balloc.c 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771
  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 "group.h"
  22. #include "mballoc.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. {
  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. struct ext4_group_desc *gdp;
  59. struct buffer_head *bh;
  60. gdp = ext4_get_group_desc(sb, block_group, &bh);
  61. if (!ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp),
  62. block_group))
  63. used_blocks--;
  64. if (!ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp),
  65. block_group))
  66. used_blocks--;
  67. tmp = ext4_inode_table(sb, gdp);
  68. for (; tmp < ext4_inode_table(sb, gdp) +
  69. sbi->s_itb_per_group; tmp++) {
  70. if (!ext4_block_in_group(sb, tmp, block_group))
  71. used_blocks -= 1;
  72. }
  73. }
  74. return used_blocks;
  75. }
  76. /* Initializes an uninitialized block bitmap if given, and returns the
  77. * number of blocks free in the group. */
  78. unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
  79. ext4_group_t block_group, struct ext4_group_desc *gdp)
  80. {
  81. int bit, bit_max;
  82. unsigned free_blocks, group_blocks;
  83. struct ext4_sb_info *sbi = EXT4_SB(sb);
  84. if (bh) {
  85. J_ASSERT_BH(bh, buffer_locked(bh));
  86. /* If checksum is bad mark all blocks used to prevent allocation
  87. * essentially implementing a per-group read-only flag. */
  88. if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) {
  89. ext4_error(sb, __func__,
  90. "Checksum bad for group %u", block_group);
  91. gdp->bg_free_blocks_count = 0;
  92. gdp->bg_free_inodes_count = 0;
  93. gdp->bg_itable_unused = 0;
  94. memset(bh->b_data, 0xff, sb->s_blocksize);
  95. return 0;
  96. }
  97. memset(bh->b_data, 0, sb->s_blocksize);
  98. }
  99. /* Check for superblock and gdt backups in this group */
  100. bit_max = ext4_bg_has_super(sb, block_group);
  101. if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
  102. block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
  103. sbi->s_desc_per_block) {
  104. if (bit_max) {
  105. bit_max += ext4_bg_num_gdb(sb, block_group);
  106. bit_max +=
  107. le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
  108. }
  109. } else { /* For META_BG_BLOCK_GROUPS */
  110. bit_max += ext4_bg_num_gdb(sb, block_group);
  111. }
  112. if (block_group == sbi->s_groups_count - 1) {
  113. /*
  114. * Even though mke2fs always initialize first and last group
  115. * if some other tool enabled the EXT4_BG_BLOCK_UNINIT we need
  116. * to make sure we calculate the right free blocks
  117. */
  118. group_blocks = ext4_blocks_count(sbi->s_es) -
  119. le32_to_cpu(sbi->s_es->s_first_data_block) -
  120. (EXT4_BLOCKS_PER_GROUP(sb) * (sbi->s_groups_count - 1));
  121. } else {
  122. group_blocks = EXT4_BLOCKS_PER_GROUP(sb);
  123. }
  124. free_blocks = group_blocks - bit_max;
  125. if (bh) {
  126. ext4_fsblk_t start, tmp;
  127. int flex_bg = 0;
  128. for (bit = 0; bit < bit_max; bit++)
  129. ext4_set_bit(bit, bh->b_data);
  130. start = ext4_group_first_block_no(sb, block_group);
  131. if (EXT4_HAS_INCOMPAT_FEATURE(sb,
  132. EXT4_FEATURE_INCOMPAT_FLEX_BG))
  133. flex_bg = 1;
  134. /* Set bits for block and inode bitmaps, and inode table */
  135. tmp = ext4_block_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_bitmap(sb, gdp);
  139. if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
  140. ext4_set_bit(tmp - start, bh->b_data);
  141. tmp = ext4_inode_table(sb, gdp);
  142. for (; tmp < ext4_inode_table(sb, gdp) +
  143. sbi->s_itb_per_group; tmp++) {
  144. if (!flex_bg ||
  145. ext4_block_in_group(sb, tmp, block_group))
  146. ext4_set_bit(tmp - start, bh->b_data);
  147. }
  148. /*
  149. * Also if the number of blocks within the group is
  150. * less than the blocksize * 8 ( which is the size
  151. * of bitmap ), set rest of the block bitmap to 1
  152. */
  153. mark_bitmap_end(group_blocks, sb->s_blocksize * 8, bh->b_data);
  154. }
  155. return free_blocks - ext4_group_used_meta_blocks(sb, block_group);
  156. }
  157. /*
  158. * The free blocks are managed by bitmaps. A file system contains several
  159. * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
  160. * block for inodes, N blocks for the inode table and data blocks.
  161. *
  162. * The file system contains group descriptors which are located after the
  163. * super block. Each descriptor contains the number of the bitmap block and
  164. * the free blocks count in the block. The descriptors are loaded in memory
  165. * when a file system is mounted (see ext4_fill_super).
  166. */
  167. #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
  168. /**
  169. * ext4_get_group_desc() -- load group descriptor from disk
  170. * @sb: super block
  171. * @block_group: given block group
  172. * @bh: pointer to the buffer head to store the block
  173. * group descriptor
  174. */
  175. struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
  176. ext4_group_t block_group,
  177. struct buffer_head **bh)
  178. {
  179. unsigned int group_desc;
  180. unsigned int offset;
  181. struct ext4_group_desc *desc;
  182. struct ext4_sb_info *sbi = EXT4_SB(sb);
  183. if (block_group >= sbi->s_groups_count) {
  184. ext4_error(sb, "ext4_get_group_desc",
  185. "block_group >= groups_count - "
  186. "block_group = %u, groups_count = %u",
  187. block_group, sbi->s_groups_count);
  188. return NULL;
  189. }
  190. smp_rmb();
  191. group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
  192. offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
  193. if (!sbi->s_group_desc[group_desc]) {
  194. ext4_error(sb, "ext4_get_group_desc",
  195. "Group descriptor not loaded - "
  196. "block_group = %u, group_desc = %u, desc = %u",
  197. block_group, group_desc, offset);
  198. return NULL;
  199. }
  200. desc = (struct ext4_group_desc *)(
  201. (__u8 *)sbi->s_group_desc[group_desc]->b_data +
  202. offset * EXT4_DESC_SIZE(sb));
  203. if (bh)
  204. *bh = sbi->s_group_desc[group_desc];
  205. return desc;
  206. }
  207. static int ext4_valid_block_bitmap(struct super_block *sb,
  208. struct ext4_group_desc *desc,
  209. unsigned int block_group,
  210. struct buffer_head *bh)
  211. {
  212. ext4_grpblk_t offset;
  213. ext4_grpblk_t next_zero_bit;
  214. ext4_fsblk_t bitmap_blk;
  215. ext4_fsblk_t group_first_block;
  216. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
  217. /* with FLEX_BG, the inode/block bitmaps and itable
  218. * blocks may not be in the group at all
  219. * so the bitmap validation will be skipped for those groups
  220. * or it has to also read the block group where the bitmaps
  221. * are located to verify they are set.
  222. */
  223. return 1;
  224. }
  225. group_first_block = ext4_group_first_block_no(sb, block_group);
  226. /* check whether block bitmap block number is set */
  227. bitmap_blk = ext4_block_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 bitmap block number is set */
  233. bitmap_blk = ext4_inode_bitmap(sb, desc);
  234. offset = bitmap_blk - group_first_block;
  235. if (!ext4_test_bit(offset, bh->b_data))
  236. /* bad block bitmap */
  237. goto err_out;
  238. /* check whether the inode table block number is set */
  239. bitmap_blk = ext4_inode_table(sb, desc);
  240. offset = bitmap_blk - group_first_block;
  241. next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
  242. offset + EXT4_SB(sb)->s_itb_per_group,
  243. offset);
  244. if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group)
  245. /* good bitmap for inode tables */
  246. return 1;
  247. err_out:
  248. ext4_error(sb, __func__,
  249. "Invalid block bitmap - "
  250. "block_group = %d, block = %llu",
  251. block_group, bitmap_blk);
  252. return 0;
  253. }
  254. /**
  255. * ext4_read_block_bitmap()
  256. * @sb: super block
  257. * @block_group: given block group
  258. *
  259. * Read the bitmap for a given block_group,and validate the
  260. * bits for block/inode/inode tables are set in the bitmaps
  261. *
  262. * Return buffer_head on success or NULL in case of failure.
  263. */
  264. struct buffer_head *
  265. ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
  266. {
  267. struct ext4_group_desc *desc;
  268. struct buffer_head *bh = NULL;
  269. ext4_fsblk_t bitmap_blk;
  270. desc = ext4_get_group_desc(sb, block_group, NULL);
  271. if (!desc)
  272. return NULL;
  273. bitmap_blk = ext4_block_bitmap(sb, desc);
  274. bh = sb_getblk(sb, bitmap_blk);
  275. if (unlikely(!bh)) {
  276. ext4_error(sb, __func__,
  277. "Cannot read block bitmap - "
  278. "block_group = %u, block_bitmap = %llu",
  279. block_group, bitmap_blk);
  280. return NULL;
  281. }
  282. if (buffer_uptodate(bh) &&
  283. !(desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)))
  284. return bh;
  285. lock_buffer(bh);
  286. spin_lock(sb_bgl_lock(EXT4_SB(sb), block_group));
  287. if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
  288. ext4_init_block_bitmap(sb, bh, block_group, desc);
  289. set_buffer_uptodate(bh);
  290. unlock_buffer(bh);
  291. spin_unlock(sb_bgl_lock(EXT4_SB(sb), block_group));
  292. return bh;
  293. }
  294. spin_unlock(sb_bgl_lock(EXT4_SB(sb), block_group));
  295. if (bh_submit_read(bh) < 0) {
  296. put_bh(bh);
  297. ext4_error(sb, __func__,
  298. "Cannot read block bitmap - "
  299. "block_group = %u, block_bitmap = %llu",
  300. block_group, bitmap_blk);
  301. return NULL;
  302. }
  303. ext4_valid_block_bitmap(sb, desc, block_group, bh);
  304. /*
  305. * file system mounted not to panic on error,
  306. * continue with corrupt bitmap
  307. */
  308. return bh;
  309. }
  310. /**
  311. * ext4_add_groupblocks() -- Add given blocks to an existing group
  312. * @handle: handle to this transaction
  313. * @sb: super block
  314. * @block: start physcial block to add to the block group
  315. * @count: number of blocks to free
  316. *
  317. * This marks the blocks as free in the bitmap. We ask the
  318. * mballoc to reload the buddy after this by setting group
  319. * EXT4_GROUP_INFO_NEED_INIT_BIT flag
  320. */
  321. void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
  322. ext4_fsblk_t block, unsigned long count)
  323. {
  324. struct buffer_head *bitmap_bh = NULL;
  325. struct buffer_head *gd_bh;
  326. ext4_group_t block_group;
  327. ext4_grpblk_t bit;
  328. unsigned int i;
  329. struct ext4_group_desc *desc;
  330. struct ext4_super_block *es;
  331. struct ext4_sb_info *sbi;
  332. int err = 0, ret;
  333. ext4_grpblk_t blocks_freed;
  334. struct ext4_group_info *grp;
  335. sbi = EXT4_SB(sb);
  336. es = sbi->s_es;
  337. ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1);
  338. ext4_get_group_no_and_offset(sb, block, &block_group, &bit);
  339. grp = ext4_get_group_info(sb, block_group);
  340. /*
  341. * Check to see if we are freeing blocks across a group
  342. * boundary.
  343. */
  344. if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) {
  345. goto error_return;
  346. }
  347. bitmap_bh = ext4_read_block_bitmap(sb, block_group);
  348. if (!bitmap_bh)
  349. goto error_return;
  350. desc = ext4_get_group_desc(sb, block_group, &gd_bh);
  351. if (!desc)
  352. goto error_return;
  353. if (in_range(ext4_block_bitmap(sb, desc), block, count) ||
  354. in_range(ext4_inode_bitmap(sb, desc), block, count) ||
  355. in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) ||
  356. in_range(block + count - 1, ext4_inode_table(sb, desc),
  357. sbi->s_itb_per_group)) {
  358. ext4_error(sb, __func__,
  359. "Adding blocks in system zones - "
  360. "Block = %llu, count = %lu",
  361. block, count);
  362. goto error_return;
  363. }
  364. /*
  365. * We are about to add blocks to the bitmap,
  366. * so we need undo access.
  367. */
  368. BUFFER_TRACE(bitmap_bh, "getting undo access");
  369. err = ext4_journal_get_undo_access(handle, bitmap_bh);
  370. if (err)
  371. goto error_return;
  372. /*
  373. * We are about to modify some metadata. Call the journal APIs
  374. * to unshare ->b_data if a currently-committing transaction is
  375. * using it
  376. */
  377. BUFFER_TRACE(gd_bh, "get_write_access");
  378. err = ext4_journal_get_write_access(handle, gd_bh);
  379. if (err)
  380. goto error_return;
  381. /*
  382. * make sure we don't allow a parallel init on other groups in the
  383. * same buddy cache
  384. */
  385. down_write(&grp->alloc_sem);
  386. for (i = 0, blocks_freed = 0; i < count; i++) {
  387. BUFFER_TRACE(bitmap_bh, "clear bit");
  388. if (!ext4_clear_bit_atomic(sb_bgl_lock(sbi, block_group),
  389. bit + i, bitmap_bh->b_data)) {
  390. ext4_error(sb, __func__,
  391. "bit already cleared for block %llu",
  392. (ext4_fsblk_t)(block + i));
  393. BUFFER_TRACE(bitmap_bh, "bit already cleared");
  394. } else {
  395. blocks_freed++;
  396. }
  397. }
  398. spin_lock(sb_bgl_lock(sbi, block_group));
  399. le16_add_cpu(&desc->bg_free_blocks_count, blocks_freed);
  400. desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc);
  401. spin_unlock(sb_bgl_lock(sbi, block_group));
  402. percpu_counter_add(&sbi->s_freeblocks_counter, blocks_freed);
  403. if (sbi->s_log_groups_per_flex) {
  404. ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
  405. spin_lock(sb_bgl_lock(sbi, flex_group));
  406. sbi->s_flex_groups[flex_group].free_blocks += blocks_freed;
  407. spin_unlock(sb_bgl_lock(sbi, flex_group));
  408. }
  409. /*
  410. * request to reload the buddy with the
  411. * new bitmap information
  412. */
  413. set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state));
  414. ext4_mb_update_group_info(grp, blocks_freed);
  415. up_write(&grp->alloc_sem);
  416. /* We dirtied the bitmap block */
  417. BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
  418. err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
  419. /* And the group descriptor block */
  420. BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
  421. ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh);
  422. if (!err)
  423. err = ret;
  424. sb->s_dirt = 1;
  425. error_return:
  426. brelse(bitmap_bh);
  427. ext4_std_error(sb, err);
  428. return;
  429. }
  430. /**
  431. * ext4_free_blocks() -- Free given blocks and update quota
  432. * @handle: handle for this transaction
  433. * @inode: inode
  434. * @block: start physical block to free
  435. * @count: number of blocks to count
  436. * @metadata: Are these metadata blocks
  437. */
  438. void ext4_free_blocks(handle_t *handle, struct inode *inode,
  439. ext4_fsblk_t block, unsigned long count,
  440. int metadata)
  441. {
  442. struct super_block *sb;
  443. unsigned long dquot_freed_blocks;
  444. /* this isn't the right place to decide whether block is metadata
  445. * inode.c/extents.c knows better, but for safety ... */
  446. if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
  447. metadata = 1;
  448. /* We need to make sure we don't reuse
  449. * block released untill the transaction commit.
  450. * writeback mode have weak data consistency so
  451. * don't force data as metadata when freeing block
  452. * for writeback mode.
  453. */
  454. if (metadata == 0 && !ext4_should_writeback_data(inode))
  455. metadata = 1;
  456. sb = inode->i_sb;
  457. ext4_mb_free_blocks(handle, inode, block, count,
  458. metadata, &dquot_freed_blocks);
  459. if (dquot_freed_blocks)
  460. DQUOT_FREE_BLOCK(inode, dquot_freed_blocks);
  461. return;
  462. }
  463. /**
  464. * ext4_has_free_blocks()
  465. * @sbi: in-core super block structure.
  466. * @nblocks: number of needed blocks
  467. *
  468. * Check if filesystem has nblocks free & available for allocation.
  469. * On success return 1, return 0 on failure.
  470. */
  471. int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
  472. {
  473. s64 free_blocks, dirty_blocks, root_blocks;
  474. struct percpu_counter *fbc = &sbi->s_freeblocks_counter;
  475. struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter;
  476. free_blocks = percpu_counter_read_positive(fbc);
  477. dirty_blocks = percpu_counter_read_positive(dbc);
  478. root_blocks = ext4_r_blocks_count(sbi->s_es);
  479. if (free_blocks - (nblocks + root_blocks + dirty_blocks) <
  480. EXT4_FREEBLOCKS_WATERMARK) {
  481. free_blocks = percpu_counter_sum_positive(fbc);
  482. dirty_blocks = percpu_counter_sum_positive(dbc);
  483. if (dirty_blocks < 0) {
  484. printk(KERN_CRIT "Dirty block accounting "
  485. "went wrong %lld\n",
  486. (long long)dirty_blocks);
  487. }
  488. }
  489. /* Check whether we have space after
  490. * accounting for current dirty blocks & root reserved blocks.
  491. */
  492. if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks))
  493. return 1;
  494. /* Hm, nope. Are (enough) root reserved blocks available? */
  495. if (sbi->s_resuid == current_fsuid() ||
  496. ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) ||
  497. capable(CAP_SYS_RESOURCE)) {
  498. if (free_blocks >= (nblocks + dirty_blocks))
  499. return 1;
  500. }
  501. return 0;
  502. }
  503. int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
  504. s64 nblocks)
  505. {
  506. if (ext4_has_free_blocks(sbi, nblocks)) {
  507. percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks);
  508. return 0;
  509. } else
  510. return -ENOSPC;
  511. }
  512. /**
  513. * ext4_should_retry_alloc()
  514. * @sb: super block
  515. * @retries number of attemps has been made
  516. *
  517. * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
  518. * it is profitable to retry the operation, this function will wait
  519. * for the current or commiting transaction to complete, and then
  520. * return TRUE.
  521. *
  522. * if the total number of retries exceed three times, return FALSE.
  523. */
  524. int ext4_should_retry_alloc(struct super_block *sb, int *retries)
  525. {
  526. if (!ext4_has_free_blocks(EXT4_SB(sb), 1) || (*retries)++ > 3)
  527. return 0;
  528. jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
  529. return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
  530. }
  531. /*
  532. * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
  533. *
  534. * @handle: handle to this transaction
  535. * @inode: file inode
  536. * @goal: given target block(filesystem wide)
  537. * @count: pointer to total number of blocks needed
  538. * @errp: error code
  539. *
  540. * Return 1st allocated block number on success, *count stores total account
  541. * error stores in errp pointer
  542. */
  543. ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
  544. ext4_fsblk_t goal, unsigned long *count, int *errp)
  545. {
  546. struct ext4_allocation_request ar;
  547. ext4_fsblk_t ret;
  548. memset(&ar, 0, sizeof(ar));
  549. /* Fill with neighbour allocated blocks */
  550. ar.inode = inode;
  551. ar.goal = goal;
  552. ar.len = count ? *count : 1;
  553. ret = ext4_mb_new_blocks(handle, &ar, errp);
  554. if (count)
  555. *count = ar.len;
  556. /*
  557. * Account for the allocated meta blocks
  558. */
  559. if (!(*errp) && EXT4_I(inode)->i_delalloc_reserved_flag) {
  560. spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
  561. EXT4_I(inode)->i_allocated_meta_blocks += ar.len;
  562. spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
  563. }
  564. return ret;
  565. }
  566. /**
  567. * ext4_count_free_blocks() -- count filesystem free blocks
  568. * @sb: superblock
  569. *
  570. * Adds up the number of free blocks from each block group.
  571. */
  572. ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb)
  573. {
  574. ext4_fsblk_t desc_count;
  575. struct ext4_group_desc *gdp;
  576. ext4_group_t i;
  577. ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
  578. #ifdef EXT4FS_DEBUG
  579. struct ext4_super_block *es;
  580. ext4_fsblk_t bitmap_count;
  581. unsigned int x;
  582. struct buffer_head *bitmap_bh = NULL;
  583. es = EXT4_SB(sb)->s_es;
  584. desc_count = 0;
  585. bitmap_count = 0;
  586. gdp = NULL;
  587. smp_rmb();
  588. for (i = 0; i < ngroups; i++) {
  589. gdp = ext4_get_group_desc(sb, i, NULL);
  590. if (!gdp)
  591. continue;
  592. desc_count += le16_to_cpu(gdp->bg_free_blocks_count);
  593. brelse(bitmap_bh);
  594. bitmap_bh = ext4_read_block_bitmap(sb, i);
  595. if (bitmap_bh == NULL)
  596. continue;
  597. x = ext4_count_free(bitmap_bh, sb->s_blocksize);
  598. printk(KERN_DEBUG "group %lu: stored = %d, counted = %u\n",
  599. i, le16_to_cpu(gdp->bg_free_blocks_count), x);
  600. bitmap_count += x;
  601. }
  602. brelse(bitmap_bh);
  603. printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu"
  604. ", computed = %llu, %llu\n", ext4_free_blocks_count(es),
  605. desc_count, bitmap_count);
  606. return bitmap_count;
  607. #else
  608. desc_count = 0;
  609. smp_rmb();
  610. for (i = 0; i < ngroups; i++) {
  611. gdp = ext4_get_group_desc(sb, i, NULL);
  612. if (!gdp)
  613. continue;
  614. desc_count += le16_to_cpu(gdp->bg_free_blocks_count);
  615. }
  616. return desc_count;
  617. #endif
  618. }
  619. static inline int test_root(ext4_group_t a, int b)
  620. {
  621. int num = b;
  622. while (a > num)
  623. num *= b;
  624. return num == a;
  625. }
  626. static int ext4_group_sparse(ext4_group_t group)
  627. {
  628. if (group <= 1)
  629. return 1;
  630. if (!(group & 1))
  631. return 0;
  632. return (test_root(group, 7) || test_root(group, 5) ||
  633. test_root(group, 3));
  634. }
  635. /**
  636. * ext4_bg_has_super - number of blocks used by the superblock in group
  637. * @sb: superblock for filesystem
  638. * @group: group number to check
  639. *
  640. * Return the number of blocks used by the superblock (primary or backup)
  641. * in this group. Currently this will be only 0 or 1.
  642. */
  643. int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
  644. {
  645. if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
  646. EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
  647. !ext4_group_sparse(group))
  648. return 0;
  649. return 1;
  650. }
  651. static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
  652. ext4_group_t group)
  653. {
  654. unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
  655. ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
  656. ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
  657. if (group == first || group == first + 1 || group == last)
  658. return 1;
  659. return 0;
  660. }
  661. static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
  662. ext4_group_t group)
  663. {
  664. return ext4_bg_has_super(sb, group) ? EXT4_SB(sb)->s_gdb_count : 0;
  665. }
  666. /**
  667. * ext4_bg_num_gdb - number of blocks used by the group table in group
  668. * @sb: superblock for filesystem
  669. * @group: group number to check
  670. *
  671. * Return the number of blocks used by the group descriptor table
  672. * (primary or backup) in this group. In the future there may be a
  673. * different number of descriptor blocks in each group.
  674. */
  675. unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
  676. {
  677. unsigned long first_meta_bg =
  678. le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
  679. unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
  680. if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) ||
  681. metagroup < first_meta_bg)
  682. return ext4_bg_num_gdb_nometa(sb, group);
  683. return ext4_bg_num_gdb_meta(sb,group);
  684. }