bitmap_count);
  625. return bitmap_count;
  626. #else
  627. desc_count = 0;
  628. for (i = 0; i < ngroups; i++) {
  629. gdp = ext4_get_group_desc(sb, i, NULL);
  630. if (!gdp)
  631. continue;
  632. desc_count += ext4_free_blks_count(sb, gdp);
  633. }
  634. return desc_count;
  635. #endif
  636. }
  637. static inline int test_root(ext4_group_t a, int b)
  638. {
  639. int num = b;
  640. while (a > num)
  641. num *= b;
  642. return num == a;
  643. }
  644. static int ext4_group_sparse(ext4_group_t group)
  645. {
  646. if (group <= 1)
  647. return 1;
  648. if (!(group & 1))
  649. return 0;
  650. return (test_root(group, 7) || test_root(group, 5) ||
  651. test_root(group, 3));
  652. }
  653. /**
  654. * ext4_bg_has_super - number of blocks used by the superblock in group
  655. * @sb: superblock for filesystem
  656. * @group: group number to check
  657. *
  658. * Return the number of blocks used by the superblock (primary or backup)
  659. * in this group. Currently this will be only 0 or 1.
  660. */
  661. int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
  662. {
  663. if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
  664. EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
  665. !ext4_group_sparse(group))
  666. return 0;
  667. return 1;
  668. }
  669. static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
  670. ext4_group_t group)
  671. {
  672. unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
  673. ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
  674. ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
  675. if (group == first || group == first + 1 || group == last)
  676. return 1;
  677. return 0;
  678. }
  679. static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
  680. ext4_group_t group)
  681. {
  682. return ext4_bg_has_super(sb, group) ? EXT4_SB(sb)->s_gdb_count : 0;
  683. }
  684. /**
  685. * ext4_bg_num_gdb - number of blocks used by the group table in group
  686. * @sb: superblock for filesystem
  687. * @group: group number to check
  688. *
  689. * Return the number of blocks used by the group descriptor table
  690. * (primary or backup) in this group. In the future there may be a
  691. * different number of descriptor blocks in each group.
  692. */
  693. unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
  694. {
  695. unsigned long first_meta_bg =
  696. le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
  697. unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
  698. if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) ||
  699. metagroup < first_meta_bg)
  700. return ext4_bg_num_gdb_nometa(sb, group);
  701. return ext4_bg_num_gdb_meta(sb,group);
  702. }