git.oblomov.eu Git - linux-2.6/blob - fs/xfs/xfs_dir2_leaf.c

   1 /*
   2  * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or modify it
   5  * under the terms of version 2 of the GNU General Public License as
   6  * published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it would be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  11  *
  12  * Further, this software is distributed without any warranty that it is
  13  * free of the rightful claim of any third person regarding infringement
  14  * or the like.  Any license provided herein, whether implied or
  15  * otherwise, applies only to this software file.  Patent licenses, if
  16  * any, provided herein do not apply to combinations of this program with
  17  * other software, or any other product whatsoever.
  18  *
  19  * You should have received a copy of the GNU General Public License along
  20  * with this program; if not, write the Free Software Foundation, Inc., 59
  21  * Temple Place - Suite 330, Boston MA 02111-1307, USA.
  22  *
  23  * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
  24  * Mountain View, CA  94043, or:
  25  *
  26  * http://www.sgi.com
  27  *
  28  * For further information regarding this notice, see:
  29  *
  30  * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
  31  */
  32
  33 /*
  34  * xfs_dir2_leaf.c
  35  * XFS directory version 2 implementation - single leaf form
  36  * see xfs_dir2_leaf.h for data structures.
  37  * These directories have multiple XFS_DIR2_DATA blocks and one
  38  * XFS_DIR2_LEAF1 block containing the hash table and freespace map.
  39  */
  40
  41 #include "xfs.h"
  42
  43 #include "xfs_macros.h"
  44 #include "xfs_types.h"
  45 #include "xfs_inum.h"
  46 #include "xfs_log.h"
  47 #include "xfs_trans.h"
  48 #include "xfs_sb.h"
  49 #include "xfs_ag.h"
  50 #include "xfs_dir.h"
  51 #include "xfs_dir2.h"
  52 #include "xfs_dmapi.h"
  53 #include "xfs_mount.h"
  54 #include "xfs_bmap_btree.h"
  55 #include "xfs_attr_sf.h"
  56 #include "xfs_dir_sf.h"
  57 #include "xfs_dir2_sf.h"
  58 #include "xfs_dinode.h"
  59 #include "xfs_inode.h"
  60 #include "xfs_bmap.h"
  61 #include "xfs_da_btree.h"
  62 #include "xfs_dir2_data.h"
  63 #include "xfs_dir2_leaf.h"
  64 #include "xfs_dir2_block.h"
  65 #include "xfs_dir2_node.h"
  66 #include "xfs_dir2_trace.h"
  67 #include "xfs_error.h"
  68 #include "xfs_bit.h"
  69
  70 /*
  71  * Local function declarations.
  72  */
  73 #ifdef DEBUG
  74 static void xfs_dir2_leaf_check(xfs_inode_t *dp, xfs_dabuf_t *bp);
  75 #else
  76 #define xfs_dir2_leaf_check(dp, bp)
  77 #endif
  78 static int xfs_dir2_leaf_lookup_int(xfs_da_args_t *args, xfs_dabuf_t **lbpp,
  79                                     int *indexp, xfs_dabuf_t **dbpp);
  80
  81 /*
  82  * Convert a block form directory to a leaf form directory.
  83  */
  84 int                                             /* error */
  85 xfs_dir2_block_to_leaf(
  86         xfs_da_args_t           *args,          /* operation arguments */
  87         xfs_dabuf_t             *dbp)           /* input block's buffer */
  88 {
  89         xfs_dir2_data_off_t     *bestsp;        /* leaf's bestsp entries */
  90         xfs_dablk_t             blkno;          /* leaf block's bno */
  91         xfs_dir2_block_t        *block;         /* block structure */
  92         xfs_dir2_leaf_entry_t   *blp;           /* block's leaf entries */
  93         xfs_dir2_block_tail_t   *btp;           /* block's tail */
  94         xfs_inode_t             *dp;            /* incore directory inode */
  95         int                     error;          /* error return code */
  96         xfs_dabuf_t             *lbp;           /* leaf block's buffer */
  97         xfs_dir2_db_t           ldb;            /* leaf block's bno */
  98         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
  99         xfs_dir2_leaf_tail_t    *ltp;           /* leaf's tail */
 100         xfs_mount_t             *mp;            /* filesystem mount point */
 101         int                     needlog;        /* need to log block header */
 102         int                     needscan;       /* need to rescan bestfree */
 103         xfs_trans_t             *tp;            /* transaction pointer */
 104
 105         xfs_dir2_trace_args_b("block_to_leaf", args, dbp);
 106         dp = args->dp;
 107         mp = dp->i_mount;
 108         tp = args->trans;
 109         /*
 110          * Add the leaf block to the inode.
 111          * This interface will only put blocks in the leaf/node range.
 112          * Since that's empty now, we'll get the root (block 0 in range).
 113          */
 114         if ((error = xfs_da_grow_inode(args, &blkno))) {
 115                 return error;
 116         }
 117         ldb = XFS_DIR2_DA_TO_DB(mp, blkno);
 118         ASSERT(ldb == XFS_DIR2_LEAF_FIRSTDB(mp));
 119         /*
 120          * Initialize the leaf block, get a buffer for it.
 121          */
 122         if ((error = xfs_dir2_leaf_init(args, ldb, &lbp, XFS_DIR2_LEAF1_MAGIC))) {
 123                 return error;
 124         }
 125         ASSERT(lbp != NULL);
 126         leaf = lbp->data;
 127         block = dbp->data;
 128         xfs_dir2_data_check(dp, dbp);
 129         btp = XFS_DIR2_BLOCK_TAIL_P(mp, block);
 130         blp = XFS_DIR2_BLOCK_LEAF_P(btp);
 131         /*
 132          * Set the counts in the leaf header.
 133          */
 134         INT_COPY(leaf->hdr.count, btp->count, ARCH_CONVERT); /* INT_: type change */
 135         INT_COPY(leaf->hdr.stale, btp->stale, ARCH_CONVERT); /* INT_: type change */
 136         /*
 137          * Could compact these but I think we always do the conversion
 138          * after squeezing out stale entries.
 139          */
 140         memcpy(leaf->ents, blp, INT_GET(btp->count, ARCH_CONVERT) * sizeof(xfs_dir2_leaf_entry_t));
 141         xfs_dir2_leaf_log_ents(tp, lbp, 0, INT_GET(leaf->hdr.count, ARCH_CONVERT) - 1);
 142         needscan = 0;
 143         needlog = 1;
 144         /*
 145          * Make the space formerly occupied by the leaf entries and block
 146          * tail be free.
 147          */
 148         xfs_dir2_data_make_free(tp, dbp,
 149                 (xfs_dir2_data_aoff_t)((char *)blp - (char *)block),
 150                 (xfs_dir2_data_aoff_t)((char *)block + mp->m_dirblksize -
 151                                        (char *)blp),
 152                 &needlog, &needscan);
 153         /*
 154          * Fix up the block header, make it a data block.
 155          */
 156         INT_SET(block->hdr.magic, ARCH_CONVERT, XFS_DIR2_DATA_MAGIC);
 157         if (needscan)
 158                 xfs_dir2_data_freescan(mp, (xfs_dir2_data_t *)block, &needlog,
 159                         NULL);
 160         /*
 161          * Set up leaf tail and bests table.
 162          */
 163         ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
 164         INT_SET(ltp->bestcount, ARCH_CONVERT, 1);
 165         bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
 166         INT_COPY(bestsp[0], block->hdr.bestfree[0].length, ARCH_CONVERT);
 167         /*
 168          * Log the data header and leaf bests table.
 169          */
 170         if (needlog)
 171                 xfs_dir2_data_log_header(tp, dbp);
 172         xfs_dir2_leaf_check(dp, lbp);
 173         xfs_dir2_data_check(dp, dbp);
 174         xfs_dir2_leaf_log_bests(tp, lbp, 0, 0);
 175         xfs_da_buf_done(lbp);
 176         return 0;
 177 }
 178
 179 /*
 180  * Add an entry to a leaf form directory.
 181  */
 182 int                                             /* error */
 183 xfs_dir2_leaf_addname(
 184         xfs_da_args_t           *args)          /* operation arguments */
 185 {
 186         xfs_dir2_data_off_t     *bestsp;        /* freespace table in leaf */
 187         int                     compact;        /* need to compact leaves */
 188         xfs_dir2_data_t         *data;          /* data block structure */
 189         xfs_dabuf_t             *dbp;           /* data block buffer */
 190         xfs_dir2_data_entry_t   *dep;           /* data block entry */
 191         xfs_inode_t             *dp;            /* incore directory inode */
 192         xfs_dir2_data_unused_t  *dup;           /* data unused entry */
 193         int                     error;          /* error return value */
 194         int                     grown;          /* allocated new data block */
 195         int                     highstale;      /* index of next stale leaf */
 196         int                     i;              /* temporary, index */
 197         int                     index;          /* leaf table position */
 198         xfs_dabuf_t             *lbp;           /* leaf's buffer */
 199         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
 200         int                     length;         /* length of new entry */
 201         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry table pointer */
 202         int                     lfloglow;       /* low leaf logging index */
 203         int                     lfloghigh;      /* high leaf logging index */
 204         int                     lowstale;       /* index of prev stale leaf */
 205         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail pointer */
 206         xfs_mount_t             *mp;            /* filesystem mount point */
 207         int                     needbytes;      /* leaf block bytes needed */
 208         int                     needlog;        /* need to log data header */
 209         int                     needscan;       /* need to rescan data free */
 210         xfs_dir2_data_off_t     *tagp;          /* end of data entry */
 211         xfs_trans_t             *tp;            /* transaction pointer */
 212         xfs_dir2_db_t           use_block;      /* data block number */
 213
 214         xfs_dir2_trace_args("leaf_addname", args);
 215         dp = args->dp;
 216         tp = args->trans;
 217         mp = dp->i_mount;
 218         /*
 219          * Read the leaf block.
 220          */
 221         error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
 222                 XFS_DATA_FORK);
 223         if (error) {
 224                 return error;
 225         }
 226         ASSERT(lbp != NULL);
 227         /*
 228          * Look up the entry by hash value and name.
 229          * We know it's not there, our caller has already done a lookup.
 230          * So the index is of the entry to insert in front of.
 231          * But if there are dup hash values the index is of the first of those.
 232          */
 233         index = xfs_dir2_leaf_search_hash(args, lbp);
 234         leaf = lbp->data;
 235         ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
 236         bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
 237         length = XFS_DIR2_DATA_ENTSIZE(args->namelen);
 238         /*
 239          * See if there are any entries with the same hash value
 240          * and space in their block for the new entry.
 241          * This is good because it puts multiple same-hash value entries
 242          * in a data block, improving the lookup of those entries.
 243          */
 244         for (use_block = -1, lep = &leaf->ents[index];
 245              index < INT_GET(leaf->hdr.count, ARCH_CONVERT) && INT_GET(lep->hashval, ARCH_CONVERT) == args->hashval;
 246              index++, lep++) {
 247                 if (INT_GET(lep->address, ARCH_CONVERT) == XFS_DIR2_NULL_DATAPTR)
 248                         continue;
 249                 i = XFS_DIR2_DATAPTR_TO_DB(mp, INT_GET(lep->address, ARCH_CONVERT));
 250                 ASSERT(i < INT_GET(ltp->bestcount, ARCH_CONVERT));
 251                 ASSERT(INT_GET(bestsp[i], ARCH_CONVERT) != NULLDATAOFF);
 252                 if (INT_GET(bestsp[i], ARCH_CONVERT) >= length) {
 253                         use_block = i;
 254                         break;
 255                 }
 256         }
 257         /*
 258          * Didn't find a block yet, linear search all the data blocks.
 259          */
 260         if (use_block == -1) {
 261                 for (i = 0; i < INT_GET(ltp->bestcount, ARCH_CONVERT); i++) {
 262                         /*
 263                          * Remember a block we see that's missing.
 264                          */
 265                         if (INT_GET(bestsp[i], ARCH_CONVERT) == NULLDATAOFF && use_block == -1)
 266                                 use_block = i;
 267                         else if (INT_GET(bestsp[i], ARCH_CONVERT) >= length) {
 268                                 use_block = i;
 269                                 break;
 270                         }
 271                 }
 272         }
 273         /*
 274          * How many bytes do we need in the leaf block?
 275          */
 276         needbytes =
 277                 (leaf->hdr.stale ? 0 : (uint)sizeof(leaf->ents[0])) +
 278                 (use_block != -1 ? 0 : (uint)sizeof(leaf->bests[0]));
 279         /*
 280          * Now kill use_block if it refers to a missing block, so we
 281          * can use it as an indication of allocation needed.
 282          */
 283         if (use_block != -1 && INT_GET(bestsp[use_block], ARCH_CONVERT) == NULLDATAOFF)
 284                 use_block = -1;
 285         /*
 286          * If we don't have enough free bytes but we can make enough
 287          * by compacting out stale entries, we'll do that.
 288          */
 289         if ((char *)bestsp - (char *)&leaf->ents[INT_GET(leaf->hdr.count, ARCH_CONVERT)] < needbytes &&
 290             INT_GET(leaf->hdr.stale, ARCH_CONVERT) > 1) {
 291                 compact = 1;
 292         }
 293         /*
 294          * Otherwise if we don't have enough free bytes we need to
 295          * convert to node form.
 296          */
 297         else if ((char *)bestsp - (char *)&leaf->ents[INT_GET(leaf->hdr.count, ARCH_CONVERT)] <
 298                  needbytes) {
 299                 /*
 300                  * Just checking or no space reservation, give up.
 301                  */
 302                 if (args->justcheck || args->total == 0) {
 303                         xfs_da_brelse(tp, lbp);
 304                         return XFS_ERROR(ENOSPC);
 305                 }
 306                 /*
 307                  * Convert to node form.
 308                  */
 309                 error = xfs_dir2_leaf_to_node(args, lbp);
 310                 xfs_da_buf_done(lbp);
 311                 if (error)
 312                         return error;
 313                 /*
 314                  * Then add the new entry.
 315                  */
 316                 return xfs_dir2_node_addname(args);
 317         }
 318         /*
 319          * Otherwise it will fit without compaction.
 320          */
 321         else
 322                 compact = 0;
 323         /*
 324          * If just checking, then it will fit unless we needed to allocate
 325          * a new data block.
 326          */
 327         if (args->justcheck) {
 328                 xfs_da_brelse(tp, lbp);
 329                 return use_block == -1 ? XFS_ERROR(ENOSPC) : 0;
 330         }
 331         /*
 332          * If no allocations are allowed, return now before we've
 333          * changed anything.
 334          */
 335         if (args->total == 0 && use_block == -1) {
 336                 xfs_da_brelse(tp, lbp);
 337                 return XFS_ERROR(ENOSPC);
 338         }
 339         /*
 340          * Need to compact the leaf entries, removing stale ones.
 341          * Leave one stale entry behind - the one closest to our
 342          * insertion index - and we'll shift that one to our insertion
 343          * point later.
 344          */
 345         if (compact) {
 346                 xfs_dir2_leaf_compact_x1(lbp, &index, &lowstale, &highstale,
 347                         &lfloglow, &lfloghigh);
 348         }
 349         /*
 350          * There are stale entries, so we'll need log-low and log-high
 351          * impossibly bad values later.
 352          */
 353         else if (INT_GET(leaf->hdr.stale, ARCH_CONVERT)) {
 354                 lfloglow = INT_GET(leaf->hdr.count, ARCH_CONVERT);
 355                 lfloghigh = -1;
 356         }
 357         /*
 358          * If there was no data block space found, we need to allocate
 359          * a new one.
 360          */
 361         if (use_block == -1) {
 362                 /*
 363                  * Add the new data block.
 364                  */
 365                 if ((error = xfs_dir2_grow_inode(args, XFS_DIR2_DATA_SPACE,
 366                                 &use_block))) {
 367                         xfs_da_brelse(tp, lbp);
 368                         return error;
 369                 }
 370                 /*
 371                  * Initialize the block.
 372                  */
 373                 if ((error = xfs_dir2_data_init(args, use_block, &dbp))) {
 374                         xfs_da_brelse(tp, lbp);
 375                         return error;
 376                 }
 377                 /*
 378                  * If we're adding a new data block on the end we need to
 379                  * extend the bests table.  Copy it up one entry.
 380                  */
 381                 if (use_block >= INT_GET(ltp->bestcount, ARCH_CONVERT)) {
 382                         bestsp--;
 383                         memmove(&bestsp[0], &bestsp[1],
 384                                 INT_GET(ltp->bestcount, ARCH_CONVERT) * sizeof(bestsp[0]));
 385                         INT_MOD(ltp->bestcount, ARCH_CONVERT, +1);
 386                         xfs_dir2_leaf_log_tail(tp, lbp);
 387                         xfs_dir2_leaf_log_bests(tp, lbp, 0, INT_GET(ltp->bestcount, ARCH_CONVERT) - 1);
 388                 }
 389                 /*
 390                  * If we're filling in a previously empty block just log it.
 391                  */
 392                 else
 393                         xfs_dir2_leaf_log_bests(tp, lbp, use_block, use_block);
 394                 data = dbp->data;
 395                 INT_COPY(bestsp[use_block], data->hdr.bestfree[0].length, ARCH_CONVERT);
 396                 grown = 1;
 397         }
 398         /*
 399          * Already had space in some data block.
 400          * Just read that one in.
 401          */
 402         else {
 403                 if ((error =
 404                     xfs_da_read_buf(tp, dp, XFS_DIR2_DB_TO_DA(mp, use_block),
 405                             -1, &dbp, XFS_DATA_FORK))) {
 406                         xfs_da_brelse(tp, lbp);
 407                         return error;
 408                 }
 409                 data = dbp->data;
 410                 grown = 0;
 411         }
 412         xfs_dir2_data_check(dp, dbp);
 413         /*
 414          * Point to the biggest freespace in our data block.
 415          */
 416         dup = (xfs_dir2_data_unused_t *)
 417               ((char *)data + INT_GET(data->hdr.bestfree[0].offset, ARCH_CONVERT));
 418         ASSERT(INT_GET(dup->length, ARCH_CONVERT) >= length);
 419         needscan = needlog = 0;
 420         /*
 421          * Mark the initial part of our freespace in use for the new entry.
 422          */
 423         xfs_dir2_data_use_free(tp, dbp, dup,
 424                 (xfs_dir2_data_aoff_t)((char *)dup - (char *)data), length,
 425                 &needlog, &needscan);
 426         /*
 427          * Initialize our new entry (at last).
 428          */
 429         dep = (xfs_dir2_data_entry_t *)dup;
 430         INT_SET(dep->inumber, ARCH_CONVERT, args->inumber);
 431         dep->namelen = args->namelen;
 432         memcpy(dep->name, args->name, dep->namelen);
 433         tagp = XFS_DIR2_DATA_ENTRY_TAG_P(dep);
 434         INT_SET(*tagp, ARCH_CONVERT, (xfs_dir2_data_off_t)((char *)dep - (char *)data));
 435         /*
 436          * Need to scan fix up the bestfree table.
 437          */
 438         if (needscan)
 439                 xfs_dir2_data_freescan(mp, data, &needlog, NULL);
 440         /*
 441          * Need to log the data block's header.
 442          */
 443         if (needlog)
 444                 xfs_dir2_data_log_header(tp, dbp);
 445         xfs_dir2_data_log_entry(tp, dbp, dep);
 446         /*
 447          * If the bests table needs to be changed, do it.
 448          * Log the change unless we've already done that.
 449          */
 450         if (INT_GET(bestsp[use_block], ARCH_CONVERT) != INT_GET(data->hdr.bestfree[0].length, ARCH_CONVERT)) {
 451                 INT_COPY(bestsp[use_block], data->hdr.bestfree[0].length, ARCH_CONVERT);
 452                 if (!grown)
 453                         xfs_dir2_leaf_log_bests(tp, lbp, use_block, use_block);
 454         }
 455         /*
 456          * Now we need to make room to insert the leaf entry.
 457          * If there are no stale entries, we just insert a hole at index.
 458          */
 459         if (!leaf->hdr.stale) {
 460                 /*
 461                  * lep is still good as the index leaf entry.
 462                  */
 463                 if (index < INT_GET(leaf->hdr.count, ARCH_CONVERT))
 464                         memmove(lep + 1, lep,
 465                                 (INT_GET(leaf->hdr.count, ARCH_CONVERT) - index) * sizeof(*lep));
 466                 /*
 467                  * Record low and high logging indices for the leaf.
 468                  */
 469                 lfloglow = index;
 470                 lfloghigh = INT_GET(leaf->hdr.count, ARCH_CONVERT);
 471                 INT_MOD(leaf->hdr.count, ARCH_CONVERT, +1);
 472         }
 473         /*
 474          * There are stale entries.
 475          * We will use one of them for the new entry.
 476          * It's probably not at the right location, so we'll have to
 477          * shift some up or down first.
 478          */
 479         else {
 480                 /*
 481                  * If we didn't compact before, we need to find the nearest
 482                  * stale entries before and after our insertion point.
 483                  */
 484                 if (compact == 0) {
 485                         /*
 486                          * Find the first stale entry before the insertion
 487                          * point, if any.
 488                          */
 489                         for (lowstale = index - 1;
 490                              lowstale >= 0 &&
 491                                 INT_GET(leaf->ents[lowstale].address, ARCH_CONVERT) !=
 492                                 XFS_DIR2_NULL_DATAPTR;
 493                              lowstale--)
 494                                 continue;
 495                         /*
 496                          * Find the next stale entry at or after the insertion
 497                          * point, if any.   Stop if we go so far that the
 498                          * lowstale entry would be better.
 499                          */
 500                         for (highstale = index;
 501                              highstale < INT_GET(leaf->hdr.count, ARCH_CONVERT) &&
 502                                 INT_GET(leaf->ents[highstale].address, ARCH_CONVERT) !=
 503                                 XFS_DIR2_NULL_DATAPTR &&
 504                                 (lowstale < 0 ||
 505                                  index - lowstale - 1 >= highstale - index);
 506                              highstale++)
 507                                 continue;
 508                 }
 509                 /*
 510                  * If the low one is better, use it.
 511                  */
 512                 if (lowstale >= 0 &&
 513                     (highstale == INT_GET(leaf->hdr.count, ARCH_CONVERT) ||
 514                      index - lowstale - 1 < highstale - index)) {
 515                         ASSERT(index - lowstale - 1 >= 0);
 516                         ASSERT(INT_GET(leaf->ents[lowstale].address, ARCH_CONVERT) ==
 517                                XFS_DIR2_NULL_DATAPTR);
 518                         /*
 519                          * Copy entries up to cover the stale entry
 520                          * and make room for the new entry.
 521                          */
 522                         if (index - lowstale - 1 > 0)
 523                                 memmove(&leaf->ents[lowstale],
 524                                         &leaf->ents[lowstale + 1],
 525                                         (index - lowstale - 1) * sizeof(*lep));
 526                         lep = &leaf->ents[index - 1];
 527                         lfloglow = MIN(lowstale, lfloglow);
 528                         lfloghigh = MAX(index - 1, lfloghigh);
 529                 }
 530                 /*
 531                  * The high one is better, so use that one.
 532                  */
 533                 else {
 534                         ASSERT(highstale - index >= 0);
 535                         ASSERT(INT_GET(leaf->ents[highstale].address, ARCH_CONVERT) ==
 536                                XFS_DIR2_NULL_DATAPTR);
 537                         /*
 538                          * Copy entries down to copver the stale entry
 539                          * and make room for the new entry.
 540                          */
 541                         if (highstale - index > 0)
 542                                 memmove(&leaf->ents[index + 1],
 543                                         &leaf->ents[index],
 544                                         (highstale - index) * sizeof(*lep));
 545                         lep = &leaf->ents[index];
 546                         lfloglow = MIN(index, lfloglow);
 547                         lfloghigh = MAX(highstale, lfloghigh);
 548                 }
 549                 INT_MOD(leaf->hdr.stale, ARCH_CONVERT, -1);
 550         }
 551         /*
 552          * Fill in the new leaf entry.
 553          */
 554         INT_SET(lep->hashval, ARCH_CONVERT, args->hashval);
 555         INT_SET(lep->address, ARCH_CONVERT, XFS_DIR2_DB_OFF_TO_DATAPTR(mp, use_block, INT_GET(*tagp, ARCH_CONVERT)));
 556         /*
 557          * Log the leaf fields and give up the buffers.
 558          */
 559         xfs_dir2_leaf_log_header(tp, lbp);
 560         xfs_dir2_leaf_log_ents(tp, lbp, lfloglow, lfloghigh);
 561         xfs_dir2_leaf_check(dp, lbp);
 562         xfs_da_buf_done(lbp);
 563         xfs_dir2_data_check(dp, dbp);
 564         xfs_da_buf_done(dbp);
 565         return 0;
 566 }
 567
 568 #ifdef DEBUG
 569 /*
 570  * Check the internal consistency of a leaf1 block.
 571  * Pop an assert if something is wrong.
 572  */
 573 void
 574 xfs_dir2_leaf_check(
 575         xfs_inode_t             *dp,            /* incore directory inode */
 576         xfs_dabuf_t             *bp)            /* leaf's buffer */
 577 {
 578         int                     i;              /* leaf index */
 579         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
 580         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail pointer */
 581         xfs_mount_t             *mp;            /* filesystem mount point */
 582         int                     stale;          /* count of stale leaves */
 583
 584         leaf = bp->data;
 585         mp = dp->i_mount;
 586         ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAF1_MAGIC);
 587         /*
 588          * This value is not restrictive enough.
 589          * Should factor in the size of the bests table as well.
 590          * We can deduce a value for that from di_size.
 591          */
 592         ASSERT(INT_GET(leaf->hdr.count, ARCH_CONVERT) <= XFS_DIR2_MAX_LEAF_ENTS(mp));
 593         ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
 594         /*
 595          * Leaves and bests don't overlap.
 596          */
 597         ASSERT((char *)&leaf->ents[INT_GET(leaf->hdr.count, ARCH_CONVERT)] <=
 598                (char *)XFS_DIR2_LEAF_BESTS_P(ltp));
 599         /*
 600          * Check hash value order, count stale entries.
 601          */
 602         for (i = stale = 0; i < INT_GET(leaf->hdr.count, ARCH_CONVERT); i++) {
 603                 if (i + 1 < INT_GET(leaf->hdr.count, ARCH_CONVERT))
 604                         ASSERT(INT_GET(leaf->ents[i].hashval, ARCH_CONVERT) <=
 605                                INT_GET(leaf->ents[i + 1].hashval, ARCH_CONVERT));
 606                 if (INT_GET(leaf->ents[i].address, ARCH_CONVERT) == XFS_DIR2_NULL_DATAPTR)
 607                         stale++;
 608         }
 609         ASSERT(INT_GET(leaf->hdr.stale, ARCH_CONVERT) == stale);
 610 }
 611 #endif  /* DEBUG */
 612
 613 /*
 614  * Compact out any stale entries in the leaf.
 615  * Log the header and changed leaf entries, if any.
 616  */
 617 void
 618 xfs_dir2_leaf_compact(
 619         xfs_da_args_t   *args,          /* operation arguments */
 620         xfs_dabuf_t     *bp)            /* leaf buffer */
 621 {
 622         int             from;           /* source leaf index */
 623         xfs_dir2_leaf_t *leaf;          /* leaf structure */
 624         int             loglow;         /* first leaf entry to log */
 625         int             to;             /* target leaf index */
 626
 627         leaf = bp->data;
 628         if (!leaf->hdr.stale) {
 629                 return;
 630         }
 631         /*
 632          * Compress out the stale entries in place.
 633          */
 634         for (from = to = 0, loglow = -1; from < INT_GET(leaf->hdr.count, ARCH_CONVERT); from++) {
 635                 if (INT_GET(leaf->ents[from].address, ARCH_CONVERT) == XFS_DIR2_NULL_DATAPTR)
 636                         continue;
 637                 /*
 638                  * Only actually copy the entries that are different.
 639                  */
 640                 if (from > to) {
 641                         if (loglow == -1)
 642                                 loglow = to;
 643                         leaf->ents[to] = leaf->ents[from];
 644                 }
 645                 to++;
 646         }
 647         /*
 648          * Update and log the header, log the leaf entries.
 649          */
 650         ASSERT(INT_GET(leaf->hdr.stale, ARCH_CONVERT) == from - to);
 651         INT_MOD(leaf->hdr.count, ARCH_CONVERT, -(INT_GET(leaf->hdr.stale, ARCH_CONVERT)));
 652         leaf->hdr.stale = 0;
 653         xfs_dir2_leaf_log_header(args->trans, bp);
 654         if (loglow != -1)
 655                 xfs_dir2_leaf_log_ents(args->trans, bp, loglow, to - 1);
 656 }
 657
 658 /*
 659  * Compact the leaf entries, removing stale ones.
 660  * Leave one stale entry behind - the one closest to our
 661  * insertion index - and the caller will shift that one to our insertion
 662  * point later.
 663  * Return new insertion index, where the remaining stale entry is,
 664  * and leaf logging indices.
 665  */
 666 void
 667 xfs_dir2_leaf_compact_x1(
 668         xfs_dabuf_t     *bp,            /* leaf buffer */
 669         int             *indexp,        /* insertion index */
 670         int             *lowstalep,     /* out: stale entry before us */
 671         int             *highstalep,    /* out: stale entry after us */
 672         int             *lowlogp,       /* out: low log index */
 673         int             *highlogp)      /* out: high log index */
 674 {
 675         int             from;           /* source copy index */
 676         int             highstale;      /* stale entry at/after index */
 677         int             index;          /* insertion index */
 678         int             keepstale;      /* source index of kept stale */
 679         xfs_dir2_leaf_t *leaf;          /* leaf structure */
 680         int             lowstale;       /* stale entry before index */
 681         int             newindex=0;     /* new insertion index */
 682         int             to;             /* destination copy index */
 683
 684         leaf = bp->data;
 685         ASSERT(INT_GET(leaf->hdr.stale, ARCH_CONVERT) > 1);
 686         index = *indexp;
 687         /*
 688          * Find the first stale entry before our index, if any.
 689          */
 690         for (lowstale = index - 1;
 691              lowstale >= 0 &&
 692                 INT_GET(leaf->ents[lowstale].address, ARCH_CONVERT) != XFS_DIR2_NULL_DATAPTR;
 693              lowstale--)
 694                 continue;
 695         /*
 696          * Find the first stale entry at or after our index, if any.
 697          * Stop if the answer would be worse than lowstale.
 698          */
 699         for (highstale = index;
 700              highstale < INT_GET(leaf->hdr.count, ARCH_CONVERT) &&
 701                 INT_GET(leaf->ents[highstale].address, ARCH_CONVERT) != XFS_DIR2_NULL_DATAPTR &&
 702                 (lowstale < 0 || index - lowstale > highstale - index);
 703              highstale++)
 704                 continue;
 705         /*
 706          * Pick the better of lowstale and highstale.
 707          */
 708         if (lowstale >= 0 &&
 709             (highstale == INT_GET(leaf->hdr.count, ARCH_CONVERT) ||
 710              index - lowstale <= highstale - index))
 711                 keepstale = lowstale;
 712         else
 713                 keepstale = highstale;
 714         /*
 715          * Copy the entries in place, removing all the stale entries
 716          * except keepstale.
 717          */
 718         for (from = to = 0; from < INT_GET(leaf->hdr.count, ARCH_CONVERT); from++) {
 719                 /*
 720                  * Notice the new value of index.
 721                  */
 722                 if (index == from)
 723                         newindex = to;
 724                 if (from != keepstale &&
 725                     INT_GET(leaf->ents[from].address, ARCH_CONVERT) == XFS_DIR2_NULL_DATAPTR) {
 726                         if (from == to)
 727                                 *lowlogp = to;
 728                         continue;
 729                 }
 730                 /*
 731                  * Record the new keepstale value for the insertion.
 732                  */
 733                 if (from == keepstale)
 734                         lowstale = highstale = to;
 735                 /*
 736                  * Copy only the entries that have moved.
 737                  */
 738                 if (from > to)
 739                         leaf->ents[to] = leaf->ents[from];
 740                 to++;
 741         }
 742         ASSERT(from > to);
 743         /*
 744          * If the insertion point was past the last entry,
 745          * set the new insertion point accordingly.
 746          */
 747         if (index == from)
 748                 newindex = to;
 749         *indexp = newindex;
 750         /*
 751          * Adjust the leaf header values.
 752          */
 753         INT_MOD(leaf->hdr.count, ARCH_CONVERT, -(from - to));
 754         INT_SET(leaf->hdr.stale, ARCH_CONVERT, 1);
 755         /*
 756          * Remember the low/high stale value only in the "right"
 757          * direction.
 758          */
 759         if (lowstale >= newindex)
 760                 lowstale = -1;
 761         else
 762                 highstale = INT_GET(leaf->hdr.count, ARCH_CONVERT);
 763         *highlogp = INT_GET(leaf->hdr.count, ARCH_CONVERT) - 1;
 764         *lowstalep = lowstale;
 765         *highstalep = highstale;
 766 }
 767
 768 /*
 769  * Getdents (readdir) for leaf and node directories.
 770  * This reads the data blocks only, so is the same for both forms.
 771  */
 772 int                                             /* error */
 773 xfs_dir2_leaf_getdents(
 774         xfs_trans_t             *tp,            /* transaction pointer */
 775         xfs_inode_t             *dp,            /* incore directory inode */
 776         uio_t                   *uio,           /* I/O control & vectors */
 777         int                     *eofp,          /* out: reached end of dir */
 778         xfs_dirent_t            *dbp,           /* caller's buffer */
 779         xfs_dir2_put_t          put)            /* ABI formatting routine */
 780 {
 781         xfs_dabuf_t             *bp;            /* data block buffer */
 782         int                     byteoff;        /* offset in current block */
 783         xfs_dir2_db_t           curdb;          /* db for current block */
 784         xfs_dir2_off_t          curoff;         /* current overall offset */
 785         xfs_dir2_data_t         *data;          /* data block structure */
 786         xfs_dir2_data_entry_t   *dep;           /* data entry */
 787         xfs_dir2_data_unused_t  *dup;           /* unused entry */
 788         int                     eof;            /* reached end of directory */
 789         int                     error=0;                /* error return value */
 790         int                     i;              /* temporary loop index */
 791         int                     j;              /* temporary loop index */
 792         int                     length;         /* temporary length value */
 793         xfs_bmbt_irec_t         *map;           /* map vector for blocks */
 794         xfs_extlen_t            map_blocks;     /* number of fsbs in map */
 795         xfs_dablk_t             map_off;        /* last mapped file offset */
 796         int                     map_size;       /* total entries in *map */
 797         int                     map_valid;      /* valid entries in *map */
 798         xfs_mount_t             *mp;            /* filesystem mount point */
 799         xfs_dir2_off_t          newoff;         /* new curoff after new blk */
 800         int                     nmap;           /* mappings to ask xfs_bmapi */
 801         xfs_dir2_put_args_t     p;              /* formatting arg bundle */
 802         char                    *ptr=NULL;              /* pointer to current data */
 803         int                     ra_current;     /* number of read-ahead blks */
 804         int                     ra_index;       /* *map index for read-ahead */
 805         int                     ra_offset;      /* map entry offset for ra */
 806         int                     ra_want;        /* readahead count wanted */
 807
 808         /*
 809          * If the offset is at or past the largest allowed value,
 810          * give up right away, return eof.
 811          */
 812         if (uio->uio_offset >= XFS_DIR2_MAX_DATAPTR) {
 813                 *eofp = 1;
 814                 return 0;
 815         }
 816         mp = dp->i_mount;
 817         /*
 818          * Setup formatting arguments.
 819          */
 820         p.dbp = dbp;
 821         p.put = put;
 822         p.uio = uio;
 823         /*
 824          * Set up to bmap a number of blocks based on the caller's
 825          * buffer size, the directory block size, and the filesystem
 826          * block size.
 827          */
 828         map_size =
 829                 howmany(uio->uio_resid + mp->m_dirblksize,
 830                         mp->m_sb.sb_blocksize);
 831         map = kmem_alloc(map_size * sizeof(*map), KM_SLEEP);
 832         map_valid = ra_index = ra_offset = ra_current = map_blocks = 0;
 833         bp = NULL;
 834         eof = 1;
 835         /*
 836          * Inside the loop we keep the main offset value as a byte offset
 837          * in the directory file.
 838          */
 839         curoff = XFS_DIR2_DATAPTR_TO_BYTE(mp, uio->uio_offset);
 840         /*
 841          * Force this conversion through db so we truncate the offset
 842          * down to get the start of the data block.
 843          */
 844         map_off = XFS_DIR2_DB_TO_DA(mp, XFS_DIR2_BYTE_TO_DB(mp, curoff));
 845         /*
 846          * Loop over directory entries until we reach the end offset.
 847          * Get more blocks and readahead as necessary.
 848          */
 849         while (curoff < XFS_DIR2_LEAF_OFFSET) {
 850                 /*
 851                  * If we have no buffer, or we're off the end of the
 852                  * current buffer, need to get another one.
 853                  */
 854                 if (!bp || ptr >= (char *)bp->data + mp->m_dirblksize) {
 855                         /*
 856                          * If we have a buffer, we need to release it and
 857                          * take it out of the mapping.
 858                          */
 859                         if (bp) {
 860                                 xfs_da_brelse(tp, bp);
 861                                 bp = NULL;
 862                                 map_blocks -= mp->m_dirblkfsbs;
 863                                 /*
 864                                  * Loop to get rid of the extents for the
 865                                  * directory block.
 866                                  */
 867                                 for (i = mp->m_dirblkfsbs; i > 0; ) {
 868                                         j = MIN((int)map->br_blockcount, i);
 869                                         map->br_blockcount -= j;
 870                                         map->br_startblock += j;
 871                                         map->br_startoff += j;
 872                                         /*
 873                                          * If mapping is done, pitch it from
 874                                          * the table.
 875                                          */
 876                                         if (!map->br_blockcount && --map_valid)
 877                                                 memmove(&map[0], &map[1],
 878                                                         sizeof(map[0]) *
 879                                                         map_valid);
 880                                         i -= j;
 881                                 }
 882                         }
 883                         /*
 884                          * Recalculate the readahead blocks wanted.
 885                          */
 886                         ra_want = howmany(uio->uio_resid + mp->m_dirblksize,
 887                                           mp->m_sb.sb_blocksize) - 1;
 888                         /*
 889                          * If we don't have as many as we want, and we haven't
 890                          * run out of data blocks, get some more mappings.
 891                          */
 892                         if (1 + ra_want > map_blocks &&
 893                             map_off <
 894                             XFS_DIR2_BYTE_TO_DA(mp, XFS_DIR2_LEAF_OFFSET)) {
 895                                 /*
 896                                  * Get more bmaps, fill in after the ones
 897                                  * we already have in the table.
 898                                  */
 899                                 nmap = map_size - map_valid;
 900                                 error = xfs_bmapi(tp, dp,
 901                                         map_off,
 902                                         XFS_DIR2_BYTE_TO_DA(mp,
 903                                                 XFS_DIR2_LEAF_OFFSET) - map_off,
 904                                         XFS_BMAPI_METADATA, NULL, 0,
 905                                         &map[map_valid], &nmap, NULL);
 906                                 /*
 907                                  * Don't know if we should ignore this or
 908                                  * try to return an error.
 909                                  * The trouble with returning errors
 910                                  * is that readdir will just stop without
 911                                  * actually passing the error through.
 912                                  */
 913                                 if (error)
 914                                         break;  /* XXX */
 915                                 /*
 916                                  * If we got all the mappings we asked for,
 917                                  * set the final map offset based on the
 918                                  * last bmap value received.
 919                                  * Otherwise, we've reached the end.
 920                                  */
 921                                 if (nmap == map_size - map_valid)
 922                                         map_off =
 923                                         map[map_valid + nmap - 1].br_startoff +
 924                                         map[map_valid + nmap - 1].br_blockcount;
 925                                 else
 926                                         map_off =
 927                                                 XFS_DIR2_BYTE_TO_DA(mp,
 928                                                         XFS_DIR2_LEAF_OFFSET);
 929                                 /*
 930                                  * Look for holes in the mapping, and
 931                                  * eliminate them.  Count up the valid blocks.
 932                                  */
 933                                 for (i = map_valid; i < map_valid + nmap; ) {
 934                                         if (map[i].br_startblock ==
 935                                             HOLESTARTBLOCK) {
 936                                                 nmap--;
 937                                                 length = map_valid + nmap - i;
 938                                                 if (length)
 939                                                         memmove(&map[i],
 940                                                                 &map[i + 1],
 941                                                                 sizeof(map[i]) *
 942                                                                 length);
 943                                         } else {
 944                                                 map_blocks +=
 945                                                         map[i].br_blockcount;
 946                                                 i++;
 947                                         }
 948                                 }
 949                                 map_valid += nmap;
 950                         }
 951                         /*
 952                          * No valid mappings, so no more data blocks.
 953                          */
 954                         if (!map_valid) {
 955                                 curoff = XFS_DIR2_DA_TO_BYTE(mp, map_off);
 956                                 break;
 957                         }
 958                         /*
 959                          * Read the directory block starting at the first
 960                          * mapping.
 961                          */
 962                         curdb = XFS_DIR2_DA_TO_DB(mp, map->br_startoff);
 963                         error = xfs_da_read_buf(tp, dp, map->br_startoff,
 964                                 map->br_blockcount >= mp->m_dirblkfsbs ?
 965                                     XFS_FSB_TO_DADDR(mp, map->br_startblock) :
 966                                     -1,
 967                                 &bp, XFS_DATA_FORK);
 968                         /*
 969                          * Should just skip over the data block instead
 970                          * of giving up.
 971                          */
 972                         if (error)
 973                                 break;  /* XXX */
 974                         /*
 975                          * Adjust the current amount of read-ahead: we just
 976                          * read a block that was previously ra.
 977                          */
 978                         if (ra_current)
 979                                 ra_current -= mp->m_dirblkfsbs;
 980                         /*
 981                          * Do we need more readahead?
 982                          */
 983                         for (ra_index = ra_offset = i = 0;
 984                              ra_want > ra_current && i < map_blocks;
 985                              i += mp->m_dirblkfsbs) {
 986                                 ASSERT(ra_index < map_valid);
 987                                 /*
 988                                  * Read-ahead a contiguous directory block.
 989                                  */
 990                                 if (i > ra_current &&
 991                                     map[ra_index].br_blockcount >=
 992                                     mp->m_dirblkfsbs) {
 993                                         xfs_baread(mp->m_ddev_targp,
 994                                                 XFS_FSB_TO_DADDR(mp,
 995                                                    map[ra_index].br_startblock +
 996                                                    ra_offset),
 997                                                 (int)BTOBB(mp->m_dirblksize));
 998                                         ra_current = i;
 999                                 }
1000                                 /*
1001                                  * Read-ahead a non-contiguous directory block.
1002                                  * This doesn't use our mapping, but this
1003                                  * is a very rare case.
1004                                  */
1005                                 else if (i > ra_current) {
1006                                         (void)xfs_da_reada_buf(tp, dp,
1007                                                 map[ra_index].br_startoff +
1008                                                 ra_offset, XFS_DATA_FORK);
1009                                         ra_current = i;
1010                                 }
1011                                 /*
1012                                  * Advance offset through the mapping table.
1013                                  */
1014                                 for (j = 0; j < mp->m_dirblkfsbs; j++) {
1015                                         /*
1016                                          * The rest of this extent but not
1017                                          * more than a dir block.
1018                                          */
1019                                         length = MIN(mp->m_dirblkfsbs,
1020                                                 (int)(map[ra_index].br_blockcount -
1021                                                 ra_offset));
1022                                         j += length;
1023                                         ra_offset += length;
1024                                         /*
1025                                          * Advance to the next mapping if
1026                                          * this one is used up.
1027                                          */
1028                                         if (ra_offset ==
1029                                             map[ra_index].br_blockcount) {
1030                                                 ra_offset = 0;
1031                                                 ra_index++;
1032                                         }
1033                                 }
1034                         }
1035                         /*
1036                          * Having done a read, we need to set a new offset.
1037                          */
1038                         newoff = XFS_DIR2_DB_OFF_TO_BYTE(mp, curdb, 0);
1039                         /*
1040                          * Start of the current block.
1041                          */
1042                         if (curoff < newoff)
1043                                 curoff = newoff;
1044                         /*
1045                          * Make sure we're in the right block.
1046                          */
1047                         else if (curoff > newoff)
1048                                 ASSERT(XFS_DIR2_BYTE_TO_DB(mp, curoff) ==
1049                                        curdb);
1050                         data = bp->data;
1051                         xfs_dir2_data_check(dp, bp);
1052                         /*
1053                          * Find our position in the block.
1054                          */
1055                         ptr = (char *)&data->u;
1056                         byteoff = XFS_DIR2_BYTE_TO_OFF(mp, curoff);
1057                         /*
1058                          * Skip past the header.
1059                          */
1060                         if (byteoff == 0)
1061                                 curoff += (uint)sizeof(data->hdr);
1062                         /*
1063                          * Skip past entries until we reach our offset.
1064                          */
1065                         else {
1066                                 while ((char *)ptr - (char *)data < byteoff) {
1067                                         dup = (xfs_dir2_data_unused_t *)ptr;
1068
1069                                         if (INT_GET(dup->freetag, ARCH_CONVERT)
1070                                                   == XFS_DIR2_DATA_FREE_TAG) {
1071
1072                                                 length = INT_GET(dup->length,
1073                                                                  ARCH_CONVERT);
1074                                                 ptr += length;
1075                                                 continue;
1076                                         }
1077                                         dep = (xfs_dir2_data_entry_t *)ptr;
1078                                         length =
1079                                            XFS_DIR2_DATA_ENTSIZE(dep->namelen);
1080                                         ptr += length;
1081                                 }
1082                                 /*
1083                                  * Now set our real offset.
1084                                  */
1085                                 curoff =
1086                                         XFS_DIR2_DB_OFF_TO_BYTE(mp,
1087                                             XFS_DIR2_BYTE_TO_DB(mp, curoff),
1088                                             (char *)ptr - (char *)data);
1089                                 if (ptr >= (char *)data + mp->m_dirblksize) {
1090                                         continue;
1091                                 }
1092                         }
1093                 }
1094                 /*
1095                  * We have a pointer to an entry.
1096                  * Is it a live one?
1097                  */
1098                 dup = (xfs_dir2_data_unused_t *)ptr;
1099                 /*
1100                  * No, it's unused, skip over it.
1101                  */
1102                 if (INT_GET(dup->freetag, ARCH_CONVERT)
1103                                                 == XFS_DIR2_DATA_FREE_TAG) {
1104                         length = INT_GET(dup->length, ARCH_CONVERT);
1105                         ptr += length;
1106                         curoff += length;
1107                         continue;
1108                 }
1109
1110                 /*
1111                  * Copy the entry into the putargs, and try formatting it.
1112                  */
1113                 dep = (xfs_dir2_data_entry_t *)ptr;
1114
1115                 p.namelen = dep->namelen;
1116
1117                 length = XFS_DIR2_DATA_ENTSIZE(p.namelen);
1118
1119                 p.cook = XFS_DIR2_BYTE_TO_DATAPTR(mp, curoff + length);
1120
1121                 p.ino = INT_GET(dep->inumber, ARCH_CONVERT);
1122 #if XFS_BIG_INUMS
1123                 p.ino += mp->m_inoadd;
1124 #endif
1125                 p.name = (char *)dep->name;
1126
1127                 error = p.put(&p);
1128
1129                 /*
1130                  * Won't fit.  Return to caller.
1131                  */
1132                 if (!p.done) {
1133                         eof = 0;
1134                         break;
1135                 }
1136                 /*
1137                  * Advance to next entry in the block.
1138                  */
1139                 ptr += length;
1140                 curoff += length;
1141         }
1142
1143         /*
1144          * All done.  Set output offset value to current offset.
1145          */
1146         *eofp = eof;
1147         if (curoff > XFS_DIR2_DATAPTR_TO_BYTE(mp, XFS_DIR2_MAX_DATAPTR))
1148                 uio->uio_offset = XFS_DIR2_MAX_DATAPTR;
1149         else
1150                 uio->uio_offset = XFS_DIR2_BYTE_TO_DATAPTR(mp, curoff);
1151         kmem_free(map, map_size * sizeof(*map));
1152         if (bp)
1153                 xfs_da_brelse(tp, bp);
1154         return error;
1155 }
1156
1157 /*
1158  * Initialize a new leaf block, leaf1 or leafn magic accepted.
1159  */
1160 int
1161 xfs_dir2_leaf_init(
1162         xfs_da_args_t           *args,          /* operation arguments */
1163         xfs_dir2_db_t           bno,            /* directory block number */
1164         xfs_dabuf_t             **bpp,          /* out: leaf buffer */
1165         int                     magic)          /* magic number for block */
1166 {
1167         xfs_dabuf_t             *bp;            /* leaf buffer */
1168         xfs_inode_t             *dp;            /* incore directory inode */
1169         int                     error;          /* error return code */
1170         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1171         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1172         xfs_mount_t             *mp;            /* filesystem mount point */
1173         xfs_trans_t             *tp;            /* transaction pointer */
1174
1175         dp = args->dp;
1176         ASSERT(dp != NULL);
1177         tp = args->trans;
1178         mp = dp->i_mount;
1179         ASSERT(bno >= XFS_DIR2_LEAF_FIRSTDB(mp) &&
1180                bno < XFS_DIR2_FREE_FIRSTDB(mp));
1181         /*
1182          * Get the buffer for the block.
1183          */
1184         error = xfs_da_get_buf(tp, dp, XFS_DIR2_DB_TO_DA(mp, bno), -1, &bp,
1185                 XFS_DATA_FORK);
1186         if (error) {
1187                 return error;
1188         }
1189         ASSERT(bp != NULL);
1190         leaf = bp->data;
1191         /*
1192          * Initialize the header.
1193          */
1194         INT_SET(leaf->hdr.info.magic, ARCH_CONVERT, magic);
1195         leaf->hdr.info.forw = 0;
1196         leaf->hdr.info.back = 0;
1197         leaf->hdr.count = 0;
1198         leaf->hdr.stale = 0;
1199         xfs_dir2_leaf_log_header(tp, bp);
1200         /*
1201          * If it's a leaf-format directory initialize the tail.
1202          * In this case our caller has the real bests table to copy into
1203          * the block.
1204          */
1205         if (magic == XFS_DIR2_LEAF1_MAGIC) {
1206                 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1207                 ltp->bestcount = 0;
1208                 xfs_dir2_leaf_log_tail(tp, bp);
1209         }
1210         *bpp = bp;
1211         return 0;
1212 }
1213
1214 /*
1215  * Log the bests entries indicated from a leaf1 block.
1216  */
1217 void
1218 xfs_dir2_leaf_log_bests(
1219         xfs_trans_t             *tp,            /* transaction pointer */
1220         xfs_dabuf_t             *bp,            /* leaf buffer */
1221         int                     first,          /* first entry to log */
1222         int                     last)           /* last entry to log */
1223 {
1224         xfs_dir2_data_off_t     *firstb;        /* pointer to first entry */
1225         xfs_dir2_data_off_t     *lastb;         /* pointer to last entry */
1226         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1227         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1228
1229         leaf = bp->data;
1230         ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAF1_MAGIC);
1231         ltp = XFS_DIR2_LEAF_TAIL_P(tp->t_mountp, leaf);
1232         firstb = XFS_DIR2_LEAF_BESTS_P(ltp) + first;
1233         lastb = XFS_DIR2_LEAF_BESTS_P(ltp) + last;
1234         xfs_da_log_buf(tp, bp, (uint)((char *)firstb - (char *)leaf),
1235                 (uint)((char *)lastb - (char *)leaf + sizeof(*lastb) - 1));
1236 }
1237
1238 /*
1239  * Log the leaf entries indicated from a leaf1 or leafn block.
1240  */
1241 void
1242 xfs_dir2_leaf_log_ents(
1243         xfs_trans_t             *tp,            /* transaction pointer */
1244         xfs_dabuf_t             *bp,            /* leaf buffer */
1245         int                     first,          /* first entry to log */
1246         int                     last)           /* last entry to log */
1247 {
1248         xfs_dir2_leaf_entry_t   *firstlep;      /* pointer to first entry */
1249         xfs_dir2_leaf_entry_t   *lastlep;       /* pointer to last entry */
1250         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1251
1252         leaf = bp->data;
1253         ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAF1_MAGIC ||
1254                INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAFN_MAGIC);
1255         firstlep = &leaf->ents[first];
1256         lastlep = &leaf->ents[last];
1257         xfs_da_log_buf(tp, bp, (uint)((char *)firstlep - (char *)leaf),
1258                 (uint)((char *)lastlep - (char *)leaf + sizeof(*lastlep) - 1));
1259 }
1260
1261 /*
1262  * Log the header of the leaf1 or leafn block.
1263  */
1264 void
1265 xfs_dir2_leaf_log_header(
1266         xfs_trans_t             *tp,            /* transaction pointer */
1267         xfs_dabuf_t             *bp)            /* leaf buffer */
1268 {
1269         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1270
1271         leaf = bp->data;
1272         ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAF1_MAGIC ||
1273                INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAFN_MAGIC);
1274         xfs_da_log_buf(tp, bp, (uint)((char *)&leaf->hdr - (char *)leaf),
1275                 (uint)(sizeof(leaf->hdr) - 1));
1276 }
1277
1278 /*
1279  * Log the tail of the leaf1 block.
1280  */
1281 void
1282 xfs_dir2_leaf_log_tail(
1283         xfs_trans_t             *tp,            /* transaction pointer */
1284         xfs_dabuf_t             *bp)            /* leaf buffer */
1285 {
1286         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1287         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1288         xfs_mount_t             *mp;            /* filesystem mount point */
1289
1290         mp = tp->t_mountp;
1291         leaf = bp->data;
1292         ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAF1_MAGIC);
1293         ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1294         xfs_da_log_buf(tp, bp, (uint)((char *)ltp - (char *)leaf),
1295                 (uint)(mp->m_dirblksize - 1));
1296 }
1297
1298 /*
1299  * Look up the entry referred to by args in the leaf format directory.
1300  * Most of the work is done by the xfs_dir2_leaf_lookup_int routine which
1301  * is also used by the node-format code.
1302  */
1303 int
1304 xfs_dir2_leaf_lookup(
1305         xfs_da_args_t           *args)          /* operation arguments */
1306 {
1307         xfs_dabuf_t             *dbp;           /* data block buffer */
1308         xfs_dir2_data_entry_t   *dep;           /* data block entry */
1309         xfs_inode_t             *dp;            /* incore directory inode */
1310         int                     error;          /* error return code */
1311         int                     index;          /* found entry index */
1312         xfs_dabuf_t             *lbp;           /* leaf buffer */
1313         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1314         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1315         xfs_trans_t             *tp;            /* transaction pointer */
1316
1317         xfs_dir2_trace_args("leaf_lookup", args);
1318         /*
1319          * Look up name in the leaf block, returning both buffers and index.
1320          */
1321         if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1322                 return error;
1323         }
1324         tp = args->trans;
1325         dp = args->dp;
1326         xfs_dir2_leaf_check(dp, lbp);
1327         leaf = lbp->data;
1328         /*
1329          * Get to the leaf entry and contained data entry address.
1330          */
1331         lep = &leaf->ents[index];
1332         /*
1333          * Point to the data entry.
1334          */
1335         dep = (xfs_dir2_data_entry_t *)
1336               ((char *)dbp->data +
1337                XFS_DIR2_DATAPTR_TO_OFF(dp->i_mount, INT_GET(lep->address, ARCH_CONVERT)));
1338         /*
1339          * Return the found inode number.
1340          */
1341         args->inumber = INT_GET(dep->inumber, ARCH_CONVERT);
1342         xfs_da_brelse(tp, dbp);
1343         xfs_da_brelse(tp, lbp);
1344         return XFS_ERROR(EEXIST);
1345 }
1346
1347 /*
1348  * Look up name/hash in the leaf block.
1349  * Fill in indexp with the found index, and dbpp with the data buffer.
1350  * If not found dbpp will be NULL, and ENOENT comes back.
1351  * lbpp will always be filled in with the leaf buffer unless there's an error.
1352  */
1353 static int                                      /* error */
1354 xfs_dir2_leaf_lookup_int(
1355         xfs_da_args_t           *args,          /* operation arguments */
1356         xfs_dabuf_t             **lbpp,         /* out: leaf buffer */
1357         int                     *indexp,        /* out: index in leaf block */
1358         xfs_dabuf_t             **dbpp)         /* out: data buffer */
1359 {
1360         xfs_dir2_db_t           curdb;          /* current data block number */
1361         xfs_dabuf_t             *dbp;           /* data buffer */
1362         xfs_dir2_data_entry_t   *dep;           /* data entry */
1363         xfs_inode_t             *dp;            /* incore directory inode */
1364         int                     error;          /* error return code */
1365         int                     index;          /* index in leaf block */
1366         xfs_dabuf_t             *lbp;           /* leaf buffer */
1367         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1368         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1369         xfs_mount_t             *mp;            /* filesystem mount point */
1370         xfs_dir2_db_t           newdb;          /* new data block number */
1371         xfs_trans_t             *tp;            /* transaction pointer */
1372
1373         dp = args->dp;
1374         tp = args->trans;
1375         mp = dp->i_mount;
1376         /*
1377          * Read the leaf block into the buffer.
1378          */
1379         if ((error =
1380             xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
1381                     XFS_DATA_FORK))) {
1382                 return error;
1383         }
1384         *lbpp = lbp;
1385         leaf = lbp->data;
1386         xfs_dir2_leaf_check(dp, lbp);
1387         /*
1388          * Look for the first leaf entry with our hash value.
1389          */
1390         index = xfs_dir2_leaf_search_hash(args, lbp);
1391         /*
1392          * Loop over all the entries with the right hash value
1393          * looking to match the name.
1394          */
1395         for (lep = &leaf->ents[index], dbp = NULL, curdb = -1;
1396              index < INT_GET(leaf->hdr.count, ARCH_CONVERT) && INT_GET(lep->hashval, ARCH_CONVERT) == args->hashval;
1397              lep++, index++) {
1398                 /*
1399                  * Skip over stale leaf entries.
1400                  */
1401                 if (INT_GET(lep->address, ARCH_CONVERT) == XFS_DIR2_NULL_DATAPTR)
1402                         continue;
1403                 /*
1404                  * Get the new data block number.
1405                  */
1406                 newdb = XFS_DIR2_DATAPTR_TO_DB(mp, INT_GET(lep->address, ARCH_CONVERT));
1407                 /*
1408                  * If it's not the same as the old data block number,
1409                  * need to pitch the old one and read the new one.
1410                  */
1411                 if (newdb != curdb) {
1412                         if (dbp)
1413                                 xfs_da_brelse(tp, dbp);
1414                         if ((error =
1415                             xfs_da_read_buf(tp, dp,
1416                                     XFS_DIR2_DB_TO_DA(mp, newdb), -1, &dbp,
1417                                     XFS_DATA_FORK))) {
1418                                 xfs_da_brelse(tp, lbp);
1419                                 return error;
1420                         }
1421                         xfs_dir2_data_check(dp, dbp);
1422                         curdb = newdb;
1423                 }
1424                 /*
1425                  * Point to the data entry.
1426                  */
1427                 dep = (xfs_dir2_data_entry_t *)
1428                       ((char *)dbp->data +
1429                        XFS_DIR2_DATAPTR_TO_OFF(mp, INT_GET(lep->address, ARCH_CONVERT)));
1430                 /*
1431                  * If it matches then return it.
1432                  */
1433                 if (dep->namelen == args->namelen &&
1434                     dep->name[0] == args->name[0] &&
1435                     memcmp(dep->name, args->name, args->namelen) == 0) {
1436                         *dbpp = dbp;
1437                         *indexp = index;
1438                         return 0;
1439                 }
1440         }
1441         /*
1442          * No match found, return ENOENT.
1443          */
1444         ASSERT(args->oknoent);
1445         if (dbp)
1446                 xfs_da_brelse(tp, dbp);
1447         xfs_da_brelse(tp, lbp);
1448         return XFS_ERROR(ENOENT);
1449 }
1450
1451 /*
1452  * Remove an entry from a leaf format directory.
1453  */
1454 int                                             /* error */
1455 xfs_dir2_leaf_removename(
1456         xfs_da_args_t           *args)          /* operation arguments */
1457 {
1458         xfs_dir2_data_off_t     *bestsp;        /* leaf block best freespace */
1459         xfs_dir2_data_t         *data;          /* data block structure */
1460         xfs_dir2_db_t           db;             /* data block number */
1461         xfs_dabuf_t             *dbp;           /* data block buffer */
1462         xfs_dir2_data_entry_t   *dep;           /* data entry structure */
1463         xfs_inode_t             *dp;            /* incore directory inode */
1464         int                     error;          /* error return code */
1465         xfs_dir2_db_t           i;              /* temporary data block # */
1466         int                     index;          /* index into leaf entries */
1467         xfs_dabuf_t             *lbp;           /* leaf buffer */
1468         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1469         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1470         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1471         xfs_mount_t             *mp;            /* filesystem mount point */
1472         int                     needlog;        /* need to log data header */
1473         int                     needscan;       /* need to rescan data frees */
1474         xfs_dir2_data_off_t     oldbest;        /* old value of best free */
1475         xfs_trans_t             *tp;            /* transaction pointer */
1476
1477         xfs_dir2_trace_args("leaf_removename", args);
1478         /*
1479          * Lookup the leaf entry, get the leaf and data blocks read in.
1480          */
1481         if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1482                 return error;
1483         }
1484         dp = args->dp;
1485         tp = args->trans;
1486         mp = dp->i_mount;
1487         leaf = lbp->data;
1488         data = dbp->data;
1489         xfs_dir2_data_check(dp, dbp);
1490         /*
1491          * Point to the leaf entry, use that to point to the data entry.
1492          */
1493         lep = &leaf->ents[index];
1494         db = XFS_DIR2_DATAPTR_TO_DB(mp, INT_GET(lep->address, ARCH_CONVERT));
1495         dep = (xfs_dir2_data_entry_t *)
1496               ((char *)data + XFS_DIR2_DATAPTR_TO_OFF(mp, INT_GET(lep->address, ARCH_CONVERT)));
1497         needscan = needlog = 0;
1498         oldbest = INT_GET(data->hdr.bestfree[0].length, ARCH_CONVERT);
1499         ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1500         bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
1501         ASSERT(INT_GET(bestsp[db], ARCH_CONVERT) == oldbest);
1502         /*
1503          * Mark the former data entry unused.
1504          */
1505         xfs_dir2_data_make_free(tp, dbp,
1506                 (xfs_dir2_data_aoff_t)((char *)dep - (char *)data),
1507                 XFS_DIR2_DATA_ENTSIZE(dep->namelen), &needlog, &needscan);
1508         /*
1509          * We just mark the leaf entry stale by putting a null in it.
1510          */
1511         INT_MOD(leaf->hdr.stale, ARCH_CONVERT, +1);
1512         xfs_dir2_leaf_log_header(tp, lbp);
1513         INT_SET(lep->address, ARCH_CONVERT, XFS_DIR2_NULL_DATAPTR);
1514         xfs_dir2_leaf_log_ents(tp, lbp, index, index);
1515         /*
1516          * Scan the freespace in the data block again if necessary,
1517          * log the data block header if necessary.
1518          */
1519         if (needscan)
1520                 xfs_dir2_data_freescan(mp, data, &needlog, NULL);
1521         if (needlog)
1522                 xfs_dir2_data_log_header(tp, dbp);
1523         /*
1524          * If the longest freespace in the data block has changed,
1525          * put the new value in the bests table and log that.
1526          */
1527         if (INT_GET(data->hdr.bestfree[0].length, ARCH_CONVERT) != oldbest) {
1528                 INT_COPY(bestsp[db], data->hdr.bestfree[0].length, ARCH_CONVERT);
1529                 xfs_dir2_leaf_log_bests(tp, lbp, db, db);
1530         }
1531         xfs_dir2_data_check(dp, dbp);
1532         /*
1533          * If the data block is now empty then get rid of the data block.
1534          */
1535         if (INT_GET(data->hdr.bestfree[0].length, ARCH_CONVERT) ==
1536             mp->m_dirblksize - (uint)sizeof(data->hdr)) {
1537                 ASSERT(db != mp->m_dirdatablk);
1538                 if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1539                         /*
1540                          * Nope, can't get rid of it because it caused
1541                          * allocation of a bmap btree block to do so.
1542                          * Just go on, returning success, leaving the
1543                          * empty block in place.
1544                          */
1545                         if (error == ENOSPC && args->total == 0) {
1546                                 xfs_da_buf_done(dbp);
1547                                 error = 0;
1548                         }
1549                         xfs_dir2_leaf_check(dp, lbp);
1550                         xfs_da_buf_done(lbp);
1551                         return error;
1552                 }
1553                 dbp = NULL;
1554                 /*
1555                  * If this is the last data block then compact the
1556                  * bests table by getting rid of entries.
1557                  */
1558                 if (db == INT_GET(ltp->bestcount, ARCH_CONVERT) - 1) {
1559                         /*
1560                          * Look for the last active entry (i).
1561                          */
1562                         for (i = db - 1; i > 0; i--) {
1563                                 if (INT_GET(bestsp[i], ARCH_CONVERT) != NULLDATAOFF)
1564                                         break;
1565                         }
1566                         /*
1567                          * Copy the table down so inactive entries at the
1568                          * end are removed.
1569                          */
1570                         memmove(&bestsp[db - i], bestsp,
1571                                 (INT_GET(ltp->bestcount, ARCH_CONVERT) - (db - i)) * sizeof(*bestsp));
1572                         INT_MOD(ltp->bestcount, ARCH_CONVERT, -(db - i));
1573                         xfs_dir2_leaf_log_tail(tp, lbp);
1574                         xfs_dir2_leaf_log_bests(tp, lbp, 0, INT_GET(ltp->bestcount, ARCH_CONVERT) - 1);
1575                 } else
1576                         INT_SET(bestsp[db], ARCH_CONVERT, NULLDATAOFF);
1577         }
1578         /*
1579          * If the data block was not the first one, drop it.
1580          */
1581         else if (db != mp->m_dirdatablk && dbp != NULL) {
1582                 xfs_da_buf_done(dbp);
1583                 dbp = NULL;
1584         }
1585         xfs_dir2_leaf_check(dp, lbp);
1586         /*
1587          * See if we can convert to block form.
1588          */
1589         return xfs_dir2_leaf_to_block(args, lbp, dbp);
1590 }
1591
1592 /*
1593  * Replace the inode number in a leaf format directory entry.
1594  */
1595 int                                             /* error */
1596 xfs_dir2_leaf_replace(
1597         xfs_da_args_t           *args)          /* operation arguments */
1598 {
1599         xfs_dabuf_t             *dbp;           /* data block buffer */
1600         xfs_dir2_data_entry_t   *dep;           /* data block entry */
1601         xfs_inode_t             *dp;            /* incore directory inode */
1602         int                     error;          /* error return code */
1603         int                     index;          /* index of leaf entry */
1604         xfs_dabuf_t             *lbp;           /* leaf buffer */
1605         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1606         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1607         xfs_trans_t             *tp;            /* transaction pointer */
1608
1609         xfs_dir2_trace_args("leaf_replace", args);
1610         /*
1611          * Look up the entry.
1612          */
1613         if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1614                 return error;
1615         }
1616         dp = args->dp;
1617         leaf = lbp->data;
1618         /*
1619          * Point to the leaf entry, get data address from it.
1620          */
1621         lep = &leaf->ents[index];
1622         /*
1623          * Point to the data entry.
1624          */
1625         dep = (xfs_dir2_data_entry_t *)
1626               ((char *)dbp->data +
1627                XFS_DIR2_DATAPTR_TO_OFF(dp->i_mount, INT_GET(lep->address, ARCH_CONVERT)));
1628         ASSERT(args->inumber != INT_GET(dep->inumber, ARCH_CONVERT));
1629         /*
1630          * Put the new inode number in, log it.
1631          */
1632         INT_SET(dep->inumber, ARCH_CONVERT, args->inumber);
1633         tp = args->trans;
1634         xfs_dir2_data_log_entry(tp, dbp, dep);
1635         xfs_da_buf_done(dbp);
1636         xfs_dir2_leaf_check(dp, lbp);
1637         xfs_da_brelse(tp, lbp);
1638         return 0;
1639 }
1640
1641 /*
1642  * Return index in the leaf block (lbp) which is either the first
1643  * one with this hash value, or if there are none, the insert point
1644  * for that hash value.
1645  */
1646 int                                             /* index value */
1647 xfs_dir2_leaf_search_hash(
1648         xfs_da_args_t           *args,          /* operation arguments */
1649         xfs_dabuf_t             *lbp)           /* leaf buffer */
1650 {
1651         xfs_dahash_t            hash=0;         /* hash from this entry */
1652         xfs_dahash_t            hashwant;       /* hash value looking for */
1653         int                     high;           /* high leaf index */
1654         int                     low;            /* low leaf index */
1655         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1656         xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1657         int                     mid=0;          /* current leaf index */
1658
1659         leaf = lbp->data;
1660 #ifndef __KERNEL__
1661         if (!leaf->hdr.count)
1662                 return 0;
1663 #endif
1664         /*
1665          * Note, the table cannot be empty, so we have to go through the loop.
1666          * Binary search the leaf entries looking for our hash value.
1667          */
1668         for (lep = leaf->ents, low = 0, high = INT_GET(leaf->hdr.count, ARCH_CONVERT) - 1,
1669                 hashwant = args->hashval;
1670              low <= high; ) {
1671                 mid = (low + high) >> 1;
1672                 if ((hash = INT_GET(lep[mid].hashval, ARCH_CONVERT)) == hashwant)
1673                         break;
1674                 if (hash < hashwant)
1675                         low = mid + 1;
1676                 else
1677                         high = mid - 1;
1678         }
1679         /*
1680          * Found one, back up through all the equal hash values.
1681          */
1682         if (hash == hashwant) {
1683                 while (mid > 0 && INT_GET(lep[mid - 1].hashval, ARCH_CONVERT) == hashwant) {
1684                         mid--;
1685                 }
1686         }
1687         /*
1688          * Need to point to an entry higher than ours.
1689          */
1690         else if (hash < hashwant)
1691                 mid++;
1692         return mid;
1693 }
1694
1695 /*
1696  * Trim off a trailing data block.  We know it's empty since the leaf
1697  * freespace table says so.
1698  */
1699 int                                             /* error */
1700 xfs_dir2_leaf_trim_data(
1701         xfs_da_args_t           *args,          /* operation arguments */
1702         xfs_dabuf_t             *lbp,           /* leaf buffer */
1703         xfs_dir2_db_t           db)             /* data block number */
1704 {
1705         xfs_dir2_data_off_t     *bestsp;        /* leaf bests table */
1706 #ifdef DEBUG
1707         xfs_dir2_data_t         *data;          /* data block structure */
1708 #endif
1709         xfs_dabuf_t             *dbp;           /* data block buffer */
1710         xfs_inode_t             *dp;            /* incore directory inode */
1711         int                     error;          /* error return value */
1712         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1713         xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1714         xfs_mount_t             *mp;            /* filesystem mount point */
1715         xfs_trans_t             *tp;            /* transaction pointer */
1716
1717         dp = args->dp;
1718         mp = dp->i_mount;
1719         tp = args->trans;
1720         /*
1721          * Read the offending data block.  We need its buffer.
1722          */
1723         if ((error = xfs_da_read_buf(tp, dp, XFS_DIR2_DB_TO_DA(mp, db), -1, &dbp,
1724                         XFS_DATA_FORK))) {
1725                 return error;
1726         }
1727 #ifdef DEBUG
1728         data = dbp->data;
1729         ASSERT(INT_GET(data->hdr.magic, ARCH_CONVERT) == XFS_DIR2_DATA_MAGIC);
1730 #endif
1731         /* this seems to be an error
1732          * data is only valid if DEBUG is defined?
1733          * RMC 09/08/1999
1734          */
1735
1736         leaf = lbp->data;
1737         ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1738         ASSERT(INT_GET(data->hdr.bestfree[0].length, ARCH_CONVERT) ==
1739                mp->m_dirblksize - (uint)sizeof(data->hdr));
1740         ASSERT(db == INT_GET(ltp->bestcount, ARCH_CONVERT) - 1);
1741         /*
1742          * Get rid of the data block.
1743          */
1744         if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1745                 ASSERT(error != ENOSPC);
1746                 xfs_da_brelse(tp, dbp);
1747                 return error;
1748         }
1749         /*
1750          * Eliminate the last bests entry from the table.
1751          */
1752         bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
1753         INT_MOD(ltp->bestcount, ARCH_CONVERT, -1);
1754         memmove(&bestsp[1], &bestsp[0], INT_GET(ltp->bestcount, ARCH_CONVERT) * sizeof(*bestsp));
1755         xfs_dir2_leaf_log_tail(tp, lbp);
1756         xfs_dir2_leaf_log_bests(tp, lbp, 0, INT_GET(ltp->bestcount, ARCH_CONVERT) - 1);
1757         return 0;
1758 }
1759
1760 /*
1761  * Convert node form directory to leaf form directory.
1762  * The root of the node form dir needs to already be a LEAFN block.
1763  * Just return if we can't do anything.
1764  */
1765 int                                             /* error */
1766 xfs_dir2_node_to_leaf(
1767         xfs_da_state_t          *state)         /* directory operation state */
1768 {
1769         xfs_da_args_t           *args;          /* operation arguments */
1770         xfs_inode_t             *dp;            /* incore directory inode */
1771         int                     error;          /* error return code */
1772         xfs_dabuf_t             *fbp;           /* buffer for freespace block */
1773         xfs_fileoff_t           fo;             /* freespace file offset */
1774         xfs_dir2_free_t         *free;          /* freespace structure */
1775         xfs_dabuf_t             *lbp;           /* buffer for leaf block */
1776         xfs_dir2_leaf_tail_t    *ltp;           /* tail of leaf structure */
1777         xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1778         xfs_mount_t             *mp;            /* filesystem mount point */
1779         int                     rval;           /* successful free trim? */
1780         xfs_trans_t             *tp;            /* transaction pointer */
1781
1782         /*
1783          * There's more than a leaf level in the btree, so there must
1784          * be multiple leafn blocks.  Give up.
1785          */
1786         if (state->path.active > 1)
1787                 return 0;
1788         args = state->args;
1789         xfs_dir2_trace_args("node_to_leaf", args);
1790         mp = state->mp;
1791         dp = args->dp;
1792         tp = args->trans;
1793         /*
1794          * Get the last offset in the file.
1795          */
1796         if ((error = xfs_bmap_last_offset(tp, dp, &fo, XFS_DATA_FORK))) {
1797                 return error;
1798         }
1799         fo -= mp->m_dirblkfsbs;
1800         /*
1801          * If there are freespace blocks other than the first one,
1802          * take this opportunity to remove trailing empty freespace blocks
1803          * that may have been left behind during no-space-reservation
1804          * operations.
1805          */
1806         while (fo > mp->m_dirfreeblk) {
1807                 if ((error = xfs_dir2_node_trim_free(args, fo, &rval))) {
1808                         return error;
1809                 }
1810                 if (rval)
1811                         fo -= mp->m_dirblkfsbs;
1812                 else
1813                         return 0;
1814         }
1815         /*
1816          * Now find the block just before the freespace block.
1817          */
1818         if ((error = xfs_bmap_last_before(tp, dp, &fo, XFS_DATA_FORK))) {
1819                 return error;
1820         }
1821         /*
1822          * If it's not the single leaf block, give up.
1823          */
1824         if (XFS_FSB_TO_B(mp, fo) > XFS_DIR2_LEAF_OFFSET + mp->m_dirblksize)
1825                 return 0;
1826         lbp = state->path.blk[0].bp;
1827         leaf = lbp->data;
1828         ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAFN_MAGIC);
1829         /*
1830          * Read the freespace block.
1831          */
1832         if ((error = xfs_da_read_buf(tp, dp, mp->m_dirfreeblk, -1, &fbp,
1833                         XFS_DATA_FORK))) {
1834                 return error;
1835         }
1836         free = fbp->data;
1837         ASSERT(INT_GET(free->hdr.magic, ARCH_CONVERT) == XFS_DIR2_FREE_MAGIC);
1838         ASSERT(!free->hdr.firstdb);
1839         /*
1840          * Now see if the leafn and free data will fit in a leaf1.
1841          * If not, release the buffer and give up.
1842          */
1843         if ((uint)sizeof(leaf->hdr) +
1844             (INT_GET(leaf->hdr.count, ARCH_CONVERT) - INT_GET(leaf->hdr.stale, ARCH_CONVERT)) * (uint)sizeof(leaf->ents[0]) +
1845             INT_GET(free->hdr.nvalid, ARCH_CONVERT) * (uint)sizeof(leaf->bests[0]) +
1846             (uint)sizeof(leaf->tail) >
1847             mp->m_dirblksize) {
1848                 xfs_da_brelse(tp, fbp);
1849                 return 0;
1850         }
1851         /*
1852          * If the leaf has any stale entries in it, compress them out.
1853          * The compact routine will log the header.
1854          */
1855         if (INT_GET(leaf->hdr.stale, ARCH_CONVERT))
1856                 xfs_dir2_leaf_compact(args, lbp);
1857         else
1858                 xfs_dir2_leaf_log_header(tp, lbp);
1859         INT_SET(leaf->hdr.info.magic, ARCH_CONVERT, XFS_DIR2_LEAF1_MAGIC);
1860         /*
1861          * Set up the leaf tail from the freespace block.
1862          */
1863         ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1864         INT_COPY(ltp->bestcount, free->hdr.nvalid, ARCH_CONVERT);
1865         /*
1866          * Set up the leaf bests table.
1867          */
1868         memcpy(XFS_DIR2_LEAF_BESTS_P(ltp), free->bests,
1869                 INT_GET(ltp->bestcount, ARCH_CONVERT) * sizeof(leaf->bests[0]));
1870         xfs_dir2_leaf_log_bests(tp, lbp, 0, INT_GET(ltp->bestcount, ARCH_CONVERT) - 1);
1871         xfs_dir2_leaf_log_tail(tp, lbp);
1872         xfs_dir2_leaf_check(dp, lbp);
1873         /*
1874          * Get rid of the freespace block.
1875          */
1876         error = xfs_dir2_shrink_inode(args, XFS_DIR2_FREE_FIRSTDB(mp), fbp);
1877         if (error) {
1878                 /*
1879                  * This can't fail here because it can only happen when
1880                  * punching out the middle of an extent, and this is an
1881                  * isolated block.
1882                  */
1883                 ASSERT(error != ENOSPC);
1884                 return error;
1885         }
1886         fbp = NULL;
1887         /*
1888          * Now see if we can convert the single-leaf directory
1889          * down to a block form directory.
1890          * This routine always kills the dabuf for the leaf, so
1891          * eliminate it from the path.
1892          */
1893         error = xfs_dir2_leaf_to_block(args, lbp, NULL);
1894         state->path.blk[0].bp = NULL;
1895         return error;
1896 }