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
  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 {
  90         xfs_alloc_arg_t         args;           /* block allocation args */
  91         int                     error;          /* error return value */
  92         xfs_agblock_t           sbno = be32_to_cpu(start->s);
  93
  94         XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
  95
  96         memset(&args, 0, sizeof(args));
  97         args.tp = cur->bc_tp;
  98         args.mp = cur->bc_mp;
  99         args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_private.a.agno, sbno);
 100         args.minlen = 1;
 101         args.maxlen = 1;
 102         args.prod = 1;
 103         args.type = XFS_ALLOCTYPE_NEAR_BNO;
 104
 105         error = xfs_alloc_vextent(&args);
 106         if (error) {
 107                 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
 108                 return error;
 109         }
 110         if (args.fsbno == NULLFSBLOCK) {
 111                 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 112                 *stat = 0;
 113                 return 0;
 114         }
 115         ASSERT(args.len == 1);
 116         XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 117
 118         new->s = cpu_to_be32(XFS_FSB_TO_AGBNO(args.mp, args.fsbno));
 119         *stat = 1;
 120         return 0;
 121 }
 122
 123 STATIC int
 124 xfs_inobt_free_block(
 125         struct xfs_btree_cur    *cur,
 126         struct xfs_buf          *bp)
 127 {
 128         xfs_fsblock_t           fsbno;
 129         int                     error;
 130
 131         fsbno = XFS_DADDR_TO_FSB(cur->bc_mp, XFS_BUF_ADDR(bp));
 132         error = xfs_free_extent(cur->bc_tp, fsbno, 1);
 133         if (error)
 134                 return error;
 135
 136         xfs_trans_binval(cur->bc_tp, bp);
 137         return error;
 138 }
 139
 140 STATIC int
 141 xfs_inobt_get_maxrecs(
 142         struct xfs_btree_cur    *cur,
 143         int                     level)
 144 {
 145         return cur->bc_mp->m_inobt_mxr[level != 0];
 146 }
 147
 148 STATIC void
 149 xfs_inobt_init_key_from_rec(
 150         union xfs_btree_key     *key,
 151         union xfs_btree_rec     *rec)
 152 {
 153         key->inobt.ir_startino = rec->inobt.ir_startino;
 154 }
 155
 156 STATIC void
 157 xfs_inobt_init_rec_from_key(
 158         union xfs_btree_key     *key,
 159         union xfs_btree_rec     *rec)
 160 {
 161         rec->inobt.ir_startino = key->inobt.ir_startino;
 162 }
 163
 164 STATIC void
 165 xfs_inobt_init_rec_from_cur(
 166         struct xfs_btree_cur    *cur,
 167         union xfs_btree_rec     *rec)
 168 {
 169         rec->inobt.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
 170         if (xfs_sb_version_hassparseinodes(&cur->bc_mp->m_sb)) {
 171                 rec->inobt.ir_u.sp.ir_holemask =
 172                                         cpu_to_be16(cur->bc_rec.i.ir_holemask);
 173                 rec->inobt.ir_u.sp.ir_count = cur->bc_rec.i.ir_count;
 174                 rec->inobt.ir_u.sp.ir_freecount = cur->bc_rec.i.ir_freecount;
 175         } else {
 176                 /* ir_holemask/ir_count not supported on-disk */
 177                 rec->inobt.ir_u.f.ir_freecount =
 178                                         cpu_to_be32(cur->bc_rec.i.ir_freecount);
 179         }
 180         rec->inobt.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
 181 }
 182
 183 /*
 184  * initial value of ptr for lookup
 185  */
 186 STATIC void
 187 xfs_inobt_init_ptr_from_cur(
 188         struct xfs_btree_cur    *cur,
 189         union xfs_btree_ptr     *ptr)
 190 {
 191         struct xfs_agi          *agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
 192
 193         ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));
 194
 195         ptr->s = agi->agi_root;
 196 }
 197
 198 STATIC void
 199 xfs_finobt_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         ptr->s = agi->agi_free_root;
 207 }
 208
 209 STATIC __int64_t
 210 xfs_inobt_key_diff(
 211         struct xfs_btree_cur    *cur,
 212         union xfs_btree_key     *key)
 213 {
 214         return (__int64_t)be32_to_cpu(key->inobt.ir_startino) -
 215                           cur->bc_rec.i.ir_startino;
 216 }
 217
 218 static int
 219 xfs_inobt_verify(
 220         struct xfs_buf          *bp)
 221 {
 222         struct xfs_mount        *mp = bp->b_target->bt_mount;
 223         struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
 224         unsigned int            level;
 225
 226         /*
 227          * During growfs operations, we can't verify the exact owner as the
 228          * perag is not fully initialised and hence not attached to the buffer.
 229          *
 230          * Similarly, during log recovery we will have a perag structure
 231          * attached, but the agi information will not yet have been initialised
 232          * from the on disk AGI. We don't currently use any of this information,
 233          * but beware of the landmine (i.e. need to check pag->pagi_init) if we
 234          * ever do.
 235          */
 236         switch (block->bb_magic) {
 237         case cpu_to_be32(XFS_IBT_CRC_MAGIC):
 238         case cpu_to_be32(XFS_FIBT_CRC_MAGIC):
 239                 if (!xfs_btree_sblock_v5hdr_verify(bp))
 240                         return false;
 241                 /* fall through */
 242         case cpu_to_be32(XFS_IBT_MAGIC):
 243         case cpu_to_be32(XFS_FIBT_MAGIC):
 244                 break;
 245         default:
 246                 return 0;
 247         }
 248
 249         /* level verification */
 250         level = be16_to_cpu(block->bb_level);
 251         if (level >= mp->m_in_maxlevels)
 252                 return false;
 253
 254         return xfs_btree_sblock_verify(bp, mp->m_inobt_mxr[level != 0]);
 255 }
 256
 257 static void
 258 xfs_inobt_read_verify(
 259         struct xfs_buf  *bp)
 260 {
 261         if (!xfs_btree_sblock_verify_crc(bp))
 262                 xfs_buf_ioerror(bp, -EFSBADCRC);
 263         else if (!xfs_inobt_verify(bp))
 264                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
 265
 266         if (bp->b_error) {
 267                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 268                 xfs_verifier_error(bp);
 269         }
 270 }
 271
 272 static void
 273 xfs_inobt_write_verify(
 274         struct xfs_buf  *bp)
 275 {
 276         if (!xfs_inobt_verify(bp)) {
 277                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 278                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
 279                 xfs_verifier_error(bp);
 280                 return;
 281         }
 282         xfs_btree_sblock_calc_crc(bp);
 283
 284 }
 285
 286 const struct xfs_buf_ops xfs_inobt_buf_ops = {
 287         .name = "xfs_inobt",
 288         .verify_read = xfs_inobt_read_verify,
 289         .verify_write = xfs_inobt_write_verify,
 290 };
 291
 292 #if defined(DEBUG) || defined(XFS_WARN)
 293 STATIC int
 294 xfs_inobt_keys_inorder(
 295         struct xfs_btree_cur    *cur,
 296         union xfs_btree_key     *k1,
 297         union xfs_btree_key     *k2)
 298 {
 299         return be32_to_cpu(k1->inobt.ir_startino) <
 300                 be32_to_cpu(k2->inobt.ir_startino);
 301 }
 302
 303 STATIC int
 304 xfs_inobt_recs_inorder(
 305         struct xfs_btree_cur    *cur,
 306         union xfs_btree_rec     *r1,
 307         union xfs_btree_rec     *r2)
 308 {
 309         return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
 310                 be32_to_cpu(r2->inobt.ir_startino);
 311 }
 312 #endif  /* DEBUG */
 313
 314 static const struct xfs_btree_ops xfs_inobt_ops = {
 315         .rec_len                = sizeof(xfs_inobt_rec_t),
 316         .key_len                = sizeof(xfs_inobt_key_t),
 317
 318         .dup_cursor             = xfs_inobt_dup_cursor,
 319         .set_root               = xfs_inobt_set_root,
 320         .alloc_block            = xfs_inobt_alloc_block,
 321         .free_block             = xfs_inobt_free_block,
 322         .get_minrecs            = xfs_inobt_get_minrecs,
 323         .get_maxrecs            = xfs_inobt_get_maxrecs,
 324         .init_key_from_rec      = xfs_inobt_init_key_from_rec,
 325         .init_rec_from_key      = xfs_inobt_init_rec_from_key,
 326         .init_rec_from_cur      = xfs_inobt_init_rec_from_cur,
 327         .init_ptr_from_cur      = xfs_inobt_init_ptr_from_cur,
 328         .key_diff               = xfs_inobt_key_diff,
 329         .buf_ops                = &xfs_inobt_buf_ops,
 330 #if defined(DEBUG) || defined(XFS_WARN)
 331         .keys_inorder           = xfs_inobt_keys_inorder,
 332         .recs_inorder           = xfs_inobt_recs_inorder,
 333 #endif
 334 };
 335
 336 static const struct xfs_btree_ops xfs_finobt_ops = {
 337         .rec_len                = sizeof(xfs_inobt_rec_t),
 338         .key_len                = sizeof(xfs_inobt_key_t),
 339
 340         .dup_cursor             = xfs_inobt_dup_cursor,
 341         .set_root               = xfs_finobt_set_root,
 342         .alloc_block            = xfs_inobt_alloc_block,
 343         .free_block             = xfs_inobt_free_block,
 344         .get_minrecs            = xfs_inobt_get_minrecs,
 345         .get_maxrecs            = xfs_inobt_get_maxrecs,
 346         .init_key_from_rec      = xfs_inobt_init_key_from_rec,
 347         .init_rec_from_key      = xfs_inobt_init_rec_from_key,
 348         .init_rec_from_cur      = xfs_inobt_init_rec_from_cur,
 349         .init_ptr_from_cur      = xfs_finobt_init_ptr_from_cur,
 350         .key_diff               = xfs_inobt_key_diff,
 351         .buf_ops                = &xfs_inobt_buf_ops,
 352 #if defined(DEBUG) || defined(XFS_WARN)
 353         .keys_inorder           = xfs_inobt_keys_inorder,
 354         .recs_inorder           = xfs_inobt_recs_inorder,
 355 #endif
 356 };
 357
 358 /*
 359  * Allocate a new inode btree cursor.
 360  */
 361 struct xfs_btree_cur *                          /* new inode btree cursor */
 362 xfs_inobt_init_cursor(
 363         struct xfs_mount        *mp,            /* file system mount point */
 364         struct xfs_trans        *tp,            /* transaction pointer */
 365         struct xfs_buf          *agbp,          /* buffer for agi structure */
 366         xfs_agnumber_t          agno,           /* allocation group number */
 367         xfs_btnum_t             btnum)          /* ialloc or free ino btree */
 368 {
 369         struct xfs_agi          *agi = XFS_BUF_TO_AGI(agbp);
 370         struct xfs_btree_cur    *cur;
 371
 372         cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
 373
 374         cur->bc_tp = tp;
 375         cur->bc_mp = mp;
 376         cur->bc_btnum = btnum;
 377         if (btnum == XFS_BTNUM_INO) {
 378                 cur->bc_nlevels = be32_to_cpu(agi->agi_level);
 379                 cur->bc_ops = &xfs_inobt_ops;
 380         } else {
 381                 cur->bc_nlevels = be32_to_cpu(agi->agi_free_level);
 382                 cur->bc_ops = &xfs_finobt_ops;
 383         }
 384
 385         cur->bc_blocklog = mp->m_sb.sb_blocklog;
 386
 387         if (xfs_sb_version_hascrc(&mp->m_sb))
 388                 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
 389
 390         cur->bc_private.a.agbp = agbp;
 391         cur->bc_private.a.agno = agno;
 392
 393         return cur;
 394 }
 395
 396 /*
 397  * Calculate number of records in an inobt btree block.
 398  */
 399 int
 400 xfs_inobt_maxrecs(
 401         struct xfs_mount        *mp,
 402         int                     blocklen,
 403         int                     leaf)
 404 {
 405         blocklen -= XFS_INOBT_BLOCK_LEN(mp);
 406
 407         if (leaf)
 408                 return blocklen / sizeof(xfs_inobt_rec_t);
 409         return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
 410 }
 411
 412 /*
 413  * Convert the inode record holemask to an inode allocation bitmap. The inode
 414  * allocation bitmap is inode granularity and specifies whether an inode is
 415  * physically allocated on disk (not whether the inode is considered allocated
 416  * or free by the fs).
 417  *
 418  * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
 419  */
 420 uint64_t
 421 xfs_inobt_irec_to_allocmask(
 422         struct xfs_inobt_rec_incore     *rec)
 423 {
 424         uint64_t                        bitmap = 0;
 425         uint64_t                        inodespbit;
 426         int                             nextbit;
 427         uint                            allocbitmap;
 428
 429         /*
 430          * The holemask has 16-bits for a 64 inode record. Therefore each
 431          * holemask bit represents multiple inodes. Create a mask of bits to set
 432          * in the allocmask for each holemask bit.
 433          */
 434         inodespbit = (1 << XFS_INODES_PER_HOLEMASK_BIT) - 1;
 435
 436         /*
 437          * Allocated inodes are represented by 0 bits in holemask. Invert the 0
 438          * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
 439          * anything beyond the 16 holemask bits since this casts to a larger
 440          * type.
 441          */
 442         allocbitmap = ~rec->ir_holemask & ((1 << XFS_INOBT_HOLEMASK_BITS) - 1);
 443
 444         /*
 445          * allocbitmap is the inverted holemask so every set bit represents
 446          * allocated inodes. To expand from 16-bit holemask granularity to
 447          * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
 448          * bitmap for every holemask bit.
 449          */
 450         nextbit = xfs_next_bit(&allocbitmap, 1, 0);
 451         while (nextbit != -1) {
 452                 ASSERT(nextbit < (sizeof(rec->ir_holemask) * NBBY));
 453
 454                 bitmap |= (inodespbit <<
 455                            (nextbit * XFS_INODES_PER_HOLEMASK_BIT));
 456
 457                 nextbit = xfs_next_bit(&allocbitmap, 1, nextbit + 1);
 458         }
 459
 460         return bitmap;
 461 }
 462
 463 #if defined(DEBUG) || defined(XFS_WARN)
 464 /*
 465  * Verify that an in-core inode record has a valid inode count.
 466  */
 467 int
 468 xfs_inobt_rec_check_count(
 469         struct xfs_mount                *mp,
 470         struct xfs_inobt_rec_incore     *rec)
 471 {
 472         int                             inocount = 0;
 473         int                             nextbit = 0;
 474         uint64_t                        allocbmap;
 475         int                             wordsz;
 476
 477         wordsz = sizeof(allocbmap) / sizeof(unsigned int);
 478         allocbmap = xfs_inobt_irec_to_allocmask(rec);
 479
 480         nextbit = xfs_next_bit((uint *) &allocbmap, wordsz, nextbit);
 481         while (nextbit != -1) {
 482                 inocount++;
 483                 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz,
 484                                        nextbit + 1);
 485         }
 486
 487         if (inocount != rec->ir_count)
 488                 return -EFSCORRUPTED;
 489
 490         return 0;
 491 }
 492 #endif  /* DEBUG */