libxfs/xfs_ialloc_btree.c

   1 /*
   2  * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
   3  * All Rights Reserved.
   4  *
   5  * This program is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU General Public License as
   7  * published by the Free Software Foundation.
   8  *
   9  * This program is distributed in the hope that it would be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12  * GNU General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU General Public License
  15  * along with this program; if not, write the Free Software Foundation,
  16  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  17  */
  18 #include "libxfs_priv.h"
  19 #include "xfs_fs.h"
  20 #include "xfs_shared.h"
  21 #include "xfs_format.h"
  22 #include "xfs_log_format.h"
  23 #include "xfs_trans_resv.h"
  24 #include "xfs_bit.h"
  25 #include "xfs_mount.h"
  26 #include "xfs_inode.h"
  27 #include "xfs_btree.h"
  28 #include "xfs_ialloc.h"
  29 #include "xfs_ialloc_btree.h"
  30 #include "xfs_alloc.h"
  31 #include "xfs_trace.h"
  32 #include "xfs_cksum.h"
  33 #include "xfs_trans.h"
  34 #include "xfs_rmap.h"
  35
  36
  37 STATIC int
  38 xfs_inobt_get_minrecs(
  39         struct xfs_btree_cur    *cur,
  40         int                     level)
  41 {
  42         return cur->bc_mp->m_inobt_mnr[level != 0];
  43 }
  44
  45 STATIC struct xfs_btree_cur *
  46 xfs_inobt_dup_cursor(
  47         struct xfs_btree_cur    *cur)
  48 {
  49         return xfs_inobt_init_cursor(cur->bc_mp, cur->bc_tp,
  50                         cur->bc_private.a.agbp, cur->bc_private.a.agno,
  51                         cur->bc_btnum);
  52 }
  53
  54 STATIC void
  55 xfs_inobt_set_root(
  56         struct xfs_btree_cur    *cur,
  57         union xfs_btree_ptr     *nptr,
  58         int                     inc)    /* level change */
  59 {
  60         struct xfs_buf          *agbp = cur->bc_private.a.agbp;
  61         struct xfs_agi          *agi = XFS_BUF_TO_AGI(agbp);
  62
  63         agi->agi_root = nptr->s;
  64         be32_add_cpu(&agi->agi_level, inc);
  65         xfs_ialloc_log_agi(cur->bc_tp, agbp, XFS_AGI_ROOT | XFS_AGI_LEVEL);
  66 }
  67
  68 STATIC void
  69 xfs_finobt_set_root(
  70         struct xfs_btree_cur    *cur,
  71         union xfs_btree_ptr     *nptr,
  72         int                     inc)    /* level change */
  73 {
  74         struct xfs_buf          *agbp = cur->bc_private.a.agbp;
  75         struct xfs_agi          *agi = XFS_BUF_TO_AGI(agbp);
  76
  77         agi->agi_free_root = nptr->s;
  78         be32_add_cpu(&agi->agi_free_level, inc);
  79         xfs_ialloc_log_agi(cur->bc_tp, agbp,
  80                            XFS_AGI_FREE_ROOT | XFS_AGI_FREE_LEVEL);
  81 }
  82
  83 STATIC int
  84 __xfs_inobt_alloc_block(
  85         struct xfs_btree_cur    *cur,
  86         union xfs_btree_ptr     *start,
  87         union xfs_btree_ptr     *new,
  88         int                     *stat,
  89         enum xfs_ag_resv_type   resv)
  90 {
  91         xfs_alloc_arg_t         args;           /* block allocation args */
  92         int                     error;          /* error return value */
  93         xfs_agblock_t           sbno = be32_to_cpu(start->s);
  94
  95         XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
  96
  97         memset(&args, 0, sizeof(args));
  98         args.tp = cur->bc_tp;
  99         args.mp = cur->bc_mp;
 100         xfs_rmap_ag_owner(&args.oinfo, XFS_RMAP_OWN_INOBT);
 101         args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_private.a.agno, sbno);
 102         args.minlen = 1;
 103         args.maxlen = 1;
 104         args.prod = 1;
 105         args.type = XFS_ALLOCTYPE_NEAR_BNO;
 106         args.resv = resv;
 107
 108         error = xfs_alloc_vextent(&args);
 109         if (error) {
 110                 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
 111                 return error;
 112         }
 113         if (args.fsbno == NULLFSBLOCK) {
 114                 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 115                 *stat = 0;
 116                 return 0;
 117         }
 118         ASSERT(args.len == 1);
 119         XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 120
 121         new->s = cpu_to_be32(XFS_FSB_TO_AGBNO(args.mp, args.fsbno));
 122         *stat = 1;
 123         return 0;
 124 }
 125
 126 STATIC int
 127 xfs_inobt_alloc_block(
 128         struct xfs_btree_cur    *cur,
 129         union xfs_btree_ptr     *start,
 130         union xfs_btree_ptr     *new,
 131         int                     *stat)
 132 {
 133         return __xfs_inobt_alloc_block(cur, start, new, stat, XFS_AG_RESV_NONE);
 134 }
 135
 136 STATIC int
 137 xfs_finobt_alloc_block(
 138         struct xfs_btree_cur    *cur,
 139         union xfs_btree_ptr     *start,
 140         union xfs_btree_ptr     *new,
 141         int                     *stat)
 142 {
 143         return __xfs_inobt_alloc_block(cur, start, new, stat,
 144                         XFS_AG_RESV_METADATA);
 145 }
 146
 147 STATIC int
 148 xfs_inobt_free_block(
 149         struct xfs_btree_cur    *cur,
 150         struct xfs_buf          *bp)
 151 {
 152         struct xfs_owner_info   oinfo;
 153
 154         xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INOBT);
 155         return xfs_free_extent(cur->bc_tp,
 156                         XFS_DADDR_TO_FSB(cur->bc_mp, XFS_BUF_ADDR(bp)), 1,
 157                         &oinfo, XFS_AG_RESV_NONE);
 158 }
 159
 160 STATIC int
 161 xfs_inobt_get_maxrecs(
 162         struct xfs_btree_cur    *cur,
 163         int                     level)
 164 {
 165         return cur->bc_mp->m_inobt_mxr[level != 0];
 166 }
 167
 168 STATIC void
 169 xfs_inobt_init_key_from_rec(
 170         union xfs_btree_key     *key,
 171         union xfs_btree_rec     *rec)
 172 {
 173         key->inobt.ir_startino = rec->inobt.ir_startino;
 174 }
 175
 176 STATIC void
 177 xfs_inobt_init_rec_from_cur(
 178         struct xfs_btree_cur    *cur,
 179         union xfs_btree_rec     *rec)
 180 {
 181         rec->inobt.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
 182         if (xfs_sb_version_hassparseinodes(&cur->bc_mp->m_sb)) {
 183                 rec->inobt.ir_u.sp.ir_holemask =
 184                                         cpu_to_be16(cur->bc_rec.i.ir_holemask);
 185                 rec->inobt.ir_u.sp.ir_count = cur->bc_rec.i.ir_count;
 186                 rec->inobt.ir_u.sp.ir_freecount = cur->bc_rec.i.ir_freecount;
 187         } else {
 188                 /* ir_holemask/ir_count not supported on-disk */
 189                 rec->inobt.ir_u.f.ir_freecount =
 190                                         cpu_to_be32(cur->bc_rec.i.ir_freecount);
 191         }
 192         rec->inobt.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
 193 }
 194
 195 /*
 196  * initial value of ptr for lookup
 197  */
 198 STATIC void
 199 xfs_inobt_init_ptr_from_cur(
 200         struct xfs_btree_cur    *cur,
 201         union xfs_btree_ptr     *ptr)
 202 {
 203         struct xfs_agi          *agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
 204
 205         ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));
 206
 207         ptr->s = agi->agi_root;
 208 }
 209
 210 STATIC void
 211 xfs_finobt_init_ptr_from_cur(
 212         struct xfs_btree_cur    *cur,
 213         union xfs_btree_ptr     *ptr)
 214 {
 215         struct xfs_agi          *agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
 216
 217         ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));
 218         ptr->s = agi->agi_free_root;
 219 }
 220
 221 STATIC int64_t
 222 xfs_inobt_key_diff(
 223         struct xfs_btree_cur    *cur,
 224         union xfs_btree_key     *key)
 225 {
 226         return (int64_t)be32_to_cpu(key->inobt.ir_startino) -
 227                           cur->bc_rec.i.ir_startino;
 228 }
 229
 230 static int
 231 xfs_inobt_verify(
 232         struct xfs_buf          *bp)
 233 {
 234         struct xfs_mount        *mp = bp->b_target->bt_mount;
 235         struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
 236         unsigned int            level;
 237
 238         /*
 239          * During growfs operations, we can't verify the exact owner as the
 240          * perag is not fully initialised and hence not attached to the buffer.
 241          *
 242          * Similarly, during log recovery we will have a perag structure
 243          * attached, but the agi information will not yet have been initialised
 244          * from the on disk AGI. We don't currently use any of this information,
 245          * but beware of the landmine (i.e. need to check pag->pagi_init) if we
 246          * ever do.
 247          */
 248         switch (block->bb_magic) {
 249         case cpu_to_be32(XFS_IBT_CRC_MAGIC):
 250         case cpu_to_be32(XFS_FIBT_CRC_MAGIC):
 251                 if (!xfs_btree_sblock_v5hdr_verify(bp))
 252                         return false;
 253                 /* fall through */
 254         case cpu_to_be32(XFS_IBT_MAGIC):
 255         case cpu_to_be32(XFS_FIBT_MAGIC):
 256                 break;
 257         default:
 258                 return 0;
 259         }
 260
 261         /* level verification */
 262         level = be16_to_cpu(block->bb_level);
 263         if (level >= mp->m_in_maxlevels)
 264                 return false;
 265
 266         return xfs_btree_sblock_verify(bp, mp->m_inobt_mxr[level != 0]);
 267 }
 268
 269 static void
 270 xfs_inobt_read_verify(
 271         struct xfs_buf  *bp)
 272 {
 273         if (!xfs_btree_sblock_verify_crc(bp))
 274                 xfs_buf_ioerror(bp, -EFSBADCRC);
 275         else if (!xfs_inobt_verify(bp))
 276                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
 277
 278         if (bp->b_error) {
 279                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 280                 xfs_verifier_error(bp);
 281         }
 282 }
 283
 284 static void
 285 xfs_inobt_write_verify(
 286         struct xfs_buf  *bp)
 287 {
 288         if (!xfs_inobt_verify(bp)) {
 289                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 290                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
 291                 xfs_verifier_error(bp);
 292                 return;
 293         }
 294         xfs_btree_sblock_calc_crc(bp);
 295
 296 }
 297
 298 const struct xfs_buf_ops xfs_inobt_buf_ops = {
 299         .name = "xfs_inobt",
 300         .verify_read = xfs_inobt_read_verify,
 301         .verify_write = xfs_inobt_write_verify,
 302 };
 303
 304 #if defined(DEBUG) || defined(XFS_WARN)
 305 STATIC int
 306 xfs_inobt_keys_inorder(
 307         struct xfs_btree_cur    *cur,
 308         union xfs_btree_key     *k1,
 309         union xfs_btree_key     *k2)
 310 {
 311         return be32_to_cpu(k1->inobt.ir_startino) <
 312                 be32_to_cpu(k2->inobt.ir_startino);
 313 }
 314
 315 STATIC int
 316 xfs_inobt_recs_inorder(
 317         struct xfs_btree_cur    *cur,
 318         union xfs_btree_rec     *r1,
 319         union xfs_btree_rec     *r2)
 320 {
 321         return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
 322                 be32_to_cpu(r2->inobt.ir_startino);
 323 }
 324 #endif  /* DEBUG */
 325
 326 static const struct xfs_btree_ops xfs_inobt_ops = {
 327         .rec_len                = sizeof(xfs_inobt_rec_t),
 328         .key_len                = sizeof(xfs_inobt_key_t),
 329
 330         .dup_cursor             = xfs_inobt_dup_cursor,
 331         .set_root               = xfs_inobt_set_root,
 332         .alloc_block            = xfs_inobt_alloc_block,
 333         .free_block             = xfs_inobt_free_block,
 334         .get_minrecs            = xfs_inobt_get_minrecs,
 335         .get_maxrecs            = xfs_inobt_get_maxrecs,
 336         .init_key_from_rec      = xfs_inobt_init_key_from_rec,
 337         .init_rec_from_cur      = xfs_inobt_init_rec_from_cur,
 338         .init_ptr_from_cur      = xfs_inobt_init_ptr_from_cur,
 339         .key_diff               = xfs_inobt_key_diff,
 340         .buf_ops                = &xfs_inobt_buf_ops,
 341 #if defined(DEBUG) || defined(XFS_WARN)
 342         .keys_inorder           = xfs_inobt_keys_inorder,
 343         .recs_inorder           = xfs_inobt_recs_inorder,
 344 #endif
 345 };
 346
 347 static const struct xfs_btree_ops xfs_finobt_ops = {
 348         .rec_len                = sizeof(xfs_inobt_rec_t),
 349         .key_len                = sizeof(xfs_inobt_key_t),
 350
 351         .dup_cursor             = xfs_inobt_dup_cursor,
 352         .set_root               = xfs_finobt_set_root,
 353         .alloc_block            = xfs_finobt_alloc_block,
 354         .free_block             = xfs_inobt_free_block,
 355         .get_minrecs            = xfs_inobt_get_minrecs,
 356         .get_maxrecs            = xfs_inobt_get_maxrecs,
 357         .init_key_from_rec      = xfs_inobt_init_key_from_rec,
 358         .init_rec_from_cur      = xfs_inobt_init_rec_from_cur,
 359         .init_ptr_from_cur      = xfs_finobt_init_ptr_from_cur,
 360         .key_diff               = xfs_inobt_key_diff,
 361         .buf_ops                = &xfs_inobt_buf_ops,
 362 #if defined(DEBUG) || defined(XFS_WARN)
 363         .keys_inorder           = xfs_inobt_keys_inorder,
 364         .recs_inorder           = xfs_inobt_recs_inorder,
 365 #endif
 366 };
 367
 368 /*
 369  * Allocate a new inode btree cursor.
 370  */
 371 struct xfs_btree_cur *                          /* new inode btree cursor */
 372 xfs_inobt_init_cursor(
 373         struct xfs_mount        *mp,            /* file system mount point */
 374         struct xfs_trans        *tp,            /* transaction pointer */
 375         struct xfs_buf          *agbp,          /* buffer for agi structure */
 376         xfs_agnumber_t          agno,           /* allocation group number */
 377         xfs_btnum_t             btnum)          /* ialloc or free ino btree */
 378 {
 379         struct xfs_agi          *agi = XFS_BUF_TO_AGI(agbp);
 380         struct xfs_btree_cur    *cur;
 381
 382         cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
 383
 384         cur->bc_tp = tp;
 385         cur->bc_mp = mp;
 386         cur->bc_btnum = btnum;
 387         if (btnum == XFS_BTNUM_INO) {
 388                 cur->bc_nlevels = be32_to_cpu(agi->agi_level);
 389                 cur->bc_ops = &xfs_inobt_ops;
 390                 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_ibt_2);
 391         } else {
 392                 cur->bc_nlevels = be32_to_cpu(agi->agi_free_level);
 393                 cur->bc_ops = &xfs_finobt_ops;
 394                 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_fibt_2);
 395         }
 396
 397         cur->bc_blocklog = mp->m_sb.sb_blocklog;
 398
 399         if (xfs_sb_version_hascrc(&mp->m_sb))
 400                 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
 401
 402         cur->bc_private.a.agbp = agbp;
 403         cur->bc_private.a.agno = agno;
 404
 405         return cur;
 406 }
 407
 408 /*
 409  * Calculate number of records in an inobt btree block.
 410  */
 411 int
 412 xfs_inobt_maxrecs(
 413         struct xfs_mount        *mp,
 414         int                     blocklen,
 415         int                     leaf)
 416 {
 417         blocklen -= XFS_INOBT_BLOCK_LEN(mp);
 418
 419         if (leaf)
 420                 return blocklen / sizeof(xfs_inobt_rec_t);
 421         return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
 422 }
 423
 424 /*
 425  * Convert the inode record holemask to an inode allocation bitmap. The inode
 426  * allocation bitmap is inode granularity and specifies whether an inode is
 427  * physically allocated on disk (not whether the inode is considered allocated
 428  * or free by the fs).
 429  *
 430  * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
 431  */
 432 uint64_t
 433 xfs_inobt_irec_to_allocmask(
 434         struct xfs_inobt_rec_incore     *rec)
 435 {
 436         uint64_t                        bitmap = 0;
 437         uint64_t                        inodespbit;
 438         int                             nextbit;
 439         uint                            allocbitmap;
 440
 441         /*
 442          * The holemask has 16-bits for a 64 inode record. Therefore each
 443          * holemask bit represents multiple inodes. Create a mask of bits to set
 444          * in the allocmask for each holemask bit.
 445          */
 446         inodespbit = (1 << XFS_INODES_PER_HOLEMASK_BIT) - 1;
 447
 448         /*
 449          * Allocated inodes are represented by 0 bits in holemask. Invert the 0
 450          * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
 451          * anything beyond the 16 holemask bits since this casts to a larger
 452          * type.
 453          */
 454         allocbitmap = ~rec->ir_holemask & ((1 << XFS_INOBT_HOLEMASK_BITS) - 1);
 455
 456         /*
 457          * allocbitmap is the inverted holemask so every set bit represents
 458          * allocated inodes. To expand from 16-bit holemask granularity to
 459          * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
 460          * bitmap for every holemask bit.
 461          */
 462         nextbit = xfs_next_bit(&allocbitmap, 1, 0);
 463         while (nextbit != -1) {
 464                 ASSERT(nextbit < (sizeof(rec->ir_holemask) * NBBY));
 465
 466                 bitmap |= (inodespbit <<
 467                            (nextbit * XFS_INODES_PER_HOLEMASK_BIT));
 468
 469                 nextbit = xfs_next_bit(&allocbitmap, 1, nextbit + 1);
 470         }
 471
 472         return bitmap;
 473 }
 474
 475 #if defined(DEBUG) || defined(XFS_WARN)
 476 /*
 477  * Verify that an in-core inode record has a valid inode count.
 478  */
 479 int
 480 xfs_inobt_rec_check_count(
 481         struct xfs_mount                *mp,
 482         struct xfs_inobt_rec_incore     *rec)
 483 {
 484         int                             inocount = 0;
 485         int                             nextbit = 0;
 486         uint64_t                        allocbmap;
 487         int                             wordsz;
 488
 489         wordsz = sizeof(allocbmap) / sizeof(unsigned int);
 490         allocbmap = xfs_inobt_irec_to_allocmask(rec);
 491
 492         nextbit = xfs_next_bit((uint *) &allocbmap, wordsz, nextbit);
 493         while (nextbit != -1) {
 494                 inocount++;
 495                 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz,
 496                                        nextbit + 1);
 497         }
 498
 499         if (inocount != rec->ir_count)
 500                 return -EFSCORRUPTED;
 501
 502         return 0;
 503 }
 504 #endif  /* DEBUG */
 505
 506 static xfs_extlen_t
 507 xfs_inobt_max_size(
 508         struct xfs_mount        *mp)
 509 {
 510         /* Bail out if we're uninitialized, which can happen in mkfs. */
 511         if (mp->m_inobt_mxr[0] == 0)
 512                 return 0;
 513
 514         return xfs_btree_calc_size(mp, mp->m_inobt_mnr,
 515                 (uint64_t)mp->m_sb.sb_agblocks * mp->m_sb.sb_inopblock /
 516                                 XFS_INODES_PER_CHUNK);
 517 }
 518
 519 static int
 520 xfs_inobt_count_blocks(
 521         struct xfs_mount        *mp,
 522         xfs_agnumber_t          agno,
 523         xfs_btnum_t             btnum,
 524         xfs_extlen_t            *tree_blocks)
 525 {
 526         struct xfs_buf          *agbp;
 527         struct xfs_btree_cur    *cur;
 528         int                     error;
 529
 530         error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
 531         if (error)
 532                 return error;
 533
 534         cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno, btnum);
 535         error = xfs_btree_count_blocks(cur, tree_blocks);
 536         xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
 537         xfs_buf_relse(agbp);
 538
 539         return error;
 540 }
 541
 542 /*
 543  * Figure out how many blocks to reserve and how many are used by this btree.
 544  */
 545 int
 546 xfs_finobt_calc_reserves(
 547         struct xfs_mount        *mp,
 548         xfs_agnumber_t          agno,
 549         xfs_extlen_t            *ask,
 550         xfs_extlen_t            *used)
 551 {
 552         xfs_extlen_t            tree_len = 0;
 553         int                     error;
 554
 555         if (!xfs_sb_version_hasfinobt(&mp->m_sb))
 556                 return 0;
 557
 558         error = xfs_inobt_count_blocks(mp, agno, XFS_BTNUM_FINO, &tree_len);
 559         if (error)
 560                 return error;
 561
 562         *ask += xfs_inobt_max_size(mp);
 563         *used += tree_len;
 564         return 0;
 565 }