fs/xfs/xfs_dir2_leaf.c

   1 /*
   2  * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
   3  * Copyright (c) 2013 Red Hat, Inc.
   4  * All Rights Reserved.
   5  *
   6  * This program is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU General Public License as
   8  * published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it would be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License
  16  * along with this program; if not, write the Free Software Foundation,
  17  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  18  */
  19 #include "xfs.h"
  20 #include "xfs_fs.h"
  21 #include "xfs_format.h"
  22 #include "xfs_log_format.h"
  23 #include "xfs_trans_resv.h"
  24 #include "xfs_sb.h"
  25 #include "xfs_ag.h"
  26 #include "xfs_mount.h"
  27 #include "xfs_da_format.h"
  28 #include "xfs_da_btree.h"
  29 #include "xfs_inode.h"
  30 #include "xfs_bmap.h"
  31 #include "xfs_dir2.h"
  32 #include "xfs_dir2_priv.h"
  33 #include "xfs_error.h"
  34 #include "xfs_trace.h"
  35 #include "xfs_trans.h"
  36 #include "xfs_buf_item.h"
  37 #include "xfs_cksum.h"
  38
  39 /*
  40  * Local function declarations.
  41  */
  42 static int xfs_dir2_leaf_lookup_int(xfs_da_args_t *args, struct xfs_buf **lbpp,
  43                                     int *indexp, struct xfs_buf **dbpp);
  44 static void xfs_dir3_leaf_log_bests(struct xfs_da_args *args,
  45                                     struct xfs_buf *bp, int first, int last);
  46 static void xfs_dir3_leaf_log_tail(struct xfs_da_args *args,
  47                                    struct xfs_buf *bp);
  48
  49 /*
  50  * Check the internal consistency of a leaf1 block.
  51  * Pop an assert if something is wrong.
  52  */
  53 #ifdef DEBUG
  54 #define xfs_dir3_leaf_check(dp, bp) \
  55 do { \
  56         if (!xfs_dir3_leaf1_check((dp), (bp))) \
  57                 ASSERT(0); \
  58 } while (0);
  59
  60 STATIC bool
  61 xfs_dir3_leaf1_check(
  62         struct xfs_inode        *dp,
  63         struct xfs_buf          *bp)
  64 {
  65         struct xfs_dir2_leaf    *leaf = bp->b_addr;
  66         struct xfs_dir3_icleaf_hdr leafhdr;
  67
  68         dp->d_ops->leaf_hdr_from_disk(&leafhdr, leaf);
  69
  70         if (leafhdr.magic == XFS_DIR3_LEAF1_MAGIC) {
  71                 struct xfs_dir3_leaf_hdr *leaf3 = bp->b_addr;
  72                 if (be64_to_cpu(leaf3->info.blkno) != bp->b_bn)
  73                         return false;
  74         } else if (leafhdr.magic != XFS_DIR2_LEAF1_MAGIC)
  75                 return false;
  76
  77         return xfs_dir3_leaf_check_int(dp->i_mount, dp, &leafhdr, leaf);
  78 }
  79 #else
  80 #define xfs_dir3_leaf_check(dp, bp)
  81 #endif
  82
  83 bool
  84 xfs_dir3_leaf_check_int(
  85         struct xfs_mount        *mp,
  86         struct xfs_inode        *dp,
  87         struct xfs_dir3_icleaf_hdr *hdr,
  88         struct xfs_dir2_leaf    *leaf)
  89 {
  90         struct xfs_dir2_leaf_entry *ents;
  91         xfs_dir2_leaf_tail_t    *ltp;
  92         int                     stale;
  93         int                     i;
  94         const struct xfs_dir_ops *ops;
  95         struct xfs_dir3_icleaf_hdr leafhdr;
  96         struct xfs_da_geometry  *geo = mp->m_dir_geo;
  97
  98         /*
  99          * we can be passed a null dp here from a verifier, so we need to go the
 100          * hard way to get them.
 101          */
 102         ops = xfs_dir_get_ops(mp, dp);
 103
 104         if (!hdr) {
 105                 ops->leaf_hdr_from_disk(&leafhdr, leaf);
 106                 hdr = &leafhdr;
 107         }
 108
 109         ents = ops->leaf_ents_p(leaf);
 110         ltp = xfs_dir2_leaf_tail_p(geo, leaf);
 111
 112         /*
 113          * XXX (dgc): This value is not restrictive enough.
 114          * Should factor in the size of the bests table as well.
 115          * We can deduce a value for that from di_size.
 116          */
 117         if (hdr->count > ops->leaf_max_ents(geo))
 118                 return false;
 119
 120         /* Leaves and bests don't overlap in leaf format. */
 121         if ((hdr->magic == XFS_DIR2_LEAF1_MAGIC ||
 122              hdr->magic == XFS_DIR3_LEAF1_MAGIC) &&
 123             (char *)&ents[hdr->count] > (char *)xfs_dir2_leaf_bests_p(ltp))
 124                 return false;
 125
 126         /* Check hash value order, count stale entries.  */
 127         for (i = stale = 0; i < hdr->count; i++) {
 128                 if (i + 1 < hdr->count) {
 129                         if (be32_to_cpu(ents[i].hashval) >
 130                                         be32_to_cpu(ents[i + 1].hashval))
 131                                 return false;
 132                 }
 133                 if (ents[i].address == cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 134                         stale++;
 135         }
 136         if (hdr->stale != stale)
 137                 return false;
 138         return true;
 139 }
 140
 141 /*
 142  * We verify the magic numbers before decoding the leaf header so that on debug
 143  * kernels we don't get assertion failures in xfs_dir3_leaf_hdr_from_disk() due
 144  * to incorrect magic numbers.
 145  */
 146 static bool
 147 xfs_dir3_leaf_verify(
 148         struct xfs_buf          *bp,
 149         __uint16_t              magic)
 150 {
 151         struct xfs_mount        *mp = bp->b_target->bt_mount;
 152         struct xfs_dir2_leaf    *leaf = bp->b_addr;
 153
 154         ASSERT(magic == XFS_DIR2_LEAF1_MAGIC || magic == XFS_DIR2_LEAFN_MAGIC);
 155
 156         if (xfs_sb_version_hascrc(&mp->m_sb)) {
 157                 struct xfs_dir3_leaf_hdr *leaf3 = bp->b_addr;
 158                 __uint16_t              magic3;
 159
 160                 magic3 = (magic == XFS_DIR2_LEAF1_MAGIC) ? XFS_DIR3_LEAF1_MAGIC
 161                                                          : XFS_DIR3_LEAFN_MAGIC;
 162
 163                 if (leaf3->info.hdr.magic != cpu_to_be16(magic3))
 164                         return false;
 165                 if (!uuid_equal(&leaf3->info.uuid, &mp->m_sb.sb_uuid))
 166                         return false;
 167                 if (be64_to_cpu(leaf3->info.blkno) != bp->b_bn)
 168                         return false;
 169         } else {
 170                 if (leaf->hdr.info.magic != cpu_to_be16(magic))
 171                         return false;
 172         }
 173
 174         return xfs_dir3_leaf_check_int(mp, NULL, NULL, leaf);
 175 }
 176
 177 static void
 178 __read_verify(
 179         struct xfs_buf  *bp,
 180         __uint16_t      magic)
 181 {
 182         struct xfs_mount        *mp = bp->b_target->bt_mount;
 183
 184         if (xfs_sb_version_hascrc(&mp->m_sb) &&
 185              !xfs_buf_verify_cksum(bp, XFS_DIR3_LEAF_CRC_OFF))
 186                 xfs_buf_ioerror(bp, EFSBADCRC);
 187         else if (!xfs_dir3_leaf_verify(bp, magic))
 188                 xfs_buf_ioerror(bp, EFSCORRUPTED);
 189
 190         if (bp->b_error)
 191                 xfs_verifier_error(bp);
 192 }
 193
 194 static void
 195 __write_verify(
 196         struct xfs_buf  *bp,
 197         __uint16_t      magic)
 198 {
 199         struct xfs_mount        *mp = bp->b_target->bt_mount;
 200         struct xfs_buf_log_item *bip = bp->b_fspriv;
 201         struct xfs_dir3_leaf_hdr *hdr3 = bp->b_addr;
 202
 203         if (!xfs_dir3_leaf_verify(bp, magic)) {
 204                 xfs_buf_ioerror(bp, EFSCORRUPTED);
 205                 xfs_verifier_error(bp);
 206                 return;
 207         }
 208
 209         if (!xfs_sb_version_hascrc(&mp->m_sb))
 210                 return;
 211
 212         if (bip)
 213                 hdr3->info.lsn = cpu_to_be64(bip->bli_item.li_lsn);
 214
 215         xfs_buf_update_cksum(bp, XFS_DIR3_LEAF_CRC_OFF);
 216 }
 217
 218 static void
 219 xfs_dir3_leaf1_read_verify(
 220         struct xfs_buf  *bp)
 221 {
 222         __read_verify(bp, XFS_DIR2_LEAF1_MAGIC);
 223 }
 224
 225 static void
 226 xfs_dir3_leaf1_write_verify(
 227         struct xfs_buf  *bp)
 228 {
 229         __write_verify(bp, XFS_DIR2_LEAF1_MAGIC);
 230 }
 231
 232 static void
 233 xfs_dir3_leafn_read_verify(
 234         struct xfs_buf  *bp)
 235 {
 236         __read_verify(bp, XFS_DIR2_LEAFN_MAGIC);
 237 }
 238
 239 static void
 240 xfs_dir3_leafn_write_verify(
 241         struct xfs_buf  *bp)
 242 {
 243         __write_verify(bp, XFS_DIR2_LEAFN_MAGIC);
 244 }
 245
 246 const struct xfs_buf_ops xfs_dir3_leaf1_buf_ops = {
 247         .verify_read = xfs_dir3_leaf1_read_verify,
 248         .verify_write = xfs_dir3_leaf1_write_verify,
 249 };
 250
 251 const struct xfs_buf_ops xfs_dir3_leafn_buf_ops = {
 252         .verify_read = xfs_dir3_leafn_read_verify,
 253         .verify_write = xfs_dir3_leafn_write_verify,
 254 };
 255
 256 static int
 257 xfs_dir3_leaf_read(
 258         struct xfs_trans        *tp,
 259         struct xfs_inode        *dp,
 260         xfs_dablk_t             fbno,
 261         xfs_daddr_t             mappedbno,
 262         struct xfs_buf          **bpp)
 263 {
 264         int                     err;
 265
 266         err = xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
 267                                 XFS_DATA_FORK, &xfs_dir3_leaf1_buf_ops);
 268         if (!err && tp)
 269                 xfs_trans_buf_set_type(tp, *bpp, XFS_BLFT_DIR_LEAF1_BUF);
 270         return err;
 271 }
 272
 273 int
 274 xfs_dir3_leafn_read(
 275         struct xfs_trans        *tp,
 276         struct xfs_inode        *dp,
 277         xfs_dablk_t             fbno,
 278         xfs_daddr_t             mappedbno,
 279         struct xfs_buf          **bpp)
 280 {
 281         int                     err;
 282
 283         err = xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
 284                                 XFS_DATA_FORK, &xfs_dir3_leafn_buf_ops);
 285         if (!err && tp)
 286                 xfs_trans_buf_set_type(tp, *bpp, XFS_BLFT_DIR_LEAFN_BUF);
 287         return err;
 288 }
 289
 290 /*
 291  * Initialize a new leaf block, leaf1 or leafn magic accepted.
 292  */
 293 static void
 294 xfs_dir3_leaf_init(
 295         struct xfs_mount        *mp,
 296         struct xfs_trans        *tp,
 297         struct xfs_buf          *bp,
 298         xfs_ino_t               owner,
 299         __uint16_t              type)
 300 {
 301         struct xfs_dir2_leaf    *leaf = bp->b_addr;
 302
 303         ASSERT(type == XFS_DIR2_LEAF1_MAGIC || type == XFS_DIR2_LEAFN_MAGIC);
 304
 305         if (xfs_sb_version_hascrc(&mp->m_sb)) {
 306                 struct xfs_dir3_leaf_hdr *leaf3 = bp->b_addr;
 307
 308                 memset(leaf3, 0, sizeof(*leaf3));
 309
 310                 leaf3->info.hdr.magic = (type == XFS_DIR2_LEAF1_MAGIC)
 311                                          ? cpu_to_be16(XFS_DIR3_LEAF1_MAGIC)
 312                                          : cpu_to_be16(XFS_DIR3_LEAFN_MAGIC);
 313                 leaf3->info.blkno = cpu_to_be64(bp->b_bn);
 314                 leaf3->info.owner = cpu_to_be64(owner);
 315                 uuid_copy(&leaf3->info.uuid, &mp->m_sb.sb_uuid);
 316         } else {
 317                 memset(leaf, 0, sizeof(*leaf));
 318                 leaf->hdr.info.magic = cpu_to_be16(type);
 319         }
 320
 321         /*
 322          * If it's a leaf-format directory initialize the tail.
 323          * Caller is responsible for initialising the bests table.
 324          */
 325         if (type == XFS_DIR2_LEAF1_MAGIC) {
 326                 struct xfs_dir2_leaf_tail *ltp;
 327
 328                 ltp = xfs_dir2_leaf_tail_p(mp->m_dir_geo, leaf);
 329                 ltp->bestcount = 0;
 330                 bp->b_ops = &xfs_dir3_leaf1_buf_ops;
 331                 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_DIR_LEAF1_BUF);
 332         } else {
 333                 bp->b_ops = &xfs_dir3_leafn_buf_ops;
 334                 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_DIR_LEAFN_BUF);
 335         }
 336 }
 337
 338 int
 339 xfs_dir3_leaf_get_buf(
 340         xfs_da_args_t           *args,
 341         xfs_dir2_db_t           bno,
 342         struct xfs_buf          **bpp,
 343         __uint16_t              magic)
 344 {
 345         struct xfs_inode        *dp = args->dp;
 346         struct xfs_trans        *tp = args->trans;
 347         struct xfs_mount        *mp = dp->i_mount;
 348         struct xfs_buf          *bp;
 349         int                     error;
 350
 351         ASSERT(magic == XFS_DIR2_LEAF1_MAGIC || magic == XFS_DIR2_LEAFN_MAGIC);
 352         ASSERT(bno >= xfs_dir2_byte_to_db(args->geo, XFS_DIR2_LEAF_OFFSET) &&
 353                bno < xfs_dir2_byte_to_db(args->geo, XFS_DIR2_FREE_OFFSET));
 354
 355         error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(args->geo, bno),
 356                                -1, &bp, XFS_DATA_FORK);
 357         if (error)
 358                 return error;
 359
 360         xfs_dir3_leaf_init(mp, tp, bp, dp->i_ino, magic);
 361         xfs_dir3_leaf_log_header(args, bp);
 362         if (magic == XFS_DIR2_LEAF1_MAGIC)
 363                 xfs_dir3_leaf_log_tail(args, bp);
 364         *bpp = bp;
 365         return 0;
 366 }
 367
 368 /*
 369  * Convert a block form directory to a leaf form directory.
 370  */
 371 int                                             /* error */
 372 xfs_dir2_block_to_leaf(
 373         xfs_da_args_t           *args,          /* operation arguments */
 374         struct xfs_buf          *dbp)           /* input block's buffer */
 375 {
 376         __be16                  *bestsp;        /* leaf's bestsp entries */
 377         xfs_dablk_t             blkno;          /* leaf block's bno */
 378         xfs_dir2_data_hdr_t     *hdr;           /* block header */
 379         xfs_dir2_leaf_entry_t   *blp;           /* block's leaf entries */
 380         xfs_dir2_block_tail_t   *btp;           /* block's tail */
 381         xfs_inode_t             *dp;            /* incore directory inode */
 382         int                     error;          /* error return code */
 383         struct xfs_buf          *lbp;           /* leaf block's buffer */
 384         xfs_dir2_db_t           ldb;            /* leaf block's bno */
 385         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
 386         xfs_dir2_leaf_tail_t    *ltp;           /* leaf's tail */
 387         xfs_mount_t             *mp;            /* filesystem mount point */
 388         int                     needlog;        /* need to log block header */
 389         int                     needscan;       /* need to rescan bestfree */
 390         xfs_trans_t             *tp;            /* transaction pointer */
 391         struct xfs_dir2_data_free *bf;
 392         struct xfs_dir2_leaf_entry *ents;
 393         struct xfs_dir3_icleaf_hdr leafhdr;
 394
 395         trace_xfs_dir2_block_to_leaf(args);
 396
 397         dp = args->dp;
 398         mp = dp->i_mount;
 399         tp = args->trans;
 400         /*
 401          * Add the leaf block to the inode.
 402          * This interface will only put blocks in the leaf/node range.
 403          * Since that's empty now, we'll get the root (block 0 in range).
 404          */
 405         if ((error = xfs_da_grow_inode(args, &blkno))) {
 406                 return error;
 407         }
 408         ldb = xfs_dir2_da_to_db(args->geo, blkno);
 409         ASSERT(ldb == xfs_dir2_byte_to_db(args->geo, XFS_DIR2_LEAF_OFFSET));
 410         /*
 411          * Initialize the leaf block, get a buffer for it.
 412          */
 413         error = xfs_dir3_leaf_get_buf(args, ldb, &lbp, XFS_DIR2_LEAF1_MAGIC);
 414         if (error)
 415                 return error;
 416
 417         leaf = lbp->b_addr;
 418         hdr = dbp->b_addr;
 419         xfs_dir3_data_check(dp, dbp);
 420         btp = xfs_dir2_block_tail_p(args->geo, hdr);
 421         blp = xfs_dir2_block_leaf_p(btp);
 422         bf = dp->d_ops->data_bestfree_p(hdr);
 423         ents = dp->d_ops->leaf_ents_p(leaf);
 424
 425         /*
 426          * Set the counts in the leaf header.
 427          */
 428         dp->d_ops->leaf_hdr_from_disk(&leafhdr, leaf);
 429         leafhdr.count = be32_to_cpu(btp->count);
 430         leafhdr.stale = be32_to_cpu(btp->stale);
 431         dp->d_ops->leaf_hdr_to_disk(leaf, &leafhdr);
 432         xfs_dir3_leaf_log_header(args, lbp);
 433
 434         /*
 435          * Could compact these but I think we always do the conversion
 436          * after squeezing out stale entries.
 437          */
 438         memcpy(ents, blp, be32_to_cpu(btp->count) * sizeof(xfs_dir2_leaf_entry_t));
 439         xfs_dir3_leaf_log_ents(args, lbp, 0, leafhdr.count - 1);
 440         needscan = 0;
 441         needlog = 1;
 442         /*
 443          * Make the space formerly occupied by the leaf entries and block
 444          * tail be free.
 445          */
 446         xfs_dir2_data_make_free(args, dbp,
 447                 (xfs_dir2_data_aoff_t)((char *)blp - (char *)hdr),
 448                 (xfs_dir2_data_aoff_t)((char *)hdr + args->geo->blksize -
 449                                        (char *)blp),
 450                 &needlog, &needscan);
 451         /*
 452          * Fix up the block header, make it a data block.
 453          */
 454         dbp->b_ops = &xfs_dir3_data_buf_ops;
 455         xfs_trans_buf_set_type(tp, dbp, XFS_BLFT_DIR_DATA_BUF);
 456         if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC))
 457                 hdr->magic = cpu_to_be32(XFS_DIR2_DATA_MAGIC);
 458         else
 459                 hdr->magic = cpu_to_be32(XFS_DIR3_DATA_MAGIC);
 460
 461         if (needscan)
 462                 xfs_dir2_data_freescan(dp, hdr, &needlog);
 463         /*
 464          * Set up leaf tail and bests table.
 465          */
 466         ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
 467         ltp->bestcount = cpu_to_be32(1);
 468         bestsp = xfs_dir2_leaf_bests_p(ltp);
 469         bestsp[0] =  bf[0].length;
 470         /*
 471          * Log the data header and leaf bests table.
 472          */
 473         if (needlog)
 474                 xfs_dir2_data_log_header(args, dbp);
 475         xfs_dir3_leaf_check(dp, lbp);
 476         xfs_dir3_data_check(dp, dbp);
 477         xfs_dir3_leaf_log_bests(args, lbp, 0, 0);
 478         return 0;
 479 }
 480
 481 STATIC void
 482 xfs_dir3_leaf_find_stale(
 483         struct xfs_dir3_icleaf_hdr *leafhdr,
 484         struct xfs_dir2_leaf_entry *ents,
 485         int                     index,
 486         int                     *lowstale,
 487         int                     *highstale)
 488 {
 489         /*
 490          * Find the first stale entry before our index, if any.
 491          */
 492         for (*lowstale = index - 1; *lowstale >= 0; --*lowstale) {
 493                 if (ents[*lowstale].address ==
 494                     cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 495                         break;
 496         }
 497
 498         /*
 499          * Find the first stale entry at or after our index, if any.
 500          * Stop if the result would require moving more entries than using
 501          * lowstale.
 502          */
 503         for (*highstale = index; *highstale < leafhdr->count; ++*highstale) {
 504                 if (ents[*highstale].address ==
 505                     cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 506                         break;
 507                 if (*lowstale >= 0 && index - *lowstale <= *highstale - index)
 508                         break;
 509         }
 510 }
 511
 512 struct xfs_dir2_leaf_entry *
 513 xfs_dir3_leaf_find_entry(
 514         struct xfs_dir3_icleaf_hdr *leafhdr,
 515         struct xfs_dir2_leaf_entry *ents,
 516         int                     index,          /* leaf table position */
 517         int                     compact,        /* need to compact leaves */
 518         int                     lowstale,       /* index of prev stale leaf */
 519         int                     highstale,      /* index of next stale leaf */
 520         int                     *lfloglow,      /* low leaf logging index */
 521         int                     *lfloghigh)     /* high leaf logging index */
 522 {
 523         if (!leafhdr->stale) {
 524                 xfs_dir2_leaf_entry_t   *lep;   /* leaf entry table pointer */
 525
 526                 /*
 527                  * Now we need to make room to insert the leaf entry.
 528                  *
 529                  * If there are no stale entries, just insert a hole at index.
 530                  */
 531                 lep = &ents[index];
 532                 if (index < leafhdr->count)
 533                         memmove(lep + 1, lep,
 534                                 (leafhdr->count - index) * sizeof(*lep));
 535
 536                 /*
 537                  * Record low and high logging indices for the leaf.
 538                  */
 539                 *lfloglow = index;
 540                 *lfloghigh = leafhdr->count++;
 541                 return lep;
 542         }
 543
 544         /*
 545          * There are stale entries.
 546          *
 547          * We will use one of them for the new entry.  It's probably not at
 548          * the right location, so we'll have to shift some up or down first.
 549          *
 550          * If we didn't compact before, we need to find the nearest stale
 551          * entries before and after our insertion point.
 552          */
 553         if (compact == 0)
 554                 xfs_dir3_leaf_find_stale(leafhdr, ents, index,
 555                                          &lowstale, &highstale);
 556
 557         /*
 558          * If the low one is better, use it.
 559          */
 560         if (lowstale >= 0 &&
 561             (highstale == leafhdr->count ||
 562              index - lowstale - 1 < highstale - index)) {
 563                 ASSERT(index - lowstale - 1 >= 0);
 564                 ASSERT(ents[lowstale].address ==
 565                        cpu_to_be32(XFS_DIR2_NULL_DATAPTR));
 566
 567                 /*
 568                  * Copy entries up to cover the stale entry and make room
 569                  * for the new entry.
 570                  */
 571                 if (index - lowstale - 1 > 0) {
 572                         memmove(&ents[lowstale], &ents[lowstale + 1],
 573                                 (index - lowstale - 1) *
 574                                         sizeof(xfs_dir2_leaf_entry_t));
 575                 }
 576                 *lfloglow = MIN(lowstale, *lfloglow);
 577                 *lfloghigh = MAX(index - 1, *lfloghigh);
 578                 leafhdr->stale--;
 579                 return &ents[index - 1];
 580         }
 581
 582         /*
 583          * The high one is better, so use that one.
 584          */
 585         ASSERT(highstale - index >= 0);
 586         ASSERT(ents[highstale].address == cpu_to_be32(XFS_DIR2_NULL_DATAPTR));
 587
 588         /*
 589          * Copy entries down to cover the stale entry and make room for the
 590          * new entry.
 591          */
 592         if (highstale - index > 0) {
 593                 memmove(&ents[index + 1], &ents[index],
 594                         (highstale - index) * sizeof(xfs_dir2_leaf_entry_t));
 595         }
 596         *lfloglow = MIN(index, *lfloglow);
 597         *lfloghigh = MAX(highstale, *lfloghigh);
 598         leafhdr->stale--;
 599         return &ents[index];
 600 }
 601
 602 /*
 603  * Add an entry to a leaf form directory.
 604  */
 605 int                                             /* error */
 606 xfs_dir2_leaf_addname(
 607         xfs_da_args_t           *args)          /* operation arguments */
 608 {
 609         __be16                  *bestsp;        /* freespace table in leaf */
 610         int                     compact;        /* need to compact leaves */
 611         xfs_dir2_data_hdr_t     *hdr;           /* data block header */
 612         struct xfs_buf          *dbp;           /* data block buffer */
 613         xfs_dir2_data_entry_t   *dep;           /* data block entry */
 614         xfs_inode_t             *dp;            /* incore directory inode */
 615         xfs_dir2_data_unused_t  *dup;           /* data unused entry */
 616         int                     error;          /* error return value */
 617         int                     grown;          /* allocated new data block */
 618         int                     highstale;      /* index of next stale leaf */
 619         int                     i;              /* temporary, index */
 620         int                     index;          /* leaf table position */
 621         struct xfs_buf          *lbp;           /* leaf's buffer */
 622         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
 623         int                     length;         /* length of new entry */
 624         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry table pointer */
 625         int                     lfloglow;       /* low leaf logging index */
 626         int                     lfloghigh;      /* high leaf logging index */
 627         int                     lowstale;       /* index of prev stale leaf */
 628         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail pointer */
 629         xfs_mount_t             *mp;            /* filesystem mount point */
 630         int                     needbytes;      /* leaf block bytes needed */
 631         int                     needlog;        /* need to log data header */
 632         int                     needscan;       /* need to rescan data free */
 633         __be16                  *tagp;          /* end of data entry */
 634         xfs_trans_t             *tp;            /* transaction pointer */
 635         xfs_dir2_db_t           use_block;      /* data block number */
 636         struct xfs_dir2_data_free *bf;          /* bestfree table */
 637         struct xfs_dir2_leaf_entry *ents;
 638         struct xfs_dir3_icleaf_hdr leafhdr;
 639
 640         trace_xfs_dir2_leaf_addname(args);
 641
 642         dp = args->dp;
 643         tp = args->trans;
 644         mp = dp->i_mount;
 645
 646         error = xfs_dir3_leaf_read(tp, dp, args->geo->leafblk, -1, &lbp);
 647         if (error)
 648                 return error;
 649
 650         /*
 651          * Look up the entry by hash value and name.
 652          * We know it's not there, our caller has already done a lookup.
 653          * So the index is of the entry to insert in front of.
 654          * But if there are dup hash values the index is of the first of those.
 655          */
 656         index = xfs_dir2_leaf_search_hash(args, lbp);
 657         leaf = lbp->b_addr;
 658         ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
 659         ents = dp->d_ops->leaf_ents_p(leaf);
 660         dp->d_ops->leaf_hdr_from_disk(&leafhdr, leaf);
 661         bestsp = xfs_dir2_leaf_bests_p(ltp);
 662         length = dp->d_ops->data_entsize(args->namelen);
 663
 664         /*
 665          * See if there are any entries with the same hash value
 666          * and space in their block for the new entry.
 667          * This is good because it puts multiple same-hash value entries
 668          * in a data block, improving the lookup of those entries.
 669          */
 670         for (use_block = -1, lep = &ents[index];
 671              index < leafhdr.count && be32_to_cpu(lep->hashval) == args->hashval;
 672              index++, lep++) {
 673                 if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
 674                         continue;
 675                 i = xfs_dir2_dataptr_to_db(args->geo, be32_to_cpu(lep->address));
 676                 ASSERT(i < be32_to_cpu(ltp->bestcount));
 677                 ASSERT(bestsp[i] != cpu_to_be16(NULLDATAOFF));
 678                 if (be16_to_cpu(bestsp[i]) >= length) {
 679                         use_block = i;
 680                         break;
 681                 }
 682         }
 683         /*
 684          * Didn't find a block yet, linear search all the data blocks.
 685          */
 686         if (use_block == -1) {
 687                 for (i = 0; i < be32_to_cpu(ltp->bestcount); i++) {
 688                         /*
 689                          * Remember a block we see that's missing.
 690                          */
 691                         if (bestsp[i] == cpu_to_be16(NULLDATAOFF) &&
 692                             use_block == -1)
 693                                 use_block = i;
 694                         else if (be16_to_cpu(bestsp[i]) >= length) {
 695                                 use_block = i;
 696                                 break;
 697                         }
 698                 }
 699         }
 700         /*
 701          * How many bytes do we need in the leaf block?
 702          */
 703         needbytes = 0;
 704         if (!leafhdr.stale)
 705                 needbytes += sizeof(xfs_dir2_leaf_entry_t);
 706         if (use_block == -1)
 707                 needbytes += sizeof(xfs_dir2_data_off_t);
 708
 709         /*
 710          * Now kill use_block if it refers to a missing block, so we
 711          * can use it as an indication of allocation needed.
 712          */
 713         if (use_block != -1 && bestsp[use_block] == cpu_to_be16(NULLDATAOFF))
 714                 use_block = -1;
 715         /*
 716          * If we don't have enough free bytes but we can make enough
 717          * by compacting out stale entries, we'll do that.
 718          */
 719         if ((char *)bestsp - (char *)&ents[leafhdr.count] < needbytes &&
 720             leafhdr.stale > 1)
 721                 compact = 1;
 722
 723         /*
 724          * Otherwise if we don't have enough free bytes we need to
 725          * convert to node form.
 726          */
 727         else if ((char *)bestsp - (char *)&ents[leafhdr.count] < needbytes) {
 728                 /*
 729                  * Just checking or no space reservation, give up.
 730                  */
 731                 if ((args->op_flags & XFS_DA_OP_JUSTCHECK) ||
 732                                                         args->total == 0) {
 733                         xfs_trans_brelse(tp, lbp);
 734                         return XFS_ERROR(ENOSPC);
 735                 }
 736                 /*
 737                  * Convert to node form.
 738                  */
 739                 error = xfs_dir2_leaf_to_node(args, lbp);
 740                 if (error)
 741                         return error;
 742                 /*
 743                  * Then add the new entry.
 744                  */
 745                 return xfs_dir2_node_addname(args);
 746         }
 747         /*
 748          * Otherwise it will fit without compaction.
 749          */
 750         else
 751                 compact = 0;
 752         /*
 753          * If just checking, then it will fit unless we needed to allocate
 754          * a new data block.
 755          */
 756         if (args->op_flags & XFS_DA_OP_JUSTCHECK) {
 757                 xfs_trans_brelse(tp, lbp);
 758                 return use_block == -1 ? XFS_ERROR(ENOSPC) : 0;
 759         }
 760         /*
 761          * If no allocations are allowed, return now before we've
 762          * changed anything.
 763          */
 764         if (args->total == 0 && use_block == -1) {
 765                 xfs_trans_brelse(tp, lbp);
 766                 return XFS_ERROR(ENOSPC);
 767         }
 768         /*
 769          * Need to compact the leaf entries, removing stale ones.
 770          * Leave one stale entry behind - the one closest to our
 771          * insertion index - and we'll shift that one to our insertion
 772          * point later.
 773          */
 774         if (compact) {
 775                 xfs_dir3_leaf_compact_x1(&leafhdr, ents, &index, &lowstale,
 776                         &highstale, &lfloglow, &lfloghigh);
 777         }
 778         /*
 779          * There are stale entries, so we'll need log-low and log-high
 780          * impossibly bad values later.
 781          */
 782         else if (leafhdr.stale) {
 783                 lfloglow = leafhdr.count;
 784                 lfloghigh = -1;
 785         }
 786         /*
 787          * If there was no data block space found, we need to allocate
 788          * a new one.
 789          */
 790         if (use_block == -1) {
 791                 /*
 792                  * Add the new data block.
 793                  */
 794                 if ((error = xfs_dir2_grow_inode(args, XFS_DIR2_DATA_SPACE,
 795                                 &use_block))) {
 796                         xfs_trans_brelse(tp, lbp);
 797                         return error;
 798                 }
 799                 /*
 800                  * Initialize the block.
 801                  */
 802                 if ((error = xfs_dir3_data_init(args, use_block, &dbp))) {
 803                         xfs_trans_brelse(tp, lbp);
 804                         return error;
 805                 }
 806                 /*
 807                  * If we're adding a new data block on the end we need to
 808                  * extend the bests table.  Copy it up one entry.
 809                  */
 810                 if (use_block >= be32_to_cpu(ltp->bestcount)) {
 811                         bestsp--;
 812                         memmove(&bestsp[0], &bestsp[1],
 813                                 be32_to_cpu(ltp->bestcount) * sizeof(bestsp[0]));
 814                         be32_add_cpu(&ltp->bestcount, 1);
 815                         xfs_dir3_leaf_log_tail(args, lbp);
 816                         xfs_dir3_leaf_log_bests(args, lbp, 0,
 817                                                 be32_to_cpu(ltp->bestcount) - 1);
 818                 }
 819                 /*
 820                  * If we're filling in a previously empty block just log it.
 821                  */
 822                 else
 823                         xfs_dir3_leaf_log_bests(args, lbp, use_block, use_block);
 824                 hdr = dbp->b_addr;
 825                 bf = dp->d_ops->data_bestfree_p(hdr);
 826                 bestsp[use_block] = bf[0].length;
 827                 grown = 1;
 828         } else {
 829                 /*
 830                  * Already had space in some data block.
 831                  * Just read that one in.
 832                  */
 833                 error = xfs_dir3_data_read(tp, dp,
 834                                    xfs_dir2_db_to_da(args->geo, use_block),
 835                                    -1, &dbp);
 836                 if (error) {
 837                         xfs_trans_brelse(tp, lbp);
 838                         return error;
 839                 }
 840                 hdr = dbp->b_addr;
 841                 bf = dp->d_ops->data_bestfree_p(hdr);
 842                 grown = 0;
 843         }
 844         /*
 845          * Point to the biggest freespace in our data block.
 846          */
 847         dup = (xfs_dir2_data_unused_t *)
 848               ((char *)hdr + be16_to_cpu(bf[0].offset));
 849         ASSERT(be16_to_cpu(dup->length) >= length);
 850         needscan = needlog = 0;
 851         /*
 852          * Mark the initial part of our freespace in use for the new entry.
 853          */
 854         xfs_dir2_data_use_free(args, dbp, dup,
 855                 (xfs_dir2_data_aoff_t)((char *)dup - (char *)hdr), length,
 856                 &needlog, &needscan);
 857         /*
 858          * Initialize our new entry (at last).
 859          */
 860         dep = (xfs_dir2_data_entry_t *)dup;
 861         dep->inumber = cpu_to_be64(args->inumber);
 862         dep->namelen = args->namelen;
 863         memcpy(dep->name, args->name, dep->namelen);
 864         dp->d_ops->data_put_ftype(dep, args->filetype);
 865         tagp = dp->d_ops->data_entry_tag_p(dep);
 866         *tagp = cpu_to_be16((char *)dep - (char *)hdr);
 867         /*
 868          * Need to scan fix up the bestfree table.
 869          */
 870         if (needscan)
 871                 xfs_dir2_data_freescan(dp, hdr, &needlog);
 872         /*
 873          * Need to log the data block's header.
 874          */
 875         if (needlog)
 876                 xfs_dir2_data_log_header(args, dbp);
 877         xfs_dir2_data_log_entry(args, dbp, dep);
 878         /*
 879          * If the bests table needs to be changed, do it.
 880          * Log the change unless we've already done that.
 881          */
 882         if (be16_to_cpu(bestsp[use_block]) != be16_to_cpu(bf[0].length)) {
 883                 bestsp[use_block] = bf[0].length;
 884                 if (!grown)
 885                         xfs_dir3_leaf_log_bests(args, lbp, use_block, use_block);
 886         }
 887
 888         lep = xfs_dir3_leaf_find_entry(&leafhdr, ents, index, compact, lowstale,
 889                                        highstale, &lfloglow, &lfloghigh);
 890
 891         /*
 892          * Fill in the new leaf entry.
 893          */
 894         lep->hashval = cpu_to_be32(args->hashval);
 895         lep->address = cpu_to_be32(
 896                                 xfs_dir2_db_off_to_dataptr(args->geo, use_block,
 897                                 be16_to_cpu(*tagp)));
 898         /*
 899          * Log the leaf fields and give up the buffers.
 900          */
 901         dp->d_ops->leaf_hdr_to_disk(leaf, &leafhdr);
 902         xfs_dir3_leaf_log_header(args, lbp);
 903         xfs_dir3_leaf_log_ents(args, lbp, lfloglow, lfloghigh);
 904         xfs_dir3_leaf_check(dp, lbp);
 905         xfs_dir3_data_check(dp, dbp);
 906         return 0;
 907 }
 908
 909 /*
 910  * Compact out any stale entries in the leaf.
 911  * Log the header and changed leaf entries, if any.
 912  */
 913 void
 914 xfs_dir3_leaf_compact(
 915         xfs_da_args_t   *args,          /* operation arguments */
 916         struct xfs_dir3_icleaf_hdr *leafhdr,
 917         struct xfs_buf  *bp)            /* leaf buffer */
 918 {
 919         int             from;           /* source leaf index */
 920         xfs_dir2_leaf_t *leaf;          /* leaf structure */
 921         int             loglow;         /* first leaf entry to log */
 922         int             to;             /* target leaf index */
 923         struct xfs_dir2_leaf_entry *ents;
 924         struct xfs_inode *dp = args->dp;
 925
 926         leaf = bp->b_addr;
 927         if (!leafhdr->stale)
 928                 return;
 929
 930         /*
 931          * Compress out the stale entries in place.
 932          */
 933         ents = dp->d_ops->leaf_ents_p(leaf);
 934         for (from = to = 0, loglow = -1; from < leafhdr->count; from++) {
 935                 if (ents[from].address == cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 936                         continue;
 937                 /*
 938                  * Only actually copy the entries that are different.
 939                  */
 940                 if (from > to) {
 941                         if (loglow == -1)
 942                                 loglow = to;
 943                         ents[to] = ents[from];
 944                 }
 945                 to++;
 946         }
 947         /*
 948          * Update and log the header, log the leaf entries.
 949          */
 950         ASSERT(leafhdr->stale == from - to);
 951         leafhdr->count -= leafhdr->stale;
 952         leafhdr->stale = 0;
 953
 954         dp->d_ops->leaf_hdr_to_disk(leaf, leafhdr);
 955         xfs_dir3_leaf_log_header(args, bp);
 956         if (loglow != -1)
 957                 xfs_dir3_leaf_log_ents(args, bp, loglow, to - 1);
 958 }
 959
 960 /*
 961  * Compact the leaf entries, removing stale ones.
 962  * Leave one stale entry behind - the one closest to our
 963  * insertion index - and the caller will shift that one to our insertion
 964  * point later.
 965  * Return new insertion index, where the remaining stale entry is,
 966  * and leaf logging indices.
 967  */
 968 void
 969 xfs_dir3_leaf_compact_x1(
 970         struct xfs_dir3_icleaf_hdr *leafhdr,
 971         struct xfs_dir2_leaf_entry *ents,
 972         int             *indexp,        /* insertion index */
 973         int             *lowstalep,     /* out: stale entry before us */
 974         int             *highstalep,    /* out: stale entry after us */
 975         int             *lowlogp,       /* out: low log index */
 976         int             *highlogp)      /* out: high log index */
 977 {
 978         int             from;           /* source copy index */
 979         int             highstale;      /* stale entry at/after index */
 980         int             index;          /* insertion index */
 981         int             keepstale;      /* source index of kept stale */
 982         int             lowstale;       /* stale entry before index */
 983         int             newindex=0;     /* new insertion index */
 984         int             to;             /* destination copy index */
 985
 986         ASSERT(leafhdr->stale > 1);
 987         index = *indexp;
 988
 989         xfs_dir3_leaf_find_stale(leafhdr, ents, index, &lowstale, &highstale);
 990
 991         /*
 992          * Pick the better of lowstale and highstale.
 993          */
 994         if (lowstale >= 0 &&
 995             (highstale == leafhdr->count ||
 996              index - lowstale <= highstale - index))
 997                 keepstale = lowstale;
 998         else
 999                 keepstale = highstale;
1000         /*
1001          * Copy the entries in place, removing all the stale entries
1002          * except keepstale.
1003          */
1004         for (from = to = 0; from < leafhdr->count; from++) {
1005                 /*
1006                  * Notice the new value of index.
1007                  */
1008                 if (index == from)
1009                         newindex = to;
1010                 if (from != keepstale &&
1011                     ents[from].address == cpu_to_be32(XFS_DIR2_NULL_DATAPTR)) {
1012                         if (from == to)
1013                                 *lowlogp = to;
1014                         continue;
1015                 }
1016                 /*
1017                  * Record the new keepstale value for the insertion.
1018                  */
1019                 if (from == keepstale)
1020                         lowstale = highstale = to;
1021                 /*
1022                  * Copy only the entries that have moved.
1023                  */
1024                 if (from > to)
1025                         ents[to] = ents[from];
1026                 to++;
1027         }
1028         ASSERT(from > to);
1029         /*
1030          * If the insertion point was past the last entry,
1031          * set the new insertion point accordingly.
1032          */
1033         if (index == from)
1034                 newindex = to;
1035         *indexp = newindex;
1036         /*
1037          * Adjust the leaf header values.
1038          */
1039         leafhdr->count -= from - to;
1040         leafhdr->stale = 1;
1041         /*
1042          * Remember the low/high stale value only in the "right"
1043          * direction.
1044          */
1045         if (lowstale >= newindex)
1046                 lowstale = -1;
1047         else
1048                 highstale = leafhdr->count;
1049         *highlogp = leafhdr->count - 1;
1050         *lowstalep = lowstale;
1051         *highstalep = highstale;
1052 }
1053
1054 /*
1055  * Log the bests entries indicated from a leaf1 block.
1056  */
1057 static void
1058 xfs_dir3_leaf_log_bests(
1059         struct xfs_da_args      *args,
1060         struct xfs_buf          *bp,            /* leaf buffer */
1061         int                     first,          /* first entry to log */
1062         int                     last)           /* last entry to log */
1063 {
1064         __be16                  *firstb;        /* pointer to first entry */
1065         __be16                  *lastb;         /* pointer to last entry */
1066         struct xfs_dir2_leaf    *leaf = bp->b_addr;
1067         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1068
1069         ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC) ||
1070                leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAF1_MAGIC));
1071
1072         ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
1073         firstb = xfs_dir2_leaf_bests_p(ltp) + first;
1074         lastb = xfs_dir2_leaf_bests_p(ltp) + last;
1075         xfs_trans_log_buf(args->trans, bp,
1076                 (uint)((char *)firstb - (char *)leaf),
1077                 (uint)((char *)lastb - (char *)leaf + sizeof(*lastb) - 1));
1078 }
1079
1080 /*
1081  * Log the leaf entries indicated from a leaf1 or leafn block.
1082  */
1083 void
1084 xfs_dir3_leaf_log_ents(
1085         struct xfs_da_args      *args,
1086         struct xfs_buf          *bp,
1087         int                     first,
1088         int                     last)
1089 {
1090         xfs_dir2_leaf_entry_t   *firstlep;      /* pointer to first entry */
1091         xfs_dir2_leaf_entry_t   *lastlep;       /* pointer to last entry */
1092         struct xfs_dir2_leaf    *leaf = bp->b_addr;
1093         struct xfs_dir2_leaf_entry *ents;
1094
1095         ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC) ||
1096                leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAF1_MAGIC) ||
1097                leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC) ||
1098                leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC));
1099
1100         ents = args->dp->d_ops->leaf_ents_p(leaf);
1101         firstlep = &ents[first];
1102         lastlep = &ents[last];
1103         xfs_trans_log_buf(args->trans, bp,
1104                 (uint)((char *)firstlep - (char *)leaf),
1105                 (uint)((char *)lastlep - (char *)leaf + sizeof(*lastlep) - 1));
1106 }
1107
1108 /*
1109  * Log the header of the leaf1 or leafn block.
1110  */
1111 void
1112 xfs_dir3_leaf_log_header(
1113         struct xfs_da_args      *args,
1114         struct xfs_buf          *bp)
1115 {
1116         struct xfs_dir2_leaf    *leaf = bp->b_addr;
1117
1118         ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC) ||
1119                leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAF1_MAGIC) ||
1120                leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC) ||
1121                leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC));
1122
1123         xfs_trans_log_buf(args->trans, bp,
1124                           (uint)((char *)&leaf->hdr - (char *)leaf),
1125                           args->dp->d_ops->leaf_hdr_size - 1);
1126 }
1127
1128 /*
1129  * Log the tail of the leaf1 block.
1130  */
1131 STATIC void
1132 xfs_dir3_leaf_log_tail(
1133         struct xfs_da_args      *args,
1134         struct xfs_buf          *bp)
1135 {
1136         struct xfs_dir2_leaf    *leaf = bp->b_addr;
1137         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1138
1139         ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC) ||
1140                leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAF1_MAGIC) ||
1141                leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC) ||
1142                leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC));
1143
1144         ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
1145         xfs_trans_log_buf(args->trans, bp, (uint)((char *)ltp - (char *)leaf),
1146                 (uint)(args->geo->blksize - 1));
1147 }
1148
1149 /*
1150  * Look up the entry referred to by args in the leaf format directory.
1151  * Most of the work is done by the xfs_dir2_leaf_lookup_int routine which
1152  * is also used by the node-format code.
1153  */
1154 int
1155 xfs_dir2_leaf_lookup(
1156         xfs_da_args_t           *args)          /* operation arguments */
1157 {
1158         struct xfs_buf          *dbp;           /* data block buffer */
1159         xfs_dir2_data_entry_t   *dep;           /* data block entry */
1160         xfs_inode_t             *dp;            /* incore directory inode */
1161         int                     error;          /* error return code */
1162         int                     index;          /* found entry index */
1163         struct xfs_buf          *lbp;           /* leaf buffer */
1164         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1165         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1166         xfs_trans_t             *tp;            /* transaction pointer */
1167         struct xfs_dir2_leaf_entry *ents;
1168
1169         trace_xfs_dir2_leaf_lookup(args);
1170
1171         /*
1172          * Look up name in the leaf block, returning both buffers and index.
1173          */
1174         if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1175                 return error;
1176         }
1177         tp = args->trans;
1178         dp = args->dp;
1179         xfs_dir3_leaf_check(dp, lbp);
1180         leaf = lbp->b_addr;
1181         ents = dp->d_ops->leaf_ents_p(leaf);
1182         /*
1183          * Get to the leaf entry and contained data entry address.
1184          */
1185         lep = &ents[index];
1186
1187         /*
1188          * Point to the data entry.
1189          */
1190         dep = (xfs_dir2_data_entry_t *)
1191               ((char *)dbp->b_addr +
1192                xfs_dir2_dataptr_to_off(args->geo, be32_to_cpu(lep->address)));
1193         /*
1194          * Return the found inode number & CI name if appropriate
1195          */
1196         args->inumber = be64_to_cpu(dep->inumber);
1197         args->filetype = dp->d_ops->data_get_ftype(dep);
1198         error = xfs_dir_cilookup_result(args, dep->name, dep->namelen);
1199         xfs_trans_brelse(tp, dbp);
1200         xfs_trans_brelse(tp, lbp);
1201         return XFS_ERROR(error);
1202 }
1203
1204 /*
1205  * Look up name/hash in the leaf block.
1206  * Fill in indexp with the found index, and dbpp with the data buffer.
1207  * If not found dbpp will be NULL, and ENOENT comes back.
1208  * lbpp will always be filled in with the leaf buffer unless there's an error.
1209  */
1210 static int                                      /* error */
1211 xfs_dir2_leaf_lookup_int(
1212         xfs_da_args_t           *args,          /* operation arguments */
1213         struct xfs_buf          **lbpp,         /* out: leaf buffer */
1214         int                     *indexp,        /* out: index in leaf block */
1215         struct xfs_buf          **dbpp)         /* out: data buffer */
1216 {
1217         xfs_dir2_db_t           curdb = -1;     /* current data block number */
1218         struct xfs_buf          *dbp = NULL;    /* data buffer */
1219         xfs_dir2_data_entry_t   *dep;           /* data entry */
1220         xfs_inode_t             *dp;            /* incore directory inode */
1221         int                     error;          /* error return code */
1222         int                     index;          /* index in leaf block */
1223         struct xfs_buf          *lbp;           /* leaf buffer */
1224         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1225         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1226         xfs_mount_t             *mp;            /* filesystem mount point */
1227         xfs_dir2_db_t           newdb;          /* new data block number */
1228         xfs_trans_t             *tp;            /* transaction pointer */
1229         xfs_dir2_db_t           cidb = -1;      /* case match data block no. */
1230         enum xfs_dacmp          cmp;            /* name compare result */
1231         struct xfs_dir2_leaf_entry *ents;
1232         struct xfs_dir3_icleaf_hdr leafhdr;
1233
1234         dp = args->dp;
1235         tp = args->trans;
1236         mp = dp->i_mount;
1237
1238         error = xfs_dir3_leaf_read(tp, dp, args->geo->leafblk, -1, &lbp);
1239         if (error)
1240                 return error;
1241
1242         *lbpp = lbp;
1243         leaf = lbp->b_addr;
1244         xfs_dir3_leaf_check(dp, lbp);
1245         ents = dp->d_ops->leaf_ents_p(leaf);
1246         dp->d_ops->leaf_hdr_from_disk(&leafhdr, leaf);
1247
1248         /*
1249          * Look for the first leaf entry with our hash value.
1250          */
1251         index = xfs_dir2_leaf_search_hash(args, lbp);
1252         /*
1253          * Loop over all the entries with the right hash value
1254          * looking to match the name.
1255          */
1256         for (lep = &ents[index];
1257              index < leafhdr.count && be32_to_cpu(lep->hashval) == args->hashval;
1258              lep++, index++) {
1259                 /*
1260                  * Skip over stale leaf entries.
1261                  */
1262                 if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
1263                         continue;
1264                 /*
1265                  * Get the new data block number.
1266                  */
1267                 newdb = xfs_dir2_dataptr_to_db(args->geo,
1268                                                be32_to_cpu(lep->address));
1269                 /*
1270                  * If it's not the same as the old data block number,
1271                  * need to pitch the old one and read the new one.
1272                  */
1273                 if (newdb != curdb) {
1274                         if (dbp)
1275                                 xfs_trans_brelse(tp, dbp);
1276                         error = xfs_dir3_data_read(tp, dp,
1277                                            xfs_dir2_db_to_da(args->geo, newdb),
1278                                            -1, &dbp);
1279                         if (error) {
1280                                 xfs_trans_brelse(tp, lbp);
1281                                 return error;
1282                         }
1283                         curdb = newdb;
1284                 }
1285                 /*
1286                  * Point to the data entry.
1287                  */
1288                 dep = (xfs_dir2_data_entry_t *)((char *)dbp->b_addr +
1289                         xfs_dir2_dataptr_to_off(args->geo,
1290                                                 be32_to_cpu(lep->address)));
1291                 /*
1292                  * Compare name and if it's an exact match, return the index
1293                  * and buffer. If it's the first case-insensitive match, store
1294                  * the index and buffer and continue looking for an exact match.
1295                  */
1296                 cmp = mp->m_dirnameops->compname(args, dep->name, dep->namelen);
1297                 if (cmp != XFS_CMP_DIFFERENT && cmp != args->cmpresult) {
1298                         args->cmpresult = cmp;
1299                         *indexp = index;
1300                         /* case exact match: return the current buffer. */
1301                         if (cmp == XFS_CMP_EXACT) {
1302                                 *dbpp = dbp;
1303                                 return 0;
1304                         }
1305                         cidb = curdb;
1306                 }
1307         }
1308         ASSERT(args->op_flags & XFS_DA_OP_OKNOENT);
1309         /*
1310          * Here, we can only be doing a lookup (not a rename or remove).
1311          * If a case-insensitive match was found earlier, re-read the
1312          * appropriate data block if required and return it.
1313          */
1314         if (args->cmpresult == XFS_CMP_CASE) {
1315                 ASSERT(cidb != -1);
1316                 if (cidb != curdb) {
1317                         xfs_trans_brelse(tp, dbp);
1318                         error = xfs_dir3_data_read(tp, dp,
1319                                            xfs_dir2_db_to_da(args->geo, cidb),
1320                                            -1, &dbp);
1321                         if (error) {
1322                                 xfs_trans_brelse(tp, lbp);
1323                                 return error;
1324                         }
1325                 }
1326                 *dbpp = dbp;
1327                 return 0;
1328         }
1329         /*
1330          * No match found, return ENOENT.
1331          */
1332         ASSERT(cidb == -1);
1333         if (dbp)
1334                 xfs_trans_brelse(tp, dbp);
1335         xfs_trans_brelse(tp, lbp);
1336         return XFS_ERROR(ENOENT);
1337 }
1338
1339 /*
1340  * Remove an entry from a leaf format directory.
1341  */
1342 int                                             /* error */
1343 xfs_dir2_leaf_removename(
1344         xfs_da_args_t           *args)          /* operation arguments */
1345 {
1346         __be16                  *bestsp;        /* leaf block best freespace */
1347         xfs_dir2_data_hdr_t     *hdr;           /* data block header */
1348         xfs_dir2_db_t           db;             /* data block number */
1349         struct xfs_buf          *dbp;           /* data block buffer */
1350         xfs_dir2_data_entry_t   *dep;           /* data entry structure */
1351         xfs_inode_t             *dp;            /* incore directory inode */
1352         int                     error;          /* error return code */
1353         xfs_dir2_db_t           i;              /* temporary data block # */
1354         int                     index;          /* index into leaf entries */
1355         struct xfs_buf          *lbp;           /* leaf buffer */
1356         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1357         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1358         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1359         xfs_mount_t             *mp;            /* filesystem mount point */
1360         int                     needlog;        /* need to log data header */
1361         int                     needscan;       /* need to rescan data frees */
1362         xfs_dir2_data_off_t     oldbest;        /* old value of best free */
1363         xfs_trans_t             *tp;            /* transaction pointer */
1364         struct xfs_dir2_data_free *bf;          /* bestfree table */
1365         struct xfs_dir2_leaf_entry *ents;
1366         struct xfs_dir3_icleaf_hdr leafhdr;
1367
1368         trace_xfs_dir2_leaf_removename(args);
1369
1370         /*
1371          * Lookup the leaf entry, get the leaf and data blocks read in.
1372          */
1373         if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1374                 return error;
1375         }
1376         dp = args->dp;
1377         tp = args->trans;
1378         mp = dp->i_mount;
1379         leaf = lbp->b_addr;
1380         hdr = dbp->b_addr;
1381         xfs_dir3_data_check(dp, dbp);
1382         bf = dp->d_ops->data_bestfree_p(hdr);
1383         dp->d_ops->leaf_hdr_from_disk(&leafhdr, leaf);
1384         ents = dp->d_ops->leaf_ents_p(leaf);
1385         /*
1386          * Point to the leaf entry, use that to point to the data entry.
1387          */
1388         lep = &ents[index];
1389         db = xfs_dir2_dataptr_to_db(args->geo, be32_to_cpu(lep->address));
1390         dep = (xfs_dir2_data_entry_t *)((char *)hdr +
1391                 xfs_dir2_dataptr_to_off(args->geo, be32_to_cpu(lep->address)));
1392         needscan = needlog = 0;
1393         oldbest = be16_to_cpu(bf[0].length);
1394         ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
1395         bestsp = xfs_dir2_leaf_bests_p(ltp);
1396         ASSERT(be16_to_cpu(bestsp[db]) == oldbest);
1397         /*
1398          * Mark the former data entry unused.
1399          */
1400         xfs_dir2_data_make_free(args, dbp,
1401                 (xfs_dir2_data_aoff_t)((char *)dep - (char *)hdr),
1402                 dp->d_ops->data_entsize(dep->namelen), &needlog, &needscan);
1403         /*
1404          * We just mark the leaf entry stale by putting a null in it.
1405          */
1406         leafhdr.stale++;
1407         dp->d_ops->leaf_hdr_to_disk(leaf, &leafhdr);
1408         xfs_dir3_leaf_log_header(args, lbp);
1409
1410         lep->address = cpu_to_be32(XFS_DIR2_NULL_DATAPTR);
1411         xfs_dir3_leaf_log_ents(args, lbp, index, index);
1412
1413         /*
1414          * Scan the freespace in the data block again if necessary,
1415          * log the data block header if necessary.
1416          */
1417         if (needscan)
1418                 xfs_dir2_data_freescan(dp, hdr, &needlog);
1419         if (needlog)
1420                 xfs_dir2_data_log_header(args, dbp);
1421         /*
1422          * If the longest freespace in the data block has changed,
1423          * put the new value in the bests table and log that.
1424          */
1425         if (be16_to_cpu(bf[0].length) != oldbest) {
1426                 bestsp[db] = bf[0].length;
1427                 xfs_dir3_leaf_log_bests(args, lbp, db, db);
1428         }
1429         xfs_dir3_data_check(dp, dbp);
1430         /*
1431          * If the data block is now empty then get rid of the data block.
1432          */
1433         if (be16_to_cpu(bf[0].length) ==
1434                         args->geo->blksize - dp->d_ops->data_entry_offset) {
1435                 ASSERT(db != args->geo->datablk);
1436                 if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1437                         /*
1438                          * Nope, can't get rid of it because it caused
1439                          * allocation of a bmap btree block to do so.
1440                          * Just go on, returning success, leaving the
1441                          * empty block in place.
1442                          */
1443                         if (error == ENOSPC && args->total == 0)
1444                                 error = 0;
1445                         xfs_dir3_leaf_check(dp, lbp);
1446                         return error;
1447                 }
1448                 dbp = NULL;
1449                 /*
1450                  * If this is the last data block then compact the
1451                  * bests table by getting rid of entries.
1452                  */
1453                 if (db == be32_to_cpu(ltp->bestcount) - 1) {
1454                         /*
1455                          * Look for the last active entry (i).
1456                          */
1457                         for (i = db - 1; i > 0; i--) {
1458                                 if (bestsp[i] != cpu_to_be16(NULLDATAOFF))
1459                                         break;
1460                         }
1461                         /*
1462                          * Copy the table down so inactive entries at the
1463                          * end are removed.
1464                          */
1465                         memmove(&bestsp[db - i], bestsp,
1466                                 (be32_to_cpu(ltp->bestcount) - (db - i)) * sizeof(*bestsp));
1467                         be32_add_cpu(&ltp->bestcount, -(db - i));
1468                         xfs_dir3_leaf_log_tail(args, lbp);
1469                         xfs_dir3_leaf_log_bests(args, lbp, 0,
1470                                                 be32_to_cpu(ltp->bestcount) - 1);
1471                 } else
1472                         bestsp[db] = cpu_to_be16(NULLDATAOFF);
1473         }
1474         /*
1475          * If the data block was not the first one, drop it.
1476          */
1477         else if (db != args->geo->datablk)
1478                 dbp = NULL;
1479
1480         xfs_dir3_leaf_check(dp, lbp);
1481         /*
1482          * See if we can convert to block form.
1483          */
1484         return xfs_dir2_leaf_to_block(args, lbp, dbp);
1485 }
1486
1487 /*
1488  * Replace the inode number in a leaf format directory entry.
1489  */
1490 int                                             /* error */
1491 xfs_dir2_leaf_replace(
1492         xfs_da_args_t           *args)          /* operation arguments */
1493 {
1494         struct xfs_buf          *dbp;           /* data block buffer */
1495         xfs_dir2_data_entry_t   *dep;           /* data block entry */
1496         xfs_inode_t             *dp;            /* incore directory inode */
1497         int                     error;          /* error return code */
1498         int                     index;          /* index of leaf entry */
1499         struct xfs_buf          *lbp;           /* leaf buffer */
1500         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1501         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1502         xfs_trans_t             *tp;            /* transaction pointer */
1503         struct xfs_dir2_leaf_entry *ents;
1504
1505         trace_xfs_dir2_leaf_replace(args);
1506
1507         /*
1508          * Look up the entry.
1509          */
1510         if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1511                 return error;
1512         }
1513         dp = args->dp;
1514         leaf = lbp->b_addr;
1515         ents = dp->d_ops->leaf_ents_p(leaf);
1516         /*
1517          * Point to the leaf entry, get data address from it.
1518          */
1519         lep = &ents[index];
1520         /*
1521          * Point to the data entry.
1522          */
1523         dep = (xfs_dir2_data_entry_t *)
1524               ((char *)dbp->b_addr +
1525                xfs_dir2_dataptr_to_off(args->geo, be32_to_cpu(lep->address)));
1526         ASSERT(args->inumber != be64_to_cpu(dep->inumber));
1527         /*
1528          * Put the new inode number in, log it.
1529          */
1530         dep->inumber = cpu_to_be64(args->inumber);
1531         dp->d_ops->data_put_ftype(dep, args->filetype);
1532         tp = args->trans;
1533         xfs_dir2_data_log_entry(args, dbp, dep);
1534         xfs_dir3_leaf_check(dp, lbp);
1535         xfs_trans_brelse(tp, lbp);
1536         return 0;
1537 }
1538
1539 /*
1540  * Return index in the leaf block (lbp) which is either the first
1541  * one with this hash value, or if there are none, the insert point
1542  * for that hash value.
1543  */
1544 int                                             /* index value */
1545 xfs_dir2_leaf_search_hash(
1546         xfs_da_args_t           *args,          /* operation arguments */
1547         struct xfs_buf          *lbp)           /* leaf buffer */
1548 {
1549         xfs_dahash_t            hash=0;         /* hash from this entry */
1550         xfs_dahash_t            hashwant;       /* hash value looking for */
1551         int                     high;           /* high leaf index */
1552         int                     low;            /* low leaf index */
1553         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1554         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1555         int                     mid=0;          /* current leaf index */
1556         struct xfs_dir2_leaf_entry *ents;
1557         struct xfs_dir3_icleaf_hdr leafhdr;
1558
1559         leaf = lbp->b_addr;
1560         ents = args->dp->d_ops->leaf_ents_p(leaf);
1561         args->dp->d_ops->leaf_hdr_from_disk(&leafhdr, leaf);
1562
1563         /*
1564          * Note, the table cannot be empty, so we have to go through the loop.
1565          * Binary search the leaf entries looking for our hash value.
1566          */
1567         for (lep = ents, low = 0, high = leafhdr.count - 1,
1568                 hashwant = args->hashval;
1569              low <= high; ) {
1570                 mid = (low + high) >> 1;
1571                 if ((hash = be32_to_cpu(lep[mid].hashval)) == hashwant)
1572                         break;
1573                 if (hash < hashwant)
1574                         low = mid + 1;
1575                 else
1576                         high = mid - 1;
1577         }
1578         /*
1579          * Found one, back up through all the equal hash values.
1580          */
1581         if (hash == hashwant) {
1582                 while (mid > 0 && be32_to_cpu(lep[mid - 1].hashval) == hashwant) {
1583                         mid--;
1584                 }
1585         }
1586         /*
1587          * Need to point to an entry higher than ours.
1588          */
1589         else if (hash < hashwant)
1590                 mid++;
1591         return mid;
1592 }
1593
1594 /*
1595  * Trim off a trailing data block.  We know it's empty since the leaf
1596  * freespace table says so.
1597  */
1598 int                                             /* error */
1599 xfs_dir2_leaf_trim_data(
1600         xfs_da_args_t           *args,          /* operation arguments */
1601         struct xfs_buf          *lbp,           /* leaf buffer */
1602         xfs_dir2_db_t           db)             /* data block number */
1603 {
1604         __be16                  *bestsp;        /* leaf bests table */
1605         struct xfs_buf          *dbp;           /* data block buffer */
1606         xfs_inode_t             *dp;            /* incore directory inode */
1607         int                     error;          /* error return value */
1608         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1609         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1610         xfs_mount_t             *mp;            /* filesystem mount point */
1611         xfs_trans_t             *tp;            /* transaction pointer */
1612
1613         dp = args->dp;
1614         mp = dp->i_mount;
1615         tp = args->trans;
1616         /*
1617          * Read the offending data block.  We need its buffer.
1618          */
1619         error = xfs_dir3_data_read(tp, dp, xfs_dir2_db_to_da(args->geo, db),
1620                                    -1, &dbp);
1621         if (error)
1622                 return error;
1623
1624         leaf = lbp->b_addr;
1625         ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
1626
1627 #ifdef DEBUG
1628 {
1629         struct xfs_dir2_data_hdr *hdr = dbp->b_addr;
1630         struct xfs_dir2_data_free *bf = dp->d_ops->data_bestfree_p(hdr);
1631
1632         ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
1633                hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC));
1634         ASSERT(be16_to_cpu(bf[0].length) ==
1635                args->geo->blksize - dp->d_ops->data_entry_offset);
1636         ASSERT(db == be32_to_cpu(ltp->bestcount) - 1);
1637 }
1638 #endif
1639
1640         /*
1641          * Get rid of the data block.
1642          */
1643         if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1644                 ASSERT(error != ENOSPC);
1645                 xfs_trans_brelse(tp, dbp);
1646                 return error;
1647         }
1648         /*
1649          * Eliminate the last bests entry from the table.
1650          */
1651         bestsp = xfs_dir2_leaf_bests_p(ltp);
1652         be32_add_cpu(&ltp->bestcount, -1);
1653         memmove(&bestsp[1], &bestsp[0], be32_to_cpu(ltp->bestcount) * sizeof(*bestsp));
1654         xfs_dir3_leaf_log_tail(args, lbp);
1655         xfs_dir3_leaf_log_bests(args, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
1656         return 0;
1657 }
1658
1659 static inline size_t
1660 xfs_dir3_leaf_size(
1661         struct xfs_dir3_icleaf_hdr      *hdr,
1662         int                             counts)
1663 {
1664         int     entries;
1665         int     hdrsize;
1666
1667         entries = hdr->count - hdr->stale;
1668         if (hdr->magic == XFS_DIR2_LEAF1_MAGIC ||
1669             hdr->magic == XFS_DIR2_LEAFN_MAGIC)
1670                 hdrsize = sizeof(struct xfs_dir2_leaf_hdr);
1671         else
1672                 hdrsize = sizeof(struct xfs_dir3_leaf_hdr);
1673
1674         return hdrsize + entries * sizeof(xfs_dir2_leaf_entry_t)
1675                        + counts * sizeof(xfs_dir2_data_off_t)
1676                        + sizeof(xfs_dir2_leaf_tail_t);
1677 }
1678
1679 /*
1680  * Convert node form directory to leaf form directory.
1681  * The root of the node form dir needs to already be a LEAFN block.
1682  * Just return if we can't do anything.
1683  */
1684 int                                             /* error */
1685 xfs_dir2_node_to_leaf(
1686         xfs_da_state_t          *state)         /* directory operation state */
1687 {
1688         xfs_da_args_t           *args;          /* operation arguments */
1689         xfs_inode_t             *dp;            /* incore directory inode */
1690         int                     error;          /* error return code */
1691         struct xfs_buf          *fbp;           /* buffer for freespace block */
1692         xfs_fileoff_t           fo;             /* freespace file offset */
1693         xfs_dir2_free_t         *free;          /* freespace structure */
1694         struct xfs_buf          *lbp;           /* buffer for leaf block */
1695         xfs_dir2_leaf_tail_t    *ltp;           /* tail of leaf structure */
1696         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1697         xfs_mount_t             *mp;            /* filesystem mount point */
1698         int                     rval;           /* successful free trim? */
1699         xfs_trans_t             *tp;            /* transaction pointer */
1700         struct xfs_dir3_icleaf_hdr leafhdr;
1701         struct xfs_dir3_icfree_hdr freehdr;
1702
1703         /*
1704          * There's more than a leaf level in the btree, so there must
1705          * be multiple leafn blocks.  Give up.
1706          */
1707         if (state->path.active > 1)
1708                 return 0;
1709         args = state->args;
1710
1711         trace_xfs_dir2_node_to_leaf(args);
1712
1713         mp = state->mp;
1714         dp = args->dp;
1715         tp = args->trans;
1716         /*
1717          * Get the last offset in the file.
1718          */
1719         if ((error = xfs_bmap_last_offset(dp, &fo, XFS_DATA_FORK))) {
1720                 return error;
1721         }
1722         fo -= args->geo->fsbcount;
1723         /*
1724          * If there are freespace blocks other than the first one,
1725          * take this opportunity to remove trailing empty freespace blocks
1726          * that may have been left behind during no-space-reservation
1727          * operations.
1728          */
1729         while (fo > args->geo->freeblk) {
1730                 if ((error = xfs_dir2_node_trim_free(args, fo, &rval))) {
1731                         return error;
1732                 }
1733                 if (rval)
1734                         fo -= args->geo->fsbcount;
1735                 else
1736                         return 0;
1737         }
1738         /*
1739          * Now find the block just before the freespace block.
1740          */
1741         if ((error = xfs_bmap_last_before(tp, dp, &fo, XFS_DATA_FORK))) {
1742                 return error;
1743         }
1744         /*
1745          * If it's not the single leaf block, give up.
1746          */
1747         if (XFS_FSB_TO_B(mp, fo) > XFS_DIR2_LEAF_OFFSET + args->geo->blksize)
1748                 return 0;
1749         lbp = state->path.blk[0].bp;
1750         leaf = lbp->b_addr;
1751         dp->d_ops->leaf_hdr_from_disk(&leafhdr, leaf);
1752
1753         ASSERT(leafhdr.magic == XFS_DIR2_LEAFN_MAGIC ||
1754                leafhdr.magic == XFS_DIR3_LEAFN_MAGIC);
1755
1756         /*
1757          * Read the freespace block.
1758          */
1759         error = xfs_dir2_free_read(tp, dp,  args->geo->freeblk, &fbp);
1760         if (error)
1761                 return error;
1762         free = fbp->b_addr;
1763         dp->d_ops->free_hdr_from_disk(&freehdr, free);
1764
1765         ASSERT(!freehdr.firstdb);
1766
1767         /*
1768          * Now see if the leafn and free data will fit in a leaf1.
1769          * If not, release the buffer and give up.
1770          */
1771         if (xfs_dir3_leaf_size(&leafhdr, freehdr.nvalid) > args->geo->blksize) {
1772                 xfs_trans_brelse(tp, fbp);
1773                 return 0;
1774         }
1775
1776         /*
1777          * If the leaf has any stale entries in it, compress them out.
1778          */
1779         if (leafhdr.stale)
1780                 xfs_dir3_leaf_compact(args, &leafhdr, lbp);
1781
1782         lbp->b_ops = &xfs_dir3_leaf1_buf_ops;
1783         xfs_trans_buf_set_type(tp, lbp, XFS_BLFT_DIR_LEAF1_BUF);
1784         leafhdr.magic = (leafhdr.magic == XFS_DIR2_LEAFN_MAGIC)
1785                                         ? XFS_DIR2_LEAF1_MAGIC
1786                                         : XFS_DIR3_LEAF1_MAGIC;
1787
1788         /*
1789          * Set up the leaf tail from the freespace block.
1790          */
1791         ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
1792         ltp->bestcount = cpu_to_be32(freehdr.nvalid);
1793
1794         /*
1795          * Set up the leaf bests table.
1796          */
1797         memcpy(xfs_dir2_leaf_bests_p(ltp), dp->d_ops->free_bests_p(free),
1798                 freehdr.nvalid * sizeof(xfs_dir2_data_off_t));
1799
1800         dp->d_ops->leaf_hdr_to_disk(leaf, &leafhdr);
1801         xfs_dir3_leaf_log_header(args, lbp);
1802         xfs_dir3_leaf_log_bests(args, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
1803         xfs_dir3_leaf_log_tail(args, lbp);
1804         xfs_dir3_leaf_check(dp, lbp);
1805
1806         /*
1807          * Get rid of the freespace block.
1808          */
1809         error = xfs_dir2_shrink_inode(args,
1810                         xfs_dir2_byte_to_db(args->geo, XFS_DIR2_FREE_OFFSET),
1811                         fbp);
1812         if (error) {
1813                 /*
1814                  * This can't fail here because it can only happen when
1815                  * punching out the middle of an extent, and this is an
1816                  * isolated block.
1817                  */
1818                 ASSERT(error != ENOSPC);
1819                 return error;
1820         }
1821         fbp = NULL;
1822         /*
1823          * Now see if we can convert the single-leaf directory
1824          * down to a block form directory.
1825          * This routine always kills the dabuf for the leaf, so
1826          * eliminate it from the path.
1827          */
1828         error = xfs_dir2_leaf_to_block(args, lbp, NULL);
1829         state->path.blk[0].bp = NULL;
1830         return error;
1831 }