linux/fs/xfs/libxfs/xfs_dir2_leaf.c
<<
>>
Prefs
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
   4 * Copyright (c) 2013 Red Hat, Inc.
   5 * All Rights Reserved.
   6 */
   7#include "xfs.h"
   8#include "xfs_fs.h"
   9#include "xfs_shared.h"
  10#include "xfs_format.h"
  11#include "xfs_log_format.h"
  12#include "xfs_trans_resv.h"
  13#include "xfs_mount.h"
  14#include "xfs_inode.h"
  15#include "xfs_bmap.h"
  16#include "xfs_dir2.h"
  17#include "xfs_dir2_priv.h"
  18#include "xfs_error.h"
  19#include "xfs_trace.h"
  20#include "xfs_trans.h"
  21#include "xfs_buf_item.h"
  22
  23/*
  24 * Local function declarations.
  25 */
  26static int xfs_dir2_leaf_lookup_int(xfs_da_args_t *args, struct xfs_buf **lbpp,
  27                                    int *indexp, struct xfs_buf **dbpp,
  28                                    struct xfs_dir3_icleaf_hdr *leafhdr);
  29static void xfs_dir3_leaf_log_bests(struct xfs_da_args *args,
  30                                    struct xfs_buf *bp, int first, int last);
  31static void xfs_dir3_leaf_log_tail(struct xfs_da_args *args,
  32                                   struct xfs_buf *bp);
  33
  34void
  35xfs_dir2_leaf_hdr_from_disk(
  36        struct xfs_mount                *mp,
  37        struct xfs_dir3_icleaf_hdr      *to,
  38        struct xfs_dir2_leaf            *from)
  39{
  40        if (xfs_has_crc(mp)) {
  41                struct xfs_dir3_leaf *from3 = (struct xfs_dir3_leaf *)from;
  42
  43                to->forw = be32_to_cpu(from3->hdr.info.hdr.forw);
  44                to->back = be32_to_cpu(from3->hdr.info.hdr.back);
  45                to->magic = be16_to_cpu(from3->hdr.info.hdr.magic);
  46                to->count = be16_to_cpu(from3->hdr.count);
  47                to->stale = be16_to_cpu(from3->hdr.stale);
  48                to->ents = from3->__ents;
  49
  50                ASSERT(to->magic == XFS_DIR3_LEAF1_MAGIC ||
  51                       to->magic == XFS_DIR3_LEAFN_MAGIC);
  52        } else {
  53                to->forw = be32_to_cpu(from->hdr.info.forw);
  54                to->back = be32_to_cpu(from->hdr.info.back);
  55                to->magic = be16_to_cpu(from->hdr.info.magic);
  56                to->count = be16_to_cpu(from->hdr.count);
  57                to->stale = be16_to_cpu(from->hdr.stale);
  58                to->ents = from->__ents;
  59
  60                ASSERT(to->magic == XFS_DIR2_LEAF1_MAGIC ||
  61                       to->magic == XFS_DIR2_LEAFN_MAGIC);
  62        }
  63}
  64
  65void
  66xfs_dir2_leaf_hdr_to_disk(
  67        struct xfs_mount                *mp,
  68        struct xfs_dir2_leaf            *to,
  69        struct xfs_dir3_icleaf_hdr      *from)
  70{
  71        if (xfs_has_crc(mp)) {
  72                struct xfs_dir3_leaf *to3 = (struct xfs_dir3_leaf *)to;
  73
  74                ASSERT(from->magic == XFS_DIR3_LEAF1_MAGIC ||
  75                       from->magic == XFS_DIR3_LEAFN_MAGIC);
  76
  77                to3->hdr.info.hdr.forw = cpu_to_be32(from->forw);
  78                to3->hdr.info.hdr.back = cpu_to_be32(from->back);
  79                to3->hdr.info.hdr.magic = cpu_to_be16(from->magic);
  80                to3->hdr.count = cpu_to_be16(from->count);
  81                to3->hdr.stale = cpu_to_be16(from->stale);
  82        } else {
  83                ASSERT(from->magic == XFS_DIR2_LEAF1_MAGIC ||
  84                       from->magic == XFS_DIR2_LEAFN_MAGIC);
  85
  86                to->hdr.info.forw = cpu_to_be32(from->forw);
  87                to->hdr.info.back = cpu_to_be32(from->back);
  88                to->hdr.info.magic = cpu_to_be16(from->magic);
  89                to->hdr.count = cpu_to_be16(from->count);
  90                to->hdr.stale = cpu_to_be16(from->stale);
  91        }
  92}
  93
  94/*
  95 * Check the internal consistency of a leaf1 block.
  96 * Pop an assert if something is wrong.
  97 */
  98#ifdef DEBUG
  99static xfs_failaddr_t
 100xfs_dir3_leaf1_check(
 101        struct xfs_inode        *dp,
 102        struct xfs_buf          *bp)
 103{
 104        struct xfs_dir2_leaf    *leaf = bp->b_addr;
 105        struct xfs_dir3_icleaf_hdr leafhdr;
 106
 107        xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &leafhdr, leaf);
 108
 109        if (leafhdr.magic == XFS_DIR3_LEAF1_MAGIC) {
 110                struct xfs_dir3_leaf_hdr *leaf3 = bp->b_addr;
 111                if (be64_to_cpu(leaf3->info.blkno) != xfs_buf_daddr(bp))
 112                        return __this_address;
 113        } else if (leafhdr.magic != XFS_DIR2_LEAF1_MAGIC)
 114                return __this_address;
 115
 116        return xfs_dir3_leaf_check_int(dp->i_mount, &leafhdr, leaf, false);
 117}
 118
 119static inline void
 120xfs_dir3_leaf_check(
 121        struct xfs_inode        *dp,
 122        struct xfs_buf          *bp)
 123{
 124        xfs_failaddr_t          fa;
 125
 126        fa = xfs_dir3_leaf1_check(dp, bp);
 127        if (!fa)
 128                return;
 129        xfs_corruption_error(__func__, XFS_ERRLEVEL_LOW, dp->i_mount,
 130                        bp->b_addr, BBTOB(bp->b_length), __FILE__, __LINE__,
 131                        fa);
 132        ASSERT(0);
 133}
 134#else
 135#define xfs_dir3_leaf_check(dp, bp)
 136#endif
 137
 138xfs_failaddr_t
 139xfs_dir3_leaf_check_int(
 140        struct xfs_mount                *mp,
 141        struct xfs_dir3_icleaf_hdr      *hdr,
 142        struct xfs_dir2_leaf            *leaf,
 143        bool                            expensive_checking)
 144{
 145        struct xfs_da_geometry          *geo = mp->m_dir_geo;
 146        xfs_dir2_leaf_tail_t            *ltp;
 147        int                             stale;
 148        int                             i;
 149
 150        ltp = xfs_dir2_leaf_tail_p(geo, leaf);
 151
 152        /*
 153         * XXX (dgc): This value is not restrictive enough.
 154         * Should factor in the size of the bests table as well.
 155         * We can deduce a value for that from i_disk_size.
 156         */
 157        if (hdr->count > geo->leaf_max_ents)
 158                return __this_address;
 159
 160        /* Leaves and bests don't overlap in leaf format. */
 161        if ((hdr->magic == XFS_DIR2_LEAF1_MAGIC ||
 162             hdr->magic == XFS_DIR3_LEAF1_MAGIC) &&
 163            (char *)&hdr->ents[hdr->count] > (char *)xfs_dir2_leaf_bests_p(ltp))
 164                return __this_address;
 165
 166        if (!expensive_checking)
 167                return NULL;
 168
 169        /* Check hash value order, count stale entries.  */
 170        for (i = stale = 0; i < hdr->count; i++) {
 171                if (i + 1 < hdr->count) {
 172                        if (be32_to_cpu(hdr->ents[i].hashval) >
 173                                        be32_to_cpu(hdr->ents[i + 1].hashval))
 174                                return __this_address;
 175                }
 176                if (hdr->ents[i].address == cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 177                        stale++;
 178        }
 179        if (hdr->stale != stale)
 180                return __this_address;
 181        return NULL;
 182}
 183
 184/*
 185 * We verify the magic numbers before decoding the leaf header so that on debug
 186 * kernels we don't get assertion failures in xfs_dir3_leaf_hdr_from_disk() due
 187 * to incorrect magic numbers.
 188 */
 189static xfs_failaddr_t
 190xfs_dir3_leaf_verify(
 191        struct xfs_buf                  *bp)
 192{
 193        struct xfs_mount                *mp = bp->b_mount;
 194        struct xfs_dir3_icleaf_hdr      leafhdr;
 195        xfs_failaddr_t                  fa;
 196
 197        fa = xfs_da3_blkinfo_verify(bp, bp->b_addr);
 198        if (fa)
 199                return fa;
 200
 201        xfs_dir2_leaf_hdr_from_disk(mp, &leafhdr, bp->b_addr);
 202        return xfs_dir3_leaf_check_int(mp, &leafhdr, bp->b_addr, true);
 203}
 204
 205static void
 206xfs_dir3_leaf_read_verify(
 207        struct xfs_buf  *bp)
 208{
 209        struct xfs_mount        *mp = bp->b_mount;
 210        xfs_failaddr_t          fa;
 211
 212        if (xfs_has_crc(mp) &&
 213             !xfs_buf_verify_cksum(bp, XFS_DIR3_LEAF_CRC_OFF))
 214                xfs_verifier_error(bp, -EFSBADCRC, __this_address);
 215        else {
 216                fa = xfs_dir3_leaf_verify(bp);
 217                if (fa)
 218                        xfs_verifier_error(bp, -EFSCORRUPTED, fa);
 219        }
 220}
 221
 222static void
 223xfs_dir3_leaf_write_verify(
 224        struct xfs_buf  *bp)
 225{
 226        struct xfs_mount        *mp = bp->b_mount;
 227        struct xfs_buf_log_item *bip = bp->b_log_item;
 228        struct xfs_dir3_leaf_hdr *hdr3 = bp->b_addr;
 229        xfs_failaddr_t          fa;
 230
 231        fa = xfs_dir3_leaf_verify(bp);
 232        if (fa) {
 233                xfs_verifier_error(bp, -EFSCORRUPTED, fa);
 234                return;
 235        }
 236
 237        if (!xfs_has_crc(mp))
 238                return;
 239
 240        if (bip)
 241                hdr3->info.lsn = cpu_to_be64(bip->bli_item.li_lsn);
 242
 243        xfs_buf_update_cksum(bp, XFS_DIR3_LEAF_CRC_OFF);
 244}
 245
 246const struct xfs_buf_ops xfs_dir3_leaf1_buf_ops = {
 247        .name = "xfs_dir3_leaf1",
 248        .magic16 = { cpu_to_be16(XFS_DIR2_LEAF1_MAGIC),
 249                     cpu_to_be16(XFS_DIR3_LEAF1_MAGIC) },
 250        .verify_read = xfs_dir3_leaf_read_verify,
 251        .verify_write = xfs_dir3_leaf_write_verify,
 252        .verify_struct = xfs_dir3_leaf_verify,
 253};
 254
 255const struct xfs_buf_ops xfs_dir3_leafn_buf_ops = {
 256        .name = "xfs_dir3_leafn",
 257        .magic16 = { cpu_to_be16(XFS_DIR2_LEAFN_MAGIC),
 258                     cpu_to_be16(XFS_DIR3_LEAFN_MAGIC) },
 259        .verify_read = xfs_dir3_leaf_read_verify,
 260        .verify_write = xfs_dir3_leaf_write_verify,
 261        .verify_struct = xfs_dir3_leaf_verify,
 262};
 263
 264int
 265xfs_dir3_leaf_read(
 266        struct xfs_trans        *tp,
 267        struct xfs_inode        *dp,
 268        xfs_dablk_t             fbno,
 269        struct xfs_buf          **bpp)
 270{
 271        int                     err;
 272
 273        err = xfs_da_read_buf(tp, dp, fbno, 0, bpp, XFS_DATA_FORK,
 274                        &xfs_dir3_leaf1_buf_ops);
 275        if (!err && tp && *bpp)
 276                xfs_trans_buf_set_type(tp, *bpp, XFS_BLFT_DIR_LEAF1_BUF);
 277        return err;
 278}
 279
 280int
 281xfs_dir3_leafn_read(
 282        struct xfs_trans        *tp,
 283        struct xfs_inode        *dp,
 284        xfs_dablk_t             fbno,
 285        struct xfs_buf          **bpp)
 286{
 287        int                     err;
 288
 289        err = xfs_da_read_buf(tp, dp, fbno, 0, bpp, XFS_DATA_FORK,
 290                        &xfs_dir3_leafn_buf_ops);
 291        if (!err && tp && *bpp)
 292                xfs_trans_buf_set_type(tp, *bpp, XFS_BLFT_DIR_LEAFN_BUF);
 293        return err;
 294}
 295
 296/*
 297 * Initialize a new leaf block, leaf1 or leafn magic accepted.
 298 */
 299static void
 300xfs_dir3_leaf_init(
 301        struct xfs_mount        *mp,
 302        struct xfs_trans        *tp,
 303        struct xfs_buf          *bp,
 304        xfs_ino_t               owner,
 305        uint16_t                type)
 306{
 307        struct xfs_dir2_leaf    *leaf = bp->b_addr;
 308
 309        ASSERT(type == XFS_DIR2_LEAF1_MAGIC || type == XFS_DIR2_LEAFN_MAGIC);
 310
 311        if (xfs_has_crc(mp)) {
 312                struct xfs_dir3_leaf_hdr *leaf3 = bp->b_addr;
 313
 314                memset(leaf3, 0, sizeof(*leaf3));
 315
 316                leaf3->info.hdr.magic = (type == XFS_DIR2_LEAF1_MAGIC)
 317                                         ? cpu_to_be16(XFS_DIR3_LEAF1_MAGIC)
 318                                         : cpu_to_be16(XFS_DIR3_LEAFN_MAGIC);
 319                leaf3->info.blkno = cpu_to_be64(xfs_buf_daddr(bp));
 320                leaf3->info.owner = cpu_to_be64(owner);
 321                uuid_copy(&leaf3->info.uuid, &mp->m_sb.sb_meta_uuid);
 322        } else {
 323                memset(leaf, 0, sizeof(*leaf));
 324                leaf->hdr.info.magic = cpu_to_be16(type);
 325        }
 326
 327        /*
 328         * If it's a leaf-format directory initialize the tail.
 329         * Caller is responsible for initialising the bests table.
 330         */
 331        if (type == XFS_DIR2_LEAF1_MAGIC) {
 332                struct xfs_dir2_leaf_tail *ltp;
 333
 334                ltp = xfs_dir2_leaf_tail_p(mp->m_dir_geo, leaf);
 335                ltp->bestcount = 0;
 336                bp->b_ops = &xfs_dir3_leaf1_buf_ops;
 337                xfs_trans_buf_set_type(tp, bp, XFS_BLFT_DIR_LEAF1_BUF);
 338        } else {
 339                bp->b_ops = &xfs_dir3_leafn_buf_ops;
 340                xfs_trans_buf_set_type(tp, bp, XFS_BLFT_DIR_LEAFN_BUF);
 341        }
 342}
 343
 344int
 345xfs_dir3_leaf_get_buf(
 346        xfs_da_args_t           *args,
 347        xfs_dir2_db_t           bno,
 348        struct xfs_buf          **bpp,
 349        uint16_t                magic)
 350{
 351        struct xfs_inode        *dp = args->dp;
 352        struct xfs_trans        *tp = args->trans;
 353        struct xfs_mount        *mp = dp->i_mount;
 354        struct xfs_buf          *bp;
 355        int                     error;
 356
 357        ASSERT(magic == XFS_DIR2_LEAF1_MAGIC || magic == XFS_DIR2_LEAFN_MAGIC);
 358        ASSERT(bno >= xfs_dir2_byte_to_db(args->geo, XFS_DIR2_LEAF_OFFSET) &&
 359               bno < xfs_dir2_byte_to_db(args->geo, XFS_DIR2_FREE_OFFSET));
 360
 361        error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(args->geo, bno),
 362                               &bp, XFS_DATA_FORK);
 363        if (error)
 364                return error;
 365
 366        xfs_dir3_leaf_init(mp, tp, bp, dp->i_ino, magic);
 367        xfs_dir3_leaf_log_header(args, bp);
 368        if (magic == XFS_DIR2_LEAF1_MAGIC)
 369                xfs_dir3_leaf_log_tail(args, bp);
 370        *bpp = bp;
 371        return 0;
 372}
 373
 374/*
 375 * Convert a block form directory to a leaf form directory.
 376 */
 377int                                             /* error */
 378xfs_dir2_block_to_leaf(
 379        xfs_da_args_t           *args,          /* operation arguments */
 380        struct xfs_buf          *dbp)           /* input block's buffer */
 381{
 382        __be16                  *bestsp;        /* leaf's bestsp entries */
 383        xfs_dablk_t             blkno;          /* leaf block's bno */
 384        xfs_dir2_data_hdr_t     *hdr;           /* block header */
 385        xfs_dir2_leaf_entry_t   *blp;           /* block's leaf entries */
 386        xfs_dir2_block_tail_t   *btp;           /* block's tail */
 387        xfs_inode_t             *dp;            /* incore directory inode */
 388        int                     error;          /* error return code */
 389        struct xfs_buf          *lbp;           /* leaf block's buffer */
 390        xfs_dir2_db_t           ldb;            /* leaf block's bno */
 391        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
 392        xfs_dir2_leaf_tail_t    *ltp;           /* leaf's tail */
 393        int                     needlog;        /* need to log block header */
 394        int                     needscan;       /* need to rescan bestfree */
 395        xfs_trans_t             *tp;            /* transaction pointer */
 396        struct xfs_dir2_data_free *bf;
 397        struct xfs_dir3_icleaf_hdr leafhdr;
 398
 399        trace_xfs_dir2_block_to_leaf(args);
 400
 401        dp = args->dp;
 402        tp = args->trans;
 403        /*
 404         * Add the leaf block to the inode.
 405         * This interface will only put blocks in the leaf/node range.
 406         * Since that's empty now, we'll get the root (block 0 in range).
 407         */
 408        if ((error = xfs_da_grow_inode(args, &blkno))) {
 409                return error;
 410        }
 411        ldb = xfs_dir2_da_to_db(args->geo, blkno);
 412        ASSERT(ldb == xfs_dir2_byte_to_db(args->geo, XFS_DIR2_LEAF_OFFSET));
 413        /*
 414         * Initialize the leaf block, get a buffer for it.
 415         */
 416        error = xfs_dir3_leaf_get_buf(args, ldb, &lbp, XFS_DIR2_LEAF1_MAGIC);
 417        if (error)
 418                return error;
 419
 420        leaf = lbp->b_addr;
 421        hdr = dbp->b_addr;
 422        xfs_dir3_data_check(dp, dbp);
 423        btp = xfs_dir2_block_tail_p(args->geo, hdr);
 424        blp = xfs_dir2_block_leaf_p(btp);
 425        bf = xfs_dir2_data_bestfree_p(dp->i_mount, hdr);
 426
 427        /*
 428         * Set the counts in the leaf header.
 429         */
 430        xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &leafhdr, leaf);
 431        leafhdr.count = be32_to_cpu(btp->count);
 432        leafhdr.stale = be32_to_cpu(btp->stale);
 433        xfs_dir2_leaf_hdr_to_disk(dp->i_mount, leaf, &leafhdr);
 434        xfs_dir3_leaf_log_header(args, lbp);
 435
 436        /*
 437         * Could compact these but I think we always do the conversion
 438         * after squeezing out stale entries.
 439         */
 440        memcpy(leafhdr.ents, blp,
 441                be32_to_cpu(btp->count) * sizeof(struct xfs_dir2_leaf_entry));
 442        xfs_dir3_leaf_log_ents(args, &leafhdr, lbp, 0, leafhdr.count - 1);
 443        needscan = 0;
 444        needlog = 1;
 445        /*
 446         * Make the space formerly occupied by the leaf entries and block
 447         * tail be free.
 448         */
 449        xfs_dir2_data_make_free(args, dbp,
 450                (xfs_dir2_data_aoff_t)((char *)blp - (char *)hdr),
 451                (xfs_dir2_data_aoff_t)((char *)hdr + args->geo->blksize -
 452                                       (char *)blp),
 453                &needlog, &needscan);
 454        /*
 455         * Fix up the block header, make it a data block.
 456         */
 457        dbp->b_ops = &xfs_dir3_data_buf_ops;
 458        xfs_trans_buf_set_type(tp, dbp, XFS_BLFT_DIR_DATA_BUF);
 459        if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC))
 460                hdr->magic = cpu_to_be32(XFS_DIR2_DATA_MAGIC);
 461        else
 462                hdr->magic = cpu_to_be32(XFS_DIR3_DATA_MAGIC);
 463
 464        if (needscan)
 465                xfs_dir2_data_freescan(dp->i_mount, hdr, &needlog);
 466        /*
 467         * Set up leaf tail and bests table.
 468         */
 469        ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
 470        ltp->bestcount = cpu_to_be32(1);
 471        bestsp = xfs_dir2_leaf_bests_p(ltp);
 472        bestsp[0] =  bf[0].length;
 473        /*
 474         * Log the data header and leaf bests table.
 475         */
 476        if (needlog)
 477                xfs_dir2_data_log_header(args, dbp);
 478        xfs_dir3_leaf_check(dp, lbp);
 479        xfs_dir3_data_check(dp, dbp);
 480        xfs_dir3_leaf_log_bests(args, lbp, 0, 0);
 481        return 0;
 482}
 483
 484STATIC void
 485xfs_dir3_leaf_find_stale(
 486        struct xfs_dir3_icleaf_hdr *leafhdr,
 487        struct xfs_dir2_leaf_entry *ents,
 488        int                     index,
 489        int                     *lowstale,
 490        int                     *highstale)
 491{
 492        /*
 493         * Find the first stale entry before our index, if any.
 494         */
 495        for (*lowstale = index - 1; *lowstale >= 0; --*lowstale) {
 496                if (ents[*lowstale].address ==
 497                    cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 498                        break;
 499        }
 500
 501        /*
 502         * Find the first stale entry at or after our index, if any.
 503         * Stop if the result would require moving more entries than using
 504         * lowstale.
 505         */
 506        for (*highstale = index; *highstale < leafhdr->count; ++*highstale) {
 507                if (ents[*highstale].address ==
 508                    cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 509                        break;
 510                if (*lowstale >= 0 && index - *lowstale <= *highstale - index)
 511                        break;
 512        }
 513}
 514
 515struct xfs_dir2_leaf_entry *
 516xfs_dir3_leaf_find_entry(
 517        struct xfs_dir3_icleaf_hdr *leafhdr,
 518        struct xfs_dir2_leaf_entry *ents,
 519        int                     index,          /* leaf table position */
 520        int                     compact,        /* need to compact leaves */
 521        int                     lowstale,       /* index of prev stale leaf */
 522        int                     highstale,      /* index of next stale leaf */
 523        int                     *lfloglow,      /* low leaf logging index */
 524        int                     *lfloghigh)     /* high leaf logging index */
 525{
 526        if (!leafhdr->stale) {
 527                xfs_dir2_leaf_entry_t   *lep;   /* leaf entry table pointer */
 528
 529                /*
 530                 * Now we need to make room to insert the leaf entry.
 531                 *
 532                 * If there are no stale entries, just insert a hole at index.
 533                 */
 534                lep = &ents[index];
 535                if (index < leafhdr->count)
 536                        memmove(lep + 1, lep,
 537                                (leafhdr->count - index) * sizeof(*lep));
 538
 539                /*
 540                 * Record low and high logging indices for the leaf.
 541                 */
 542                *lfloglow = index;
 543                *lfloghigh = leafhdr->count++;
 544                return lep;
 545        }
 546
 547        /*
 548         * There are stale entries.
 549         *
 550         * We will use one of them for the new entry.  It's probably not at
 551         * the right location, so we'll have to shift some up or down first.
 552         *
 553         * If we didn't compact before, we need to find the nearest stale
 554         * entries before and after our insertion point.
 555         */
 556        if (compact == 0)
 557                xfs_dir3_leaf_find_stale(leafhdr, ents, index,
 558                                         &lowstale, &highstale);
 559
 560        /*
 561         * If the low one is better, use it.
 562         */
 563        if (lowstale >= 0 &&
 564            (highstale == leafhdr->count ||
 565             index - lowstale - 1 < highstale - index)) {
 566                ASSERT(index - lowstale - 1 >= 0);
 567                ASSERT(ents[lowstale].address ==
 568                       cpu_to_be32(XFS_DIR2_NULL_DATAPTR));
 569
 570                /*
 571                 * Copy entries up to cover the stale entry and make room
 572                 * for the new entry.
 573                 */
 574                if (index - lowstale - 1 > 0) {
 575                        memmove(&ents[lowstale], &ents[lowstale + 1],
 576                                (index - lowstale - 1) *
 577                                        sizeof(xfs_dir2_leaf_entry_t));
 578                }
 579                *lfloglow = min(lowstale, *lfloglow);
 580                *lfloghigh = max(index - 1, *lfloghigh);
 581                leafhdr->stale--;
 582                return &ents[index - 1];
 583        }
 584
 585        /*
 586         * The high one is better, so use that one.
 587         */
 588        ASSERT(highstale - index >= 0);
 589        ASSERT(ents[highstale].address == cpu_to_be32(XFS_DIR2_NULL_DATAPTR));
 590
 591        /*
 592         * Copy entries down to cover the stale entry and make room for the
 593         * new entry.
 594         */
 595        if (highstale - index > 0) {
 596                memmove(&ents[index + 1], &ents[index],
 597                        (highstale - index) * sizeof(xfs_dir2_leaf_entry_t));
 598        }
 599        *lfloglow = min(index, *lfloglow);
 600        *lfloghigh = max(highstale, *lfloghigh);
 601        leafhdr->stale--;
 602        return &ents[index];
 603}
 604
 605/*
 606 * Add an entry to a leaf form directory.
 607 */
 608int                                             /* error */
 609xfs_dir2_leaf_addname(
 610        struct xfs_da_args      *args)          /* operation arguments */
 611{
 612        struct xfs_dir3_icleaf_hdr leafhdr;
 613        struct xfs_trans        *tp = args->trans;
 614        __be16                  *bestsp;        /* freespace table in leaf */
 615        __be16                  *tagp;          /* end of data entry */
 616        struct xfs_buf          *dbp;           /* data block buffer */
 617        struct xfs_buf          *lbp;           /* leaf's buffer */
 618        struct xfs_dir2_leaf    *leaf;          /* leaf structure */
 619        struct xfs_inode        *dp = args->dp; /* incore directory inode */
 620        struct xfs_dir2_data_hdr *hdr;          /* data block header */
 621        struct xfs_dir2_data_entry *dep;        /* data block entry */
 622        struct xfs_dir2_leaf_entry *lep;        /* leaf entry table pointer */
 623        struct xfs_dir2_leaf_entry *ents;
 624        struct xfs_dir2_data_unused *dup;       /* data unused entry */
 625        struct xfs_dir2_leaf_tail *ltp;         /* leaf tail pointer */
 626        struct xfs_dir2_data_free *bf;          /* bestfree table */
 627        int                     compact;        /* need to compact leaves */
 628        int                     error;          /* error return value */
 629        int                     grown;          /* allocated new data block */
 630        int                     highstale = 0;  /* index of next stale leaf */
 631        int                     i;              /* temporary, index */
 632        int                     index;          /* leaf table position */
 633        int                     length;         /* length of new entry */
 634        int                     lfloglow;       /* low leaf logging index */
 635        int                     lfloghigh;      /* high leaf logging index */
 636        int                     lowstale = 0;   /* index of prev stale leaf */
 637        int                     needbytes;      /* leaf block bytes needed */
 638        int                     needlog;        /* need to log data header */
 639        int                     needscan;       /* need to rescan data free */
 640        xfs_dir2_db_t           use_block;      /* data block number */
 641
 642        trace_xfs_dir2_leaf_addname(args);
 643
 644        error = xfs_dir3_leaf_read(tp, dp, args->geo->leafblk, &lbp);
 645        if (error)
 646                return error;
 647
 648        /*
 649         * Look up the entry by hash value and name.
 650         * We know it's not there, our caller has already done a lookup.
 651         * So the index is of the entry to insert in front of.
 652         * But if there are dup hash values the index is of the first of those.
 653         */
 654        index = xfs_dir2_leaf_search_hash(args, lbp);
 655        leaf = lbp->b_addr;
 656        ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
 657        xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &leafhdr, leaf);
 658        ents = leafhdr.ents;
 659        bestsp = xfs_dir2_leaf_bests_p(ltp);
 660        length = xfs_dir2_data_entsize(dp->i_mount, args->namelen);
 661
 662        /*
 663         * See if there are any entries with the same hash value
 664         * and space in their block for the new entry.
 665         * This is good because it puts multiple same-hash value entries
 666         * in a data block, improving the lookup of those entries.
 667         */
 668        for (use_block = -1, lep = &ents[index];
 669             index < leafhdr.count && be32_to_cpu(lep->hashval) == args->hashval;
 670             index++, lep++) {
 671                if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
 672                        continue;
 673                i = xfs_dir2_dataptr_to_db(args->geo, be32_to_cpu(lep->address));
 674                ASSERT(i < be32_to_cpu(ltp->bestcount));
 675                ASSERT(bestsp[i] != cpu_to_be16(NULLDATAOFF));
 676                if (be16_to_cpu(bestsp[i]) >= length) {
 677                        use_block = i;
 678                        break;
 679                }
 680        }
 681        /*
 682         * Didn't find a block yet, linear search all the data blocks.
 683         */
 684        if (use_block == -1) {
 685                for (i = 0; i < be32_to_cpu(ltp->bestcount); i++) {
 686                        /*
 687                         * Remember a block we see that's missing.
 688                         */
 689                        if (bestsp[i] == cpu_to_be16(NULLDATAOFF) &&
 690                            use_block == -1)
 691                                use_block = i;
 692                        else if (be16_to_cpu(bestsp[i]) >= length) {
 693                                use_block = i;
 694                                break;
 695                        }
 696                }
 697        }
 698        /*
 699         * How many bytes do we need in the leaf block?
 700         */
 701        needbytes = 0;
 702        if (!leafhdr.stale)
 703                needbytes += sizeof(xfs_dir2_leaf_entry_t);
 704        if (use_block == -1)
 705                needbytes += sizeof(xfs_dir2_data_off_t);
 706
 707        /*
 708         * Now kill use_block if it refers to a missing block, so we
 709         * can use it as an indication of allocation needed.
 710         */
 711        if (use_block != -1 && bestsp[use_block] == cpu_to_be16(NULLDATAOFF))
 712                use_block = -1;
 713        /*
 714         * If we don't have enough free bytes but we can make enough
 715         * by compacting out stale entries, we'll do that.
 716         */
 717        if ((char *)bestsp - (char *)&ents[leafhdr.count] < needbytes &&
 718            leafhdr.stale > 1)
 719                compact = 1;
 720
 721        /*
 722         * Otherwise if we don't have enough free bytes we need to
 723         * convert to node form.
 724         */
 725        else if ((char *)bestsp - (char *)&ents[leafhdr.count] < needbytes) {
 726                /*
 727                 * Just checking or no space reservation, give up.
 728                 */
 729                if ((args->op_flags & XFS_DA_OP_JUSTCHECK) ||
 730                                                        args->total == 0) {
 731                        xfs_trans_brelse(tp, lbp);
 732                        return -ENOSPC;
 733                }
 734                /*
 735                 * Convert to node form.
 736                 */
 737                error = xfs_dir2_leaf_to_node(args, lbp);
 738                if (error)
 739                        return error;
 740                /*
 741                 * Then add the new entry.
 742                 */
 743                return xfs_dir2_node_addname(args);
 744        }
 745        /*
 746         * Otherwise it will fit without compaction.
 747         */
 748        else
 749                compact = 0;
 750        /*
 751         * If just checking, then it will fit unless we needed to allocate
 752         * a new data block.
 753         */
 754        if (args->op_flags & XFS_DA_OP_JUSTCHECK) {
 755                xfs_trans_brelse(tp, lbp);
 756                return use_block == -1 ? -ENOSPC : 0;
 757        }
 758        /*
 759         * If no allocations are allowed, return now before we've
 760         * changed anything.
 761         */
 762        if (args->total == 0 && use_block == -1) {
 763                xfs_trans_brelse(tp, lbp);
 764                return -ENOSPC;
 765        }
 766        /*
 767         * Need to compact the leaf entries, removing stale ones.
 768         * Leave one stale entry behind - the one closest to our
 769         * insertion index - and we'll shift that one to our insertion
 770         * point later.
 771         */
 772        if (compact) {
 773                xfs_dir3_leaf_compact_x1(&leafhdr, ents, &index, &lowstale,
 774                        &highstale, &lfloglow, &lfloghigh);
 775        }
 776        /*
 777         * There are stale entries, so we'll need log-low and log-high
 778         * impossibly bad values later.
 779         */
 780        else if (leafhdr.stale) {
 781                lfloglow = leafhdr.count;
 782                lfloghigh = -1;
 783        }
 784        /*
 785         * If there was no data block space found, we need to allocate
 786         * a new one.
 787         */
 788        if (use_block == -1) {
 789                /*
 790                 * Add the new data block.
 791                 */
 792                if ((error = xfs_dir2_grow_inode(args, XFS_DIR2_DATA_SPACE,
 793                                &use_block))) {
 794                        xfs_trans_brelse(tp, lbp);
 795                        return error;
 796                }
 797                /*
 798                 * Initialize the block.
 799                 */
 800                if ((error = xfs_dir3_data_init(args, use_block, &dbp))) {
 801                        xfs_trans_brelse(tp, lbp);
 802                        return error;
 803                }
 804                /*
 805                 * If we're adding a new data block on the end we need to
 806                 * extend the bests table.  Copy it up one entry.
 807                 */
 808                if (use_block >= be32_to_cpu(ltp->bestcount)) {
 809                        bestsp--;
 810                        memmove(&bestsp[0], &bestsp[1],
 811                                be32_to_cpu(ltp->bestcount) * sizeof(bestsp[0]));
 812                        be32_add_cpu(&ltp->bestcount, 1);
 813                        xfs_dir3_leaf_log_tail(args, lbp);
 814                        xfs_dir3_leaf_log_bests(args, lbp, 0,
 815                                                be32_to_cpu(ltp->bestcount) - 1);
 816                }
 817                /*
 818                 * If we're filling in a previously empty block just log it.
 819                 */
 820                else
 821                        xfs_dir3_leaf_log_bests(args, lbp, use_block, use_block);
 822                hdr = dbp->b_addr;
 823                bf = xfs_dir2_data_bestfree_p(dp->i_mount, hdr);
 824                bestsp[use_block] = bf[0].length;
 825                grown = 1;
 826        } else {
 827                /*
 828                 * Already had space in some data block.
 829                 * Just read that one in.
 830                 */
 831                error = xfs_dir3_data_read(tp, dp,
 832                                   xfs_dir2_db_to_da(args->geo, use_block),
 833                                   0, &dbp);
 834                if (error) {
 835                        xfs_trans_brelse(tp, lbp);
 836                        return error;
 837                }
 838                hdr = dbp->b_addr;
 839                bf = xfs_dir2_data_bestfree_p(dp->i_mount, hdr);
 840                grown = 0;
 841        }
 842        /*
 843         * Point to the biggest freespace in our data block.
 844         */
 845        dup = (xfs_dir2_data_unused_t *)
 846              ((char *)hdr + be16_to_cpu(bf[0].offset));
 847        needscan = needlog = 0;
 848        /*
 849         * Mark the initial part of our freespace in use for the new entry.
 850         */
 851        error = xfs_dir2_data_use_free(args, dbp, dup,
 852                        (xfs_dir2_data_aoff_t)((char *)dup - (char *)hdr),
 853                        length, &needlog, &needscan);
 854        if (error) {
 855                xfs_trans_brelse(tp, lbp);
 856                return error;
 857        }
 858        /*
 859         * Initialize our new entry (at last).
 860         */
 861        dep = (xfs_dir2_data_entry_t *)dup;
 862        dep->inumber = cpu_to_be64(args->inumber);
 863        dep->namelen = args->namelen;
 864        memcpy(dep->name, args->name, dep->namelen);
 865        xfs_dir2_data_put_ftype(dp->i_mount, dep, args->filetype);
 866        tagp = xfs_dir2_data_entry_tag_p(dp->i_mount, dep);
 867        *tagp = cpu_to_be16((char *)dep - (char *)hdr);
 868        /*
 869         * Need to scan fix up the bestfree table.
 870         */
 871        if (needscan)
 872                xfs_dir2_data_freescan(dp->i_mount, hdr, &needlog);
 873        /*
 874         * Need to log the data block's header.
 875         */
 876        if (needlog)
 877                xfs_dir2_data_log_header(args, dbp);
 878        xfs_dir2_data_log_entry(args, dbp, dep);
 879        /*
 880         * If the bests table needs to be changed, do it.
 881         * Log the change unless we've already done that.
 882         */
 883        if (be16_to_cpu(bestsp[use_block]) != be16_to_cpu(bf[0].length)) {
 884                bestsp[use_block] = bf[0].length;
 885                if (!grown)
 886                        xfs_dir3_leaf_log_bests(args, lbp, use_block, use_block);
 887        }
 888
 889        lep = xfs_dir3_leaf_find_entry(&leafhdr, ents, index, compact, lowstale,
 890                                       highstale, &lfloglow, &lfloghigh);
 891
 892        /*
 893         * Fill in the new leaf entry.
 894         */
 895        lep->hashval = cpu_to_be32(args->hashval);
 896        lep->address = cpu_to_be32(
 897                                xfs_dir2_db_off_to_dataptr(args->geo, use_block,
 898                                be16_to_cpu(*tagp)));
 899        /*
 900         * Log the leaf fields and give up the buffers.
 901         */
 902        xfs_dir2_leaf_hdr_to_disk(dp->i_mount, leaf, &leafhdr);
 903        xfs_dir3_leaf_log_header(args, lbp);
 904        xfs_dir3_leaf_log_ents(args, &leafhdr, lbp, lfloglow, lfloghigh);
 905        xfs_dir3_leaf_check(dp, lbp);
 906        xfs_dir3_data_check(dp, dbp);
 907        return 0;
 908}
 909
 910/*
 911 * Compact out any stale entries in the leaf.
 912 * Log the header and changed leaf entries, if any.
 913 */
 914void
 915xfs_dir3_leaf_compact(
 916        xfs_da_args_t   *args,          /* operation arguments */
 917        struct xfs_dir3_icleaf_hdr *leafhdr,
 918        struct xfs_buf  *bp)            /* leaf buffer */
 919{
 920        int             from;           /* source leaf index */
 921        xfs_dir2_leaf_t *leaf;          /* leaf structure */
 922        int             loglow;         /* first leaf entry to log */
 923        int             to;             /* target leaf index */
 924        struct xfs_inode *dp = args->dp;
 925
 926        leaf = bp->b_addr;
 927        if (!leafhdr->stale)
 928                return;
 929
 930        /*
 931         * Compress out the stale entries in place.
 932         */
 933        for (from = to = 0, loglow = -1; from < leafhdr->count; from++) {
 934                if (leafhdr->ents[from].address ==
 935                    cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 936                        continue;
 937                /*
 938                 * Only actually copy the entries that are different.
 939                 */
 940                if (from > to) {
 941                        if (loglow == -1)
 942                                loglow = to;
 943                        leafhdr->ents[to] = leafhdr->ents[from];
 944                }
 945                to++;
 946        }
 947        /*
 948         * Update and log the header, log the leaf entries.
 949         */
 950        ASSERT(leafhdr->stale == from - to);
 951        leafhdr->count -= leafhdr->stale;
 952        leafhdr->stale = 0;
 953
 954        xfs_dir2_leaf_hdr_to_disk(dp->i_mount, leaf, leafhdr);
 955        xfs_dir3_leaf_log_header(args, bp);
 956        if (loglow != -1)
 957                xfs_dir3_leaf_log_ents(args, leafhdr, bp, loglow, to - 1);
 958}
 959
 960/*
 961 * Compact the leaf entries, removing stale ones.
 962 * Leave one stale entry behind - the one closest to our
 963 * insertion index - and the caller will shift that one to our insertion
 964 * point later.
 965 * Return new insertion index, where the remaining stale entry is,
 966 * and leaf logging indices.
 967 */
 968void
 969xfs_dir3_leaf_compact_x1(
 970        struct xfs_dir3_icleaf_hdr *leafhdr,
 971        struct xfs_dir2_leaf_entry *ents,
 972        int             *indexp,        /* insertion index */
 973        int             *lowstalep,     /* out: stale entry before us */
 974        int             *highstalep,    /* out: stale entry after us */
 975        int             *lowlogp,       /* out: low log index */
 976        int             *highlogp)      /* out: high log index */
 977{
 978        int             from;           /* source copy index */
 979        int             highstale;      /* stale entry at/after index */
 980        int             index;          /* insertion index */
 981        int             keepstale;      /* source index of kept stale */
 982        int             lowstale;       /* stale entry before index */
 983        int             newindex=0;     /* new insertion index */
 984        int             to;             /* destination copy index */
 985
 986        ASSERT(leafhdr->stale > 1);
 987        index = *indexp;
 988
 989        xfs_dir3_leaf_find_stale(leafhdr, ents, index, &lowstale, &highstale);
 990
 991        /*
 992         * Pick the better of lowstale and highstale.
 993         */
 994        if (lowstale >= 0 &&
 995            (highstale == leafhdr->count ||
 996             index - lowstale <= highstale - index))
 997                keepstale = lowstale;
 998        else
 999                keepstale = highstale;
1000        /*
1001         * Copy the entries in place, removing all the stale entries
1002         * except keepstale.
1003         */
1004        for (from = to = 0; from < leafhdr->count; from++) {
1005                /*
1006                 * Notice the new value of index.
1007                 */
1008                if (index == from)
1009                        newindex = to;
1010                if (from != keepstale &&
1011                    ents[from].address == cpu_to_be32(XFS_DIR2_NULL_DATAPTR)) {
1012                        if (from == to)
1013                                *lowlogp = to;
1014                        continue;
1015                }
1016                /*
1017                 * Record the new keepstale value for the insertion.
1018                 */
1019                if (from == keepstale)
1020                        lowstale = highstale = to;
1021                /*
1022                 * Copy only the entries that have moved.
1023                 */
1024                if (from > to)
1025                        ents[to] = ents[from];
1026                to++;
1027        }
1028        ASSERT(from > to);
1029        /*
1030         * If the insertion point was past the last entry,
1031         * set the new insertion point accordingly.
1032         */
1033        if (index == from)
1034                newindex = to;
1035        *indexp = newindex;
1036        /*
1037         * Adjust the leaf header values.
1038         */
1039        leafhdr->count -= from - to;
1040        leafhdr->stale = 1;
1041        /*
1042         * Remember the low/high stale value only in the "right"
1043         * direction.
1044         */
1045        if (lowstale >= newindex)
1046                lowstale = -1;
1047        else
1048                highstale = leafhdr->count;
1049        *highlogp = leafhdr->count - 1;
1050        *lowstalep = lowstale;
1051        *highstalep = highstale;
1052}
1053
1054/*
1055 * Log the bests entries indicated from a leaf1 block.
1056 */
1057static void
1058xfs_dir3_leaf_log_bests(
1059        struct xfs_da_args      *args,
1060        struct xfs_buf          *bp,            /* leaf buffer */
1061        int                     first,          /* first entry to log */
1062        int                     last)           /* last entry to log */
1063{
1064        __be16                  *firstb;        /* pointer to first entry */
1065        __be16                  *lastb;         /* pointer to last entry */
1066        struct xfs_dir2_leaf    *leaf = bp->b_addr;
1067        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1068
1069        ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC) ||
1070               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAF1_MAGIC));
1071
1072        ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
1073        firstb = xfs_dir2_leaf_bests_p(ltp) + first;
1074        lastb = xfs_dir2_leaf_bests_p(ltp) + last;
1075        xfs_trans_log_buf(args->trans, bp,
1076                (uint)((char *)firstb - (char *)leaf),
1077                (uint)((char *)lastb - (char *)leaf + sizeof(*lastb) - 1));
1078}
1079
1080/*
1081 * Log the leaf entries indicated from a leaf1 or leafn block.
1082 */
1083void
1084xfs_dir3_leaf_log_ents(
1085        struct xfs_da_args      *args,
1086        struct xfs_dir3_icleaf_hdr *hdr,
1087        struct xfs_buf          *bp,
1088        int                     first,
1089        int                     last)
1090{
1091        xfs_dir2_leaf_entry_t   *firstlep;      /* pointer to first entry */
1092        xfs_dir2_leaf_entry_t   *lastlep;       /* pointer to last entry */
1093        struct xfs_dir2_leaf    *leaf = bp->b_addr;
1094
1095        ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC) ||
1096               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAF1_MAGIC) ||
1097               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC) ||
1098               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC));
1099
1100        firstlep = &hdr->ents[first];
1101        lastlep = &hdr->ents[last];
1102        xfs_trans_log_buf(args->trans, bp,
1103                (uint)((char *)firstlep - (char *)leaf),
1104                (uint)((char *)lastlep - (char *)leaf + sizeof(*lastlep) - 1));
1105}
1106
1107/*
1108 * Log the header of the leaf1 or leafn block.
1109 */
1110void
1111xfs_dir3_leaf_log_header(
1112        struct xfs_da_args      *args,
1113        struct xfs_buf          *bp)
1114{
1115        struct xfs_dir2_leaf    *leaf = bp->b_addr;
1116
1117        ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC) ||
1118               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAF1_MAGIC) ||
1119               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC) ||
1120               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC));
1121
1122        xfs_trans_log_buf(args->trans, bp,
1123                          (uint)((char *)&leaf->hdr - (char *)leaf),
1124                          args->geo->leaf_hdr_size - 1);
1125}
1126
1127/*
1128 * Log the tail of the leaf1 block.
1129 */
1130STATIC void
1131xfs_dir3_leaf_log_tail(
1132        struct xfs_da_args      *args,
1133        struct xfs_buf          *bp)
1134{
1135        struct xfs_dir2_leaf    *leaf = bp->b_addr;
1136        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1137
1138        ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC) ||
1139               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAF1_MAGIC) ||
1140               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC) ||
1141               leaf->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC));
1142
1143        ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
1144        xfs_trans_log_buf(args->trans, bp, (uint)((char *)ltp - (char *)leaf),
1145                (uint)(args->geo->blksize - 1));
1146}
1147
1148/*
1149 * Look up the entry referred to by args in the leaf format directory.
1150 * Most of the work is done by the xfs_dir2_leaf_lookup_int routine which
1151 * is also used by the node-format code.
1152 */
1153int
1154xfs_dir2_leaf_lookup(
1155        xfs_da_args_t           *args)          /* operation arguments */
1156{
1157        struct xfs_buf          *dbp;           /* data block buffer */
1158        xfs_dir2_data_entry_t   *dep;           /* data block entry */
1159        xfs_inode_t             *dp;            /* incore directory inode */
1160        int                     error;          /* error return code */
1161        int                     index;          /* found entry index */
1162        struct xfs_buf          *lbp;           /* leaf buffer */
1163        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1164        xfs_trans_t             *tp;            /* transaction pointer */
1165        struct xfs_dir3_icleaf_hdr leafhdr;
1166
1167        trace_xfs_dir2_leaf_lookup(args);
1168
1169        /*
1170         * Look up name in the leaf block, returning both buffers and index.
1171         */
1172        error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp, &leafhdr);
1173        if (error)
1174                return error;
1175
1176        tp = args->trans;
1177        dp = args->dp;
1178        xfs_dir3_leaf_check(dp, lbp);
1179
1180        /*
1181         * Get to the leaf entry and contained data entry address.
1182         */
1183        lep = &leafhdr.ents[index];
1184
1185        /*
1186         * Point to the data entry.
1187         */
1188        dep = (xfs_dir2_data_entry_t *)
1189              ((char *)dbp->b_addr +
1190               xfs_dir2_dataptr_to_off(args->geo, be32_to_cpu(lep->address)));
1191        /*
1192         * Return the found inode number & CI name if appropriate
1193         */
1194        args->inumber = be64_to_cpu(dep->inumber);
1195        args->filetype = xfs_dir2_data_get_ftype(dp->i_mount, dep);
1196        error = xfs_dir_cilookup_result(args, dep->name, dep->namelen);
1197        xfs_trans_brelse(tp, dbp);
1198        xfs_trans_brelse(tp, lbp);
1199        return error;
1200}
1201
1202/*
1203 * Look up name/hash in the leaf block.
1204 * Fill in indexp with the found index, and dbpp with the data buffer.
1205 * If not found dbpp will be NULL, and ENOENT comes back.
1206 * lbpp will always be filled in with the leaf buffer unless there's an error.
1207 */
1208static int                                      /* error */
1209xfs_dir2_leaf_lookup_int(
1210        xfs_da_args_t           *args,          /* operation arguments */
1211        struct xfs_buf          **lbpp,         /* out: leaf buffer */
1212        int                     *indexp,        /* out: index in leaf block */
1213        struct xfs_buf          **dbpp,         /* out: data buffer */
1214        struct xfs_dir3_icleaf_hdr *leafhdr)
1215{
1216        xfs_dir2_db_t           curdb = -1;     /* current data block number */
1217        struct xfs_buf          *dbp = NULL;    /* data buffer */
1218        xfs_dir2_data_entry_t   *dep;           /* data entry */
1219        xfs_inode_t             *dp;            /* incore directory inode */
1220        int                     error;          /* error return code */
1221        int                     index;          /* index in leaf block */
1222        struct xfs_buf          *lbp;           /* leaf buffer */
1223        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1224        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1225        xfs_mount_t             *mp;            /* filesystem mount point */
1226        xfs_dir2_db_t           newdb;          /* new data block number */
1227        xfs_trans_t             *tp;            /* transaction pointer */
1228        xfs_dir2_db_t           cidb = -1;      /* case match data block no. */
1229        enum xfs_dacmp          cmp;            /* name compare result */
1230
1231        dp = args->dp;
1232        tp = args->trans;
1233        mp = dp->i_mount;
1234
1235        error = xfs_dir3_leaf_read(tp, dp, args->geo->leafblk, &lbp);
1236        if (error)
1237                return error;
1238
1239        *lbpp = lbp;
1240        leaf = lbp->b_addr;
1241        xfs_dir3_leaf_check(dp, lbp);
1242        xfs_dir2_leaf_hdr_from_disk(mp, leafhdr, leaf);
1243
1244        /*
1245         * Look for the first leaf entry with our hash value.
1246         */
1247        index = xfs_dir2_leaf_search_hash(args, lbp);
1248        /*
1249         * Loop over all the entries with the right hash value
1250         * looking to match the name.
1251         */
1252        for (lep = &leafhdr->ents[index];
1253             index < leafhdr->count &&
1254                        be32_to_cpu(lep->hashval) == args->hashval;
1255             lep++, index++) {
1256                /*
1257                 * Skip over stale leaf entries.
1258                 */
1259                if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
1260                        continue;
1261                /*
1262                 * Get the new data block number.
1263                 */
1264                newdb = xfs_dir2_dataptr_to_db(args->geo,
1265                                               be32_to_cpu(lep->address));
1266                /*
1267                 * If it's not the same as the old data block number,
1268                 * need to pitch the old one and read the new one.
1269                 */
1270                if (newdb != curdb) {
1271                        if (dbp)
1272                                xfs_trans_brelse(tp, dbp);
1273                        error = xfs_dir3_data_read(tp, dp,
1274                                           xfs_dir2_db_to_da(args->geo, newdb),
1275                                           0, &dbp);
1276                        if (error) {
1277                                xfs_trans_brelse(tp, lbp);
1278                                return error;
1279                        }
1280                        curdb = newdb;
1281                }
1282                /*
1283                 * Point to the data entry.
1284                 */
1285                dep = (xfs_dir2_data_entry_t *)((char *)dbp->b_addr +
1286                        xfs_dir2_dataptr_to_off(args->geo,
1287                                                be32_to_cpu(lep->address)));
1288                /*
1289                 * Compare name and if it's an exact match, return the index
1290                 * and buffer. If it's the first case-insensitive match, store
1291                 * the index and buffer and continue looking for an exact match.
1292                 */
1293                cmp = xfs_dir2_compname(args, dep->name, dep->namelen);
1294                if (cmp != XFS_CMP_DIFFERENT && cmp != args->cmpresult) {
1295                        args->cmpresult = cmp;
1296                        *indexp = index;
1297                        /* case exact match: return the current buffer. */
1298                        if (cmp == XFS_CMP_EXACT) {
1299                                *dbpp = dbp;
1300                                return 0;
1301                        }
1302                        cidb = curdb;
1303                }
1304        }
1305        ASSERT(args->op_flags & XFS_DA_OP_OKNOENT);
1306        /*
1307         * Here, we can only be doing a lookup (not a rename or remove).
1308         * If a case-insensitive match was found earlier, re-read the
1309         * appropriate data block if required and return it.
1310         */
1311        if (args->cmpresult == XFS_CMP_CASE) {
1312                ASSERT(cidb != -1);
1313                if (cidb != curdb) {
1314                        xfs_trans_brelse(tp, dbp);
1315                        error = xfs_dir3_data_read(tp, dp,
1316                                           xfs_dir2_db_to_da(args->geo, cidb),
1317                                           0, &dbp);
1318                        if (error) {
1319                                xfs_trans_brelse(tp, lbp);
1320                                return error;
1321                        }
1322                }
1323                *dbpp = dbp;
1324                return 0;
1325        }
1326        /*
1327         * No match found, return -ENOENT.
1328         */
1329        ASSERT(cidb == -1);
1330        if (dbp)
1331                xfs_trans_brelse(tp, dbp);
1332        xfs_trans_brelse(tp, lbp);
1333        return -ENOENT;
1334}
1335
1336/*
1337 * Remove an entry from a leaf format directory.
1338 */
1339int                                             /* error */
1340xfs_dir2_leaf_removename(
1341        xfs_da_args_t           *args)          /* operation arguments */
1342{
1343        struct xfs_da_geometry  *geo = args->geo;
1344        __be16                  *bestsp;        /* leaf block best freespace */
1345        xfs_dir2_data_hdr_t     *hdr;           /* data block header */
1346        xfs_dir2_db_t           db;             /* data block number */
1347        struct xfs_buf          *dbp;           /* data block buffer */
1348        xfs_dir2_data_entry_t   *dep;           /* data entry structure */
1349        xfs_inode_t             *dp;            /* incore directory inode */
1350        int                     error;          /* error return code */
1351        xfs_dir2_db_t           i;              /* temporary data block # */
1352        int                     index;          /* index into leaf entries */
1353        struct xfs_buf          *lbp;           /* leaf buffer */
1354        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1355        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1356        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1357        int                     needlog;        /* need to log data header */
1358        int                     needscan;       /* need to rescan data frees */
1359        xfs_dir2_data_off_t     oldbest;        /* old value of best free */
1360        struct xfs_dir2_data_free *bf;          /* bestfree table */
1361        struct xfs_dir3_icleaf_hdr leafhdr;
1362
1363        trace_xfs_dir2_leaf_removename(args);
1364
1365        /*
1366         * Lookup the leaf entry, get the leaf and data blocks read in.
1367         */
1368        error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp, &leafhdr);
1369        if (error)
1370                return error;
1371
1372        dp = args->dp;
1373        leaf = lbp->b_addr;
1374        hdr = dbp->b_addr;
1375        xfs_dir3_data_check(dp, dbp);
1376        bf = xfs_dir2_data_bestfree_p(dp->i_mount, hdr);
1377
1378        /*
1379         * Point to the leaf entry, use that to point to the data entry.
1380         */
1381        lep = &leafhdr.ents[index];
1382        db = xfs_dir2_dataptr_to_db(geo, be32_to_cpu(lep->address));
1383        dep = (xfs_dir2_data_entry_t *)((char *)hdr +
1384                xfs_dir2_dataptr_to_off(geo, be32_to_cpu(lep->address)));
1385        needscan = needlog = 0;
1386        oldbest = be16_to_cpu(bf[0].length);
1387        ltp = xfs_dir2_leaf_tail_p(geo, leaf);
1388        bestsp = xfs_dir2_leaf_bests_p(ltp);
1389        if (be16_to_cpu(bestsp[db]) != oldbest) {
1390                xfs_buf_mark_corrupt(lbp);
1391                return -EFSCORRUPTED;
1392        }
1393        /*
1394         * Mark the former data entry unused.
1395         */
1396        xfs_dir2_data_make_free(args, dbp,
1397                (xfs_dir2_data_aoff_t)((char *)dep - (char *)hdr),
1398                xfs_dir2_data_entsize(dp->i_mount, dep->namelen), &needlog,
1399                &needscan);
1400        /*
1401         * We just mark the leaf entry stale by putting a null in it.
1402         */
1403        leafhdr.stale++;
1404        xfs_dir2_leaf_hdr_to_disk(dp->i_mount, leaf, &leafhdr);
1405        xfs_dir3_leaf_log_header(args, lbp);
1406
1407        lep->address = cpu_to_be32(XFS_DIR2_NULL_DATAPTR);
1408        xfs_dir3_leaf_log_ents(args, &leafhdr, lbp, index, index);
1409
1410        /*
1411         * Scan the freespace in the data block again if necessary,
1412         * log the data block header if necessary.
1413         */
1414        if (needscan)
1415                xfs_dir2_data_freescan(dp->i_mount, hdr, &needlog);
1416        if (needlog)
1417                xfs_dir2_data_log_header(args, dbp);
1418        /*
1419         * If the longest freespace in the data block has changed,
1420         * put the new value in the bests table and log that.
1421         */
1422        if (be16_to_cpu(bf[0].length) != oldbest) {
1423                bestsp[db] = bf[0].length;
1424                xfs_dir3_leaf_log_bests(args, lbp, db, db);
1425        }
1426        xfs_dir3_data_check(dp, dbp);
1427        /*
1428         * If the data block is now empty then get rid of the data block.
1429         */
1430        if (be16_to_cpu(bf[0].length) ==
1431            geo->blksize - geo->data_entry_offset) {
1432                ASSERT(db != geo->datablk);
1433                if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1434                        /*
1435                         * Nope, can't get rid of it because it caused
1436                         * allocation of a bmap btree block to do so.
1437                         * Just go on, returning success, leaving the
1438                         * empty block in place.
1439                         */
1440                        if (error == -ENOSPC && args->total == 0)
1441                                error = 0;
1442                        xfs_dir3_leaf_check(dp, lbp);
1443                        return error;
1444                }
1445                dbp = NULL;
1446                /*
1447                 * If this is the last data block then compact the
1448                 * bests table by getting rid of entries.
1449                 */
1450                if (db == be32_to_cpu(ltp->bestcount) - 1) {
1451                        /*
1452                         * Look for the last active entry (i).
1453                         */
1454                        for (i = db - 1; i > 0; i--) {
1455                                if (bestsp[i] != cpu_to_be16(NULLDATAOFF))
1456                                        break;
1457                        }
1458                        /*
1459                         * Copy the table down so inactive entries at the
1460                         * end are removed.
1461                         */
1462                        memmove(&bestsp[db - i], bestsp,
1463                                (be32_to_cpu(ltp->bestcount) - (db - i)) * sizeof(*bestsp));
1464                        be32_add_cpu(&ltp->bestcount, -(db - i));
1465                        xfs_dir3_leaf_log_tail(args, lbp);
1466                        xfs_dir3_leaf_log_bests(args, lbp, 0,
1467                                                be32_to_cpu(ltp->bestcount) - 1);
1468                } else
1469                        bestsp[db] = cpu_to_be16(NULLDATAOFF);
1470        }
1471        /*
1472         * If the data block was not the first one, drop it.
1473         */
1474        else if (db != geo->datablk)
1475                dbp = NULL;
1476
1477        xfs_dir3_leaf_check(dp, lbp);
1478        /*
1479         * See if we can convert to block form.
1480         */
1481        return xfs_dir2_leaf_to_block(args, lbp, dbp);
1482}
1483
1484/*
1485 * Replace the inode number in a leaf format directory entry.
1486 */
1487int                                             /* error */
1488xfs_dir2_leaf_replace(
1489        xfs_da_args_t           *args)          /* operation arguments */
1490{
1491        struct xfs_buf          *dbp;           /* data block buffer */
1492        xfs_dir2_data_entry_t   *dep;           /* data block entry */
1493        xfs_inode_t             *dp;            /* incore directory inode */
1494        int                     error;          /* error return code */
1495        int                     index;          /* index of leaf entry */
1496        struct xfs_buf          *lbp;           /* leaf buffer */
1497        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1498        xfs_trans_t             *tp;            /* transaction pointer */
1499        struct xfs_dir3_icleaf_hdr leafhdr;
1500
1501        trace_xfs_dir2_leaf_replace(args);
1502
1503        /*
1504         * Look up the entry.
1505         */
1506        error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp, &leafhdr);
1507        if (error)
1508                return error;
1509
1510        dp = args->dp;
1511        /*
1512         * Point to the leaf entry, get data address from it.
1513         */
1514        lep = &leafhdr.ents[index];
1515        /*
1516         * Point to the data entry.
1517         */
1518        dep = (xfs_dir2_data_entry_t *)
1519              ((char *)dbp->b_addr +
1520               xfs_dir2_dataptr_to_off(args->geo, be32_to_cpu(lep->address)));
1521        ASSERT(args->inumber != be64_to_cpu(dep->inumber));
1522        /*
1523         * Put the new inode number in, log it.
1524         */
1525        dep->inumber = cpu_to_be64(args->inumber);
1526        xfs_dir2_data_put_ftype(dp->i_mount, dep, args->filetype);
1527        tp = args->trans;
1528        xfs_dir2_data_log_entry(args, dbp, dep);
1529        xfs_dir3_leaf_check(dp, lbp);
1530        xfs_trans_brelse(tp, lbp);
1531        return 0;
1532}
1533
1534/*
1535 * Return index in the leaf block (lbp) which is either the first
1536 * one with this hash value, or if there are none, the insert point
1537 * for that hash value.
1538 */
1539int                                             /* index value */
1540xfs_dir2_leaf_search_hash(
1541        xfs_da_args_t           *args,          /* operation arguments */
1542        struct xfs_buf          *lbp)           /* leaf buffer */
1543{
1544        xfs_dahash_t            hash=0;         /* hash from this entry */
1545        xfs_dahash_t            hashwant;       /* hash value looking for */
1546        int                     high;           /* high leaf index */
1547        int                     low;            /* low leaf index */
1548        xfs_dir2_leaf_entry_t   *lep;           /* leaf entry */
1549        int                     mid=0;          /* current leaf index */
1550        struct xfs_dir3_icleaf_hdr leafhdr;
1551
1552        xfs_dir2_leaf_hdr_from_disk(args->dp->i_mount, &leafhdr, lbp->b_addr);
1553
1554        /*
1555         * Note, the table cannot be empty, so we have to go through the loop.
1556         * Binary search the leaf entries looking for our hash value.
1557         */
1558        for (lep = leafhdr.ents, low = 0, high = leafhdr.count - 1,
1559                hashwant = args->hashval;
1560             low <= high; ) {
1561                mid = (low + high) >> 1;
1562                if ((hash = be32_to_cpu(lep[mid].hashval)) == hashwant)
1563                        break;
1564                if (hash < hashwant)
1565                        low = mid + 1;
1566                else
1567                        high = mid - 1;
1568        }
1569        /*
1570         * Found one, back up through all the equal hash values.
1571         */
1572        if (hash == hashwant) {
1573                while (mid > 0 && be32_to_cpu(lep[mid - 1].hashval) == hashwant) {
1574                        mid--;
1575                }
1576        }
1577        /*
1578         * Need to point to an entry higher than ours.
1579         */
1580        else if (hash < hashwant)
1581                mid++;
1582        return mid;
1583}
1584
1585/*
1586 * Trim off a trailing data block.  We know it's empty since the leaf
1587 * freespace table says so.
1588 */
1589int                                             /* error */
1590xfs_dir2_leaf_trim_data(
1591        xfs_da_args_t           *args,          /* operation arguments */
1592        struct xfs_buf          *lbp,           /* leaf buffer */
1593        xfs_dir2_db_t           db)             /* data block number */
1594{
1595        struct xfs_da_geometry  *geo = args->geo;
1596        __be16                  *bestsp;        /* leaf bests table */
1597        struct xfs_buf          *dbp;           /* data block buffer */
1598        xfs_inode_t             *dp;            /* incore directory inode */
1599        int                     error;          /* error return value */
1600        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1601        xfs_dir2_leaf_tail_t    *ltp;           /* leaf tail structure */
1602        xfs_trans_t             *tp;            /* transaction pointer */
1603
1604        dp = args->dp;
1605        tp = args->trans;
1606        /*
1607         * Read the offending data block.  We need its buffer.
1608         */
1609        error = xfs_dir3_data_read(tp, dp, xfs_dir2_db_to_da(geo, db), 0, &dbp);
1610        if (error)
1611                return error;
1612
1613        leaf = lbp->b_addr;
1614        ltp = xfs_dir2_leaf_tail_p(geo, leaf);
1615
1616#ifdef DEBUG
1617{
1618        struct xfs_dir2_data_hdr *hdr = dbp->b_addr;
1619        struct xfs_dir2_data_free *bf =
1620                xfs_dir2_data_bestfree_p(dp->i_mount, hdr);
1621
1622        ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
1623               hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC));
1624        ASSERT(be16_to_cpu(bf[0].length) ==
1625               geo->blksize - geo->data_entry_offset);
1626        ASSERT(db == be32_to_cpu(ltp->bestcount) - 1);
1627}
1628#endif
1629
1630        /*
1631         * Get rid of the data block.
1632         */
1633        if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1634                ASSERT(error != -ENOSPC);
1635                xfs_trans_brelse(tp, dbp);
1636                return error;
1637        }
1638        /*
1639         * Eliminate the last bests entry from the table.
1640         */
1641        bestsp = xfs_dir2_leaf_bests_p(ltp);
1642        be32_add_cpu(&ltp->bestcount, -1);
1643        memmove(&bestsp[1], &bestsp[0], be32_to_cpu(ltp->bestcount) * sizeof(*bestsp));
1644        xfs_dir3_leaf_log_tail(args, lbp);
1645        xfs_dir3_leaf_log_bests(args, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
1646        return 0;
1647}
1648
1649static inline size_t
1650xfs_dir3_leaf_size(
1651        struct xfs_dir3_icleaf_hdr      *hdr,
1652        int                             counts)
1653{
1654        int     entries;
1655        int     hdrsize;
1656
1657        entries = hdr->count - hdr->stale;
1658        if (hdr->magic == XFS_DIR2_LEAF1_MAGIC ||
1659            hdr->magic == XFS_DIR2_LEAFN_MAGIC)
1660                hdrsize = sizeof(struct xfs_dir2_leaf_hdr);
1661        else
1662                hdrsize = sizeof(struct xfs_dir3_leaf_hdr);
1663
1664        return hdrsize + entries * sizeof(xfs_dir2_leaf_entry_t)
1665                       + counts * sizeof(xfs_dir2_data_off_t)
1666                       + sizeof(xfs_dir2_leaf_tail_t);
1667}
1668
1669/*
1670 * Convert node form directory to leaf form directory.
1671 * The root of the node form dir needs to already be a LEAFN block.
1672 * Just return if we can't do anything.
1673 */
1674int                                             /* error */
1675xfs_dir2_node_to_leaf(
1676        xfs_da_state_t          *state)         /* directory operation state */
1677{
1678        xfs_da_args_t           *args;          /* operation arguments */
1679        xfs_inode_t             *dp;            /* incore directory inode */
1680        int                     error;          /* error return code */
1681        struct xfs_buf          *fbp;           /* buffer for freespace block */
1682        xfs_fileoff_t           fo;             /* freespace file offset */
1683        struct xfs_buf          *lbp;           /* buffer for leaf block */
1684        xfs_dir2_leaf_tail_t    *ltp;           /* tail of leaf structure */
1685        xfs_dir2_leaf_t         *leaf;          /* leaf structure */
1686        xfs_mount_t             *mp;            /* filesystem mount point */
1687        int                     rval;           /* successful free trim? */
1688        xfs_trans_t             *tp;            /* transaction pointer */
1689        struct xfs_dir3_icleaf_hdr leafhdr;
1690        struct xfs_dir3_icfree_hdr freehdr;
1691
1692        /*
1693         * There's more than a leaf level in the btree, so there must
1694         * be multiple leafn blocks.  Give up.
1695         */
1696        if (state->path.active > 1)
1697                return 0;
1698        args = state->args;
1699
1700        trace_xfs_dir2_node_to_leaf(args);
1701
1702        mp = state->mp;
1703        dp = args->dp;
1704        tp = args->trans;
1705        /*
1706         * Get the last offset in the file.
1707         */
1708        if ((error = xfs_bmap_last_offset(dp, &fo, XFS_DATA_FORK))) {
1709                return error;
1710        }
1711        fo -= args->geo->fsbcount;
1712        /*
1713         * If there are freespace blocks other than the first one,
1714         * take this opportunity to remove trailing empty freespace blocks
1715         * that may have been left behind during no-space-reservation
1716         * operations.
1717         */
1718        while (fo > args->geo->freeblk) {
1719                if ((error = xfs_dir2_node_trim_free(args, fo, &rval))) {
1720                        return error;
1721                }
1722                if (rval)
1723                        fo -= args->geo->fsbcount;
1724                else
1725                        return 0;
1726        }
1727        /*
1728         * Now find the block just before the freespace block.
1729         */
1730        if ((error = xfs_bmap_last_before(tp, dp, &fo, XFS_DATA_FORK))) {
1731                return error;
1732        }
1733        /*
1734         * If it's not the single leaf block, give up.
1735         */
1736        if (XFS_FSB_TO_B(mp, fo) > XFS_DIR2_LEAF_OFFSET + args->geo->blksize)
1737                return 0;
1738        lbp = state->path.blk[0].bp;
1739        leaf = lbp->b_addr;
1740        xfs_dir2_leaf_hdr_from_disk(mp, &leafhdr, leaf);
1741
1742        ASSERT(leafhdr.magic == XFS_DIR2_LEAFN_MAGIC ||
1743               leafhdr.magic == XFS_DIR3_LEAFN_MAGIC);
1744
1745        /*
1746         * Read the freespace block.
1747         */
1748        error = xfs_dir2_free_read(tp, dp,  args->geo->freeblk, &fbp);
1749        if (error)
1750                return error;
1751        xfs_dir2_free_hdr_from_disk(mp, &freehdr, fbp->b_addr);
1752
1753        ASSERT(!freehdr.firstdb);
1754
1755        /*
1756         * Now see if the leafn and free data will fit in a leaf1.
1757         * If not, release the buffer and give up.
1758         */
1759        if (xfs_dir3_leaf_size(&leafhdr, freehdr.nvalid) > args->geo->blksize) {
1760                xfs_trans_brelse(tp, fbp);
1761                return 0;
1762        }
1763
1764        /*
1765         * If the leaf has any stale entries in it, compress them out.
1766         */
1767        if (leafhdr.stale)
1768                xfs_dir3_leaf_compact(args, &leafhdr, lbp);
1769
1770        lbp->b_ops = &xfs_dir3_leaf1_buf_ops;
1771        xfs_trans_buf_set_type(tp, lbp, XFS_BLFT_DIR_LEAF1_BUF);
1772        leafhdr.magic = (leafhdr.magic == XFS_DIR2_LEAFN_MAGIC)
1773                                        ? XFS_DIR2_LEAF1_MAGIC
1774                                        : XFS_DIR3_LEAF1_MAGIC;
1775
1776        /*
1777         * Set up the leaf tail from the freespace block.
1778         */
1779        ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
1780        ltp->bestcount = cpu_to_be32(freehdr.nvalid);
1781
1782        /*
1783         * Set up the leaf bests table.
1784         */
1785        memcpy(xfs_dir2_leaf_bests_p(ltp), freehdr.bests,
1786                freehdr.nvalid * sizeof(xfs_dir2_data_off_t));
1787
1788        xfs_dir2_leaf_hdr_to_disk(mp, leaf, &leafhdr);
1789        xfs_dir3_leaf_log_header(args, lbp);
1790        xfs_dir3_leaf_log_bests(args, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
1791        xfs_dir3_leaf_log_tail(args, lbp);
1792        xfs_dir3_leaf_check(dp, lbp);
1793
1794        /*
1795         * Get rid of the freespace block.
1796         */
1797        error = xfs_dir2_shrink_inode(args,
1798                        xfs_dir2_byte_to_db(args->geo, XFS_DIR2_FREE_OFFSET),
1799                        fbp);
1800        if (error) {
1801                /*
1802                 * This can't fail here because it can only happen when
1803                 * punching out the middle of an extent, and this is an
1804                 * isolated block.
1805                 */
1806                ASSERT(error != -ENOSPC);
1807                return error;
1808        }
1809        fbp = NULL;
1810        /*
1811         * Now see if we can convert the single-leaf directory
1812         * down to a block form directory.
1813         * This routine always kills the dabuf for the leaf, so
1814         * eliminate it from the path.
1815         */
1816        error = xfs_dir2_leaf_to_block(args, lbp, NULL);
1817        state->path.blk[0].bp = NULL;
1818        return error;
1819}
1820