| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* | 
|  | 2 | *  linux/fs/ext2/ialloc.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 | *  BSD ufs-inspired inode and directory allocation by | 
|  | 10 | *  Stephen Tweedie (sct@dcs.ed.ac.uk), 1993 | 
|  | 11 | *  Big-endian to little-endian byte-swapping/bitmaps by | 
|  | 12 | *        David S. Miller (davem@caip.rutgers.edu), 1995 | 
|  | 13 | */ | 
|  | 14 |  | 
|  | 15 | #include <linux/config.h> | 
|  | 16 | #include <linux/quotaops.h> | 
|  | 17 | #include <linux/sched.h> | 
|  | 18 | #include <linux/backing-dev.h> | 
|  | 19 | #include <linux/buffer_head.h> | 
|  | 20 | #include <linux/random.h> | 
|  | 21 | #include "ext2.h" | 
|  | 22 | #include "xattr.h" | 
|  | 23 | #include "acl.h" | 
|  | 24 |  | 
|  | 25 | /* | 
|  | 26 | * ialloc.c contains the inodes allocation and deallocation routines | 
|  | 27 | */ | 
|  | 28 |  | 
|  | 29 | /* | 
|  | 30 | * The free inodes are managed by bitmaps.  A file system contains several | 
|  | 31 | * blocks groups.  Each group contains 1 bitmap block for blocks, 1 bitmap | 
|  | 32 | * block for inodes, N blocks for the inode table and data blocks. | 
|  | 33 | * | 
|  | 34 | * The file system contains group descriptors which are located after the | 
|  | 35 | * super block.  Each descriptor contains the number of the bitmap block and | 
|  | 36 | * the free blocks count in the block. | 
|  | 37 | */ | 
|  | 38 |  | 
|  | 39 |  | 
|  | 40 | /* | 
|  | 41 | * Read the inode allocation bitmap for a given block_group, reading | 
|  | 42 | * into the specified slot in the superblock's bitmap cache. | 
|  | 43 | * | 
|  | 44 | * Return buffer_head of bitmap on success or NULL. | 
|  | 45 | */ | 
|  | 46 | static struct buffer_head * | 
|  | 47 | read_inode_bitmap(struct super_block * sb, unsigned long block_group) | 
|  | 48 | { | 
|  | 49 | struct ext2_group_desc *desc; | 
|  | 50 | struct buffer_head *bh = NULL; | 
|  | 51 |  | 
|  | 52 | desc = ext2_get_group_desc(sb, block_group, NULL); | 
|  | 53 | if (!desc) | 
|  | 54 | goto error_out; | 
|  | 55 |  | 
|  | 56 | bh = sb_bread(sb, le32_to_cpu(desc->bg_inode_bitmap)); | 
|  | 57 | if (!bh) | 
|  | 58 | ext2_error(sb, "read_inode_bitmap", | 
|  | 59 | "Cannot read inode bitmap - " | 
|  | 60 | "block_group = %lu, inode_bitmap = %u", | 
|  | 61 | block_group, le32_to_cpu(desc->bg_inode_bitmap)); | 
|  | 62 | error_out: | 
|  | 63 | return bh; | 
|  | 64 | } | 
|  | 65 |  | 
|  | 66 | static void ext2_release_inode(struct super_block *sb, int group, int dir) | 
|  | 67 | { | 
|  | 68 | struct ext2_group_desc * desc; | 
|  | 69 | struct buffer_head *bh; | 
|  | 70 |  | 
|  | 71 | desc = ext2_get_group_desc(sb, group, &bh); | 
|  | 72 | if (!desc) { | 
|  | 73 | ext2_error(sb, "ext2_release_inode", | 
|  | 74 | "can't get descriptor for group %d", group); | 
|  | 75 | return; | 
|  | 76 | } | 
|  | 77 |  | 
|  | 78 | spin_lock(sb_bgl_lock(EXT2_SB(sb), group)); | 
|  | 79 | desc->bg_free_inodes_count = | 
|  | 80 | cpu_to_le16(le16_to_cpu(desc->bg_free_inodes_count) + 1); | 
|  | 81 | if (dir) | 
|  | 82 | desc->bg_used_dirs_count = | 
|  | 83 | cpu_to_le16(le16_to_cpu(desc->bg_used_dirs_count) - 1); | 
|  | 84 | spin_unlock(sb_bgl_lock(EXT2_SB(sb), group)); | 
|  | 85 | if (dir) | 
|  | 86 | percpu_counter_dec(&EXT2_SB(sb)->s_dirs_counter); | 
|  | 87 | sb->s_dirt = 1; | 
|  | 88 | mark_buffer_dirty(bh); | 
|  | 89 | } | 
|  | 90 |  | 
|  | 91 | /* | 
|  | 92 | * NOTE! When we get the inode, we're the only people | 
|  | 93 | * that have access to it, and as such there are no | 
|  | 94 | * race conditions we have to worry about. The inode | 
|  | 95 | * is not on the hash-lists, and it cannot be reached | 
|  | 96 | * through the filesystem because the directory entry | 
|  | 97 | * has been deleted earlier. | 
|  | 98 | * | 
|  | 99 | * HOWEVER: we must make sure that we get no aliases, | 
|  | 100 | * which means that we have to call "clear_inode()" | 
|  | 101 | * _before_ we mark the inode not in use in the inode | 
|  | 102 | * bitmaps. Otherwise a newly created file might use | 
|  | 103 | * the same inode number (not actually the same pointer | 
|  | 104 | * though), and then we'd have two inodes sharing the | 
|  | 105 | * same inode number and space on the harddisk. | 
|  | 106 | */ | 
|  | 107 | void ext2_free_inode (struct inode * inode) | 
|  | 108 | { | 
|  | 109 | struct super_block * sb = inode->i_sb; | 
|  | 110 | int is_directory; | 
|  | 111 | unsigned long ino; | 
|  | 112 | struct buffer_head *bitmap_bh = NULL; | 
|  | 113 | unsigned long block_group; | 
|  | 114 | unsigned long bit; | 
|  | 115 | struct ext2_super_block * es; | 
|  | 116 |  | 
|  | 117 | ino = inode->i_ino; | 
|  | 118 | ext2_debug ("freeing inode %lu\n", ino); | 
|  | 119 |  | 
|  | 120 | /* | 
|  | 121 | * Note: we must free any quota before locking the superblock, | 
|  | 122 | * as writing the quota to disk may need the lock as well. | 
|  | 123 | */ | 
|  | 124 | if (!is_bad_inode(inode)) { | 
|  | 125 | /* Quota is already initialized in iput() */ | 
|  | 126 | ext2_xattr_delete_inode(inode); | 
|  | 127 | DQUOT_FREE_INODE(inode); | 
|  | 128 | DQUOT_DROP(inode); | 
|  | 129 | } | 
|  | 130 |  | 
|  | 131 | es = EXT2_SB(sb)->s_es; | 
|  | 132 | is_directory = S_ISDIR(inode->i_mode); | 
|  | 133 |  | 
|  | 134 | /* Do this BEFORE marking the inode not in use or returning an error */ | 
|  | 135 | clear_inode (inode); | 
|  | 136 |  | 
|  | 137 | if (ino < EXT2_FIRST_INO(sb) || | 
|  | 138 | ino > le32_to_cpu(es->s_inodes_count)) { | 
|  | 139 | ext2_error (sb, "ext2_free_inode", | 
|  | 140 | "reserved or nonexistent inode %lu", ino); | 
|  | 141 | goto error_return; | 
|  | 142 | } | 
|  | 143 | block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb); | 
|  | 144 | bit = (ino - 1) % EXT2_INODES_PER_GROUP(sb); | 
|  | 145 | brelse(bitmap_bh); | 
|  | 146 | bitmap_bh = read_inode_bitmap(sb, block_group); | 
|  | 147 | if (!bitmap_bh) | 
|  | 148 | goto error_return; | 
|  | 149 |  | 
|  | 150 | /* Ok, now we can actually update the inode bitmaps.. */ | 
|  | 151 | if (!ext2_clear_bit_atomic(sb_bgl_lock(EXT2_SB(sb), block_group), | 
|  | 152 | bit, (void *) bitmap_bh->b_data)) | 
|  | 153 | ext2_error (sb, "ext2_free_inode", | 
|  | 154 | "bit already cleared for inode %lu", ino); | 
|  | 155 | else | 
|  | 156 | ext2_release_inode(sb, block_group, is_directory); | 
|  | 157 | mark_buffer_dirty(bitmap_bh); | 
|  | 158 | if (sb->s_flags & MS_SYNCHRONOUS) | 
|  | 159 | sync_dirty_buffer(bitmap_bh); | 
|  | 160 | error_return: | 
|  | 161 | brelse(bitmap_bh); | 
|  | 162 | } | 
|  | 163 |  | 
|  | 164 | /* | 
|  | 165 | * We perform asynchronous prereading of the new inode's inode block when | 
|  | 166 | * we create the inode, in the expectation that the inode will be written | 
|  | 167 | * back soon.  There are two reasons: | 
|  | 168 | * | 
|  | 169 | * - When creating a large number of files, the async prereads will be | 
|  | 170 | *   nicely merged into large reads | 
|  | 171 | * - When writing out a large number of inodes, we don't need to keep on | 
|  | 172 | *   stalling the writes while we read the inode block. | 
|  | 173 | * | 
|  | 174 | * FIXME: ext2_get_group_desc() needs to be simplified. | 
|  | 175 | */ | 
|  | 176 | static void ext2_preread_inode(struct inode *inode) | 
|  | 177 | { | 
|  | 178 | unsigned long block_group; | 
|  | 179 | unsigned long offset; | 
|  | 180 | unsigned long block; | 
|  | 181 | struct buffer_head *bh; | 
|  | 182 | struct ext2_group_desc * gdp; | 
|  | 183 | struct backing_dev_info *bdi; | 
|  | 184 |  | 
|  | 185 | bdi = inode->i_mapping->backing_dev_info; | 
|  | 186 | if (bdi_read_congested(bdi)) | 
|  | 187 | return; | 
|  | 188 | if (bdi_write_congested(bdi)) | 
|  | 189 | return; | 
|  | 190 |  | 
|  | 191 | block_group = (inode->i_ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb); | 
|  | 192 | gdp = ext2_get_group_desc(inode->i_sb, block_group, &bh); | 
|  | 193 | if (gdp == NULL) | 
|  | 194 | return; | 
|  | 195 |  | 
|  | 196 | /* | 
|  | 197 | * Figure out the offset within the block group inode table | 
|  | 198 | */ | 
|  | 199 | offset = ((inode->i_ino - 1) % EXT2_INODES_PER_GROUP(inode->i_sb)) * | 
|  | 200 | EXT2_INODE_SIZE(inode->i_sb); | 
|  | 201 | block = le32_to_cpu(gdp->bg_inode_table) + | 
|  | 202 | (offset >> EXT2_BLOCK_SIZE_BITS(inode->i_sb)); | 
|  | 203 | sb_breadahead(inode->i_sb, block); | 
|  | 204 | } | 
|  | 205 |  | 
|  | 206 | /* | 
|  | 207 | * There are two policies for allocating an inode.  If the new inode is | 
|  | 208 | * a directory, then a forward search is made for a block group with both | 
|  | 209 | * free space and a low directory-to-inode ratio; if that fails, then of | 
|  | 210 | * the groups with above-average free space, that group with the fewest | 
|  | 211 | * directories already is chosen. | 
|  | 212 | * | 
|  | 213 | * For other inodes, search forward from the parent directory\'s block | 
|  | 214 | * group to find a free inode. | 
|  | 215 | */ | 
|  | 216 | static int find_group_dir(struct super_block *sb, struct inode *parent) | 
|  | 217 | { | 
|  | 218 | int ngroups = EXT2_SB(sb)->s_groups_count; | 
|  | 219 | int avefreei = ext2_count_free_inodes(sb) / ngroups; | 
|  | 220 | struct ext2_group_desc *desc, *best_desc = NULL; | 
|  | 221 | struct buffer_head *bh, *best_bh = NULL; | 
|  | 222 | int group, best_group = -1; | 
|  | 223 |  | 
|  | 224 | for (group = 0; group < ngroups; group++) { | 
|  | 225 | desc = ext2_get_group_desc (sb, group, &bh); | 
|  | 226 | if (!desc || !desc->bg_free_inodes_count) | 
|  | 227 | continue; | 
|  | 228 | if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei) | 
|  | 229 | continue; | 
|  | 230 | if (!best_desc || | 
|  | 231 | (le16_to_cpu(desc->bg_free_blocks_count) > | 
|  | 232 | le16_to_cpu(best_desc->bg_free_blocks_count))) { | 
|  | 233 | best_group = group; | 
|  | 234 | best_desc = desc; | 
|  | 235 | best_bh = bh; | 
|  | 236 | } | 
|  | 237 | } | 
|  | 238 | if (!best_desc) | 
|  | 239 | return -1; | 
|  | 240 |  | 
|  | 241 | return best_group; | 
|  | 242 | } | 
|  | 243 |  | 
|  | 244 | /* | 
|  | 245 | * Orlov's allocator for directories. | 
|  | 246 | * | 
|  | 247 | * We always try to spread first-level directories. | 
|  | 248 | * | 
|  | 249 | * If there are blockgroups with both free inodes and free blocks counts | 
|  | 250 | * not worse than average we return one with smallest directory count. | 
|  | 251 | * Otherwise we simply return a random group. | 
|  | 252 | * | 
|  | 253 | * For the rest rules look so: | 
|  | 254 | * | 
|  | 255 | * It's OK to put directory into a group unless | 
|  | 256 | * it has too many directories already (max_dirs) or | 
|  | 257 | * it has too few free inodes left (min_inodes) or | 
|  | 258 | * it has too few free blocks left (min_blocks) or | 
|  | 259 | * it's already running too large debt (max_debt). | 
|  | 260 | * Parent's group is prefered, if it doesn't satisfy these | 
|  | 261 | * conditions we search cyclically through the rest. If none | 
|  | 262 | * of the groups look good we just look for a group with more | 
|  | 263 | * free inodes than average (starting at parent's group). | 
|  | 264 | * | 
|  | 265 | * Debt is incremented each time we allocate a directory and decremented | 
|  | 266 | * when we allocate an inode, within 0--255. | 
|  | 267 | */ | 
|  | 268 |  | 
|  | 269 | #define INODE_COST 64 | 
|  | 270 | #define BLOCK_COST 256 | 
|  | 271 |  | 
|  | 272 | static int find_group_orlov(struct super_block *sb, struct inode *parent) | 
|  | 273 | { | 
|  | 274 | int parent_group = EXT2_I(parent)->i_block_group; | 
|  | 275 | struct ext2_sb_info *sbi = EXT2_SB(sb); | 
|  | 276 | struct ext2_super_block *es = sbi->s_es; | 
|  | 277 | int ngroups = sbi->s_groups_count; | 
|  | 278 | int inodes_per_group = EXT2_INODES_PER_GROUP(sb); | 
|  | 279 | int freei; | 
|  | 280 | int avefreei; | 
|  | 281 | int free_blocks; | 
|  | 282 | int avefreeb; | 
|  | 283 | int blocks_per_dir; | 
|  | 284 | int ndirs; | 
|  | 285 | int max_debt, max_dirs, min_blocks, min_inodes; | 
|  | 286 | int group = -1, i; | 
|  | 287 | struct ext2_group_desc *desc; | 
|  | 288 | struct buffer_head *bh; | 
|  | 289 |  | 
|  | 290 | freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter); | 
|  | 291 | avefreei = freei / ngroups; | 
|  | 292 | free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter); | 
|  | 293 | avefreeb = free_blocks / ngroups; | 
|  | 294 | ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter); | 
|  | 295 |  | 
|  | 296 | if ((parent == sb->s_root->d_inode) || | 
|  | 297 | (EXT2_I(parent)->i_flags & EXT2_TOPDIR_FL)) { | 
|  | 298 | struct ext2_group_desc *best_desc = NULL; | 
|  | 299 | struct buffer_head *best_bh = NULL; | 
|  | 300 | int best_ndir = inodes_per_group; | 
|  | 301 | int best_group = -1; | 
|  | 302 |  | 
|  | 303 | get_random_bytes(&group, sizeof(group)); | 
|  | 304 | parent_group = (unsigned)group % ngroups; | 
|  | 305 | for (i = 0; i < ngroups; i++) { | 
|  | 306 | group = (parent_group + i) % ngroups; | 
|  | 307 | desc = ext2_get_group_desc (sb, group, &bh); | 
|  | 308 | if (!desc || !desc->bg_free_inodes_count) | 
|  | 309 | continue; | 
|  | 310 | if (le16_to_cpu(desc->bg_used_dirs_count) >= best_ndir) | 
|  | 311 | continue; | 
|  | 312 | if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei) | 
|  | 313 | continue; | 
|  | 314 | if (le16_to_cpu(desc->bg_free_blocks_count) < avefreeb) | 
|  | 315 | continue; | 
|  | 316 | best_group = group; | 
|  | 317 | best_ndir = le16_to_cpu(desc->bg_used_dirs_count); | 
|  | 318 | best_desc = desc; | 
|  | 319 | best_bh = bh; | 
|  | 320 | } | 
|  | 321 | if (best_group >= 0) { | 
|  | 322 | desc = best_desc; | 
|  | 323 | bh = best_bh; | 
|  | 324 | group = best_group; | 
|  | 325 | goto found; | 
|  | 326 | } | 
|  | 327 | goto fallback; | 
|  | 328 | } | 
|  | 329 |  | 
|  | 330 | if (ndirs == 0) | 
|  | 331 | ndirs = 1;	/* percpu_counters are approximate... */ | 
|  | 332 |  | 
|  | 333 | blocks_per_dir = (le32_to_cpu(es->s_blocks_count)-free_blocks) / ndirs; | 
|  | 334 |  | 
|  | 335 | max_dirs = ndirs / ngroups + inodes_per_group / 16; | 
|  | 336 | min_inodes = avefreei - inodes_per_group / 4; | 
|  | 337 | min_blocks = avefreeb - EXT2_BLOCKS_PER_GROUP(sb) / 4; | 
|  | 338 |  | 
|  | 339 | max_debt = EXT2_BLOCKS_PER_GROUP(sb) / max(blocks_per_dir, BLOCK_COST); | 
|  | 340 | if (max_debt * INODE_COST > inodes_per_group) | 
|  | 341 | max_debt = inodes_per_group / INODE_COST; | 
|  | 342 | if (max_debt > 255) | 
|  | 343 | max_debt = 255; | 
|  | 344 | if (max_debt == 0) | 
|  | 345 | max_debt = 1; | 
|  | 346 |  | 
|  | 347 | for (i = 0; i < ngroups; i++) { | 
|  | 348 | group = (parent_group + i) % ngroups; | 
|  | 349 | desc = ext2_get_group_desc (sb, group, &bh); | 
|  | 350 | if (!desc || !desc->bg_free_inodes_count) | 
|  | 351 | continue; | 
|  | 352 | if (sbi->s_debts[group] >= max_debt) | 
|  | 353 | continue; | 
|  | 354 | if (le16_to_cpu(desc->bg_used_dirs_count) >= max_dirs) | 
|  | 355 | continue; | 
|  | 356 | if (le16_to_cpu(desc->bg_free_inodes_count) < min_inodes) | 
|  | 357 | continue; | 
|  | 358 | if (le16_to_cpu(desc->bg_free_blocks_count) < min_blocks) | 
|  | 359 | continue; | 
|  | 360 | goto found; | 
|  | 361 | } | 
|  | 362 |  | 
|  | 363 | fallback: | 
|  | 364 | for (i = 0; i < ngroups; i++) { | 
|  | 365 | group = (parent_group + i) % ngroups; | 
|  | 366 | desc = ext2_get_group_desc (sb, group, &bh); | 
|  | 367 | if (!desc || !desc->bg_free_inodes_count) | 
|  | 368 | continue; | 
|  | 369 | if (le16_to_cpu(desc->bg_free_inodes_count) >= avefreei) | 
|  | 370 | goto found; | 
|  | 371 | } | 
|  | 372 |  | 
|  | 373 | if (avefreei) { | 
|  | 374 | /* | 
|  | 375 | * The free-inodes counter is approximate, and for really small | 
|  | 376 | * filesystems the above test can fail to find any blockgroups | 
|  | 377 | */ | 
|  | 378 | avefreei = 0; | 
|  | 379 | goto fallback; | 
|  | 380 | } | 
|  | 381 |  | 
|  | 382 | return -1; | 
|  | 383 |  | 
|  | 384 | found: | 
|  | 385 | return group; | 
|  | 386 | } | 
|  | 387 |  | 
|  | 388 | static int find_group_other(struct super_block *sb, struct inode *parent) | 
|  | 389 | { | 
|  | 390 | int parent_group = EXT2_I(parent)->i_block_group; | 
|  | 391 | int ngroups = EXT2_SB(sb)->s_groups_count; | 
|  | 392 | struct ext2_group_desc *desc; | 
|  | 393 | struct buffer_head *bh; | 
|  | 394 | int group, i; | 
|  | 395 |  | 
|  | 396 | /* | 
|  | 397 | * Try to place the inode in its parent directory | 
|  | 398 | */ | 
|  | 399 | group = parent_group; | 
|  | 400 | desc = ext2_get_group_desc (sb, group, &bh); | 
|  | 401 | if (desc && le16_to_cpu(desc->bg_free_inodes_count) && | 
|  | 402 | le16_to_cpu(desc->bg_free_blocks_count)) | 
|  | 403 | goto found; | 
|  | 404 |  | 
|  | 405 | /* | 
|  | 406 | * We're going to place this inode in a different blockgroup from its | 
|  | 407 | * parent.  We want to cause files in a common directory to all land in | 
|  | 408 | * the same blockgroup.  But we want files which are in a different | 
|  | 409 | * directory which shares a blockgroup with our parent to land in a | 
|  | 410 | * different blockgroup. | 
|  | 411 | * | 
|  | 412 | * So add our directory's i_ino into the starting point for the hash. | 
|  | 413 | */ | 
|  | 414 | group = (group + parent->i_ino) % ngroups; | 
|  | 415 |  | 
|  | 416 | /* | 
|  | 417 | * Use a quadratic hash to find a group with a free inode and some | 
|  | 418 | * free blocks. | 
|  | 419 | */ | 
|  | 420 | for (i = 1; i < ngroups; i <<= 1) { | 
|  | 421 | group += i; | 
|  | 422 | if (group >= ngroups) | 
|  | 423 | group -= ngroups; | 
|  | 424 | desc = ext2_get_group_desc (sb, group, &bh); | 
|  | 425 | if (desc && le16_to_cpu(desc->bg_free_inodes_count) && | 
|  | 426 | le16_to_cpu(desc->bg_free_blocks_count)) | 
|  | 427 | goto found; | 
|  | 428 | } | 
|  | 429 |  | 
|  | 430 | /* | 
|  | 431 | * That failed: try linear search for a free inode, even if that group | 
|  | 432 | * has no free blocks. | 
|  | 433 | */ | 
|  | 434 | group = parent_group; | 
|  | 435 | for (i = 0; i < ngroups; i++) { | 
|  | 436 | if (++group >= ngroups) | 
|  | 437 | group = 0; | 
|  | 438 | desc = ext2_get_group_desc (sb, group, &bh); | 
|  | 439 | if (desc && le16_to_cpu(desc->bg_free_inodes_count)) | 
|  | 440 | goto found; | 
|  | 441 | } | 
|  | 442 |  | 
|  | 443 | return -1; | 
|  | 444 |  | 
|  | 445 | found: | 
|  | 446 | return group; | 
|  | 447 | } | 
|  | 448 |  | 
|  | 449 | struct inode *ext2_new_inode(struct inode *dir, int mode) | 
|  | 450 | { | 
|  | 451 | struct super_block *sb; | 
|  | 452 | struct buffer_head *bitmap_bh = NULL; | 
|  | 453 | struct buffer_head *bh2; | 
|  | 454 | int group, i; | 
|  | 455 | ino_t ino = 0; | 
|  | 456 | struct inode * inode; | 
|  | 457 | struct ext2_group_desc *gdp; | 
|  | 458 | struct ext2_super_block *es; | 
|  | 459 | struct ext2_inode_info *ei; | 
|  | 460 | struct ext2_sb_info *sbi; | 
|  | 461 | int err; | 
|  | 462 |  | 
|  | 463 | sb = dir->i_sb; | 
|  | 464 | inode = new_inode(sb); | 
|  | 465 | if (!inode) | 
|  | 466 | return ERR_PTR(-ENOMEM); | 
|  | 467 |  | 
|  | 468 | ei = EXT2_I(inode); | 
|  | 469 | sbi = EXT2_SB(sb); | 
|  | 470 | es = sbi->s_es; | 
|  | 471 | if (S_ISDIR(mode)) { | 
|  | 472 | if (test_opt(sb, OLDALLOC)) | 
|  | 473 | group = find_group_dir(sb, dir); | 
|  | 474 | else | 
|  | 475 | group = find_group_orlov(sb, dir); | 
|  | 476 | } else | 
|  | 477 | group = find_group_other(sb, dir); | 
|  | 478 |  | 
|  | 479 | if (group == -1) { | 
|  | 480 | err = -ENOSPC; | 
|  | 481 | goto fail; | 
|  | 482 | } | 
|  | 483 |  | 
|  | 484 | for (i = 0; i < sbi->s_groups_count; i++) { | 
|  | 485 | gdp = ext2_get_group_desc(sb, group, &bh2); | 
|  | 486 | brelse(bitmap_bh); | 
|  | 487 | bitmap_bh = read_inode_bitmap(sb, group); | 
|  | 488 | if (!bitmap_bh) { | 
|  | 489 | err = -EIO; | 
|  | 490 | goto fail; | 
|  | 491 | } | 
|  | 492 | ino = 0; | 
|  | 493 |  | 
|  | 494 | repeat_in_this_group: | 
|  | 495 | ino = ext2_find_next_zero_bit((unsigned long *)bitmap_bh->b_data, | 
|  | 496 | EXT2_INODES_PER_GROUP(sb), ino); | 
|  | 497 | if (ino >= EXT2_INODES_PER_GROUP(sb)) { | 
|  | 498 | /* | 
|  | 499 | * Rare race: find_group_xx() decided that there were | 
|  | 500 | * free inodes in this group, but by the time we tried | 
|  | 501 | * to allocate one, they're all gone.  This can also | 
|  | 502 | * occur because the counters which find_group_orlov() | 
|  | 503 | * uses are approximate.  So just go and search the | 
|  | 504 | * next block group. | 
|  | 505 | */ | 
|  | 506 | if (++group == sbi->s_groups_count) | 
|  | 507 | group = 0; | 
|  | 508 | continue; | 
|  | 509 | } | 
|  | 510 | if (ext2_set_bit_atomic(sb_bgl_lock(sbi, group), | 
|  | 511 | ino, bitmap_bh->b_data)) { | 
|  | 512 | /* we lost this inode */ | 
|  | 513 | if (++ino >= EXT2_INODES_PER_GROUP(sb)) { | 
|  | 514 | /* this group is exhausted, try next group */ | 
|  | 515 | if (++group == sbi->s_groups_count) | 
|  | 516 | group = 0; | 
|  | 517 | continue; | 
|  | 518 | } | 
|  | 519 | /* try to find free inode in the same group */ | 
|  | 520 | goto repeat_in_this_group; | 
|  | 521 | } | 
|  | 522 | goto got; | 
|  | 523 | } | 
|  | 524 |  | 
|  | 525 | /* | 
|  | 526 | * Scanned all blockgroups. | 
|  | 527 | */ | 
|  | 528 | err = -ENOSPC; | 
|  | 529 | goto fail; | 
|  | 530 | got: | 
|  | 531 | mark_buffer_dirty(bitmap_bh); | 
|  | 532 | if (sb->s_flags & MS_SYNCHRONOUS) | 
|  | 533 | sync_dirty_buffer(bitmap_bh); | 
|  | 534 | brelse(bitmap_bh); | 
|  | 535 |  | 
|  | 536 | ino += group * EXT2_INODES_PER_GROUP(sb) + 1; | 
|  | 537 | if (ino < EXT2_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) { | 
|  | 538 | ext2_error (sb, "ext2_new_inode", | 
|  | 539 | "reserved inode or inode > inodes count - " | 
|  | 540 | "block_group = %d,inode=%lu", group, | 
|  | 541 | (unsigned long) ino); | 
|  | 542 | err = -EIO; | 
|  | 543 | goto fail; | 
|  | 544 | } | 
|  | 545 |  | 
|  | 546 | percpu_counter_mod(&sbi->s_freeinodes_counter, -1); | 
|  | 547 | if (S_ISDIR(mode)) | 
|  | 548 | percpu_counter_inc(&sbi->s_dirs_counter); | 
|  | 549 |  | 
|  | 550 | spin_lock(sb_bgl_lock(sbi, group)); | 
|  | 551 | gdp->bg_free_inodes_count = | 
|  | 552 | cpu_to_le16(le16_to_cpu(gdp->bg_free_inodes_count) - 1); | 
|  | 553 | if (S_ISDIR(mode)) { | 
|  | 554 | if (sbi->s_debts[group] < 255) | 
|  | 555 | sbi->s_debts[group]++; | 
|  | 556 | gdp->bg_used_dirs_count = | 
|  | 557 | cpu_to_le16(le16_to_cpu(gdp->bg_used_dirs_count) + 1); | 
|  | 558 | } else { | 
|  | 559 | if (sbi->s_debts[group]) | 
|  | 560 | sbi->s_debts[group]--; | 
|  | 561 | } | 
|  | 562 | spin_unlock(sb_bgl_lock(sbi, group)); | 
|  | 563 |  | 
|  | 564 | sb->s_dirt = 1; | 
|  | 565 | mark_buffer_dirty(bh2); | 
|  | 566 | inode->i_uid = current->fsuid; | 
|  | 567 | if (test_opt (sb, GRPID)) | 
|  | 568 | inode->i_gid = dir->i_gid; | 
|  | 569 | else if (dir->i_mode & S_ISGID) { | 
|  | 570 | inode->i_gid = dir->i_gid; | 
|  | 571 | if (S_ISDIR(mode)) | 
|  | 572 | mode |= S_ISGID; | 
|  | 573 | } else | 
|  | 574 | inode->i_gid = current->fsgid; | 
|  | 575 | inode->i_mode = mode; | 
|  | 576 |  | 
|  | 577 | inode->i_ino = ino; | 
|  | 578 | inode->i_blksize = PAGE_SIZE;	/* This is the optimal IO size (for stat), not the fs block size */ | 
|  | 579 | inode->i_blocks = 0; | 
|  | 580 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC; | 
|  | 581 | memset(ei->i_data, 0, sizeof(ei->i_data)); | 
|  | 582 | ei->i_flags = EXT2_I(dir)->i_flags & ~EXT2_BTREE_FL; | 
|  | 583 | if (S_ISLNK(mode)) | 
|  | 584 | ei->i_flags &= ~(EXT2_IMMUTABLE_FL|EXT2_APPEND_FL); | 
|  | 585 | /* dirsync is only applied to directories */ | 
|  | 586 | if (!S_ISDIR(mode)) | 
|  | 587 | ei->i_flags &= ~EXT2_DIRSYNC_FL; | 
|  | 588 | ei->i_faddr = 0; | 
|  | 589 | ei->i_frag_no = 0; | 
|  | 590 | ei->i_frag_size = 0; | 
|  | 591 | ei->i_file_acl = 0; | 
|  | 592 | ei->i_dir_acl = 0; | 
|  | 593 | ei->i_dtime = 0; | 
|  | 594 | ei->i_block_group = group; | 
|  | 595 | ei->i_next_alloc_block = 0; | 
|  | 596 | ei->i_next_alloc_goal = 0; | 
|  | 597 | ei->i_prealloc_block = 0; | 
|  | 598 | ei->i_prealloc_count = 0; | 
|  | 599 | ei->i_dir_start_lookup = 0; | 
|  | 600 | ei->i_state = EXT2_STATE_NEW; | 
|  | 601 | ext2_set_inode_flags(inode); | 
|  | 602 | spin_lock(&sbi->s_next_gen_lock); | 
|  | 603 | inode->i_generation = sbi->s_next_generation++; | 
|  | 604 | spin_unlock(&sbi->s_next_gen_lock); | 
|  | 605 | insert_inode_hash(inode); | 
|  | 606 |  | 
|  | 607 | if (DQUOT_ALLOC_INODE(inode)) { | 
|  | 608 | DQUOT_DROP(inode); | 
|  | 609 | err = -ENOSPC; | 
|  | 610 | goto fail2; | 
|  | 611 | } | 
|  | 612 | err = ext2_init_acl(inode, dir); | 
|  | 613 | if (err) { | 
|  | 614 | DQUOT_FREE_INODE(inode); | 
|  | 615 | goto fail2; | 
|  | 616 | } | 
|  | 617 | mark_inode_dirty(inode); | 
|  | 618 | ext2_debug("allocating inode %lu\n", inode->i_ino); | 
|  | 619 | ext2_preread_inode(inode); | 
|  | 620 | return inode; | 
|  | 621 |  | 
|  | 622 | fail2: | 
|  | 623 | inode->i_flags |= S_NOQUOTA; | 
|  | 624 | inode->i_nlink = 0; | 
|  | 625 | iput(inode); | 
|  | 626 | return ERR_PTR(err); | 
|  | 627 |  | 
|  | 628 | fail: | 
|  | 629 | make_bad_inode(inode); | 
|  | 630 | iput(inode); | 
|  | 631 | return ERR_PTR(err); | 
|  | 632 | } | 
|  | 633 |  | 
|  | 634 | unsigned long ext2_count_free_inodes (struct super_block * sb) | 
|  | 635 | { | 
|  | 636 | struct ext2_group_desc *desc; | 
|  | 637 | unsigned long desc_count = 0; | 
|  | 638 | int i; | 
|  | 639 |  | 
|  | 640 | #ifdef EXT2FS_DEBUG | 
|  | 641 | struct ext2_super_block *es; | 
|  | 642 | unsigned long bitmap_count = 0; | 
|  | 643 | struct buffer_head *bitmap_bh = NULL; | 
|  | 644 |  | 
|  | 645 | lock_super (sb); | 
|  | 646 | es = EXT2_SB(sb)->s_es; | 
|  | 647 | for (i = 0; i < EXT2_SB(sb)->s_groups_count; i++) { | 
|  | 648 | unsigned x; | 
|  | 649 |  | 
|  | 650 | desc = ext2_get_group_desc (sb, i, NULL); | 
|  | 651 | if (!desc) | 
|  | 652 | continue; | 
|  | 653 | desc_count += le16_to_cpu(desc->bg_free_inodes_count); | 
|  | 654 | brelse(bitmap_bh); | 
|  | 655 | bitmap_bh = read_inode_bitmap(sb, i); | 
|  | 656 | if (!bitmap_bh) | 
|  | 657 | continue; | 
|  | 658 |  | 
|  | 659 | x = ext2_count_free(bitmap_bh, EXT2_INODES_PER_GROUP(sb) / 8); | 
|  | 660 | printk("group %d: stored = %d, counted = %u\n", | 
|  | 661 | i, le16_to_cpu(desc->bg_free_inodes_count), x); | 
|  | 662 | bitmap_count += x; | 
|  | 663 | } | 
|  | 664 | brelse(bitmap_bh); | 
|  | 665 | printk("ext2_count_free_inodes: stored = %lu, computed = %lu, %lu\n", | 
|  | 666 | percpu_counter_read(&EXT2_SB(sb)->s_freeinodes_counter), | 
|  | 667 | desc_count, bitmap_count); | 
|  | 668 | unlock_super(sb); | 
|  | 669 | return desc_count; | 
|  | 670 | #else | 
|  | 671 | for (i = 0; i < EXT2_SB(sb)->s_groups_count; i++) { | 
|  | 672 | desc = ext2_get_group_desc (sb, i, NULL); | 
|  | 673 | if (!desc) | 
|  | 674 | continue; | 
|  | 675 | desc_count += le16_to_cpu(desc->bg_free_inodes_count); | 
|  | 676 | } | 
|  | 677 | return desc_count; | 
|  | 678 | #endif | 
|  | 679 | } | 
|  | 680 |  | 
|  | 681 | /* Called at mount-time, super-block is locked */ | 
|  | 682 | unsigned long ext2_count_dirs (struct super_block * sb) | 
|  | 683 | { | 
|  | 684 | unsigned long count = 0; | 
|  | 685 | int i; | 
|  | 686 |  | 
|  | 687 | for (i = 0; i < EXT2_SB(sb)->s_groups_count; i++) { | 
|  | 688 | struct ext2_group_desc *gdp = ext2_get_group_desc (sb, i, NULL); | 
|  | 689 | if (!gdp) | 
|  | 690 | continue; | 
|  | 691 | count += le16_to_cpu(gdp->bg_used_dirs_count); | 
|  | 692 | } | 
|  | 693 | return count; | 
|  | 694 | } | 
|  | 695 |  | 
|  | 696 | #ifdef CONFIG_EXT2_CHECK | 
|  | 697 | /* Called at mount-time, super-block is locked */ | 
|  | 698 | void ext2_check_inodes_bitmap (struct super_block * sb) | 
|  | 699 | { | 
|  | 700 | struct ext2_super_block * es = EXT2_SB(sb)->s_es; | 
|  | 701 | unsigned long desc_count = 0, bitmap_count = 0; | 
|  | 702 | struct buffer_head *bitmap_bh = NULL; | 
|  | 703 | int i; | 
|  | 704 |  | 
|  | 705 | for (i = 0; i < EXT2_SB(sb)->s_groups_count; i++) { | 
|  | 706 | struct ext2_group_desc *desc; | 
|  | 707 | unsigned x; | 
|  | 708 |  | 
|  | 709 | desc = ext2_get_group_desc(sb, i, NULL); | 
|  | 710 | if (!desc) | 
|  | 711 | continue; | 
|  | 712 | desc_count += le16_to_cpu(desc->bg_free_inodes_count); | 
|  | 713 | brelse(bitmap_bh); | 
|  | 714 | bitmap_bh = read_inode_bitmap(sb, i); | 
|  | 715 | if (!bitmap_bh) | 
|  | 716 | continue; | 
|  | 717 |  | 
|  | 718 | x = ext2_count_free(bitmap_bh, EXT2_INODES_PER_GROUP(sb) / 8); | 
|  | 719 | if (le16_to_cpu(desc->bg_free_inodes_count) != x) | 
|  | 720 | ext2_error (sb, "ext2_check_inodes_bitmap", | 
|  | 721 | "Wrong free inodes count in group %d, " | 
|  | 722 | "stored = %d, counted = %lu", i, | 
|  | 723 | le16_to_cpu(desc->bg_free_inodes_count), x); | 
|  | 724 | bitmap_count += x; | 
|  | 725 | } | 
|  | 726 | brelse(bitmap_bh); | 
|  | 727 | if (percpu_counter_read(&EXT2_SB(sb)->s_freeinodes_counter) != | 
|  | 728 | bitmap_count) | 
|  | 729 | ext2_error(sb, "ext2_check_inodes_bitmap", | 
|  | 730 | "Wrong free inodes count in super block, " | 
|  | 731 | "stored = %lu, counted = %lu", | 
|  | 732 | (unsigned long)le32_to_cpu(es->s_free_inodes_count), | 
|  | 733 | bitmap_count); | 
|  | 734 | } | 
|  | 735 | #endif |