fs/xfs/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 "xfs.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_error.h"
  32 #include "xfs_trace.h"
  33 #include "xfs_cksum.h"
  34 #include "xfs_trans.h"
  35 #include "xfs_rmap.h"
  36
  37
  38 STATIC int
  39 xfs_inobt_get_minrecs(
  40         struct xfs_btree_cur    *cur,
  41         int                     level)
  42 {
  43         return cur->bc_mp->m_inobt_mnr[level != 0];
  44 }
  45
  46 STATIC struct xfs_btree_cur *
  47 xfs_inobt_dup_cursor(
  48         struct xfs_btree_cur    *cur)
  49 {
  50         return xfs_inobt_init_cursor(cur->bc_mp, cur->bc_tp,
  51                         cur->bc_private.a.agbp, cur->bc_private.a.agno,
  52                         cur->bc_btnum);
  53 }
  54
  55 STATIC void
  56 xfs_inobt_set_root(
  57         struct xfs_btree_cur    *cur,
  58         union xfs_btree_ptr     *nptr,
  59         int                     inc)    /* level change */
  60 {
  61         struct xfs_buf          *agbp = cur->bc_private.a.agbp;
  62         struct xfs_agi          *agi = XFS_BUF_TO_AGI(agbp);
  63
  64         agi->agi_root = nptr->s;
  65         be32_add_cpu(&agi->agi_level, inc);
  66         xfs_ialloc_log_agi(cur->bc_tp, agbp, XFS_AGI_ROOT | XFS_AGI_LEVEL);
  67 }
  68
  69 STATIC void
  70 xfs_finobt_set_root(
  71         struct xfs_btree_cur    *cur,
  72         union xfs_btree_ptr     *nptr,
  73         int                     inc)    /* level change */
  74 {
  75         struct xfs_buf          *agbp = cur->bc_private.a.agbp;
  76         struct xfs_agi          *agi = XFS_BUF_TO_AGI(agbp);
  77
  78         agi->agi_free_root = nptr->s;
  79         be32_add_cpu(&agi->agi_free_level, inc);
  80         xfs_ialloc_log_agi(cur->bc_tp, agbp,
  81                            XFS_AGI_FREE_ROOT | XFS_AGI_FREE_LEVEL);
  82 }
  83
  84 STATIC int
  85 xfs_inobt_alloc_block(
  86         struct xfs_btree_cur    *cur,
  87         union xfs_btree_ptr     *start,
  88         union xfs_btree_ptr     *new,
  89         int                     *stat)
  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
 107         error = xfs_alloc_vextent(&args);
 108         if (error) {
 109                 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
 110                 return error;
 111         }
 112         if (args.fsbno == NULLFSBLOCK) {
 113                 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 114                 *stat = 0;
 115                 return 0;
 116         }
 117         ASSERT(args.len == 1);
 118         XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 119
 120         new->s = cpu_to_be32(XFS_FSB_TO_AGBNO(args.mp, args.fsbno));
 121         *stat = 1;
 122         return 0;
 123 }
 124
 125 STATIC int
 126 xfs_inobt_free_block(
 127         struct xfs_btree_cur    *cur,
 128         struct xfs_buf          *bp)
 129 {
 130         struct xfs_owner_info   oinfo;
 131
 132         xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INOBT);
 133         return xfs_free_extent(cur->bc_tp,
 134                         XFS_DADDR_TO_FSB(cur->bc_mp, XFS_BUF_ADDR(bp)), 1,
 135                         &oinfo, XFS_AG_RESV_NONE);
 136 }
 137
 138 STATIC int
 139 xfs_inobt_get_maxrecs(
 140         struct xfs_btree_cur    *cur,
 141         int                     level)
 142 {
 143         return cur->bc_mp->m_inobt_mxr[level != 0];
 144 }
 145
 146 STATIC void
 147 xfs_inobt_init_key_from_rec(
 148         union xfs_btree_key     *key,
 149         union xfs_btree_rec     *rec)
 150 {
 151         key->inobt.ir_startino = rec->inobt.ir_startino;
 152 }
 153
 154 STATIC void
 155 xfs_inobt_init_rec_from_cur(
 156         struct xfs_btree_cur    *cur,
 157         union xfs_btree_rec     *rec)
 158 {
 159         rec->inobt.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
 160         if (xfs_sb_version_hassparseinodes(&cur->bc_mp->m_sb)) {
 161                 rec->inobt.ir_u.sp.ir_holemask =
 162                                         cpu_to_be16(cur->bc_rec.i.ir_holemask);
 163                 rec->inobt.ir_u.sp.ir_count = cur->bc_rec.i.ir_count;
 164                 rec->inobt.ir_u.sp.ir_freecount = cur->bc_rec.i.ir_freecount;
 165         } else {
 166                 /* ir_holemask/ir_count not supported on-disk */
 167                 rec->inobt.ir_u.f.ir_freecount =
 168                                         cpu_to_be32(cur->bc_rec.i.ir_freecount);
 169         }
 170         rec->inobt.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
 171 }
 172
 173 /*
 174  * initial value of ptr for lookup
 175  */
 176 STATIC void
 177 xfs_inobt_init_ptr_from_cur(
 178         struct xfs_btree_cur    *cur,
 179         union xfs_btree_ptr     *ptr)
 180 {
 181         struct xfs_agi          *agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
 182
 183         ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));
 184
 185         ptr->s = agi->agi_root;
 186 }
 187
 188 STATIC void
 189 xfs_finobt_init_ptr_from_cur(
 190         struct xfs_btree_cur    *cur,
 191         union xfs_btree_ptr     *ptr)
 192 {
 193         struct xfs_agi          *agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
 194
 195         ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));
 196         ptr->s = agi->agi_free_root;
 197 }
 198
 199 STATIC __int64_t
 200 xfs_inobt_key_diff(
 201         struct xfs_btree_cur    *cur,
 202         union xfs_btree_key     *key)
 203 {
 204         return (__int64_t)be32_to_cpu(key->inobt.ir_startino) -
 205                           cur->bc_rec.i.ir_startino;
 206 }
 207
 208 static int
 209 xfs_inobt_verify(
 210         struct xfs_buf          *bp)
 211 {
 212         struct xfs_mount        *mp = bp->b_target->bt_mount;
 213         struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
 214         unsigned int            level;
 215
 216         /*
 217          * During growfs operations, we can't verify the exact owner as the
 218          * perag is not fully initialised and hence not attached to the buffer.
 219          *
 220          * Similarly, during log recovery we will have a perag structure
 221          * attached, but the agi information will not yet have been initialised
 222          * from the on disk AGI. We don't currently use any of this information,
 223          * but beware of the landmine (i.e. need to check pag->pagi_init) if we
 224          * ever do.
 225          */
 226         switch (block->bb_magic) {
 227         case cpu_to_be32(XFS_IBT_CRC_MAGIC):
 228         case cpu_to_be32(XFS_FIBT_CRC_MAGIC):
 229                 if (!xfs_btree_sblock_v5hdr_verify(bp))
 230                         return false;
 231                 /* fall through */
 232         case cpu_to_be32(XFS_IBT_MAGIC):
 233         case cpu_to_be32(XFS_FIBT_MAGIC):
 234                 break;
 235         default:
 236                 return 0;
 237         }
 238
 239         /* level verification */
 240         level = be16_to_cpu(block->bb_level);
 241         if (level >= mp->m_in_maxlevels)
 242                 return false;
 243
 244         return xfs_btree_sblock_verify(bp, mp->m_inobt_mxr[level != 0]);
 245 }
 246
 247 static void
 248 xfs_inobt_read_verify(
 249         struct xfs_buf  *bp)
 250 {
 251         if (!xfs_btree_sblock_verify_crc(bp))
 252                 xfs_buf_ioerror(bp, -EFSBADCRC);
 253         else if (!xfs_inobt_verify(bp))
 254                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
 255
 256         if (bp->b_error) {
 257                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 258                 xfs_verifier_error(bp);
 259         }
 260 }
 261
 262 static void
 263 xfs_inobt_write_verify(
 264         struct xfs_buf  *bp)
 265 {
 266         if (!xfs_inobt_verify(bp)) {
 267                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 268                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
 269                 xfs_verifier_error(bp);
 270                 return;
 271         }
 272         xfs_btree_sblock_calc_crc(bp);
 273
 274 }
 275
 276 const struct xfs_buf_ops xfs_inobt_buf_ops = {
 277         .name = "xfs_inobt",
 278         .verify_read = xfs_inobt_read_verify,
 279         .verify_write = xfs_inobt_write_verify,
 280 };
 281
 282 #if defined(DEBUG) || defined(XFS_WARN)
 283 STATIC int
 284 xfs_inobt_keys_inorder(
 285         struct xfs_btree_cur    *cur,
 286         union xfs_btree_key     *k1,
 287         union xfs_btree_key     *k2)
 288 {
 289         return be32_to_cpu(k1->inobt.ir_startino) <
 290                 be32_to_cpu(k2->inobt.ir_startino);
 291 }
 292
 293 STATIC int
 294 xfs_inobt_recs_inorder(
 295         struct xfs_btree_cur    *cur,
 296         union xfs_btree_rec     *r1,
 297         union xfs_btree_rec     *r2)
 298 {
 299         return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
 300                 be32_to_cpu(r2->inobt.ir_startino);
 301 }
 302 #endif  /* DEBUG */
 303
 304 static const struct xfs_btree_ops xfs_inobt_ops = {
 305         .rec_len                = sizeof(xfs_inobt_rec_t),
 306         .key_len                = sizeof(xfs_inobt_key_t),
 307
 308         .dup_cursor             = xfs_inobt_dup_cursor,
 309         .set_root               = xfs_inobt_set_root,
 310         .alloc_block            = xfs_inobt_alloc_block,
 311         .free_block             = xfs_inobt_free_block,
 312         .get_minrecs            = xfs_inobt_get_minrecs,
 313         .get_maxrecs            = xfs_inobt_get_maxrecs,
 314         .init_key_from_rec      = xfs_inobt_init_key_from_rec,
 315         .init_rec_from_cur      = xfs_inobt_init_rec_from_cur,
 316         .init_ptr_from_cur      = xfs_inobt_init_ptr_from_cur,
 317         .key_diff               = xfs_inobt_key_diff,
 318         .buf_ops                = &xfs_inobt_buf_ops,
 319 #if defined(DEBUG) || defined(XFS_WARN)
 320         .keys_inorder           = xfs_inobt_keys_inorder,
 321         .recs_inorder           = xfs_inobt_recs_inorder,
 322 #endif
 323 };
 324
 325 static const struct xfs_btree_ops xfs_finobt_ops = {
 326         .rec_len                = sizeof(xfs_inobt_rec_t),
 327         .key_len                = sizeof(xfs_inobt_key_t),
 328
 329         .dup_cursor             = xfs_inobt_dup_cursor,
 330         .set_root               = xfs_finobt_set_root,
 331         .alloc_block            = xfs_inobt_alloc_block,
 332         .free_block             = xfs_inobt_free_block,
 333         .get_minrecs            = xfs_inobt_get_minrecs,
 334         .get_maxrecs            = xfs_inobt_get_maxrecs,
 335         .init_key_from_rec      = xfs_inobt_init_key_from_rec,
 336         .init_rec_from_cur      = xfs_inobt_init_rec_from_cur,
 337         .init_ptr_from_cur      = xfs_finobt_init_ptr_from_cur,
 338         .key_diff               = xfs_inobt_key_diff,
 339         .buf_ops                = &xfs_inobt_buf_ops,
 340 #if defined(DEBUG) || defined(XFS_WARN)
 341         .keys_inorder           = xfs_inobt_keys_inorder,
 342         .recs_inorder           = xfs_inobt_recs_inorder,
 343 #endif
 344 };
 345
 346 /*
 347  * Allocate a new inode btree cursor.
 348  */
 349 struct xfs_btree_cur *                          /* new inode btree cursor */
 350 xfs_inobt_init_cursor(
 351         struct xfs_mount        *mp,            /* file system mount point */
 352         struct xfs_trans        *tp,            /* transaction pointer */
 353         struct xfs_buf          *agbp,          /* buffer for agi structure */
 354         xfs_agnumber_t          agno,           /* allocation group number */
 355         xfs_btnum_t             btnum)          /* ialloc or free ino btree */
 356 {
 357         struct xfs_agi          *agi = XFS_BUF_TO_AGI(agbp);
 358         struct xfs_btree_cur    *cur;
 359
 360         cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
 361
 362         cur->bc_tp = tp;
 363         cur->bc_mp = mp;
 364         cur->bc_btnum = btnum;
 365         if (btnum == XFS_BTNUM_INO) {
 366                 cur->bc_nlevels = be32_to_cpu(agi->agi_level);
 367                 cur->bc_ops = &xfs_inobt_ops;
 368                 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_ibt_2);
 369         } else {
 370                 cur->bc_nlevels = be32_to_cpu(agi->agi_free_level);
 371                 cur->bc_ops = &xfs_finobt_ops;
 372                 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_fibt_2);
 373         }
 374
 375         cur->bc_blocklog = mp->m_sb.sb_blocklog;
 376
 377         if (xfs_sb_version_hascrc(&mp->m_sb))
 378                 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
 379
 380         cur->bc_private.a.agbp = agbp;
 381         cur->bc_private.a.agno = agno;
 382
 383         return cur;
 384 }
 385
 386 /*
 387  * Calculate number of records in an inobt btree block.
 388  */
 389 int
 390 xfs_inobt_maxrecs(
 391         struct xfs_mount        *mp,
 392         int                     blocklen,
 393         int                     leaf)
 394 {
 395         blocklen -= XFS_INOBT_BLOCK_LEN(mp);
 396
 397         if (leaf)
 398                 return blocklen / sizeof(xfs_inobt_rec_t);
 399         return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
 400 }
 401
 402 /*
 403  * Convert the inode record holemask to an inode allocation bitmap. The inode
 404  * allocation bitmap is inode granularity and specifies whether an inode is
 405  * physically allocated on disk (not whether the inode is considered allocated
 406  * or free by the fs).
 407  *
 408  * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
 409  */
 410 uint64_t
 411 xfs_inobt_irec_to_allocmask(
 412         struct xfs_inobt_rec_incore     *rec)
 413 {
 414         uint64_t                        bitmap = 0;
 415         uint64_t                        inodespbit;
 416         int                             nextbit;
 417         uint                            allocbitmap;
 418
 419         /*
 420          * The holemask has 16-bits for a 64 inode record. Therefore each
 421          * holemask bit represents multiple inodes. Create a mask of bits to set
 422          * in the allocmask for each holemask bit.
 423          */
 424         inodespbit = (1 << XFS_INODES_PER_HOLEMASK_BIT) - 1;
 425
 426         /*
 427          * Allocated inodes are represented by 0 bits in holemask. Invert the 0
 428          * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
 429          * anything beyond the 16 holemask bits since this casts to a larger
 430          * type.
 431          */
 432         allocbitmap = ~rec->ir_holemask & ((1 << XFS_INOBT_HOLEMASK_BITS) - 1);
 433
 434         /*
 435          * allocbitmap is the inverted holemask so every set bit represents
 436          * allocated inodes. To expand from 16-bit holemask granularity to
 437          * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
 438          * bitmap for every holemask bit.
 439          */
 440         nextbit = xfs_next_bit(&allocbitmap, 1, 0);
 441         while (nextbit != -1) {
 442                 ASSERT(nextbit < (sizeof(rec->ir_holemask) * NBBY));
 443
 444                 bitmap |= (inodespbit <<
 445                            (nextbit * XFS_INODES_PER_HOLEMASK_BIT));
 446
 447                 nextbit = xfs_next_bit(&allocbitmap, 1, nextbit + 1);
 448         }
 449
 450         return bitmap;
 451 }
 452
 453 #if defined(DEBUG) || defined(XFS_WARN)
 454 /*
 455  * Verify that an in-core inode record has a valid inode count.
 456  */
 457 int
 458 xfs_inobt_rec_check_count(
 459         struct xfs_mount                *mp,
 460         struct xfs_inobt_rec_incore     *rec)
 461 {
 462         int                             inocount = 0;
 463         int                             nextbit = 0;
 464         uint64_t                        allocbmap;
 465         int                             wordsz;
 466
 467         wordsz = sizeof(allocbmap) / sizeof(unsigned int);
 468         allocbmap = xfs_inobt_irec_to_allocmask(rec);
 469
 470         nextbit = xfs_next_bit((uint *) &allocbmap, wordsz, nextbit);
 471         while (nextbit != -1) {
 472                 inocount++;
 473                 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz,
 474                                        nextbit + 1);
 475         }
 476
 477         if (inocount != rec->ir_count)
 478                 return -EFSCORRUPTED;
 479
 480         return 0;
 481 }
 482 #endif  /* DEBUG */