linux/fs/xfs/xfs_dir2_leaf.c
<<
>>
Prefs
   1/*
   2 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
   3 * All Rights Reserved.
   4 *
   5 * This program is free software; you can redistribute it and/or
   6 * modify it under the terms of the GNU General Public License as
   7 * published by the Free Software Foundation.
   8 *
   9 * This program is distributed in the hope that it would be useful,
  10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12 * GNU General Public License for more details.
  13 *
  14 * You should have received a copy of the GNU General Public License
  15 * along with this program; if not, write the Free Software Foundation,
  16 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  17 */
  18#include "xfs.h"
  19#include "xfs_fs.h"
  20#include "xfs_types.h"
  21#include "xfs_bit.h"
  22#include "xfs_log.h"
  23#include "xfs_trans.h"
  24#include "xfs_sb.h"
  25#include "xfs_ag.h"
  26#include "xfs_mount.h"
  27#include "xfs_da_btree.h"
  28#include "xfs_bmap_btree.h"
  29#include "xfs_dinode.h"
  30#include "xfs_inode.h"
  31#include "xfs_bmap.h"
  32#include "xfs_dir2_format.h"
  33#include "xfs_dir2_priv.h"
  34#include "xfs_error.h"
  35#include "xfs_trace.h"
  36
  37/*
  38 * Local function declarations.
  39 */
  40#ifdef DEBUG
  41static void xfs_dir2_leaf_check(struct xfs_inode *dp, struct xfs_buf *bp);
  42#else
  43#define xfs_dir2_leaf_check(dp, bp)
  44#endif
  45static int xfs_dir2_leaf_lookup_int(xfs_da_args_t *args, struct xfs_buf **lbpp,
  46                                    int *indexp, struct xfs_buf **dbpp);
  47static void xfs_dir2_leaf_log_bests(struct xfs_trans *tp, struct xfs_buf *bp,
  48                                    int first, int last);
  49static void xfs_dir2_leaf_log_tail(struct xfs_trans *tp, struct xfs_buf *bp);
  50
  51
  52/*
  53 * Convert a block form directory to a leaf form directory.
  54 */
  55int                                             /* error */
  56xfs_dir2_block_to_leaf(
  57        xfs_da_args_t           *args,          /* operation arguments */
  58        struct xfs_buf          *dbp)           /* input block's buffer */
  59{
  60        __be16                  *bestsp;        /* leaf's bestsp entries */
  61        xfs_dablk_t             blkno;          /* leaf block's bno */
  62        xfs_dir2_data_hdr_t     *hdr;           /* block header */
  63        xfs_dir2_leaf_entry_t   *blp;           /* block's leaf entries */
  64        xfs_dir2_block_tail_t   *btp;           /* block's tail */
  65        xfs_inode_t             *dp;            /* incore directory inode */
  66        int                     error;          /* error return code */
  67        struct xfs_buf          *lbp;           /* leaf block's buffer */
  68        xfs_dir2_db_t           ldb;            /* leaf block's bno */
  69        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
  70        xfs_dir2_leaf_tail_t    *ltp;           /* leaf's tail */
  71        xfs_mount_t             *mp;            /* filesystem mount point */
  72        int                     needlog;        /* need to log block header */
  73        int                     needscan;       /* need to rescan bestfree */
  74        xfs_trans_t             *tp;            /* transaction pointer */
  75
  76        trace_xfs_dir2_block_to_leaf(args);
  77
  78        dp = args->dp;
  79        mp = dp->i_mount;
  80        tp = args->trans;
  81        /*
  82         * Add the leaf block to the inode.
  83         * This interface will only put blocks in the leaf/node range.
  84         * Since that's empty now, we'll get the root (block 0 in range).
  85         */
  86        if ((error = xfs_da_grow_inode(args, &blkno))) {
  87                return error;
  88        }
  89        ldb = xfs_dir2_da_to_db(mp, blkno);
  90        ASSERT(ldb == XFS_DIR2_LEAF_FIRSTDB(mp));
  91        /*
  92         * Initialize the leaf block, get a buffer for it.
  93         */
  94        if ((error = xfs_dir2_leaf_init(args, ldb, &lbp, XFS_DIR2_LEAF1_MAGIC))) {
  95                return error;
  96        }
  97        ASSERT(lbp != NULL);
  98        leaf = lbp->b_addr;
  99        hdr = dbp->b_addr;
 100        xfs_dir2_data_check(dp, dbp);
 101        btp = xfs_dir2_block_tail_p(mp, hdr);
 102        blp = xfs_dir2_block_leaf_p(btp);
 103        /*
 104         * Set the counts in the leaf header.
 105         */
 106        leaf->hdr.count = cpu_to_be16(be32_to_cpu(btp->count));
 107        leaf->hdr.stale = cpu_to_be16(be32_to_cpu(btp->stale));
 108        /*
 109         * Could compact these but I think we always do the conversion
 110         * after squeezing out stale entries.
 111         */
 112        memcpy(leaf->ents, blp, be32_to_cpu(btp->count) * sizeof(xfs_dir2_leaf_entry_t));
 113        xfs_dir2_leaf_log_ents(tp, lbp, 0, be16_to_cpu(leaf->hdr.count) - 1);
 114        needscan = 0;
 115        needlog = 1;
 116        /*
 117         * Make the space formerly occupied by the leaf entries and block
 118         * tail be free.
 119         */
 120        xfs_dir2_data_make_free(tp, dbp,
 121                (xfs_dir2_data_aoff_t)((char *)blp - (char *)hdr),
 122                (xfs_dir2_data_aoff_t)((char *)hdr + mp->m_dirblksize -
 123                                       (char *)blp),
 124                &needlog, &needscan);
 125        /*
 126         * Fix up the block header, make it a data block.
 127         */
 128        hdr->magic = cpu_to_be32(XFS_DIR2_DATA_MAGIC);
 129        if (needscan)
 130                xfs_dir2_data_freescan(mp, hdr, &needlog);
 131        /*
 132         * Set up leaf tail and bests table.
 133         */
 134        ltp = xfs_dir2_leaf_tail_p(mp, leaf);
 135        ltp->bestcount = cpu_to_be32(1);
 136        bestsp = xfs_dir2_leaf_bests_p(ltp);
 137        bestsp[0] =  hdr->bestfree[0].length;
 138        /*
 139         * Log the data header and leaf bests table.
 140         */
 141        if (needlog)
 142                xfs_dir2_data_log_header(tp, dbp);
 143        xfs_dir2_leaf_check(dp, lbp);
 144        xfs_dir2_data_check(dp, dbp);
 145        xfs_dir2_leaf_log_bests(tp, lbp, 0, 0);
 146        return 0;
 147}
 148
 149STATIC void
 150xfs_dir2_leaf_find_stale(
 151        struct xfs_dir2_leaf    *leaf,
 152        int                     index,
 153        int                     *lowstale,
 154        int                     *highstale)
 155{
 156        /*
 157         * Find the first stale entry before our index, if any.
 158         */
 159        for (*lowstale = index - 1; *lowstale >= 0; --*lowstale) {
 160                if (leaf->ents[*lowstale].address ==
 161                    cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 162                        break;
 163        }
 164
 165        /*
 166         * Find the first stale entry at or after our index, if any.
 167         * Stop if the result would require moving more entries than using
 168         * lowstale.
 169         */
 170        for (*highstale = index;
 171             *highstale < be16_to_cpu(leaf->hdr.count);
 172             ++*highstale) {
 173                if (leaf->ents[*highstale].address ==
 174                    cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 175                        break;
 176                if (*lowstale >= 0 && index - *lowstale <= *highstale - index)
 177                        break;
 178        }
 179}
 180
 181struct xfs_dir2_leaf_entry *
 182xfs_dir2_leaf_find_entry(
 183        xfs_dir2_leaf_t         *leaf,          /* leaf structure */
 184        int                     index,          /* leaf table position */
 185        int                     compact,        /* need to compact leaves */
 186        int                     lowstale,       /* index of prev stale leaf */
 187        int                     highstale,      /* index of next stale leaf */
 188        int                     *lfloglow,      /* low leaf logging index */
 189        int                     *lfloghigh)     /* high leaf logging index */
 190{
 191        if (!leaf->hdr.stale) {
 192                xfs_dir2_leaf_entry_t   *lep;   /* leaf entry table pointer */
 193
 194                /*
 195                 * Now we need to make room to insert the leaf entry.
 196                 *
 197                 * If there are no stale entries, just insert a hole at index.
 198                 */
 199                lep = &leaf->ents[index];
 200                if (index < be16_to_cpu(leaf->hdr.count))
 201                        memmove(lep + 1, lep,
 202                                (be16_to_cpu(leaf->hdr.count) - index) *
 203                                 sizeof(*lep));
 204
 205                /*
 206                 * Record low and high logging indices for the leaf.
 207                 */
 208                *lfloglow = index;
 209                *lfloghigh = be16_to_cpu(leaf->hdr.count);
 210                be16_add_cpu(&leaf->hdr.count, 1);
 211                return lep;
 212        }
 213
 214        /*
 215         * There are stale entries.
 216         *
 217         * We will use one of them for the new entry.  It's probably not at
 218         * the right location, so we'll have to shift some up or down first.
 219         *
 220         * If we didn't compact before, we need to find the nearest stale
 221         * entries before and after our insertion point.
 222         */
 223        if (compact == 0)
 224                xfs_dir2_leaf_find_stale(leaf, index, &lowstale, &highstale);
 225
 226        /*
 227         * If the low one is better, use it.
 228         */
 229        if (lowstale >= 0 &&
 230            (highstale == be16_to_cpu(leaf->hdr.count) ||
 231             index - lowstale - 1 < highstale - index)) {
 232                ASSERT(index - lowstale - 1 >= 0);
 233                ASSERT(leaf->ents[lowstale].address ==
 234                       cpu_to_be32(XFS_DIR2_NULL_DATAPTR));
 235
 236                /*
 237                 * Copy entries up to cover the stale entry and make room
 238                 * for the new entry.
 239                 */
 240                if (index - lowstale - 1 > 0) {
 241                        memmove(&leaf->ents[lowstale],
 242                                &leaf->ents[lowstale + 1],
 243                                (index - lowstale - 1) *
 244                                sizeof(xfs_dir2_leaf_entry_t));
 245                }
 246                *lfloglow = MIN(lowstale, *lfloglow);
 247                *lfloghigh = MAX(index - 1, *lfloghigh);
 248                be16_add_cpu(&leaf->hdr.stale, -1);
 249                return &leaf->ents[index - 1];
 250        }
 251
 252        /*
 253         * The high one is better, so use that one.
 254         */
 255        ASSERT(highstale - index >= 0);
 256        ASSERT(leaf->ents[highstale].address ==
 257               cpu_to_be32(XFS_DIR2_NULL_DATAPTR));
 258
 259        /*
 260         * Copy entries down to cover the stale entry and make room for the
 261         * new entry.
 262         */
 263        if (highstale - index > 0) {
 264                memmove(&leaf->ents[index + 1],
 265                        &leaf->ents[index],
 266                        (highstale - index) * sizeof(xfs_dir2_leaf_entry_t));
 267        }
 268        *lfloglow = MIN(index, *lfloglow);
 269        *lfloghigh = MAX(highstale, *lfloghigh);
 270        be16_add_cpu(&leaf->hdr.stale, -1);
 271        return &leaf->ents[index];
 272}
 273
 274/*
 275 * Add an entry to a leaf form directory.
 276 */
 277int                                             /* error */
 278xfs_dir2_leaf_addname(
 279        xfs_da_args_t           *args)          /* operation arguments */
 280{
 281        __be16                  *bestsp;        /* freespace table in leaf */
 282        int                     compact;        /* need to compact leaves */
 283        xfs_dir2_data_hdr_t     *hdr;           /* data block header */
 284        struct xfs_buf          *dbp;           /* data block buffer */
 285        xfs_dir2_data_entry_t   *dep;           /* data block entry */
 286        xfs_inode_t             *dp;            /* incore directory inode */
 287        xfs_dir2_data_unused_t  *dup;           /* data unused entry */
 288        int                     error;          /* error return value */
 289        int                     grown;          /* allocated new data block */
 290        int                     highstale;      /* index of next stale leaf */
 291        int                     i;              /* temporary, index */
 292        int                     index;          /* leaf table position */
 293        struct xfs_buf          *lbp;           /* leaf's buffer */
 294        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
 295        int                     length;         /* length of new entry */
 296        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry table pointer */
 297        int                     lfloglow;       /* low leaf logging index */
 298        int                     lfloghigh;      /* high leaf logging index */
 299        int                     lowstale;       /* index of prev stale leaf */
 300        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail pointer */
 301        xfs_mount_t             *mp;            /* filesystem mount point */
 302        int                     needbytes;      /* leaf block bytes needed */
 303        int                     needlog;        /* need to log data header */
 304        int                     needscan;       /* need to rescan data free */
 305        __be16                  *tagp;          /* end of data entry */
 306        xfs_trans_t             *tp;            /* transaction pointer */
 307        xfs_dir2_db_t           use_block;      /* data block number */
 308
 309        trace_xfs_dir2_leaf_addname(args);
 310
 311        dp = args->dp;
 312        tp = args->trans;
 313        mp = dp->i_mount;
 314        /*
 315         * Read the leaf block.
 316         */
 317        error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
 318                XFS_DATA_FORK);
 319        if (error) {
 320                return error;
 321        }
 322        ASSERT(lbp != NULL);
 323        /*
 324         * Look up the entry by hash value and name.
 325         * We know it's not there, our caller has already done a lookup.
 326         * So the index is of the entry to insert in front of.
 327         * But if there are dup hash values the index is of the first of those.
 328         */
 329        index = xfs_dir2_leaf_search_hash(args, lbp);
 330        leaf = lbp->b_addr;
 331        ltp = xfs_dir2_leaf_tail_p(mp, leaf);
 332        bestsp = xfs_dir2_leaf_bests_p(ltp);
 333        length = xfs_dir2_data_entsize(args->namelen);
 334        /*
 335         * See if there are any entries with the same hash value
 336         * and space in their block for the new entry.
 337         * This is good because it puts multiple same-hash value entries
 338         * in a data block, improving the lookup of those entries.
 339         */
 340        for (use_block = -1, lep = &leaf->ents[index];
 341             index < be16_to_cpu(leaf->hdr.count) && be32_to_cpu(lep->hashval) == args->hashval;
 342             index++, lep++) {
 343                if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
 344                        continue;
 345                i = xfs_dir2_dataptr_to_db(mp, be32_to_cpu(lep->address));
 346                ASSERT(i < be32_to_cpu(ltp->bestcount));
 347                ASSERT(bestsp[i] != cpu_to_be16(NULLDATAOFF));
 348                if (be16_to_cpu(bestsp[i]) >= length) {
 349                        use_block = i;
 350                        break;
 351                }
 352        }
 353        /*
 354         * Didn't find a block yet, linear search all the data blocks.
 355         */
 356        if (use_block == -1) {
 357                for (i = 0; i < be32_to_cpu(ltp->bestcount); i++) {
 358                        /*
 359                         * Remember a block we see that's missing.
 360                         */
 361                        if (bestsp[i] == cpu_to_be16(NULLDATAOFF) &&
 362                            use_block == -1)
 363                                use_block = i;
 364                        else if (be16_to_cpu(bestsp[i]) >= length) {
 365                                use_block = i;
 366                                break;
 367                        }
 368                }
 369        }
 370        /*
 371         * How many bytes do we need in the leaf block?
 372         */
 373        needbytes = 0;
 374        if (!leaf->hdr.stale)
 375                needbytes += sizeof(xfs_dir2_leaf_entry_t);
 376        if (use_block == -1)
 377                needbytes += sizeof(xfs_dir2_data_off_t);
 378
 379        /*
 380         * Now kill use_block if it refers to a missing block, so we
 381         * can use it as an indication of allocation needed.
 382         */
 383        if (use_block != -1 && bestsp[use_block] == cpu_to_be16(NULLDATAOFF))
 384                use_block = -1;
 385        /*
 386         * If we don't have enough free bytes but we can make enough
 387         * by compacting out stale entries, we'll do that.
 388         */
 389        if ((char *)bestsp - (char *)&leaf->ents[be16_to_cpu(leaf->hdr.count)] <
 390                                needbytes && be16_to_cpu(leaf->hdr.stale) > 1) {
 391                compact = 1;
 392        }
 393        /*
 394         * Otherwise if we don't have enough free bytes we need to
 395         * convert to node form.
 396         */
 397        else if ((char *)bestsp - (char *)&leaf->ents[be16_to_cpu(
 398                                                leaf->hdr.count)] < needbytes) {
 399                /*
 400                 * Just checking or no space reservation, give up.
 401                 */
 402                if ((args->op_flags & XFS_DA_OP_JUSTCHECK) ||
 403                                                        args->total == 0) {
 404                        xfs_trans_brelse(tp, lbp);
 405                        return XFS_ERROR(ENOSPC);
 406                }
 407                /*
 408                 * Convert to node form.
 409                 */
 410                error = xfs_dir2_leaf_to_node(args, lbp);
 411                if (error)
 412                        return error;
 413                /*
 414                 * Then add the new entry.
 415                 */
 416                return xfs_dir2_node_addname(args);
 417        }
 418        /*
 419         * Otherwise it will fit without compaction.
 420         */
 421        else
 422                compact = 0;
 423        /*
 424         * If just checking, then it will fit unless we needed to allocate
 425         * a new data block.
 426         */
 427        if (args->op_flags & XFS_DA_OP_JUSTCHECK) {
 428                xfs_trans_brelse(tp, lbp);
 429                return use_block == -1 ? XFS_ERROR(ENOSPC) : 0;
 430        }
 431        /*
 432         * If no allocations are allowed, return now before we've
 433         * changed anything.
 434         */
 435        if (args->total == 0 && use_block == -1) {
 436                xfs_trans_brelse(tp, lbp);
 437                return XFS_ERROR(ENOSPC);
 438        }
 439        /*
 440         * Need to compact the leaf entries, removing stale ones.
 441         * Leave one stale entry behind - the one closest to our
 442         * insertion index - and we'll shift that one to our insertion
 443         * point later.
 444         */
 445        if (compact) {
 446                xfs_dir2_leaf_compact_x1(lbp, &index, &lowstale, &highstale,
 447                        &lfloglow, &lfloghigh);
 448        }
 449        /*
 450         * There are stale entries, so we'll need log-low and log-high
 451         * impossibly bad values later.
 452         */
 453        else if (be16_to_cpu(leaf->hdr.stale)) {
 454                lfloglow = be16_to_cpu(leaf->hdr.count);
 455                lfloghigh = -1;
 456        }
 457        /*
 458         * If there was no data block space found, we need to allocate
 459         * a new one.
 460         */
 461        if (use_block == -1) {
 462                /*
 463                 * Add the new data block.
 464                 */
 465                if ((error = xfs_dir2_grow_inode(args, XFS_DIR2_DATA_SPACE,
 466                                &use_block))) {
 467                        xfs_trans_brelse(tp, lbp);
 468                        return error;
 469                }
 470                /*
 471                 * Initialize the block.
 472                 */
 473                if ((error = xfs_dir2_data_init(args, use_block, &dbp))) {
 474                        xfs_trans_brelse(tp, lbp);
 475                        return error;
 476                }
 477                /*
 478                 * If we're adding a new data block on the end we need to
 479                 * extend the bests table.  Copy it up one entry.
 480                 */
 481                if (use_block >= be32_to_cpu(ltp->bestcount)) {
 482                        bestsp--;
 483                        memmove(&bestsp[0], &bestsp[1],
 484                                be32_to_cpu(ltp->bestcount) * sizeof(bestsp[0]));
 485                        be32_add_cpu(&ltp->bestcount, 1);
 486                        xfs_dir2_leaf_log_tail(tp, lbp);
 487                        xfs_dir2_leaf_log_bests(tp, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
 488                }
 489                /*
 490                 * If we're filling in a previously empty block just log it.
 491                 */
 492                else
 493                        xfs_dir2_leaf_log_bests(tp, lbp, use_block, use_block);
 494                hdr = dbp->b_addr;
 495                bestsp[use_block] = hdr->bestfree[0].length;
 496                grown = 1;
 497        }
 498        /*
 499         * Already had space in some data block.
 500         * Just read that one in.
 501         */
 502        else {
 503                if ((error =
 504                    xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, use_block),
 505                            -1, &dbp, XFS_DATA_FORK))) {
 506                        xfs_trans_brelse(tp, lbp);
 507                        return error;
 508                }
 509                hdr = dbp->b_addr;
 510                grown = 0;
 511        }
 512        xfs_dir2_data_check(dp, dbp);
 513        /*
 514         * Point to the biggest freespace in our data block.
 515         */
 516        dup = (xfs_dir2_data_unused_t *)
 517              ((char *)hdr + be16_to_cpu(hdr->bestfree[0].offset));
 518        ASSERT(be16_to_cpu(dup->length) >= length);
 519        needscan = needlog = 0;
 520        /*
 521         * Mark the initial part of our freespace in use for the new entry.
 522         */
 523        xfs_dir2_data_use_free(tp, dbp, dup,
 524                (xfs_dir2_data_aoff_t)((char *)dup - (char *)hdr), length,
 525                &needlog, &needscan);
 526        /*
 527         * Initialize our new entry (at last).
 528         */
 529        dep = (xfs_dir2_data_entry_t *)dup;
 530        dep->inumber = cpu_to_be64(args->inumber);
 531        dep->namelen = args->namelen;
 532        memcpy(dep->name, args->name, dep->namelen);
 533        tagp = xfs_dir2_data_entry_tag_p(dep);
 534        *tagp = cpu_to_be16((char *)dep - (char *)hdr);
 535        /*
 536         * Need to scan fix up the bestfree table.
 537         */
 538        if (needscan)
 539                xfs_dir2_data_freescan(mp, hdr, &needlog);
 540        /*
 541         * Need to log the data block's header.
 542         */
 543        if (needlog)
 544                xfs_dir2_data_log_header(tp, dbp);
 545        xfs_dir2_data_log_entry(tp, dbp, dep);
 546        /*
 547         * If the bests table needs to be changed, do it.
 548         * Log the change unless we've already done that.
 549         */
 550        if (be16_to_cpu(bestsp[use_block]) != be16_to_cpu(hdr->bestfree[0].length)) {
 551                bestsp[use_block] = hdr->bestfree[0].length;
 552                if (!grown)
 553                        xfs_dir2_leaf_log_bests(tp, lbp, use_block, use_block);
 554        }
 555
 556        lep = xfs_dir2_leaf_find_entry(leaf, index, compact, lowstale,
 557                                       highstale, &lfloglow, &lfloghigh);
 558
 559        /*
 560         * Fill in the new leaf entry.
 561         */
 562        lep->hashval = cpu_to_be32(args->hashval);
 563        lep->address = cpu_to_be32(xfs_dir2_db_off_to_dataptr(mp, use_block,
 564                                be16_to_cpu(*tagp)));
 565        /*
 566         * Log the leaf fields and give up the buffers.
 567         */
 568        xfs_dir2_leaf_log_header(tp, lbp);
 569        xfs_dir2_leaf_log_ents(tp, lbp, lfloglow, lfloghigh);
 570        xfs_dir2_leaf_check(dp, lbp);
 571        xfs_dir2_data_check(dp, dbp);
 572        return 0;
 573}
 574
 575#ifdef DEBUG
 576/*
 577 * Check the internal consistency of a leaf1 block.
 578 * Pop an assert if something is wrong.
 579 */
 580STATIC void
 581xfs_dir2_leaf_check(
 582        struct xfs_inode        *dp,            /* incore directory inode */
 583        struct xfs_buf          *bp)            /* leaf's buffer */
 584{
 585        int                     i;              /* leaf index */
 586        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
 587        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail pointer */
 588        xfs_mount_t             *mp;            /* filesystem mount point */
 589        int                     stale;          /* count of stale leaves */
 590
 591        leaf = bp->b_addr;
 592        mp = dp->i_mount;
 593        ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
 594        /*
 595         * This value is not restrictive enough.
 596         * Should factor in the size of the bests table as well.
 597         * We can deduce a value for that from di_size.
 598         */
 599        ASSERT(be16_to_cpu(leaf->hdr.count) <= xfs_dir2_max_leaf_ents(mp));
 600        ltp = xfs_dir2_leaf_tail_p(mp, leaf);
 601        /*
 602         * Leaves and bests don't overlap.
 603         */
 604        ASSERT((char *)&leaf->ents[be16_to_cpu(leaf->hdr.count)] <=
 605               (char *)xfs_dir2_leaf_bests_p(ltp));
 606        /*
 607         * Check hash value order, count stale entries.
 608         */
 609        for (i = stale = 0; i < be16_to_cpu(leaf->hdr.count); i++) {
 610                if (i + 1 < be16_to_cpu(leaf->hdr.count))
 611                        ASSERT(be32_to_cpu(leaf->ents[i].hashval) <=
 612                               be32_to_cpu(leaf->ents[i + 1].hashval));
 613                if (leaf->ents[i].address == cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 614                        stale++;
 615        }
 616        ASSERT(be16_to_cpu(leaf->hdr.stale) == stale);
 617}
 618#endif  /* DEBUG */
 619
 620/*
 621 * Compact out any stale entries in the leaf.
 622 * Log the header and changed leaf entries, if any.
 623 */
 624void
 625xfs_dir2_leaf_compact(
 626        xfs_da_args_t   *args,          /* operation arguments */
 627        struct xfs_buf  *bp)            /* leaf buffer */
 628{
 629        int             from;           /* source leaf index */
 630        xfs_dir2_leaf_t *leaf;          /* leaf structure */
 631        int             loglow;         /* first leaf entry to log */
 632        int             to;             /* target leaf index */
 633
 634        leaf = bp->b_addr;
 635        if (!leaf->hdr.stale) {
 636                return;
 637        }
 638        /*
 639         * Compress out the stale entries in place.
 640         */
 641        for (from = to = 0, loglow = -1; from < be16_to_cpu(leaf->hdr.count); from++) {
 642                if (leaf->ents[from].address ==
 643                    cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 644                        continue;
 645                /*
 646                 * Only actually copy the entries that are different.
 647                 */
 648                if (from > to) {
 649                        if (loglow == -1)
 650                                loglow = to;
 651                        leaf->ents[to] = leaf->ents[from];
 652                }
 653                to++;
 654        }
 655        /*
 656         * Update and log the header, log the leaf entries.
 657         */
 658        ASSERT(be16_to_cpu(leaf->hdr.stale) == from - to);
 659        be16_add_cpu(&leaf->hdr.count, -(be16_to_cpu(leaf->hdr.stale)));
 660        leaf->hdr.stale = 0;
 661        xfs_dir2_leaf_log_header(args->trans, bp);
 662        if (loglow != -1)
 663                xfs_dir2_leaf_log_ents(args->trans, bp, loglow, to - 1);
 664}
 665
 666/*
 667 * Compact the leaf entries, removing stale ones.
 668 * Leave one stale entry behind - the one closest to our
 669 * insertion index - and the caller will shift that one to our insertion
 670 * point later.
 671 * Return new insertion index, where the remaining stale entry is,
 672 * and leaf logging indices.
 673 */
 674void
 675xfs_dir2_leaf_compact_x1(
 676        struct xfs_buf  *bp,            /* leaf buffer */
 677        int             *indexp,        /* insertion index */
 678        int             *lowstalep,     /* out: stale entry before us */
 679        int             *highstalep,    /* out: stale entry after us */
 680        int             *lowlogp,       /* out: low log index */
 681        int             *highlogp)      /* out: high log index */
 682{
 683        int             from;           /* source copy index */
 684        int             highstale;      /* stale entry at/after index */
 685        int             index;          /* insertion index */
 686        int             keepstale;      /* source index of kept stale */
 687        xfs_dir2_leaf_t *leaf;          /* leaf structure */
 688        int             lowstale;       /* stale entry before index */
 689        int             newindex=0;     /* new insertion index */
 690        int             to;             /* destination copy index */
 691
 692        leaf = bp->b_addr;
 693        ASSERT(be16_to_cpu(leaf->hdr.stale) > 1);
 694        index = *indexp;
 695
 696        xfs_dir2_leaf_find_stale(leaf, index, &lowstale, &highstale);
 697
 698        /*
 699         * Pick the better of lowstale and highstale.
 700         */
 701        if (lowstale >= 0 &&
 702            (highstale == be16_to_cpu(leaf->hdr.count) ||
 703             index - lowstale <= highstale - index))
 704                keepstale = lowstale;
 705        else
 706                keepstale = highstale;
 707        /*
 708         * Copy the entries in place, removing all the stale entries
 709         * except keepstale.
 710         */
 711        for (from = to = 0; from < be16_to_cpu(leaf->hdr.count); from++) {
 712                /*
 713                 * Notice the new value of index.
 714                 */
 715                if (index == from)
 716                        newindex = to;
 717                if (from != keepstale &&
 718                    leaf->ents[from].address ==
 719                    cpu_to_be32(XFS_DIR2_NULL_DATAPTR)) {
 720                        if (from == to)
 721                                *lowlogp = to;
 722                        continue;
 723                }
 724                /*
 725                 * Record the new keepstale value for the insertion.
 726                 */
 727                if (from == keepstale)
 728                        lowstale = highstale = to;
 729                /*
 730                 * Copy only the entries that have moved.
 731                 */
 732                if (from > to)
 733                        leaf->ents[to] = leaf->ents[from];
 734                to++;
 735        }
 736        ASSERT(from > to);
 737        /*
 738         * If the insertion point was past the last entry,
 739         * set the new insertion point accordingly.
 740         */
 741        if (index == from)
 742                newindex = to;
 743        *indexp = newindex;
 744        /*
 745         * Adjust the leaf header values.
 746         */
 747        be16_add_cpu(&leaf->hdr.count, -(from - to));
 748        leaf->hdr.stale = cpu_to_be16(1);
 749        /*
 750         * Remember the low/high stale value only in the "right"
 751         * direction.
 752         */
 753        if (lowstale >= newindex)
 754                lowstale = -1;
 755        else
 756                highstale = be16_to_cpu(leaf->hdr.count);
 757        *highlogp = be16_to_cpu(leaf->hdr.count) - 1;
 758        *lowstalep = lowstale;
 759        *highstalep = highstale;
 760}
 761
 762struct xfs_dir2_leaf_map_info {
 763        xfs_extlen_t    map_blocks;     /* number of fsbs in map */
 764        xfs_dablk_t     map_off;        /* last mapped file offset */
 765        int             map_size;       /* total entries in *map */
 766        int             map_valid;      /* valid entries in *map */
 767        int             nmap;           /* mappings to ask xfs_bmapi */
 768        xfs_dir2_db_t   curdb;          /* db for current block */
 769        int             ra_current;     /* number of read-ahead blks */
 770        int             ra_index;       /* *map index for read-ahead */
 771        int             ra_offset;      /* map entry offset for ra */
 772        int             ra_want;        /* readahead count wanted */
 773        struct xfs_bmbt_irec map[];     /* map vector for blocks */
 774};
 775
 776STATIC int
 777xfs_dir2_leaf_readbuf(
 778        struct xfs_inode        *dp,
 779        size_t                  bufsize,
 780        struct xfs_dir2_leaf_map_info *mip,
 781        xfs_dir2_off_t          *curoff,
 782        struct xfs_buf          **bpp)
 783{
 784        struct xfs_mount        *mp = dp->i_mount;
 785        struct xfs_buf          *bp = *bpp;
 786        struct xfs_bmbt_irec    *map = mip->map;
 787        int                     error = 0;
 788        int                     length;
 789        int                     i;
 790        int                     j;
 791
 792        /*
 793         * If we have a buffer, we need to release it and
 794         * take it out of the mapping.
 795         */
 796
 797        if (bp) {
 798                xfs_trans_brelse(NULL, bp);
 799                bp = NULL;
 800                mip->map_blocks -= mp->m_dirblkfsbs;
 801                /*
 802                 * Loop to get rid of the extents for the
 803                 * directory block.
 804                 */
 805                for (i = mp->m_dirblkfsbs; i > 0; ) {
 806                        j = min_t(int, map->br_blockcount, i);
 807                        map->br_blockcount -= j;
 808                        map->br_startblock += j;
 809                        map->br_startoff += j;
 810                        /*
 811                         * If mapping is done, pitch it from
 812                         * the table.
 813                         */
 814                        if (!map->br_blockcount && --mip->map_valid)
 815                                memmove(&map[0], &map[1],
 816                                        sizeof(map[0]) * mip->map_valid);
 817                        i -= j;
 818                }
 819        }
 820
 821        /*
 822         * Recalculate the readahead blocks wanted.
 823         */
 824        mip->ra_want = howmany(bufsize + mp->m_dirblksize,
 825                               mp->m_sb.sb_blocksize) - 1;
 826        ASSERT(mip->ra_want >= 0);
 827
 828        /*
 829         * If we don't have as many as we want, and we haven't
 830         * run out of data blocks, get some more mappings.
 831         */
 832        if (1 + mip->ra_want > mip->map_blocks &&
 833            mip->map_off < xfs_dir2_byte_to_da(mp, XFS_DIR2_LEAF_OFFSET)) {
 834                /*
 835                 * Get more bmaps, fill in after the ones
 836                 * we already have in the table.
 837                 */
 838                mip->nmap = mip->map_size - mip->map_valid;
 839                error = xfs_bmapi_read(dp, mip->map_off,
 840                                xfs_dir2_byte_to_da(mp, XFS_DIR2_LEAF_OFFSET) -
 841                                                                mip->map_off,
 842                                &map[mip->map_valid], &mip->nmap, 0);
 843
 844                /*
 845                 * Don't know if we should ignore this or try to return an
 846                 * error.  The trouble with returning errors is that readdir
 847                 * will just stop without actually passing the error through.
 848                 */
 849                if (error)
 850                        goto out;       /* XXX */
 851
 852                /*
 853                 * If we got all the mappings we asked for, set the final map
 854                 * offset based on the last bmap value received.  Otherwise,
 855                 * we've reached the end.
 856                 */
 857                if (mip->nmap == mip->map_size - mip->map_valid) {
 858                        i = mip->map_valid + mip->nmap - 1;
 859                        mip->map_off = map[i].br_startoff + map[i].br_blockcount;
 860                } else
 861                        mip->map_off = xfs_dir2_byte_to_da(mp,
 862                                                        XFS_DIR2_LEAF_OFFSET);
 863
 864                /*
 865                 * Look for holes in the mapping, and eliminate them.  Count up
 866                 * the valid blocks.
 867                 */
 868                for (i = mip->map_valid; i < mip->map_valid + mip->nmap; ) {
 869                        if (map[i].br_startblock == HOLESTARTBLOCK) {
 870                                mip->nmap--;
 871                                length = mip->map_valid + mip->nmap - i;
 872                                if (length)
 873                                        memmove(&map[i], &map[i + 1],
 874                                                sizeof(map[i]) * length);
 875                        } else {
 876                                mip->map_blocks += map[i].br_blockcount;
 877                                i++;
 878                        }
 879                }
 880                mip->map_valid += mip->nmap;
 881        }
 882
 883        /*
 884         * No valid mappings, so no more data blocks.
 885         */
 886        if (!mip->map_valid) {
 887                *curoff = xfs_dir2_da_to_byte(mp, mip->map_off);
 888                goto out;
 889        }
 890
 891        /*
 892         * Read the directory block starting at the first mapping.
 893         */
 894        mip->curdb = xfs_dir2_da_to_db(mp, map->br_startoff);
 895        error = xfs_da_read_buf(NULL, dp, map->br_startoff,
 896                        map->br_blockcount >= mp->m_dirblkfsbs ?
 897                            XFS_FSB_TO_DADDR(mp, map->br_startblock) : -1,
 898                        &bp, XFS_DATA_FORK);
 899
 900        /*
 901         * Should just skip over the data block instead of giving up.
 902         */
 903        if (error)
 904                goto out;       /* XXX */
 905
 906        /*
 907         * Adjust the current amount of read-ahead: we just read a block that
 908         * was previously ra.
 909         */
 910        if (mip->ra_current)
 911                mip->ra_current -= mp->m_dirblkfsbs;
 912
 913        /*
 914         * Do we need more readahead?
 915         */
 916        for (mip->ra_index = mip->ra_offset = i = 0;
 917             mip->ra_want > mip->ra_current && i < mip->map_blocks;
 918             i += mp->m_dirblkfsbs) {
 919                ASSERT(mip->ra_index < mip->map_valid);
 920                /*
 921                 * Read-ahead a contiguous directory block.
 922                 */
 923                if (i > mip->ra_current &&
 924                    map[mip->ra_index].br_blockcount >= mp->m_dirblkfsbs) {
 925                        xfs_buf_readahead(mp->m_ddev_targp,
 926                                XFS_FSB_TO_DADDR(mp,
 927                                        map[mip->ra_index].br_startblock +
 928                                                        mip->ra_offset),
 929                                (int)BTOBB(mp->m_dirblksize));
 930                        mip->ra_current = i;
 931                }
 932
 933                /*
 934                 * Read-ahead a non-contiguous directory block.  This doesn't
 935                 * use our mapping, but this is a very rare case.
 936                 */
 937                else if (i > mip->ra_current) {
 938                        xfs_da_reada_buf(NULL, dp,
 939                                        map[mip->ra_index].br_startoff +
 940                                                        mip->ra_offset,
 941                                        XFS_DATA_FORK);
 942                        mip->ra_current = i;
 943                }
 944
 945                /*
 946                 * Advance offset through the mapping table.
 947                 */
 948                for (j = 0; j < mp->m_dirblkfsbs; j++) {
 949                        /*
 950                         * The rest of this extent but not more than a dir
 951                         * block.
 952                         */
 953                        length = min_t(int, mp->m_dirblkfsbs,
 954                                        map[mip->ra_index].br_blockcount -
 955                                                        mip->ra_offset);
 956                        j += length;
 957                        mip->ra_offset += length;
 958
 959                        /*
 960                         * Advance to the next mapping if this one is used up.
 961                         */
 962                        if (mip->ra_offset == map[mip->ra_index].br_blockcount) {
 963                                mip->ra_offset = 0;
 964                                mip->ra_index++;
 965                        }
 966                }
 967        }
 968
 969out:
 970        *bpp = bp;
 971        return error;
 972}
 973
 974/*
 975 * Getdents (readdir) for leaf and node directories.
 976 * This reads the data blocks only, so is the same for both forms.
 977 */
 978int                                             /* error */
 979xfs_dir2_leaf_getdents(
 980        xfs_inode_t             *dp,            /* incore directory inode */
 981        void                    *dirent,
 982        size_t                  bufsize,
 983        xfs_off_t               *offset,
 984        filldir_t               filldir)
 985{
 986        struct xfs_buf          *bp = NULL;     /* data block buffer */
 987        xfs_dir2_data_hdr_t     *hdr;           /* data block header */
 988        xfs_dir2_data_entry_t   *dep;           /* data entry */
 989        xfs_dir2_data_unused_t  *dup;           /* unused entry */
 990        int                     error = 0;      /* error return value */
 991        int                     length;         /* temporary length value */
 992        xfs_mount_t             *mp;            /* filesystem mount point */
 993        int                     byteoff;        /* offset in current block */
 994        xfs_dir2_off_t          curoff;         /* current overall offset */
 995        xfs_dir2_off_t          newoff;         /* new curoff after new blk */
 996        char                    *ptr = NULL;    /* pointer to current data */
 997        struct xfs_dir2_leaf_map_info *map_info;
 998
 999        /*
1000         * If the offset is at or past the largest allowed value,
1001         * give up right away.
1002         */
1003        if (*offset >= XFS_DIR2_MAX_DATAPTR)
1004                return 0;
1005
1006        mp = dp->i_mount;
1007
1008        /*
1009         * Set up to bmap a number of blocks based on the caller's
1010         * buffer size, the directory block size, and the filesystem
1011         * block size.
1012         */
1013        length = howmany(bufsize + mp->m_dirblksize,
1014                                     mp->m_sb.sb_blocksize);
1015        map_info = kmem_zalloc(offsetof(struct xfs_dir2_leaf_map_info, map) +
1016                                (length * sizeof(struct xfs_bmbt_irec)),
1017                               KM_SLEEP);
1018        map_info->map_size = length;
1019
1020        /*
1021         * Inside the loop we keep the main offset value as a byte offset
1022         * in the directory file.
1023         */
1024        curoff = xfs_dir2_dataptr_to_byte(mp, *offset);
1025
1026        /*
1027         * Force this conversion through db so we truncate the offset
1028         * down to get the start of the data block.
1029         */
1030        map_info->map_off = xfs_dir2_db_to_da(mp,
1031                                              xfs_dir2_byte_to_db(mp, curoff));
1032
1033        /*
1034         * Loop over directory entries until we reach the end offset.
1035         * Get more blocks and readahead as necessary.
1036         */
1037        while (curoff < XFS_DIR2_LEAF_OFFSET) {
1038                /*
1039                 * If we have no buffer, or we're off the end of the
1040                 * current buffer, need to get another one.
1041                 */
1042                if (!bp || ptr >= (char *)bp->b_addr + mp->m_dirblksize) {
1043
1044                        error = xfs_dir2_leaf_readbuf(dp, bufsize, map_info,
1045                                                      &curoff, &bp);
1046                        if (error || !map_info->map_valid)
1047                                break;
1048
1049                        /*
1050                         * Having done a read, we need to set a new offset.
1051                         */
1052                        newoff = xfs_dir2_db_off_to_byte(mp, map_info->curdb, 0);
1053                        /*
1054                         * Start of the current block.
1055                         */
1056                        if (curoff < newoff)
1057                                curoff = newoff;
1058                        /*
1059                         * Make sure we're in the right block.
1060                         */
1061                        else if (curoff > newoff)
1062                                ASSERT(xfs_dir2_byte_to_db(mp, curoff) ==
1063                                       map_info->curdb);
1064                        hdr = bp->b_addr;
1065                        xfs_dir2_data_check(dp, bp);
1066                        /*
1067                         * Find our position in the block.
1068                         */
1069                        ptr = (char *)(hdr + 1);
1070                        byteoff = xfs_dir2_byte_to_off(mp, curoff);
1071                        /*
1072                         * Skip past the header.
1073                         */
1074                        if (byteoff == 0)
1075                                curoff += (uint)sizeof(*hdr);
1076                        /*
1077                         * Skip past entries until we reach our offset.
1078                         */
1079                        else {
1080                                while ((char *)ptr - (char *)hdr < byteoff) {
1081                                        dup = (xfs_dir2_data_unused_t *)ptr;
1082
1083                                        if (be16_to_cpu(dup->freetag)
1084                                                  == XFS_DIR2_DATA_FREE_TAG) {
1085
1086                                                length = be16_to_cpu(dup->length);
1087                                                ptr += length;
1088                                                continue;
1089                                        }
1090                                        dep = (xfs_dir2_data_entry_t *)ptr;
1091                                        length =
1092                                           xfs_dir2_data_entsize(dep->namelen);
1093                                        ptr += length;
1094                                }
1095                                /*
1096                                 * Now set our real offset.
1097                                 */
1098                                curoff =
1099                                        xfs_dir2_db_off_to_byte(mp,
1100                                            xfs_dir2_byte_to_db(mp, curoff),
1101                                            (char *)ptr - (char *)hdr);
1102                                if (ptr >= (char *)hdr + mp->m_dirblksize) {
1103                                        continue;
1104                                }
1105                        }
1106                }
1107                /*
1108                 * We have a pointer to an entry.
1109                 * Is it a live one?
1110                 */
1111                dup = (xfs_dir2_data_unused_t *)ptr;
1112                /*
1113                 * No, it's unused, skip over it.
1114                 */
1115                if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
1116                        length = be16_to_cpu(dup->length);
1117                        ptr += length;
1118                        curoff += length;
1119                        continue;
1120                }
1121
1122                dep = (xfs_dir2_data_entry_t *)ptr;
1123                length = xfs_dir2_data_entsize(dep->namelen);
1124
1125                if (filldir(dirent, (char *)dep->name, dep->namelen,
1126                            xfs_dir2_byte_to_dataptr(mp, curoff) & 0x7fffffff,
1127                            be64_to_cpu(dep->inumber), DT_UNKNOWN))
1128                        break;
1129
1130                /*
1131                 * Advance to next entry in the block.
1132                 */
1133                ptr += length;
1134                curoff += length;
1135                /* bufsize may have just been a guess; don't go negative */
1136                bufsize = bufsize > length ? bufsize - length : 0;
1137        }
1138
1139        /*
1140         * All done.  Set output offset value to current offset.
1141         */
1142        if (curoff > xfs_dir2_dataptr_to_byte(mp, XFS_DIR2_MAX_DATAPTR))
1143                *offset = XFS_DIR2_MAX_DATAPTR & 0x7fffffff;
1144        else
1145                *offset = xfs_dir2_byte_to_dataptr(mp, curoff) & 0x7fffffff;
1146        kmem_free(map_info);
1147        if (bp)
1148                xfs_trans_brelse(NULL, bp);
1149        return error;
1150}
1151
1152/*
1153 * Initialize a new leaf block, leaf1 or leafn magic accepted.
1154 */
1155int
1156xfs_dir2_leaf_init(
1157        xfs_da_args_t           *args,          /* operation arguments */
1158        xfs_dir2_db_t           bno,            /* directory block number */
1159        struct xfs_buf          **bpp,          /* out: leaf buffer */
1160        int                     magic)          /* magic number for block */
1161{
1162        struct xfs_buf          *bp;            /* leaf buffer */
1163        xfs_inode_t             *dp;            /* incore directory inode */
1164        int                     error;          /* error return code */
1165        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1166        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1167        xfs_mount_t             *mp;            /* filesystem mount point */
1168        xfs_trans_t             *tp;            /* transaction pointer */
1169
1170        dp = args->dp;
1171        ASSERT(dp != NULL);
1172        tp = args->trans;
1173        mp = dp->i_mount;
1174        ASSERT(bno >= XFS_DIR2_LEAF_FIRSTDB(mp) &&
1175               bno < XFS_DIR2_FREE_FIRSTDB(mp));
1176        /*
1177         * Get the buffer for the block.
1178         */
1179        error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(mp, bno), -1, &bp,
1180                XFS_DATA_FORK);
1181        if (error) {
1182                return error;
1183        }
1184        ASSERT(bp != NULL);
1185        leaf = bp->b_addr;
1186        /*
1187         * Initialize the header.
1188         */
1189        leaf->hdr.info.magic = cpu_to_be16(magic);
1190        leaf->hdr.info.forw = 0;
1191        leaf->hdr.info.back = 0;
1192        leaf->hdr.count = 0;
1193        leaf->hdr.stale = 0;
1194        xfs_dir2_leaf_log_header(tp, bp);
1195        /*
1196         * If it's a leaf-format directory initialize the tail.
1197         * In this case our caller has the real bests table to copy into
1198         * the block.
1199         */
1200        if (magic == XFS_DIR2_LEAF1_MAGIC) {
1201                ltp = xfs_dir2_leaf_tail_p(mp, leaf);
1202                ltp->bestcount = 0;
1203                xfs_dir2_leaf_log_tail(tp, bp);
1204        }
1205        *bpp = bp;
1206        return 0;
1207}
1208
1209/*
1210 * Log the bests entries indicated from a leaf1 block.
1211 */
1212static void
1213xfs_dir2_leaf_log_bests(
1214        xfs_trans_t             *tp,            /* transaction pointer */
1215        struct xfs_buf          *bp,            /* leaf buffer */
1216        int                     first,          /* first entry to log */
1217        int                     last)           /* last entry to log */
1218{
1219        __be16                  *firstb;        /* pointer to first entry */
1220        __be16                  *lastb;         /* pointer to last entry */
1221        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1222        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1223
1224        leaf = bp->b_addr;
1225        ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
1226        ltp = xfs_dir2_leaf_tail_p(tp->t_mountp, leaf);
1227        firstb = xfs_dir2_leaf_bests_p(ltp) + first;
1228        lastb = xfs_dir2_leaf_bests_p(ltp) + last;
1229        xfs_trans_log_buf(tp, bp, (uint)((char *)firstb - (char *)leaf),
1230                (uint)((char *)lastb - (char *)leaf + sizeof(*lastb) - 1));
1231}
1232
1233/*
1234 * Log the leaf entries indicated from a leaf1 or leafn block.
1235 */
1236void
1237xfs_dir2_leaf_log_ents(
1238        xfs_trans_t             *tp,            /* transaction pointer */
1239        struct xfs_buf          *bp,            /* leaf buffer */
1240        int                     first,          /* first entry to log */
1241        int                     last)           /* last entry to log */
1242{
1243        xfs_dir2_leaf_entry_t   *firstlep;      /* pointer to first entry */
1244        xfs_dir2_leaf_entry_t   *lastlep;       /* pointer to last entry */
1245        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1246
1247        leaf = bp->b_addr;
1248        ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC) ||
1249               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
1250        firstlep = &leaf->ents[first];
1251        lastlep = &leaf->ents[last];
1252        xfs_trans_log_buf(tp, bp, (uint)((char *)firstlep - (char *)leaf),
1253                (uint)((char *)lastlep - (char *)leaf + sizeof(*lastlep) - 1));
1254}
1255
1256/*
1257 * Log the header of the leaf1 or leafn block.
1258 */
1259void
1260xfs_dir2_leaf_log_header(
1261        struct xfs_trans        *tp,
1262        struct xfs_buf          *bp)
1263{
1264        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1265
1266        leaf = bp->b_addr;
1267        ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC) ||
1268               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
1269        xfs_trans_log_buf(tp, bp, (uint)((char *)&leaf->hdr - (char *)leaf),
1270                (uint)(sizeof(leaf->hdr) - 1));
1271}
1272
1273/*
1274 * Log the tail of the leaf1 block.
1275 */
1276STATIC void
1277xfs_dir2_leaf_log_tail(
1278        struct xfs_trans        *tp,
1279        struct xfs_buf          *bp)
1280{
1281        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1282        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1283        xfs_mount_t             *mp;            /* filesystem mount point */
1284
1285        mp = tp->t_mountp;
1286        leaf = bp->b_addr;
1287        ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
1288        ltp = xfs_dir2_leaf_tail_p(mp, leaf);
1289        xfs_trans_log_buf(tp, bp, (uint)((char *)ltp - (char *)leaf),
1290                (uint)(mp->m_dirblksize - 1));
1291}
1292
1293/*
1294 * Look up the entry referred to by args in the leaf format directory.
1295 * Most of the work is done by the xfs_dir2_leaf_lookup_int routine which
1296 * is also used by the node-format code.
1297 */
1298int
1299xfs_dir2_leaf_lookup(
1300        xfs_da_args_t           *args)          /* operation arguments */
1301{
1302        struct xfs_buf          *dbp;           /* data block buffer */
1303        xfs_dir2_data_entry_t   *dep;           /* data block entry */
1304        xfs_inode_t             *dp;            /* incore directory inode */
1305        int                     error;          /* error return code */
1306        int                     index;          /* found entry index */
1307        struct xfs_buf          *lbp;           /* leaf buffer */
1308        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1309        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1310        xfs_trans_t             *tp;            /* transaction pointer */
1311
1312        trace_xfs_dir2_leaf_lookup(args);
1313
1314        /*
1315         * Look up name in the leaf block, returning both buffers and index.
1316         */
1317        if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1318                return error;
1319        }
1320        tp = args->trans;
1321        dp = args->dp;
1322        xfs_dir2_leaf_check(dp, lbp);
1323        leaf = lbp->b_addr;
1324        /*
1325         * Get to the leaf entry and contained data entry address.
1326         */
1327        lep = &leaf->ents[index];
1328        /*
1329         * Point to the data entry.
1330         */
1331        dep = (xfs_dir2_data_entry_t *)
1332              ((char *)dbp->b_addr +
1333               xfs_dir2_dataptr_to_off(dp->i_mount, be32_to_cpu(lep->address)));
1334        /*
1335         * Return the found inode number & CI name if appropriate
1336         */
1337        args->inumber = be64_to_cpu(dep->inumber);
1338        error = xfs_dir_cilookup_result(args, dep->name, dep->namelen);
1339        xfs_trans_brelse(tp, dbp);
1340        xfs_trans_brelse(tp, lbp);
1341        return XFS_ERROR(error);
1342}
1343
1344/*
1345 * Look up name/hash in the leaf block.
1346 * Fill in indexp with the found index, and dbpp with the data buffer.
1347 * If not found dbpp will be NULL, and ENOENT comes back.
1348 * lbpp will always be filled in with the leaf buffer unless there's an error.
1349 */
1350static int                                      /* error */
1351xfs_dir2_leaf_lookup_int(
1352        xfs_da_args_t           *args,          /* operation arguments */
1353        struct xfs_buf          **lbpp,         /* out: leaf buffer */
1354        int                     *indexp,        /* out: index in leaf block */
1355        struct xfs_buf          **dbpp)         /* out: data buffer */
1356{
1357        xfs_dir2_db_t           curdb = -1;     /* current data block number */
1358        struct xfs_buf          *dbp = NULL;    /* data buffer */
1359        xfs_dir2_data_entry_t   *dep;           /* data entry */
1360        xfs_inode_t             *dp;            /* incore directory inode */
1361        int                     error;          /* error return code */
1362        int                     index;          /* index in leaf block */
1363        struct xfs_buf          *lbp;           /* leaf buffer */
1364        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1365        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1366        xfs_mount_t             *mp;            /* filesystem mount point */
1367        xfs_dir2_db_t           newdb;          /* new data block number */
1368        xfs_trans_t             *tp;            /* transaction pointer */
1369        xfs_dir2_db_t           cidb = -1;      /* case match data block no. */
1370        enum xfs_dacmp          cmp;            /* name compare result */
1371
1372        dp = args->dp;
1373        tp = args->trans;
1374        mp = dp->i_mount;
1375        /*
1376         * Read the leaf block into the buffer.
1377         */
1378        error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
1379                                                        XFS_DATA_FORK);
1380        if (error)
1381                return error;
1382        *lbpp = lbp;
1383        leaf = lbp->b_addr;
1384        xfs_dir2_leaf_check(dp, lbp);
1385        /*
1386         * Look for the first leaf entry with our hash value.
1387         */
1388        index = xfs_dir2_leaf_search_hash(args, lbp);
1389        /*
1390         * Loop over all the entries with the right hash value
1391         * looking to match the name.
1392         */
1393        for (lep = &leaf->ents[index]; index < be16_to_cpu(leaf->hdr.count) &&
1394                                be32_to_cpu(lep->hashval) == args->hashval;
1395                                lep++, index++) {
1396                /*
1397                 * Skip over stale leaf entries.
1398                 */
1399                if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
1400                        continue;
1401                /*
1402                 * Get the new data block number.
1403                 */
1404                newdb = xfs_dir2_dataptr_to_db(mp, be32_to_cpu(lep->address));
1405                /*
1406                 * If it's not the same as the old data block number,
1407                 * need to pitch the old one and read the new one.
1408                 */
1409                if (newdb != curdb) {
1410                        if (dbp)
1411                                xfs_trans_brelse(tp, dbp);
1412                        error = xfs_da_read_buf(tp, dp,
1413                                                xfs_dir2_db_to_da(mp, newdb),
1414                                                -1, &dbp, XFS_DATA_FORK);
1415                        if (error) {
1416                                xfs_trans_brelse(tp, lbp);
1417                                return error;
1418                        }
1419                        xfs_dir2_data_check(dp, dbp);
1420                        curdb = newdb;
1421                }
1422                /*
1423                 * Point to the data entry.
1424                 */
1425                dep = (xfs_dir2_data_entry_t *)((char *)dbp->b_addr +
1426                        xfs_dir2_dataptr_to_off(mp, be32_to_cpu(lep->address)));
1427                /*
1428                 * Compare name and if it's an exact match, return the index
1429                 * and buffer. If it's the first case-insensitive match, store
1430                 * the index and buffer and continue looking for an exact match.
1431                 */
1432                cmp = mp->m_dirnameops->compname(args, dep->name, dep->namelen);
1433                if (cmp != XFS_CMP_DIFFERENT && cmp != args->cmpresult) {
1434                        args->cmpresult = cmp;
1435                        *indexp = index;
1436                        /* case exact match: return the current buffer. */
1437                        if (cmp == XFS_CMP_EXACT) {
1438                                *dbpp = dbp;
1439                                return 0;
1440                        }
1441                        cidb = curdb;
1442                }
1443        }
1444        ASSERT(args->op_flags & XFS_DA_OP_OKNOENT);
1445        /*
1446         * Here, we can only be doing a lookup (not a rename or remove).
1447         * If a case-insensitive match was found earlier, re-read the
1448         * appropriate data block if required and return it.
1449         */
1450        if (args->cmpresult == XFS_CMP_CASE) {
1451                ASSERT(cidb != -1);
1452                if (cidb != curdb) {
1453                        xfs_trans_brelse(tp, dbp);
1454                        error = xfs_da_read_buf(tp, dp,
1455                                                xfs_dir2_db_to_da(mp, cidb),
1456                                                -1, &dbp, XFS_DATA_FORK);
1457                        if (error) {
1458                                xfs_trans_brelse(tp, lbp);
1459                                return error;
1460                        }
1461                }
1462                *dbpp = dbp;
1463                return 0;
1464        }
1465        /*
1466         * No match found, return ENOENT.
1467         */
1468        ASSERT(cidb == -1);
1469        if (dbp)
1470                xfs_trans_brelse(tp, dbp);
1471        xfs_trans_brelse(tp, lbp);
1472        return XFS_ERROR(ENOENT);
1473}
1474
1475/*
1476 * Remove an entry from a leaf format directory.
1477 */
1478int                                             /* error */
1479xfs_dir2_leaf_removename(
1480        xfs_da_args_t           *args)          /* operation arguments */
1481{
1482        __be16                  *bestsp;        /* leaf block best freespace */
1483        xfs_dir2_data_hdr_t     *hdr;           /* data block header */
1484        xfs_dir2_db_t           db;             /* data block number */
1485        struct xfs_buf          *dbp;           /* data block buffer */
1486        xfs_dir2_data_entry_t   *dep;           /* data entry structure */
1487        xfs_inode_t             *dp;            /* incore directory inode */
1488        int                     error;          /* error return code */
1489        xfs_dir2_db_t           i;              /* temporary data block # */
1490        int                     index;          /* index into leaf entries */
1491        struct xfs_buf          *lbp;           /* leaf buffer */
1492        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1493        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1494        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1495        xfs_mount_t             *mp;            /* filesystem mount point */
1496        int                     needlog;        /* need to log data header */
1497        int                     needscan;       /* need to rescan data frees */
1498        xfs_dir2_data_off_t     oldbest;        /* old value of best free */
1499        xfs_trans_t             *tp;            /* transaction pointer */
1500
1501        trace_xfs_dir2_leaf_removename(args);
1502
1503        /*
1504         * Lookup the leaf entry, get the leaf and data blocks read in.
1505         */
1506        if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1507                return error;
1508        }
1509        dp = args->dp;
1510        tp = args->trans;
1511        mp = dp->i_mount;
1512        leaf = lbp->b_addr;
1513        hdr = dbp->b_addr;
1514        xfs_dir2_data_check(dp, dbp);
1515        /*
1516         * Point to the leaf entry, use that to point to the data entry.
1517         */
1518        lep = &leaf->ents[index];
1519        db = xfs_dir2_dataptr_to_db(mp, be32_to_cpu(lep->address));
1520        dep = (xfs_dir2_data_entry_t *)
1521              ((char *)hdr + xfs_dir2_dataptr_to_off(mp, be32_to_cpu(lep->address)));
1522        needscan = needlog = 0;
1523        oldbest = be16_to_cpu(hdr->bestfree[0].length);
1524        ltp = xfs_dir2_leaf_tail_p(mp, leaf);
1525        bestsp = xfs_dir2_leaf_bests_p(ltp);
1526        ASSERT(be16_to_cpu(bestsp[db]) == oldbest);
1527        /*
1528         * Mark the former data entry unused.
1529         */
1530        xfs_dir2_data_make_free(tp, dbp,
1531                (xfs_dir2_data_aoff_t)((char *)dep - (char *)hdr),
1532                xfs_dir2_data_entsize(dep->namelen), &needlog, &needscan);
1533        /*
1534         * We just mark the leaf entry stale by putting a null in it.
1535         */
1536        be16_add_cpu(&leaf->hdr.stale, 1);
1537        xfs_dir2_leaf_log_header(tp, lbp);
1538        lep->address = cpu_to_be32(XFS_DIR2_NULL_DATAPTR);
1539        xfs_dir2_leaf_log_ents(tp, lbp, index, index);
1540        /*
1541         * Scan the freespace in the data block again if necessary,
1542         * log the data block header if necessary.
1543         */
1544        if (needscan)
1545                xfs_dir2_data_freescan(mp, hdr, &needlog);
1546        if (needlog)
1547                xfs_dir2_data_log_header(tp, dbp);
1548        /*
1549         * If the longest freespace in the data block has changed,
1550         * put the new value in the bests table and log that.
1551         */
1552        if (be16_to_cpu(hdr->bestfree[0].length) != oldbest) {
1553                bestsp[db] = hdr->bestfree[0].length;
1554                xfs_dir2_leaf_log_bests(tp, lbp, db, db);
1555        }
1556        xfs_dir2_data_check(dp, dbp);
1557        /*
1558         * If the data block is now empty then get rid of the data block.
1559         */
1560        if (be16_to_cpu(hdr->bestfree[0].length) ==
1561            mp->m_dirblksize - (uint)sizeof(*hdr)) {
1562                ASSERT(db != mp->m_dirdatablk);
1563                if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1564                        /*
1565                         * Nope, can't get rid of it because it caused
1566                         * allocation of a bmap btree block to do so.
1567                         * Just go on, returning success, leaving the
1568                         * empty block in place.
1569                         */
1570                        if (error == ENOSPC && args->total == 0)
1571                                error = 0;
1572                        xfs_dir2_leaf_check(dp, lbp);
1573                        return error;
1574                }
1575                dbp = NULL;
1576                /*
1577                 * If this is the last data block then compact the
1578                 * bests table by getting rid of entries.
1579                 */
1580                if (db == be32_to_cpu(ltp->bestcount) - 1) {
1581                        /*
1582                         * Look for the last active entry (i).
1583                         */
1584                        for (i = db - 1; i > 0; i--) {
1585                                if (bestsp[i] != cpu_to_be16(NULLDATAOFF))
1586                                        break;
1587                        }
1588                        /*
1589                         * Copy the table down so inactive entries at the
1590                         * end are removed.
1591                         */
1592                        memmove(&bestsp[db - i], bestsp,
1593                                (be32_to_cpu(ltp->bestcount) - (db - i)) * sizeof(*bestsp));
1594                        be32_add_cpu(&ltp->bestcount, -(db - i));
1595                        xfs_dir2_leaf_log_tail(tp, lbp);
1596                        xfs_dir2_leaf_log_bests(tp, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
1597                } else
1598                        bestsp[db] = cpu_to_be16(NULLDATAOFF);
1599        }
1600        /*
1601         * If the data block was not the first one, drop it.
1602         */
1603        else if (db != mp->m_dirdatablk)
1604                dbp = NULL;
1605
1606        xfs_dir2_leaf_check(dp, lbp);
1607        /*
1608         * See if we can convert to block form.
1609         */
1610        return xfs_dir2_leaf_to_block(args, lbp, dbp);
1611}
1612
1613/*
1614 * Replace the inode number in a leaf format directory entry.
1615 */
1616int                                             /* error */
1617xfs_dir2_leaf_replace(
1618        xfs_da_args_t           *args)          /* operation arguments */
1619{
1620        struct xfs_buf          *dbp;           /* data block buffer */
1621        xfs_dir2_data_entry_t   *dep;           /* data block entry */
1622        xfs_inode_t             *dp;            /* incore directory inode */
1623        int                     error;          /* error return code */
1624        int                     index;          /* index of leaf entry */
1625        struct xfs_buf          *lbp;           /* leaf buffer */
1626        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1627        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1628        xfs_trans_t             *tp;            /* transaction pointer */
1629
1630        trace_xfs_dir2_leaf_replace(args);
1631
1632        /*
1633         * Look up the entry.
1634         */
1635        if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1636                return error;
1637        }
1638        dp = args->dp;
1639        leaf = lbp->b_addr;
1640        /*
1641         * Point to the leaf entry, get data address from it.
1642         */
1643        lep = &leaf->ents[index];
1644        /*
1645         * Point to the data entry.
1646         */
1647        dep = (xfs_dir2_data_entry_t *)
1648              ((char *)dbp->b_addr +
1649               xfs_dir2_dataptr_to_off(dp->i_mount, be32_to_cpu(lep->address)));
1650        ASSERT(args->inumber != be64_to_cpu(dep->inumber));
1651        /*
1652         * Put the new inode number in, log it.
1653         */
1654        dep->inumber = cpu_to_be64(args->inumber);
1655        tp = args->trans;
1656        xfs_dir2_data_log_entry(tp, dbp, dep);
1657        xfs_dir2_leaf_check(dp, lbp);
1658        xfs_trans_brelse(tp, lbp);
1659        return 0;
1660}
1661
1662/*
1663 * Return index in the leaf block (lbp) which is either the first
1664 * one with this hash value, or if there are none, the insert point
1665 * for that hash value.
1666 */
1667int                                             /* index value */
1668xfs_dir2_leaf_search_hash(
1669        xfs_da_args_t           *args,          /* operation arguments */
1670        struct xfs_buf          *lbp)           /* leaf buffer */
1671{
1672        xfs_dahash_t            hash=0;         /* hash from this entry */
1673        xfs_dahash_t            hashwant;       /* hash value looking for */
1674        int                     high;           /* high leaf index */
1675        int                     low;            /* low leaf index */
1676        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1677        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1678        int                     mid=0;          /* current leaf index */
1679
1680        leaf = lbp->b_addr;
1681#ifndef __KERNEL__
1682        if (!leaf->hdr.count)
1683                return 0;
1684#endif
1685        /*
1686         * Note, the table cannot be empty, so we have to go through the loop.
1687         * Binary search the leaf entries looking for our hash value.
1688         */
1689        for (lep = leaf->ents, low = 0, high = be16_to_cpu(leaf->hdr.count) - 1,
1690                hashwant = args->hashval;
1691             low <= high; ) {
1692                mid = (low + high) >> 1;
1693                if ((hash = be32_to_cpu(lep[mid].hashval)) == hashwant)
1694                        break;
1695                if (hash < hashwant)
1696                        low = mid + 1;
1697                else
1698                        high = mid - 1;
1699        }
1700        /*
1701         * Found one, back up through all the equal hash values.
1702         */
1703        if (hash == hashwant) {
1704                while (mid > 0 && be32_to_cpu(lep[mid - 1].hashval) == hashwant) {
1705                        mid--;
1706                }
1707        }
1708        /*
1709         * Need to point to an entry higher than ours.
1710         */
1711        else if (hash < hashwant)
1712                mid++;
1713        return mid;
1714}
1715
1716/*
1717 * Trim off a trailing data block.  We know it's empty since the leaf
1718 * freespace table says so.
1719 */
1720int                                             /* error */
1721xfs_dir2_leaf_trim_data(
1722        xfs_da_args_t           *args,          /* operation arguments */
1723        struct xfs_buf          *lbp,           /* leaf buffer */
1724        xfs_dir2_db_t           db)             /* data block number */
1725{
1726        __be16                  *bestsp;        /* leaf bests table */
1727        struct xfs_buf          *dbp;           /* data block buffer */
1728        xfs_inode_t             *dp;            /* incore directory inode */
1729        int                     error;          /* error return value */
1730        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1731        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1732        xfs_mount_t             *mp;            /* filesystem mount point */
1733        xfs_trans_t             *tp;            /* transaction pointer */
1734
1735        dp = args->dp;
1736        mp = dp->i_mount;
1737        tp = args->trans;
1738        /*
1739         * Read the offending data block.  We need its buffer.
1740         */
1741        if ((error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, db), -1, &dbp,
1742                        XFS_DATA_FORK))) {
1743                return error;
1744        }
1745
1746        leaf = lbp->b_addr;
1747        ltp = xfs_dir2_leaf_tail_p(mp, leaf);
1748
1749#ifdef DEBUG
1750{
1751        struct xfs_dir2_data_hdr *hdr = dbp->b_addr;
1752
1753        ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC));
1754        ASSERT(be16_to_cpu(hdr->bestfree[0].length) ==
1755               mp->m_dirblksize - (uint)sizeof(*hdr));
1756        ASSERT(db == be32_to_cpu(ltp->bestcount) - 1);
1757}
1758#endif
1759
1760        /*
1761         * Get rid of the data block.
1762         */
1763        if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1764                ASSERT(error != ENOSPC);
1765                xfs_trans_brelse(tp, dbp);
1766                return error;
1767        }
1768        /*
1769         * Eliminate the last bests entry from the table.
1770         */
1771        bestsp = xfs_dir2_leaf_bests_p(ltp);
1772        be32_add_cpu(&ltp->bestcount, -1);
1773        memmove(&bestsp[1], &bestsp[0], be32_to_cpu(ltp->bestcount) * sizeof(*bestsp));
1774        xfs_dir2_leaf_log_tail(tp, lbp);
1775        xfs_dir2_leaf_log_bests(tp, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
1776        return 0;
1777}
1778
1779static inline size_t
1780xfs_dir2_leaf_size(
1781        struct xfs_dir2_leaf_hdr        *hdr,
1782        int                             counts)
1783{
1784        int                     entries;
1785
1786        entries = be16_to_cpu(hdr->count) - be16_to_cpu(hdr->stale);
1787        return sizeof(xfs_dir2_leaf_hdr_t) +
1788            entries * sizeof(xfs_dir2_leaf_entry_t) +
1789            counts * sizeof(xfs_dir2_data_off_t) +
1790            sizeof(xfs_dir2_leaf_tail_t);
1791}
1792
1793/*
1794 * Convert node form directory to leaf form directory.
1795 * The root of the node form dir needs to already be a LEAFN block.
1796 * Just return if we can't do anything.
1797 */
1798int                                             /* error */
1799xfs_dir2_node_to_leaf(
1800        xfs_da_state_t          *state)         /* directory operation state */
1801{
1802        xfs_da_args_t           *args;          /* operation arguments */
1803        xfs_inode_t             *dp;            /* incore directory inode */
1804        int                     error;          /* error return code */
1805        struct xfs_buf          *fbp;           /* buffer for freespace block */
1806        xfs_fileoff_t           fo;             /* freespace file offset */
1807        xfs_dir2_free_t         *free;          /* freespace structure */
1808        struct xfs_buf          *lbp;           /* buffer for leaf block */
1809        xfs_dir2_leaf_tail_t    *ltp;           /* tail of leaf structure */
1810        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1811        xfs_mount_t             *mp;            /* filesystem mount point */
1812        int                     rval;           /* successful free trim? */
1813        xfs_trans_t             *tp;            /* transaction pointer */
1814
1815        /*
1816         * There's more than a leaf level in the btree, so there must
1817         * be multiple leafn blocks.  Give up.
1818         */
1819        if (state->path.active > 1)
1820                return 0;
1821        args = state->args;
1822
1823        trace_xfs_dir2_node_to_leaf(args);
1824
1825        mp = state->mp;
1826        dp = args->dp;
1827        tp = args->trans;
1828        /*
1829         * Get the last offset in the file.
1830         */
1831        if ((error = xfs_bmap_last_offset(tp, dp, &fo, XFS_DATA_FORK))) {
1832                return error;
1833        }
1834        fo -= mp->m_dirblkfsbs;
1835        /*
1836         * If there are freespace blocks other than the first one,
1837         * take this opportunity to remove trailing empty freespace blocks
1838         * that may have been left behind during no-space-reservation
1839         * operations.
1840         */
1841        while (fo > mp->m_dirfreeblk) {
1842                if ((error = xfs_dir2_node_trim_free(args, fo, &rval))) {
1843                        return error;
1844                }
1845                if (rval)
1846                        fo -= mp->m_dirblkfsbs;
1847                else
1848                        return 0;
1849        }
1850        /*
1851         * Now find the block just before the freespace block.
1852         */
1853        if ((error = xfs_bmap_last_before(tp, dp, &fo, XFS_DATA_FORK))) {
1854                return error;
1855        }
1856        /*
1857         * If it's not the single leaf block, give up.
1858         */
1859        if (XFS_FSB_TO_B(mp, fo) > XFS_DIR2_LEAF_OFFSET + mp->m_dirblksize)
1860                return 0;
1861        lbp = state->path.blk[0].bp;
1862        leaf = lbp->b_addr;
1863        ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
1864        /*
1865         * Read the freespace block.
1866         */
1867        if ((error = xfs_da_read_buf(tp, dp, mp->m_dirfreeblk, -1, &fbp,
1868                        XFS_DATA_FORK))) {
1869                return error;
1870        }
1871        free = fbp->b_addr;
1872        ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC));
1873        ASSERT(!free->hdr.firstdb);
1874
1875        /*
1876         * Now see if the leafn and free data will fit in a leaf1.
1877         * If not, release the buffer and give up.
1878         */
1879        if (xfs_dir2_leaf_size(&leaf->hdr, be32_to_cpu(free->hdr.nvalid)) >
1880                        mp->m_dirblksize) {
1881                xfs_trans_brelse(tp, fbp);
1882                return 0;
1883        }
1884
1885        /*
1886         * If the leaf has any stale entries in it, compress them out.
1887         * The compact routine will log the header.
1888         */
1889        if (be16_to_cpu(leaf->hdr.stale))
1890                xfs_dir2_leaf_compact(args, lbp);
1891        else
1892                xfs_dir2_leaf_log_header(tp, lbp);
1893        leaf->hdr.info.magic = cpu_to_be16(XFS_DIR2_LEAF1_MAGIC);
1894        /*
1895         * Set up the leaf tail from the freespace block.
1896         */
1897        ltp = xfs_dir2_leaf_tail_p(mp, leaf);
1898        ltp->bestcount = free->hdr.nvalid;
1899        /*
1900         * Set up the leaf bests table.
1901         */
1902        memcpy(xfs_dir2_leaf_bests_p(ltp), free->bests,
1903                be32_to_cpu(ltp->bestcount) * sizeof(xfs_dir2_data_off_t));
1904        xfs_dir2_leaf_log_bests(tp, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
1905        xfs_dir2_leaf_log_tail(tp, lbp);
1906        xfs_dir2_leaf_check(dp, lbp);
1907        /*
1908         * Get rid of the freespace block.
1909         */
1910        error = xfs_dir2_shrink_inode(args, XFS_DIR2_FREE_FIRSTDB(mp), fbp);
1911        if (error) {
1912                /*
1913                 * This can't fail here because it can only happen when
1914                 * punching out the middle of an extent, and this is an
1915                 * isolated block.
1916                 */
1917                ASSERT(error != ENOSPC);
1918                return error;
1919        }
1920        fbp = NULL;
1921        /*
1922         * Now see if we can convert the single-leaf directory
1923         * down to a block form directory.
1924         * This routine always kills the dabuf for the leaf, so
1925         * eliminate it from the path.
1926         */
1927        error = xfs_dir2_leaf_to_block(args, lbp, NULL);
1928        state->path.blk[0].bp = NULL;
1929        return error;
1930}
1931