1 /* 2 * linux/fs/ext4/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@redhat.com), 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/time.h> 16 #include <linux/fs.h> 17 #include <linux/jbd2.h> 18 #include <linux/stat.h> 19 #include <linux/string.h> 20 #include <linux/quotaops.h> 21 #include <linux/buffer_head.h> 22 #include <linux/random.h> 23 #include <linux/bitops.h> 24 #include <linux/blkdev.h> 25 #include <asm/byteorder.h> 26 27 #include "ext4.h" 28 #include "ext4_jbd2.h" 29 #include "xattr.h" 30 #include "acl.h" 31 32 #include <trace/events/ext4.h> 33 34 /* 35 * ialloc.c contains the inodes allocation and deallocation routines 36 */ 37 38 /* 39 * The free inodes are managed by bitmaps. A file system contains several 40 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap 41 * block for inodes, N blocks for the inode table and data blocks. 42 * 43 * The file system contains group descriptors which are located after the 44 * super block. Each descriptor contains the number of the bitmap block and 45 * the free blocks count in the block. 46 */ 47 48 /* 49 * To avoid calling the atomic setbit hundreds or thousands of times, we only 50 * need to use it within a single byte (to ensure we get endianness right). 51 * We can use memset for the rest of the bitmap as there are no other users. 52 */ 53 void mark_bitmap_end(int start_bit, int end_bit, char *bitmap) 54 { 55 int i; 56 57 if (start_bit >= end_bit) 58 return; 59 60 ext4_debug("mark end bits +%d through +%d used\n", start_bit, end_bit); 61 for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++) 62 ext4_set_bit(i, bitmap); 63 if (i < end_bit) 64 memset(bitmap + (i >> 3), 0xff, (end_bit - i) >> 3); 65 } 66 67 /* Initializes an uninitialized inode bitmap */ 68 unsigned ext4_init_inode_bitmap(struct super_block *sb, struct buffer_head *bh, 69 ext4_group_t block_group, 70 struct ext4_group_desc *gdp) 71 { 72 struct ext4_sb_info *sbi = EXT4_SB(sb); 73 74 J_ASSERT_BH(bh, buffer_locked(bh)); 75 76 /* If checksum is bad mark all blocks and inodes use to prevent 77 * allocation, essentially implementing a per-group read-only flag. */ 78 if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) { 79 ext4_error(sb, __func__, "Checksum bad for group %u", 80 block_group); 81 ext4_free_blks_set(sb, gdp, 0); 82 ext4_free_inodes_set(sb, gdp, 0); 83 ext4_itable_unused_set(sb, gdp, 0); 84 memset(bh->b_data, 0xff, sb->s_blocksize); 85 return 0; 86 } 87 88 memset(bh->b_data, 0, (EXT4_INODES_PER_GROUP(sb) + 7) / 8); 89 mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), sb->s_blocksize * 8, 90 bh->b_data); 91 92 return EXT4_INODES_PER_GROUP(sb); 93 } 94 95 /* 96 * Read the inode allocation bitmap for a given block_group, reading 97 * into the specified slot in the superblock's bitmap cache. 98 * 99 * Return buffer_head of bitmap on success or NULL. 100 */ 101 static struct buffer_head * 102 ext4_read_inode_bitmap(struct super_block *sb, ext4_group_t block_group) 103 { 104 struct ext4_group_desc *desc; 105 struct buffer_head *bh = NULL; 106 ext4_fsblk_t bitmap_blk; 107 108 desc = ext4_get_group_desc(sb, block_group, NULL); 109 if (!desc) 110 return NULL; 111 bitmap_blk = ext4_inode_bitmap(sb, desc); 112 bh = sb_getblk(sb, bitmap_blk); 113 if (unlikely(!bh)) { 114 ext4_error(sb, __func__, 115 "Cannot read inode bitmap - " 116 "block_group = %u, inode_bitmap = %llu", 117 block_group, bitmap_blk); 118 return NULL; 119 } 120 if (bitmap_uptodate(bh)) 121 return bh; 122 123 lock_buffer(bh); 124 if (bitmap_uptodate(bh)) { 125 unlock_buffer(bh); 126 return bh; 127 } 128 ext4_lock_group(sb, block_group); 129 if (desc->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) { 130 ext4_init_inode_bitmap(sb, bh, block_group, desc); 131 set_bitmap_uptodate(bh); 132 set_buffer_uptodate(bh); 133 ext4_unlock_group(sb, block_group); 134 unlock_buffer(bh); 135 return bh; 136 } 137 ext4_unlock_group(sb, block_group); 138 if (buffer_uptodate(bh)) { 139 /* 140 * if not uninit if bh is uptodate, 141 * bitmap is also uptodate 142 */ 143 set_bitmap_uptodate(bh); 144 unlock_buffer(bh); 145 return bh; 146 } 147 /* 148 * submit the buffer_head for read. We can 149 * safely mark the bitmap as uptodate now. 150 * We do it here so the bitmap uptodate bit 151 * get set with buffer lock held. 152 */ 153 set_bitmap_uptodate(bh); 154 if (bh_submit_read(bh) < 0) { 155 put_bh(bh); 156 ext4_error(sb, __func__, 157 "Cannot read inode bitmap - " 158 "block_group = %u, inode_bitmap = %llu", 159 block_group, bitmap_blk); 160 return NULL; 161 } 162 return bh; 163 } 164 165 /* 166 * NOTE! When we get the inode, we're the only people 167 * that have access to it, and as such there are no 168 * race conditions we have to worry about. The inode 169 * is not on the hash-lists, and it cannot be reached 170 * through the filesystem because the directory entry 171 * has been deleted earlier. 172 * 173 * HOWEVER: we must make sure that we get no aliases, 174 * which means that we have to call "clear_inode()" 175 * _before_ we mark the inode not in use in the inode 176 * bitmaps. Otherwise a newly created file might use 177 * the same inode number (not actually the same pointer 178 * though), and then we'd have two inodes sharing the 179 * same inode number and space on the harddisk. 180 */ 181 void ext4_free_inode(handle_t *handle, struct inode *inode) 182 { 183 struct super_block *sb = inode->i_sb; 184 int is_directory; 185 unsigned long ino; 186 struct buffer_head *bitmap_bh = NULL; 187 struct buffer_head *bh2; 188 ext4_group_t block_group; 189 unsigned long bit; 190 struct ext4_group_desc *gdp; 191 struct ext4_super_block *es; 192 struct ext4_sb_info *sbi; 193 int fatal = 0, err, count, cleared; 194 195 if (atomic_read(&inode->i_count) > 1) { 196 printk(KERN_ERR "ext4_free_inode: inode has count=%d\n", 197 atomic_read(&inode->i_count)); 198 return; 199 } 200 if (inode->i_nlink) { 201 printk(KERN_ERR "ext4_free_inode: inode has nlink=%d\n", 202 inode->i_nlink); 203 return; 204 } 205 if (!sb) { 206 printk(KERN_ERR "ext4_free_inode: inode on " 207 "nonexistent device\n"); 208 return; 209 } 210 sbi = EXT4_SB(sb); 211 212 ino = inode->i_ino; 213 ext4_debug("freeing inode %lu\n", ino); 214 trace_ext4_free_inode(inode); 215 216 /* 217 * Note: we must free any quota before locking the superblock, 218 * as writing the quota to disk may need the lock as well. 219 */ 220 vfs_dq_init(inode); 221 ext4_xattr_delete_inode(handle, inode); 222 vfs_dq_free_inode(inode); 223 vfs_dq_drop(inode); 224 225 is_directory = S_ISDIR(inode->i_mode); 226 227 /* Do this BEFORE marking the inode not in use or returning an error */ 228 clear_inode(inode); 229 230 es = EXT4_SB(sb)->s_es; 231 if (ino < EXT4_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) { 232 ext4_error(sb, "ext4_free_inode", 233 "reserved or nonexistent inode %lu", ino); 234 goto error_return; 235 } 236 block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb); 237 bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb); 238 bitmap_bh = ext4_read_inode_bitmap(sb, block_group); 239 if (!bitmap_bh) 240 goto error_return; 241 242 BUFFER_TRACE(bitmap_bh, "get_write_access"); 243 fatal = ext4_journal_get_write_access(handle, bitmap_bh); 244 if (fatal) 245 goto error_return; 246 247 /* Ok, now we can actually update the inode bitmaps.. */ 248 cleared = ext4_clear_bit_atomic(ext4_group_lock_ptr(sb, block_group), 249 bit, bitmap_bh->b_data); 250 if (!cleared) 251 ext4_error(sb, "ext4_free_inode", 252 "bit already cleared for inode %lu", ino); 253 else { 254 gdp = ext4_get_group_desc(sb, block_group, &bh2); 255 256 BUFFER_TRACE(bh2, "get_write_access"); 257 fatal = ext4_journal_get_write_access(handle, bh2); 258 if (fatal) goto error_return; 259 260 if (gdp) { 261 ext4_lock_group(sb, block_group); 262 count = ext4_free_inodes_count(sb, gdp) + 1; 263 ext4_free_inodes_set(sb, gdp, count); 264 if (is_directory) { 265 count = ext4_used_dirs_count(sb, gdp) - 1; 266 ext4_used_dirs_set(sb, gdp, count); 267 if (sbi->s_log_groups_per_flex) { 268 ext4_group_t f; 269 270 f = ext4_flex_group(sbi, block_group); 271 atomic_dec(&sbi->s_flex_groups[f].free_inodes); 272 } 273 274 } 275 gdp->bg_checksum = ext4_group_desc_csum(sbi, 276 block_group, gdp); 277 ext4_unlock_group(sb, block_group); 278 percpu_counter_inc(&sbi->s_freeinodes_counter); 279 if (is_directory) 280 percpu_counter_dec(&sbi->s_dirs_counter); 281 282 if (sbi->s_log_groups_per_flex) { 283 ext4_group_t f; 284 285 f = ext4_flex_group(sbi, block_group); 286 atomic_inc(&sbi->s_flex_groups[f].free_inodes); 287 } 288 } 289 BUFFER_TRACE(bh2, "call ext4_handle_dirty_metadata"); 290 err = ext4_handle_dirty_metadata(handle, NULL, bh2); 291 if (!fatal) fatal = err; 292 } 293 BUFFER_TRACE(bitmap_bh, "call ext4_handle_dirty_metadata"); 294 err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh); 295 if (!fatal) 296 fatal = err; 297 sb->s_dirt = 1; 298 error_return: 299 brelse(bitmap_bh); 300 ext4_std_error(sb, fatal); 301 } 302 303 /* 304 * There are two policies for allocating an inode. If the new inode is 305 * a directory, then a forward search is made for a block group with both 306 * free space and a low directory-to-inode ratio; if that fails, then of 307 * the groups with above-average free space, that group with the fewest 308 * directories already is chosen. 309 * 310 * For other inodes, search forward from the parent directory\'s block 311 * group to find a free inode. 312 */ 313 static int find_group_dir(struct super_block *sb, struct inode *parent, 314 ext4_group_t *best_group) 315 { 316 ext4_group_t ngroups = ext4_get_groups_count(sb); 317 unsigned int freei, avefreei; 318 struct ext4_group_desc *desc, *best_desc = NULL; 319 ext4_group_t group; 320 int ret = -1; 321 322 freei = percpu_counter_read_positive(&EXT4_SB(sb)->s_freeinodes_counter); 323 avefreei = freei / ngroups; 324 325 for (group = 0; group < ngroups; group++) { 326 desc = ext4_get_group_desc(sb, group, NULL); 327 if (!desc || !ext4_free_inodes_count(sb, desc)) 328 continue; 329 if (ext4_free_inodes_count(sb, desc) < avefreei) 330 continue; 331 if (!best_desc || 332 (ext4_free_blks_count(sb, desc) > 333 ext4_free_blks_count(sb, best_desc))) { 334 *best_group = group; 335 best_desc = desc; 336 ret = 0; 337 } 338 } 339 return ret; 340 } 341 342 #define free_block_ratio 10 343 344 static int find_group_flex(struct super_block *sb, struct inode *parent, 345 ext4_group_t *best_group) 346 { 347 struct ext4_sb_info *sbi = EXT4_SB(sb); 348 struct ext4_group_desc *desc; 349 struct flex_groups *flex_group = sbi->s_flex_groups; 350 ext4_group_t parent_group = EXT4_I(parent)->i_block_group; 351 ext4_group_t parent_fbg_group = ext4_flex_group(sbi, parent_group); 352 ext4_group_t ngroups = ext4_get_groups_count(sb); 353 int flex_size = ext4_flex_bg_size(sbi); 354 ext4_group_t best_flex = parent_fbg_group; 355 int blocks_per_flex = sbi->s_blocks_per_group * flex_size; 356 int flexbg_free_blocks; 357 int flex_freeb_ratio; 358 ext4_group_t n_fbg_groups; 359 ext4_group_t i; 360 361 n_fbg_groups = (ngroups + flex_size - 1) >> 362 sbi->s_log_groups_per_flex; 363 364 find_close_to_parent: 365 flexbg_free_blocks = atomic_read(&flex_group[best_flex].free_blocks); 366 flex_freeb_ratio = flexbg_free_blocks * 100 / blocks_per_flex; 367 if (atomic_read(&flex_group[best_flex].free_inodes) && 368 flex_freeb_ratio > free_block_ratio) 369 goto found_flexbg; 370 371 if (best_flex && best_flex == parent_fbg_group) { 372 best_flex--; 373 goto find_close_to_parent; 374 } 375 376 for (i = 0; i < n_fbg_groups; i++) { 377 if (i == parent_fbg_group || i == parent_fbg_group - 1) 378 continue; 379 380 flexbg_free_blocks = atomic_read(&flex_group[i].free_blocks); 381 flex_freeb_ratio = flexbg_free_blocks * 100 / blocks_per_flex; 382 383 if (flex_freeb_ratio > free_block_ratio && 384 (atomic_read(&flex_group[i].free_inodes))) { 385 best_flex = i; 386 goto found_flexbg; 387 } 388 389 if ((atomic_read(&flex_group[best_flex].free_inodes) == 0) || 390 ((atomic_read(&flex_group[i].free_blocks) > 391 atomic_read(&flex_group[best_flex].free_blocks)) && 392 atomic_read(&flex_group[i].free_inodes))) 393 best_flex = i; 394 } 395 396 if (!atomic_read(&flex_group[best_flex].free_inodes) || 397 !atomic_read(&flex_group[best_flex].free_blocks)) 398 return -1; 399 400 found_flexbg: 401 for (i = best_flex * flex_size; i < ngroups && 402 i < (best_flex + 1) * flex_size; i++) { 403 desc = ext4_get_group_desc(sb, i, NULL); 404 if (ext4_free_inodes_count(sb, desc)) { 405 *best_group = i; 406 goto out; 407 } 408 } 409 410 return -1; 411 out: 412 return 0; 413 } 414 415 struct orlov_stats { 416 __u32 free_inodes; 417 __u32 free_blocks; 418 __u32 used_dirs; 419 }; 420 421 /* 422 * Helper function for Orlov's allocator; returns critical information 423 * for a particular block group or flex_bg. If flex_size is 1, then g 424 * is a block group number; otherwise it is flex_bg number. 425 */ 426 void get_orlov_stats(struct super_block *sb, ext4_group_t g, 427 int flex_size, struct orlov_stats *stats) 428 { 429 struct ext4_group_desc *desc; 430 struct flex_groups *flex_group = EXT4_SB(sb)->s_flex_groups; 431 432 if (flex_size > 1) { 433 stats->free_inodes = atomic_read(&flex_group[g].free_inodes); 434 stats->free_blocks = atomic_read(&flex_group[g].free_blocks); 435 stats->used_dirs = atomic_read(&flex_group[g].used_dirs); 436 return; 437 } 438 439 desc = ext4_get_group_desc(sb, g, NULL); 440 if (desc) { 441 stats->free_inodes = ext4_free_inodes_count(sb, desc); 442 stats->free_blocks = ext4_free_blks_count(sb, desc); 443 stats->used_dirs = ext4_used_dirs_count(sb, desc); 444 } else { 445 stats->free_inodes = 0; 446 stats->free_blocks = 0; 447 stats->used_dirs = 0; 448 } 449 } 450 451 /* 452 * Orlov's allocator for directories. 453 * 454 * We always try to spread first-level directories. 455 * 456 * If there are blockgroups with both free inodes and free blocks counts 457 * not worse than average we return one with smallest directory count. 458 * Otherwise we simply return a random group. 459 * 460 * For the rest rules look so: 461 * 462 * It's OK to put directory into a group unless 463 * it has too many directories already (max_dirs) or 464 * it has too few free inodes left (min_inodes) or 465 * it has too few free blocks left (min_blocks) or 466 * Parent's group is preferred, if it doesn't satisfy these 467 * conditions we search cyclically through the rest. If none 468 * of the groups look good we just look for a group with more 469 * free inodes than average (starting at parent's group). 470 */ 471 472 static int find_group_orlov(struct super_block *sb, struct inode *parent, 473 ext4_group_t *group, int mode, 474 const struct qstr *qstr) 475 { 476 ext4_group_t parent_group = EXT4_I(parent)->i_block_group; 477 struct ext4_sb_info *sbi = EXT4_SB(sb); 478 ext4_group_t real_ngroups = ext4_get_groups_count(sb); 479 int inodes_per_group = EXT4_INODES_PER_GROUP(sb); 480 unsigned int freei, avefreei; 481 ext4_fsblk_t freeb, avefreeb; 482 unsigned int ndirs; 483 int max_dirs, min_inodes; 484 ext4_grpblk_t min_blocks; 485 ext4_group_t i, grp, g, ngroups; 486 struct ext4_group_desc *desc; 487 struct orlov_stats stats; 488 int flex_size = ext4_flex_bg_size(sbi); 489 struct dx_hash_info hinfo; 490 491 ngroups = real_ngroups; 492 if (flex_size > 1) { 493 ngroups = (real_ngroups + flex_size - 1) >> 494 sbi->s_log_groups_per_flex; 495 parent_group >>= sbi->s_log_groups_per_flex; 496 } 497 498 freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter); 499 avefreei = freei / ngroups; 500 freeb = percpu_counter_read_positive(&sbi->s_freeblocks_counter); 501 avefreeb = freeb; 502 do_div(avefreeb, ngroups); 503 ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter); 504 505 if (S_ISDIR(mode) && 506 ((parent == sb->s_root->d_inode) || 507 (EXT4_I(parent)->i_flags & EXT4_TOPDIR_FL))) { 508 int best_ndir = inodes_per_group; 509 int ret = -1; 510 511 if (qstr) { 512 hinfo.hash_version = DX_HASH_HALF_MD4; 513 hinfo.seed = sbi->s_hash_seed; 514 ext4fs_dirhash(qstr->name, qstr->len, &hinfo); 515 grp = hinfo.hash; 516 } else 517 get_random_bytes(&grp, sizeof(grp)); 518 parent_group = (unsigned)grp % ngroups; 519 for (i = 0; i < ngroups; i++) { 520 g = (parent_group + i) % ngroups; 521 get_orlov_stats(sb, g, flex_size, &stats); 522 if (!stats.free_inodes) 523 continue; 524 if (stats.used_dirs >= best_ndir) 525 continue; 526 if (stats.free_inodes < avefreei) 527 continue; 528 if (stats.free_blocks < avefreeb) 529 continue; 530 grp = g; 531 ret = 0; 532 best_ndir = stats.used_dirs; 533 } 534 if (ret) 535 goto fallback; 536 found_flex_bg: 537 if (flex_size == 1) { 538 *group = grp; 539 return 0; 540 } 541 542 /* 543 * We pack inodes at the beginning of the flexgroup's 544 * inode tables. Block allocation decisions will do 545 * something similar, although regular files will 546 * start at 2nd block group of the flexgroup. See 547 * ext4_ext_find_goal() and ext4_find_near(). 548 */ 549 grp *= flex_size; 550 for (i = 0; i < flex_size; i++) { 551 if (grp+i >= real_ngroups) 552 break; 553 desc = ext4_get_group_desc(sb, grp+i, NULL); 554 if (desc && ext4_free_inodes_count(sb, desc)) { 555 *group = grp+i; 556 return 0; 557 } 558 } 559 goto fallback; 560 } 561 562 max_dirs = ndirs / ngroups + inodes_per_group / 16; 563 min_inodes = avefreei - inodes_per_group*flex_size / 4; 564 if (min_inodes < 1) 565 min_inodes = 1; 566 min_blocks = avefreeb - EXT4_BLOCKS_PER_GROUP(sb)*flex_size / 4; 567 568 /* 569 * Start looking in the flex group where we last allocated an 570 * inode for this parent directory 571 */ 572 if (EXT4_I(parent)->i_last_alloc_group != ~0) { 573 parent_group = EXT4_I(parent)->i_last_alloc_group; 574 if (flex_size > 1) 575 parent_group >>= sbi->s_log_groups_per_flex; 576 } 577 578 for (i = 0; i < ngroups; i++) { 579 grp = (parent_group + i) % ngroups; 580 get_orlov_stats(sb, grp, flex_size, &stats); 581 if (stats.used_dirs >= max_dirs) 582 continue; 583 if (stats.free_inodes < min_inodes) 584 continue; 585 if (stats.free_blocks < min_blocks) 586 continue; 587 goto found_flex_bg; 588 } 589 590 fallback: 591 ngroups = real_ngroups; 592 avefreei = freei / ngroups; 593 fallback_retry: 594 parent_group = EXT4_I(parent)->i_block_group; 595 for (i = 0; i < ngroups; i++) { 596 grp = (parent_group + i) % ngroups; 597 desc = ext4_get_group_desc(sb, grp, NULL); 598 if (desc && ext4_free_inodes_count(sb, desc) && 599 ext4_free_inodes_count(sb, desc) >= avefreei) { 600 *group = grp; 601 return 0; 602 } 603 } 604 605 if (avefreei) { 606 /* 607 * The free-inodes counter is approximate, and for really small 608 * filesystems the above test can fail to find any blockgroups 609 */ 610 avefreei = 0; 611 goto fallback_retry; 612 } 613 614 return -1; 615 } 616 617 static int find_group_other(struct super_block *sb, struct inode *parent, 618 ext4_group_t *group, int mode) 619 { 620 ext4_group_t parent_group = EXT4_I(parent)->i_block_group; 621 ext4_group_t i, last, ngroups = ext4_get_groups_count(sb); 622 struct ext4_group_desc *desc; 623 int flex_size = ext4_flex_bg_size(EXT4_SB(sb)); 624 625 /* 626 * Try to place the inode is the same flex group as its 627 * parent. If we can't find space, use the Orlov algorithm to 628 * find another flex group, and store that information in the 629 * parent directory's inode information so that use that flex 630 * group for future allocations. 631 */ 632 if (flex_size > 1) { 633 int retry = 0; 634 635 try_again: 636 parent_group &= ~(flex_size-1); 637 last = parent_group + flex_size; 638 if (last > ngroups) 639 last = ngroups; 640 for (i = parent_group; i < last; i++) { 641 desc = ext4_get_group_desc(sb, i, NULL); 642 if (desc && ext4_free_inodes_count(sb, desc)) { 643 *group = i; 644 return 0; 645 } 646 } 647 if (!retry && EXT4_I(parent)->i_last_alloc_group != ~0) { 648 retry = 1; 649 parent_group = EXT4_I(parent)->i_last_alloc_group; 650 goto try_again; 651 } 652 /* 653 * If this didn't work, use the Orlov search algorithm 654 * to find a new flex group; we pass in the mode to 655 * avoid the topdir algorithms. 656 */ 657 *group = parent_group + flex_size; 658 if (*group > ngroups) 659 *group = 0; 660 return find_group_orlov(sb, parent, group, mode, 0); 661 } 662 663 /* 664 * Try to place the inode in its parent directory 665 */ 666 *group = parent_group; 667 desc = ext4_get_group_desc(sb, *group, NULL); 668 if (desc && ext4_free_inodes_count(sb, desc) && 669 ext4_free_blks_count(sb, desc)) 670 return 0; 671 672 /* 673 * We're going to place this inode in a different blockgroup from its 674 * parent. We want to cause files in a common directory to all land in 675 * the same blockgroup. But we want files which are in a different 676 * directory which shares a blockgroup with our parent to land in a 677 * different blockgroup. 678 * 679 * So add our directory's i_ino into the starting point for the hash. 680 */ 681 *group = (*group + parent->i_ino) % ngroups; 682 683 /* 684 * Use a quadratic hash to find a group with a free inode and some free 685 * blocks. 686 */ 687 for (i = 1; i < ngroups; i <<= 1) { 688 *group += i; 689 if (*group >= ngroups) 690 *group -= ngroups; 691 desc = ext4_get_group_desc(sb, *group, NULL); 692 if (desc && ext4_free_inodes_count(sb, desc) && 693 ext4_free_blks_count(sb, desc)) 694 return 0; 695 } 696 697 /* 698 * That failed: try linear search for a free inode, even if that group 699 * has no free blocks. 700 */ 701 *group = parent_group; 702 for (i = 0; i < ngroups; i++) { 703 if (++*group >= ngroups) 704 *group = 0; 705 desc = ext4_get_group_desc(sb, *group, NULL); 706 if (desc && ext4_free_inodes_count(sb, desc)) 707 return 0; 708 } 709 710 return -1; 711 } 712 713 /* 714 * claim the inode from the inode bitmap. If the group 715 * is uninit we need to take the groups's ext4_group_lock 716 * and clear the uninit flag. The inode bitmap update 717 * and group desc uninit flag clear should be done 718 * after holding ext4_group_lock so that ext4_read_inode_bitmap 719 * doesn't race with the ext4_claim_inode 720 */ 721 static int ext4_claim_inode(struct super_block *sb, 722 struct buffer_head *inode_bitmap_bh, 723 unsigned long ino, ext4_group_t group, int mode) 724 { 725 int free = 0, retval = 0, count; 726 struct ext4_sb_info *sbi = EXT4_SB(sb); 727 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 728 729 ext4_lock_group(sb, group); 730 if (ext4_set_bit(ino, inode_bitmap_bh->b_data)) { 731 /* not a free inode */ 732 retval = 1; 733 goto err_ret; 734 } 735 ino++; 736 if ((group == 0 && ino < EXT4_FIRST_INO(sb)) || 737 ino > EXT4_INODES_PER_GROUP(sb)) { 738 ext4_unlock_group(sb, group); 739 ext4_error(sb, __func__, 740 "reserved inode or inode > inodes count - " 741 "block_group = %u, inode=%lu", group, 742 ino + group * EXT4_INODES_PER_GROUP(sb)); 743 return 1; 744 } 745 /* If we didn't allocate from within the initialized part of the inode 746 * table then we need to initialize up to this inode. */ 747 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) { 748 749 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) { 750 gdp->bg_flags &= cpu_to_le16(~EXT4_BG_INODE_UNINIT); 751 /* When marking the block group with 752 * ~EXT4_BG_INODE_UNINIT we don't want to depend 753 * on the value of bg_itable_unused even though 754 * mke2fs could have initialized the same for us. 755 * Instead we calculated the value below 756 */ 757 758 free = 0; 759 } else { 760 free = EXT4_INODES_PER_GROUP(sb) - 761 ext4_itable_unused_count(sb, gdp); 762 } 763 764 /* 765 * Check the relative inode number against the last used 766 * relative inode number in this group. if it is greater 767 * we need to update the bg_itable_unused count 768 * 769 */ 770 if (ino > free) 771 ext4_itable_unused_set(sb, gdp, 772 (EXT4_INODES_PER_GROUP(sb) - ino)); 773 } 774 count = ext4_free_inodes_count(sb, gdp) - 1; 775 ext4_free_inodes_set(sb, gdp, count); 776 if (S_ISDIR(mode)) { 777 count = ext4_used_dirs_count(sb, gdp) + 1; 778 ext4_used_dirs_set(sb, gdp, count); 779 if (sbi->s_log_groups_per_flex) { 780 ext4_group_t f = ext4_flex_group(sbi, group); 781 782 atomic_inc(&sbi->s_flex_groups[f].free_inodes); 783 } 784 } 785 gdp->bg_checksum = ext4_group_desc_csum(sbi, group, gdp); 786 err_ret: 787 ext4_unlock_group(sb, group); 788 return retval; 789 } 790 791 /* 792 * There are two policies for allocating an inode. If the new inode is 793 * a directory, then a forward search is made for a block group with both 794 * free space and a low directory-to-inode ratio; if that fails, then of 795 * the groups with above-average free space, that group with the fewest 796 * directories already is chosen. 797 * 798 * For other inodes, search forward from the parent directory's block 799 * group to find a free inode. 800 */ 801 struct inode *ext4_new_inode(handle_t *handle, struct inode *dir, int mode, 802 const struct qstr *qstr, __u32 goal) 803 { 804 struct super_block *sb; 805 struct buffer_head *inode_bitmap_bh = NULL; 806 struct buffer_head *group_desc_bh; 807 ext4_group_t ngroups, group = 0; 808 unsigned long ino = 0; 809 struct inode *inode; 810 struct ext4_group_desc *gdp = NULL; 811 struct ext4_inode_info *ei; 812 struct ext4_sb_info *sbi; 813 int ret2, err = 0; 814 struct inode *ret; 815 ext4_group_t i; 816 int free = 0; 817 static int once = 1; 818 ext4_group_t flex_group; 819 820 /* Cannot create files in a deleted directory */ 821 if (!dir || !dir->i_nlink) 822 return ERR_PTR(-EPERM); 823 824 sb = dir->i_sb; 825 ngroups = ext4_get_groups_count(sb); 826 trace_ext4_request_inode(dir, mode); 827 inode = new_inode(sb); 828 if (!inode) 829 return ERR_PTR(-ENOMEM); 830 ei = EXT4_I(inode); 831 sbi = EXT4_SB(sb); 832 833 if (!goal) 834 goal = sbi->s_inode_goal; 835 836 if (goal && goal <= le32_to_cpu(sbi->s_es->s_inodes_count)) { 837 group = (goal - 1) / EXT4_INODES_PER_GROUP(sb); 838 ino = (goal - 1) % EXT4_INODES_PER_GROUP(sb); 839 ret2 = 0; 840 goto got_group; 841 } 842 843 if (sbi->s_log_groups_per_flex && test_opt(sb, OLDALLOC)) { 844 ret2 = find_group_flex(sb, dir, &group); 845 if (ret2 == -1) { 846 ret2 = find_group_other(sb, dir, &group, mode); 847 if (ret2 == 0 && once) { 848 once = 0; 849 printk(KERN_NOTICE "ext4: find_group_flex " 850 "failed, fallback succeeded dir %lu\n", 851 dir->i_ino); 852 } 853 } 854 goto got_group; 855 } 856 857 if (S_ISDIR(mode)) { 858 if (test_opt(sb, OLDALLOC)) 859 ret2 = find_group_dir(sb, dir, &group); 860 else 861 ret2 = find_group_orlov(sb, dir, &group, mode, qstr); 862 } else 863 ret2 = find_group_other(sb, dir, &group, mode); 864 865 got_group: 866 EXT4_I(dir)->i_last_alloc_group = group; 867 err = -ENOSPC; 868 if (ret2 == -1) 869 goto out; 870 871 for (i = 0; i < ngroups; i++, ino = 0) { 872 err = -EIO; 873 874 gdp = ext4_get_group_desc(sb, group, &group_desc_bh); 875 if (!gdp) 876 goto fail; 877 878 brelse(inode_bitmap_bh); 879 inode_bitmap_bh = ext4_read_inode_bitmap(sb, group); 880 if (!inode_bitmap_bh) 881 goto fail; 882 883 repeat_in_this_group: 884 ino = ext4_find_next_zero_bit((unsigned long *) 885 inode_bitmap_bh->b_data, 886 EXT4_INODES_PER_GROUP(sb), ino); 887 888 if (ino < EXT4_INODES_PER_GROUP(sb)) { 889 890 BUFFER_TRACE(inode_bitmap_bh, "get_write_access"); 891 err = ext4_journal_get_write_access(handle, 892 inode_bitmap_bh); 893 if (err) 894 goto fail; 895 896 BUFFER_TRACE(group_desc_bh, "get_write_access"); 897 err = ext4_journal_get_write_access(handle, 898 group_desc_bh); 899 if (err) 900 goto fail; 901 if (!ext4_claim_inode(sb, inode_bitmap_bh, 902 ino, group, mode)) { 903 /* we won it */ 904 BUFFER_TRACE(inode_bitmap_bh, 905 "call ext4_handle_dirty_metadata"); 906 err = ext4_handle_dirty_metadata(handle, 907 inode, 908 inode_bitmap_bh); 909 if (err) 910 goto fail; 911 /* zero bit is inode number 1*/ 912 ino++; 913 goto got; 914 } 915 /* we lost it */ 916 ext4_handle_release_buffer(handle, inode_bitmap_bh); 917 ext4_handle_release_buffer(handle, group_desc_bh); 918 919 if (++ino < EXT4_INODES_PER_GROUP(sb)) 920 goto repeat_in_this_group; 921 } 922 923 /* 924 * This case is possible in concurrent environment. It is very 925 * rare. We cannot repeat the find_group_xxx() call because 926 * that will simply return the same blockgroup, because the 927 * group descriptor metadata has not yet been updated. 928 * So we just go onto the next blockgroup. 929 */ 930 if (++group == ngroups) 931 group = 0; 932 } 933 err = -ENOSPC; 934 goto out; 935 936 got: 937 /* We may have to initialize the block bitmap if it isn't already */ 938 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM) && 939 gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { 940 struct buffer_head *block_bitmap_bh; 941 942 block_bitmap_bh = ext4_read_block_bitmap(sb, group); 943 BUFFER_TRACE(block_bitmap_bh, "get block bitmap access"); 944 err = ext4_journal_get_write_access(handle, block_bitmap_bh); 945 if (err) { 946 brelse(block_bitmap_bh); 947 goto fail; 948 } 949 950 free = 0; 951 ext4_lock_group(sb, group); 952 /* recheck and clear flag under lock if we still need to */ 953 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { 954 free = ext4_free_blocks_after_init(sb, group, gdp); 955 gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT); 956 ext4_free_blks_set(sb, gdp, free); 957 gdp->bg_checksum = ext4_group_desc_csum(sbi, group, 958 gdp); 959 } 960 ext4_unlock_group(sb, group); 961 962 /* Don't need to dirty bitmap block if we didn't change it */ 963 if (free) { 964 BUFFER_TRACE(block_bitmap_bh, "dirty block bitmap"); 965 err = ext4_handle_dirty_metadata(handle, 966 NULL, block_bitmap_bh); 967 } 968 969 brelse(block_bitmap_bh); 970 if (err) 971 goto fail; 972 } 973 BUFFER_TRACE(group_desc_bh, "call ext4_handle_dirty_metadata"); 974 err = ext4_handle_dirty_metadata(handle, NULL, group_desc_bh); 975 if (err) 976 goto fail; 977 978 percpu_counter_dec(&sbi->s_freeinodes_counter); 979 if (S_ISDIR(mode)) 980 percpu_counter_inc(&sbi->s_dirs_counter); 981 sb->s_dirt = 1; 982 983 if (sbi->s_log_groups_per_flex) { 984 flex_group = ext4_flex_group(sbi, group); 985 atomic_dec(&sbi->s_flex_groups[flex_group].free_inodes); 986 } 987 988 inode->i_uid = current_fsuid(); 989 if (test_opt(sb, GRPID)) 990 inode->i_gid = dir->i_gid; 991 else if (dir->i_mode & S_ISGID) { 992 inode->i_gid = dir->i_gid; 993 if (S_ISDIR(mode)) 994 mode |= S_ISGID; 995 } else 996 inode->i_gid = current_fsgid(); 997 inode->i_mode = mode; 998 999 inode->i_ino = ino + group * EXT4_INODES_PER_GROUP(sb); 1000 /* This is the optimal IO size (for stat), not the fs block size */ 1001 inode->i_blocks = 0; 1002 inode->i_mtime = inode->i_atime = inode->i_ctime = ei->i_crtime = 1003 ext4_current_time(inode); 1004 1005 memset(ei->i_data, 0, sizeof(ei->i_data)); 1006 ei->i_dir_start_lookup = 0; 1007 ei->i_disksize = 0; 1008 1009 /* 1010 * Don't inherit extent flag from directory, amongst others. We set 1011 * extent flag on newly created directory and file only if -o extent 1012 * mount option is specified 1013 */ 1014 ei->i_flags = 1015 ext4_mask_flags(mode, EXT4_I(dir)->i_flags & EXT4_FL_INHERITED); 1016 ei->i_file_acl = 0; 1017 ei->i_dtime = 0; 1018 ei->i_block_group = group; 1019 ei->i_last_alloc_group = ~0; 1020 1021 ext4_set_inode_flags(inode); 1022 if (IS_DIRSYNC(inode)) 1023 ext4_handle_sync(handle); 1024 if (insert_inode_locked(inode) < 0) { 1025 err = -EINVAL; 1026 goto fail_drop; 1027 } 1028 spin_lock(&sbi->s_next_gen_lock); 1029 inode->i_generation = sbi->s_next_generation++; 1030 spin_unlock(&sbi->s_next_gen_lock); 1031 1032 ei->i_state = EXT4_STATE_NEW; 1033 1034 ei->i_extra_isize = EXT4_SB(sb)->s_want_extra_isize; 1035 1036 ret = inode; 1037 if (vfs_dq_alloc_inode(inode)) { 1038 err = -EDQUOT; 1039 goto fail_drop; 1040 } 1041 1042 err = ext4_init_acl(handle, inode, dir); 1043 if (err) 1044 goto fail_free_drop; 1045 1046 err = ext4_init_security(handle, inode, dir); 1047 if (err) 1048 goto fail_free_drop; 1049 1050 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) { 1051 /* set extent flag only for directory, file and normal symlink*/ 1052 if (S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode)) { 1053 EXT4_I(inode)->i_flags |= EXT4_EXTENTS_FL; 1054 ext4_ext_tree_init(handle, inode); 1055 } 1056 } 1057 1058 err = ext4_mark_inode_dirty(handle, inode); 1059 if (err) { 1060 ext4_std_error(sb, err); 1061 goto fail_free_drop; 1062 } 1063 1064 ext4_debug("allocating inode %lu\n", inode->i_ino); 1065 trace_ext4_allocate_inode(inode, dir, mode); 1066 goto really_out; 1067 fail: 1068 ext4_std_error(sb, err); 1069 out: 1070 iput(inode); 1071 ret = ERR_PTR(err); 1072 really_out: 1073 brelse(inode_bitmap_bh); 1074 return ret; 1075 1076 fail_free_drop: 1077 vfs_dq_free_inode(inode); 1078 1079 fail_drop: 1080 vfs_dq_drop(inode); 1081 inode->i_flags |= S_NOQUOTA; 1082 inode->i_nlink = 0; 1083 unlock_new_inode(inode); 1084 iput(inode); 1085 brelse(inode_bitmap_bh); 1086 return ERR_PTR(err); 1087 } 1088 1089 /* Verify that we are loading a valid orphan from disk */ 1090 struct inode *ext4_orphan_get(struct super_block *sb, unsigned long ino) 1091 { 1092 unsigned long max_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count); 1093 ext4_group_t block_group; 1094 int bit; 1095 struct buffer_head *bitmap_bh; 1096 struct inode *inode = NULL; 1097 long err = -EIO; 1098 1099 /* Error cases - e2fsck has already cleaned up for us */ 1100 if (ino > max_ino) { 1101 ext4_warning(sb, __func__, 1102 "bad orphan ino %lu! e2fsck was run?", ino); 1103 goto error; 1104 } 1105 1106 block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb); 1107 bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb); 1108 bitmap_bh = ext4_read_inode_bitmap(sb, block_group); 1109 if (!bitmap_bh) { 1110 ext4_warning(sb, __func__, 1111 "inode bitmap error for orphan %lu", ino); 1112 goto error; 1113 } 1114 1115 /* Having the inode bit set should be a 100% indicator that this 1116 * is a valid orphan (no e2fsck run on fs). Orphans also include 1117 * inodes that were being truncated, so we can't check i_nlink==0. 1118 */ 1119 if (!ext4_test_bit(bit, bitmap_bh->b_data)) 1120 goto bad_orphan; 1121 1122 inode = ext4_iget(sb, ino); 1123 if (IS_ERR(inode)) 1124 goto iget_failed; 1125 1126 /* 1127 * If the orphans has i_nlinks > 0 then it should be able to be 1128 * truncated, otherwise it won't be removed from the orphan list 1129 * during processing and an infinite loop will result. 1130 */ 1131 if (inode->i_nlink && !ext4_can_truncate(inode)) 1132 goto bad_orphan; 1133 1134 if (NEXT_ORPHAN(inode) > max_ino) 1135 goto bad_orphan; 1136 brelse(bitmap_bh); 1137 return inode; 1138 1139 iget_failed: 1140 err = PTR_ERR(inode); 1141 inode = NULL; 1142 bad_orphan: 1143 ext4_warning(sb, __func__, 1144 "bad orphan inode %lu! e2fsck was run?", ino); 1145 printk(KERN_NOTICE "ext4_test_bit(bit=%d, block=%llu) = %d\n", 1146 bit, (unsigned long long)bitmap_bh->b_blocknr, 1147 ext4_test_bit(bit, bitmap_bh->b_data)); 1148 printk(KERN_NOTICE "inode=%p\n", inode); 1149 if (inode) { 1150 printk(KERN_NOTICE "is_bad_inode(inode)=%d\n", 1151 is_bad_inode(inode)); 1152 printk(KERN_NOTICE "NEXT_ORPHAN(inode)=%u\n", 1153 NEXT_ORPHAN(inode)); 1154 printk(KERN_NOTICE "max_ino=%lu\n", max_ino); 1155 printk(KERN_NOTICE "i_nlink=%u\n", inode->i_nlink); 1156 /* Avoid freeing blocks if we got a bad deleted inode */ 1157 if (inode->i_nlink == 0) 1158 inode->i_blocks = 0; 1159 iput(inode); 1160 } 1161 brelse(bitmap_bh); 1162 error: 1163 return ERR_PTR(err); 1164 } 1165 1166 unsigned long ext4_count_free_inodes(struct super_block *sb) 1167 { 1168 unsigned long desc_count; 1169 struct ext4_group_desc *gdp; 1170 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 1171 #ifdef EXT4FS_DEBUG 1172 struct ext4_super_block *es; 1173 unsigned long bitmap_count, x; 1174 struct buffer_head *bitmap_bh = NULL; 1175 1176 es = EXT4_SB(sb)->s_es; 1177 desc_count = 0; 1178 bitmap_count = 0; 1179 gdp = NULL; 1180 for (i = 0; i < ngroups; i++) { 1181 gdp = ext4_get_group_desc(sb, i, NULL); 1182 if (!gdp) 1183 continue; 1184 desc_count += ext4_free_inodes_count(sb, gdp); 1185 brelse(bitmap_bh); 1186 bitmap_bh = ext4_read_inode_bitmap(sb, i); 1187 if (!bitmap_bh) 1188 continue; 1189 1190 x = ext4_count_free(bitmap_bh, EXT4_INODES_PER_GROUP(sb) / 8); 1191 printk(KERN_DEBUG "group %lu: stored = %d, counted = %lu\n", 1192 (unsigned long) i, ext4_free_inodes_count(sb, gdp), x); 1193 bitmap_count += x; 1194 } 1195 brelse(bitmap_bh); 1196 printk(KERN_DEBUG "ext4_count_free_inodes: " 1197 "stored = %u, computed = %lu, %lu\n", 1198 le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count); 1199 return desc_count; 1200 #else 1201 desc_count = 0; 1202 for (i = 0; i < ngroups; i++) { 1203 gdp = ext4_get_group_desc(sb, i, NULL); 1204 if (!gdp) 1205 continue; 1206 desc_count += ext4_free_inodes_count(sb, gdp); 1207 cond_resched(); 1208 } 1209 return desc_count; 1210 #endif 1211 } 1212 1213 /* Called at mount-time, super-block is locked */ 1214 unsigned long ext4_count_dirs(struct super_block * sb) 1215 { 1216 unsigned long count = 0; 1217 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 1218 1219 for (i = 0; i < ngroups; i++) { 1220 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 1221 if (!gdp) 1222 continue; 1223 count += ext4_used_dirs_count(sb, gdp); 1224 } 1225 return count; 1226 } 1227