sys/gnu/fs/xfs/xfs_da_btree.c

   1 /*
   2  * Copyright (c) 2000-2005 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_inum.h"
  24 #include "xfs_trans.h"
  25 #include "xfs_sb.h"
  26 #include "xfs_ag.h"
  27 #include "xfs_dir.h"
  28 #include "xfs_dir2.h"
  29 #include "xfs_dmapi.h"
  30 #include "xfs_mount.h"
  31 #include "xfs_da_btree.h"
  32 #include "xfs_bmap_btree.h"
  33 #include "xfs_alloc_btree.h"
  34 #include "xfs_ialloc_btree.h"
  35 #include "xfs_dir_sf.h"
  36 #include "xfs_dir2_sf.h"
  37 #include "xfs_attr_sf.h"
  38 #include "xfs_dinode.h"
  39 #include "xfs_inode.h"
  40 #include "xfs_inode_item.h"
  41 #include "xfs_alloc.h"
  42 #include "xfs_btree.h"
  43 #include "xfs_bmap.h"
  44 #include "xfs_attr.h"
  45 #include "xfs_attr_leaf.h"
  46 #include "xfs_dir_leaf.h"
  47 #include "xfs_dir2_data.h"
  48 #include "xfs_dir2_leaf.h"
  49 #include "xfs_dir2_block.h"
  50 #include "xfs_dir2_node.h"
  51 #include "xfs_error.h"
  52
  53 /*
  54  * xfs_da_btree.c
  55  *
  56  * Routines to implement directories as Btrees of hashed names.
  57  */
  58
  59 /*========================================================================
  60  * Function prototypes for the kernel.
  61  *========================================================================*/
  62
  63 /*
  64  * Routines used for growing the Btree.
  65  */
  66 STATIC int xfs_da_root_split(xfs_da_state_t *state,
  67                                             xfs_da_state_blk_t *existing_root,
  68                                             xfs_da_state_blk_t *new_child);
  69 STATIC int xfs_da_node_split(xfs_da_state_t *state,
  70                                             xfs_da_state_blk_t *existing_blk,
  71                                             xfs_da_state_blk_t *split_blk,
  72                                             xfs_da_state_blk_t *blk_to_add,
  73                                             int treelevel,
  74                                             int *result);
  75 STATIC void xfs_da_node_rebalance(xfs_da_state_t *state,
  76                                          xfs_da_state_blk_t *node_blk_1,
  77                                          xfs_da_state_blk_t *node_blk_2);
  78 STATIC void xfs_da_node_add(xfs_da_state_t *state,
  79                                    xfs_da_state_blk_t *old_node_blk,
  80                                    xfs_da_state_blk_t *new_node_blk);
  81
  82 /*
  83  * Routines used for shrinking the Btree.
  84  */
  85 STATIC int xfs_da_root_join(xfs_da_state_t *state,
  86                                            xfs_da_state_blk_t *root_blk);
  87 STATIC int xfs_da_node_toosmall(xfs_da_state_t *state, int *retval);
  88 STATIC void xfs_da_node_remove(xfs_da_state_t *state,
  89                                               xfs_da_state_blk_t *drop_blk);
  90 STATIC void xfs_da_node_unbalance(xfs_da_state_t *state,
  91                                          xfs_da_state_blk_t *src_node_blk,
  92                                          xfs_da_state_blk_t *dst_node_blk);
  93
  94 /*
  95  * Utility routines.
  96  */
  97 STATIC uint     xfs_da_node_lasthash(xfs_dabuf_t *bp, int *count);
  98 STATIC int      xfs_da_node_order(xfs_dabuf_t *node1_bp, xfs_dabuf_t *node2_bp);
  99 STATIC xfs_dabuf_t *xfs_da_buf_make(int nbuf, xfs_buf_t **bps, inst_t *ra);
 100 STATIC int      xfs_da_blk_unlink(xfs_da_state_t *state,
 101                                   xfs_da_state_blk_t *drop_blk,
 102                                   xfs_da_state_blk_t *save_blk);
 103 STATIC void     xfs_da_state_kill_altpath(xfs_da_state_t *state);
 104
 105 /*========================================================================
 106  * Routines used for growing the Btree.
 107  *========================================================================*/
 108
 109 /*
 110  * Create the initial contents of an intermediate node.
 111  */
 112 int
 113 xfs_da_node_create(xfs_da_args_t *args, xfs_dablk_t blkno, int level,
 114                                  xfs_dabuf_t **bpp, int whichfork)
 115 {
 116         xfs_da_intnode_t *node;
 117         xfs_dabuf_t *bp;
 118         int error;
 119         xfs_trans_t *tp;
 120
 121         tp = args->trans;
 122         error = xfs_da_get_buf(tp, args->dp, blkno, -1, &bp, whichfork);
 123         if (error)
 124                 return(error);
 125         ASSERT(bp != NULL);
 126         node = bp->data;
 127         node->hdr.info.forw = 0;
 128         node->hdr.info.back = 0;
 129         node->hdr.info.magic = cpu_to_be16(XFS_DA_NODE_MAGIC);
 130         node->hdr.info.pad = 0;
 131         node->hdr.count = 0;
 132         node->hdr.level = cpu_to_be16(level);
 133
 134         xfs_da_log_buf(tp, bp,
 135                 XFS_DA_LOGRANGE(node, &node->hdr, sizeof(node->hdr)));
 136
 137         *bpp = bp;
 138         return(0);
 139 }
 140
 141 /*
 142  * Split a leaf node, rebalance, then possibly split
 143  * intermediate nodes, rebalance, etc.
 144  */
 145 int                                                     /* error */
 146 xfs_da_split(xfs_da_state_t *state)
 147 {
 148         xfs_da_state_blk_t *oldblk, *newblk, *addblk;
 149         xfs_da_intnode_t *node;
 150         xfs_dabuf_t *bp;
 151         int max, action, error, i;
 152
 153         /*
 154          * Walk back up the tree splitting/inserting/adjusting as necessary.
 155          * If we need to insert and there isn't room, split the node, then
 156          * decide which fragment to insert the new block from below into.
 157          * Note that we may split the root this way, but we need more fixup.
 158          */
 159         max = state->path.active - 1;
 160         ASSERT((max >= 0) && (max < XFS_DA_NODE_MAXDEPTH));
 161         ASSERT(state->path.blk[max].magic == XFS_ATTR_LEAF_MAGIC ||
 162                state->path.blk[max].magic == XFS_DIRX_LEAF_MAGIC(state->mp));
 163
 164         addblk = &state->path.blk[max];         /* initial dummy value */
 165         for (i = max; (i >= 0) && addblk; state->path.active--, i--) {
 166                 oldblk = &state->path.blk[i];
 167                 newblk = &state->altpath.blk[i];
 168
 169                 /*
 170                  * If a leaf node then
 171                  *     Allocate a new leaf node, then rebalance across them.
 172                  * else if an intermediate node then
 173                  *     We split on the last layer, must we split the node?
 174                  */
 175                 switch (oldblk->magic) {
 176                 case XFS_ATTR_LEAF_MAGIC:
 177                         error = xfs_attr_leaf_split(state, oldblk, newblk);
 178                         if ((error != 0) && (error != ENOSPC)) {
 179                                 return(error);  /* GROT: attr is inconsistent */
 180                         }
 181                         if (!error) {
 182                                 addblk = newblk;
 183                                 break;
 184                         }
 185                         /*
 186                          * Entry wouldn't fit, split the leaf again.
 187                          */
 188                         state->extravalid = 1;
 189                         if (state->inleaf) {
 190                                 state->extraafter = 0;  /* before newblk */
 191                                 error = xfs_attr_leaf_split(state, oldblk,
 192                                                             &state->extrablk);
 193                         } else {
 194                                 state->extraafter = 1;  /* after newblk */
 195                                 error = xfs_attr_leaf_split(state, newblk,
 196                                                             &state->extrablk);
 197                         }
 198                         if (error)
 199                                 return(error);  /* GROT: attr inconsistent */
 200                         addblk = newblk;
 201                         break;
 202                 case XFS_DIR_LEAF_MAGIC:
 203                         ASSERT(XFS_DIR_IS_V1(state->mp));
 204                         error = xfs_dir_leaf_split(state, oldblk, newblk);
 205                         if ((error != 0) && (error != ENOSPC)) {
 206                                 return(error);  /* GROT: dir is inconsistent */
 207                         }
 208                         if (!error) {
 209                                 addblk = newblk;
 210                                 break;
 211                         }
 212                         /*
 213                          * Entry wouldn't fit, split the leaf again.
 214                          */
 215                         state->extravalid = 1;
 216                         if (state->inleaf) {
 217                                 state->extraafter = 0;  /* before newblk */
 218                                 error = xfs_dir_leaf_split(state, oldblk,
 219                                                            &state->extrablk);
 220                                 if (error)
 221                                         return(error);  /* GROT: dir incon. */
 222                                 addblk = newblk;
 223                         } else {
 224                                 state->extraafter = 1;  /* after newblk */
 225                                 error = xfs_dir_leaf_split(state, newblk,
 226                                                            &state->extrablk);
 227                                 if (error)
 228                                         return(error);  /* GROT: dir incon. */
 229                                 addblk = newblk;
 230                         }
 231                         break;
 232                 case XFS_DIR2_LEAFN_MAGIC:
 233                         ASSERT(XFS_DIR_IS_V2(state->mp));
 234                         error = xfs_dir2_leafn_split(state, oldblk, newblk);
 235                         if (error)
 236                                 return error;
 237                         addblk = newblk;
 238                         break;
 239                 case XFS_DA_NODE_MAGIC:
 240                         error = xfs_da_node_split(state, oldblk, newblk, addblk,
 241                                                          max - i, &action);
 242                         xfs_da_buf_done(addblk->bp);
 243                         addblk->bp = NULL;
 244                         if (error)
 245                                 return(error);  /* GROT: dir is inconsistent */
 246                         /*
 247                          * Record the newly split block for the next time thru?
 248                          */
 249                         if (action)
 250                                 addblk = newblk;
 251                         else
 252                                 addblk = NULL;
 253                         break;
 254                 }
 255
 256                 /*
 257                  * Update the btree to show the new hashval for this child.
 258                  */
 259                 xfs_da_fixhashpath(state, &state->path);
 260                 /*
 261                  * If we won't need this block again, it's getting dropped
 262                  * from the active path by the loop control, so we need
 263                  * to mark it done now.
 264                  */
 265                 if (i > 0 || !addblk)
 266                         xfs_da_buf_done(oldblk->bp);
 267         }
 268         if (!addblk)
 269                 return(0);
 270
 271         /*
 272          * Split the root node.
 273          */
 274         ASSERT(state->path.active == 0);
 275         oldblk = &state->path.blk[0];
 276         error = xfs_da_root_split(state, oldblk, addblk);
 277         if (error) {
 278                 xfs_da_buf_done(oldblk->bp);
 279                 xfs_da_buf_done(addblk->bp);
 280                 addblk->bp = NULL;
 281                 return(error);  /* GROT: dir is inconsistent */
 282         }
 283
 284         /*
 285          * Update pointers to the node which used to be block 0 and
 286          * just got bumped because of the addition of a new root node.
 287          * There might be three blocks involved if a double split occurred,
 288          * and the original block 0 could be at any position in the list.
 289          */
 290
 291         node = oldblk->bp->data;
 292         if (node->hdr.info.forw) {
 293                 if (be32_to_cpu(node->hdr.info.forw) == addblk->blkno) {
 294                         bp = addblk->bp;
 295                 } else {
 296                         ASSERT(state->extravalid);
 297                         bp = state->extrablk.bp;
 298                 }
 299                 node = bp->data;
 300                 node->hdr.info.back = cpu_to_be32(oldblk->blkno);
 301                 xfs_da_log_buf(state->args->trans, bp,
 302                     XFS_DA_LOGRANGE(node, &node->hdr.info,
 303                     sizeof(node->hdr.info)));
 304         }
 305         node = oldblk->bp->data;
 306         if (node->hdr.info.back) {
 307                 if (be32_to_cpu(node->hdr.info.back) == addblk->blkno) {
 308                         bp = addblk->bp;
 309                 } else {
 310                         ASSERT(state->extravalid);
 311                         bp = state->extrablk.bp;
 312                 }
 313                 node = bp->data;
 314                 node->hdr.info.forw = cpu_to_be32(oldblk->blkno);
 315                 xfs_da_log_buf(state->args->trans, bp,
 316                     XFS_DA_LOGRANGE(node, &node->hdr.info,
 317                     sizeof(node->hdr.info)));
 318         }
 319         xfs_da_buf_done(oldblk->bp);
 320         xfs_da_buf_done(addblk->bp);
 321         addblk->bp = NULL;
 322         return(0);
 323 }
 324
 325 /*
 326  * Split the root.  We have to create a new root and point to the two
 327  * parts (the split old root) that we just created.  Copy block zero to
 328  * the EOF, extending the inode in process.
 329  */
 330 STATIC int                                              /* error */
 331 xfs_da_root_split(xfs_da_state_t *state, xfs_da_state_blk_t *blk1,
 332                                  xfs_da_state_blk_t *blk2)
 333 {
 334         xfs_da_intnode_t *node, *oldroot;
 335         xfs_da_args_t *args;
 336         xfs_dablk_t blkno;
 337         xfs_dabuf_t *bp;
 338         int error, size;
 339         xfs_inode_t *dp;
 340         xfs_trans_t *tp;
 341         xfs_mount_t *mp;
 342         xfs_dir2_leaf_t *leaf;
 343
 344         /*
 345          * Copy the existing (incorrect) block from the root node position
 346          * to a free space somewhere.
 347          */
 348         args = state->args;
 349         ASSERT(args != NULL);
 350         error = xfs_da_grow_inode(args, &blkno);
 351         if (error)
 352                 return(error);
 353         dp = args->dp;
 354         tp = args->trans;
 355         mp = state->mp;
 356         error = xfs_da_get_buf(tp, dp, blkno, -1, &bp, args->whichfork);
 357         if (error)
 358                 return(error);
 359         ASSERT(bp != NULL);
 360         node = bp->data;
 361         oldroot = blk1->bp->data;
 362         if (be16_to_cpu(oldroot->hdr.info.magic) == XFS_DA_NODE_MAGIC) {
 363                 size = (int)((char *)&oldroot->btree[be16_to_cpu(oldroot->hdr.count)] -
 364                              (char *)oldroot);
 365         } else {
 366                 ASSERT(XFS_DIR_IS_V2(mp));
 367                 ASSERT(be16_to_cpu(oldroot->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
 368                 leaf = (xfs_dir2_leaf_t *)oldroot;
 369                 size = (int)((char *)&leaf->ents[be16_to_cpu(leaf->hdr.count)] -
 370                              (char *)leaf);
 371         }
 372         memcpy(node, oldroot, size);
 373         xfs_da_log_buf(tp, bp, 0, size - 1);
 374         xfs_da_buf_done(blk1->bp);
 375         blk1->bp = bp;
 376         blk1->blkno = blkno;
 377
 378         /*
 379          * Set up the new root node.
 380          */
 381         error = xfs_da_node_create(args,
 382                 args->whichfork == XFS_DATA_FORK &&
 383                 XFS_DIR_IS_V2(mp) ? mp->m_dirleafblk : 0,
 384                 be16_to_cpu(node->hdr.level) + 1, &bp, args->whichfork);
 385         if (error)
 386                 return(error);
 387         node = bp->data;
 388         node->btree[0].hashval = cpu_to_be32(blk1->hashval);
 389         node->btree[0].before = cpu_to_be32(blk1->blkno);
 390         node->btree[1].hashval = cpu_to_be32(blk2->hashval);
 391         node->btree[1].before = cpu_to_be32(blk2->blkno);
 392         node->hdr.count = cpu_to_be16(2);
 393
 394 #ifdef DEBUG
 395         if (be16_to_cpu(oldroot->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC) {
 396                 ASSERT(blk1->blkno >= mp->m_dirleafblk &&
 397                        blk1->blkno < mp->m_dirfreeblk);
 398                 ASSERT(blk2->blkno >= mp->m_dirleafblk &&
 399                        blk2->blkno < mp->m_dirfreeblk);
 400         }
 401 #endif
 402
 403         /* Header is already logged by xfs_da_node_create */
 404         xfs_da_log_buf(tp, bp,
 405                 XFS_DA_LOGRANGE(node, node->btree,
 406                         sizeof(xfs_da_node_entry_t) * 2));
 407         xfs_da_buf_done(bp);
 408
 409         return(0);
 410 }
 411
 412 /*
 413  * Split the node, rebalance, then add the new entry.
 414  */
 415 STATIC int                                              /* error */
 416 xfs_da_node_split(xfs_da_state_t *state, xfs_da_state_blk_t *oldblk,
 417                                  xfs_da_state_blk_t *newblk,
 418                                  xfs_da_state_blk_t *addblk,
 419                                  int treelevel, int *result)
 420 {
 421         xfs_da_intnode_t *node;
 422         xfs_dablk_t blkno;
 423         int newcount, error;
 424         int useextra;
 425
 426         node = oldblk->bp->data;
 427         ASSERT(be16_to_cpu(node->hdr.info.magic) == XFS_DA_NODE_MAGIC);
 428
 429         /*
 430          * With V2 the extra block is data or freespace.
 431          */
 432         useextra = state->extravalid && (XFS_DIR_IS_V1(state->mp) ||
 433                         state->args->whichfork == XFS_ATTR_FORK);
 434         newcount = 1 + useextra;
 435         /*
 436          * Do we have to split the node?
 437          */
 438         if ((be16_to_cpu(node->hdr.count) + newcount) > state->node_ents) {
 439                 /*
 440                  * Allocate a new node, add to the doubly linked chain of
 441                  * nodes, then move some of our excess entries into it.
 442                  */
 443                 error = xfs_da_grow_inode(state->args, &blkno);
 444                 if (error)
 445                         return(error);  /* GROT: dir is inconsistent */
 446
 447                 error = xfs_da_node_create(state->args, blkno, treelevel,
 448                                            &newblk->bp, state->args->whichfork);
 449                 if (error)
 450                         return(error);  /* GROT: dir is inconsistent */
 451                 newblk->blkno = blkno;
 452                 newblk->magic = XFS_DA_NODE_MAGIC;
 453                 xfs_da_node_rebalance(state, oldblk, newblk);
 454                 error = xfs_da_blk_link(state, oldblk, newblk);
 455                 if (error)
 456                         return(error);
 457                 *result = 1;
 458         } else {
 459                 *result = 0;
 460         }
 461
 462         /*
 463          * Insert the new entry(s) into the correct block
 464          * (updating last hashval in the process).
 465          *
 466          * xfs_da_node_add() inserts BEFORE the given index,
 467          * and as a result of using node_lookup_int() we always
 468          * point to a valid entry (not after one), but a split
 469          * operation always results in a new block whose hashvals
 470          * FOLLOW the current block.
 471          *
 472          * If we had double-split op below us, then add the extra block too.
 473          */
 474         node = oldblk->bp->data;
 475         if (oldblk->index <= be16_to_cpu(node->hdr.count)) {
 476                 oldblk->index++;
 477                 xfs_da_node_add(state, oldblk, addblk);
 478                 if (useextra) {
 479                         if (state->extraafter)
 480                                 oldblk->index++;
 481                         xfs_da_node_add(state, oldblk, &state->extrablk);
 482                         state->extravalid = 0;
 483                 }
 484         } else {
 485                 newblk->index++;
 486                 xfs_da_node_add(state, newblk, addblk);
 487                 if (useextra) {
 488                         if (state->extraafter)
 489                                 newblk->index++;
 490                         xfs_da_node_add(state, newblk, &state->extrablk);
 491                         state->extravalid = 0;
 492                 }
 493         }
 494
 495         return(0);
 496 }
 497
 498 /*
 499  * Balance the btree elements between two intermediate nodes,
 500  * usually one full and one empty.
 501  *
 502  * NOTE: if blk2 is empty, then it will get the upper half of blk1.
 503  */
 504 STATIC void
 505 xfs_da_node_rebalance(xfs_da_state_t *state, xfs_da_state_blk_t *blk1,
 506                                      xfs_da_state_blk_t *blk2)
 507 {
 508         xfs_da_intnode_t *node1, *node2, *tmpnode;
 509         xfs_da_node_entry_t *btree_s, *btree_d;
 510         int count, tmp;
 511         xfs_trans_t *tp;
 512
 513         node1 = blk1->bp->data;
 514         node2 = blk2->bp->data;
 515         /*
 516          * Figure out how many entries need to move, and in which direction.
 517          * Swap the nodes around if that makes it simpler.
 518          */
 519         if ((be16_to_cpu(node1->hdr.count) > 0) && (be16_to_cpu(node2->hdr.count) > 0) &&
 520             ((be32_to_cpu(node2->btree[0].hashval) < be32_to_cpu(node1->btree[0].hashval)) ||
 521              (be32_to_cpu(node2->btree[be16_to_cpu(node2->hdr.count)-1].hashval) <
 522               be32_to_cpu(node1->btree[be16_to_cpu(node1->hdr.count)-1].hashval)))) {
 523                 tmpnode = node1;
 524                 node1 = node2;
 525                 node2 = tmpnode;
 526         }
 527         ASSERT(be16_to_cpu(node1->hdr.info.magic) == XFS_DA_NODE_MAGIC);
 528         ASSERT(be16_to_cpu(node2->hdr.info.magic) == XFS_DA_NODE_MAGIC);
 529         count = (be16_to_cpu(node1->hdr.count) - be16_to_cpu(node2->hdr.count)) / 2;
 530         if (count == 0)
 531                 return;
 532         tp = state->args->trans;
 533         /*
 534          * Two cases: high-to-low and low-to-high.
 535          */
 536         if (count > 0) {
 537                 /*
 538                  * Move elements in node2 up to make a hole.
 539                  */
 540                 if ((tmp = be16_to_cpu(node2->hdr.count)) > 0) {
 541                         tmp *= (uint)sizeof(xfs_da_node_entry_t);
 542                         btree_s = &node2->btree[0];
 543                         btree_d = &node2->btree[count];
 544                         memmove(btree_d, btree_s, tmp);
 545                 }
 546
 547                 /*
 548                  * Move the req'd B-tree elements from high in node1 to
 549                  * low in node2.
 550                  */
 551                 be16_add(&node2->hdr.count, count);
 552                 tmp = count * (uint)sizeof(xfs_da_node_entry_t);
 553                 btree_s = &node1->btree[be16_to_cpu(node1->hdr.count) - count];
 554                 btree_d = &node2->btree[0];
 555                 memcpy(btree_d, btree_s, tmp);
 556                 be16_add(&node1->hdr.count, -count);
 557         } else {
 558                 /*
 559                  * Move the req'd B-tree elements from low in node2 to
 560                  * high in node1.
 561                  */
 562                 count = -count;
 563                 tmp = count * (uint)sizeof(xfs_da_node_entry_t);
 564                 btree_s = &node2->btree[0];
 565                 btree_d = &node1->btree[be16_to_cpu(node1->hdr.count)];
 566                 memcpy(btree_d, btree_s, tmp);
 567                 be16_add(&node1->hdr.count, count);
 568                 xfs_da_log_buf(tp, blk1->bp,
 569                         XFS_DA_LOGRANGE(node1, btree_d, tmp));
 570
 571                 /*
 572                  * Move elements in node2 down to fill the hole.
 573                  */
 574                 tmp  = be16_to_cpu(node2->hdr.count) - count;
 575                 tmp *= (uint)sizeof(xfs_da_node_entry_t);
 576                 btree_s = &node2->btree[count];
 577                 btree_d = &node2->btree[0];
 578                 memmove(btree_d, btree_s, tmp);
 579                 be16_add(&node2->hdr.count, -count);
 580         }
 581
 582         /*
 583          * Log header of node 1 and all current bits of node 2.
 584          */
 585         xfs_da_log_buf(tp, blk1->bp,
 586                 XFS_DA_LOGRANGE(node1, &node1->hdr, sizeof(node1->hdr)));
 587         xfs_da_log_buf(tp, blk2->bp,
 588                 XFS_DA_LOGRANGE(node2, &node2->hdr,
 589                         sizeof(node2->hdr) +
 590                         sizeof(node2->btree[0]) * be16_to_cpu(node2->hdr.count)));
 591
 592         /*
 593          * Record the last hashval from each block for upward propagation.
 594          * (note: don't use the swapped node pointers)
 595          */
 596         node1 = blk1->bp->data;
 597         node2 = blk2->bp->data;
 598         blk1->hashval = be32_to_cpu(node1->btree[be16_to_cpu(node1->hdr.count)-1].hashval);
 599         blk2->hashval = be32_to_cpu(node2->btree[be16_to_cpu(node2->hdr.count)-1].hashval);
 600
 601         /*
 602          * Adjust the expected index for insertion.
 603          */
 604         if (blk1->index >= be16_to_cpu(node1->hdr.count)) {
 605                 blk2->index = blk1->index - be16_to_cpu(node1->hdr.count);
 606                 blk1->index = be16_to_cpu(node1->hdr.count) + 1;        /* make it invalid */
 607         }
 608 }
 609
 610 /*
 611  * Add a new entry to an intermediate node.
 612  */
 613 STATIC void
 614 xfs_da_node_add(xfs_da_state_t *state, xfs_da_state_blk_t *oldblk,
 615                                xfs_da_state_blk_t *newblk)
 616 {
 617         xfs_da_intnode_t *node;
 618         xfs_da_node_entry_t *btree;
 619         int tmp;
 620         xfs_mount_t *mp;
 621
 622         node = oldblk->bp->data;
 623         mp = state->mp;
 624         ASSERT(be16_to_cpu(node->hdr.info.magic) == XFS_DA_NODE_MAGIC);
 625         ASSERT((oldblk->index >= 0) && (oldblk->index <= be16_to_cpu(node->hdr.count)));
 626         ASSERT(newblk->blkno != 0);
 627         if (state->args->whichfork == XFS_DATA_FORK && XFS_DIR_IS_V2(mp))
 628                 ASSERT(newblk->blkno >= mp->m_dirleafblk &&
 629                        newblk->blkno < mp->m_dirfreeblk);
 630
 631         /*
 632          * We may need to make some room before we insert the new node.
 633          */
 634         tmp = 0;
 635         btree = &node->btree[ oldblk->index ];
 636         if (oldblk->index < be16_to_cpu(node->hdr.count)) {
 637                 tmp = (be16_to_cpu(node->hdr.count) - oldblk->index) * (uint)sizeof(*btree);
 638                 memmove(btree + 1, btree, tmp);
 639         }
 640         btree->hashval = cpu_to_be32(newblk->hashval);
 641         btree->before = cpu_to_be32(newblk->blkno);
 642         xfs_da_log_buf(state->args->trans, oldblk->bp,
 643                 XFS_DA_LOGRANGE(node, btree, tmp + sizeof(*btree)));
 644         be16_add(&node->hdr.count, 1);
 645         xfs_da_log_buf(state->args->trans, oldblk->bp,
 646                 XFS_DA_LOGRANGE(node, &node->hdr, sizeof(node->hdr)));
 647
 648         /*
 649          * Copy the last hash value from the oldblk to propagate upwards.
 650          */
 651         oldblk->hashval = be32_to_cpu(node->btree[be16_to_cpu(node->hdr.count)-1 ].hashval);
 652 }
 653
 654 /*========================================================================
 655  * Routines used for shrinking the Btree.
 656  *========================================================================*/
 657
 658 /*
 659  * Deallocate an empty leaf node, remove it from its parent,
 660  * possibly deallocating that block, etc...
 661  */
 662 int
 663 xfs_da_join(xfs_da_state_t *state)
 664 {
 665         xfs_da_state_blk_t *drop_blk, *save_blk;
 666         int action, error;
 667
 668         action = 0;
 669         drop_blk = &state->path.blk[ state->path.active-1 ];
 670         save_blk = &state->altpath.blk[ state->path.active-1 ];
 671         ASSERT(state->path.blk[0].magic == XFS_DA_NODE_MAGIC);
 672         ASSERT(drop_blk->magic == XFS_ATTR_LEAF_MAGIC ||
 673                drop_blk->magic == XFS_DIRX_LEAF_MAGIC(state->mp));
 674
 675         /*
 676          * Walk back up the tree joining/deallocating as necessary.
 677          * When we stop dropping blocks, break out.
 678          */
 679         for (  ; state->path.active >= 2; drop_blk--, save_blk--,
 680                  state->path.active--) {
 681                 /*
 682                  * See if we can combine the block with a neighbor.
 683                  *   (action == 0) => no options, just leave
 684                  *   (action == 1) => coalesce, then unlink
 685                  *   (action == 2) => block empty, unlink it
 686                  */
 687                 switch (drop_blk->magic) {
 688                 case XFS_ATTR_LEAF_MAGIC:
 689                         error = xfs_attr_leaf_toosmall(state, &action);
 690                         if (error)
 691                                 return(error);
 692                         if (action == 0)
 693                                 return(0);
 694                         xfs_attr_leaf_unbalance(state, drop_blk, save_blk);
 695                         break;
 696                 case XFS_DIR_LEAF_MAGIC:
 697                         ASSERT(XFS_DIR_IS_V1(state->mp));
 698                         error = xfs_dir_leaf_toosmall(state, &action);
 699                         if (error)
 700                                 return(error);
 701                         if (action == 0)
 702                                 return(0);
 703                         xfs_dir_leaf_unbalance(state, drop_blk, save_blk);
 704                         break;
 705                 case XFS_DIR2_LEAFN_MAGIC:
 706                         ASSERT(XFS_DIR_IS_V2(state->mp));
 707                         error = xfs_dir2_leafn_toosmall(state, &action);
 708                         if (error)
 709                                 return error;
 710                         if (action == 0)
 711                                 return 0;
 712                         xfs_dir2_leafn_unbalance(state, drop_blk, save_blk);
 713                         break;
 714                 case XFS_DA_NODE_MAGIC:
 715                         /*
 716                          * Remove the offending node, fixup hashvals,
 717                          * check for a toosmall neighbor.
 718                          */
 719                         xfs_da_node_remove(state, drop_blk);
 720                         xfs_da_fixhashpath(state, &state->path);
 721                         error = xfs_da_node_toosmall(state, &action);
 722                         if (error)
 723                                 return(error);
 724                         if (action == 0)
 725                                 return 0;
 726                         xfs_da_node_unbalance(state, drop_blk, save_blk);
 727                         break;
 728                 }
 729                 xfs_da_fixhashpath(state, &state->altpath);
 730                 error = xfs_da_blk_unlink(state, drop_blk, save_blk);
 731                 xfs_da_state_kill_altpath(state);
 732                 if (error)
 733                         return(error);
 734                 error = xfs_da_shrink_inode(state->args, drop_blk->blkno,
 735                                                          drop_blk->bp);
 736                 drop_blk->bp = NULL;
 737                 if (error)
 738                         return(error);
 739         }
 740         /*
 741          * We joined all the way to the top.  If it turns out that
 742          * we only have one entry in the root, make the child block
 743          * the new root.
 744          */
 745         xfs_da_node_remove(state, drop_blk);
 746         xfs_da_fixhashpath(state, &state->path);
 747         error = xfs_da_root_join(state, &state->path.blk[0]);
 748         return(error);
 749 }
 750
 751 /*
 752  * We have only one entry in the root.  Copy the only remaining child of
 753  * the old root to block 0 as the new root node.
 754  */
 755 STATIC int
 756 xfs_da_root_join(xfs_da_state_t *state, xfs_da_state_blk_t *root_blk)
 757 {
 758         xfs_da_intnode_t *oldroot;
 759         /* REFERENCED */
 760         xfs_da_blkinfo_t *blkinfo;
 761         xfs_da_args_t *args;
 762         xfs_dablk_t child;
 763         xfs_dabuf_t *bp;
 764         int error;
 765
 766         args = state->args;
 767         ASSERT(args != NULL);
 768         ASSERT(root_blk->magic == XFS_DA_NODE_MAGIC);
 769         oldroot = root_blk->bp->data;
 770         ASSERT(be16_to_cpu(oldroot->hdr.info.magic) == XFS_DA_NODE_MAGIC);
 771         ASSERT(!oldroot->hdr.info.forw);
 772         ASSERT(!oldroot->hdr.info.back);
 773
 774         /*
 775          * If the root has more than one child, then don't do anything.
 776          */
 777         if (be16_to_cpu(oldroot->hdr.count) > 1)
 778                 return(0);
 779
 780         /*
 781          * Read in the (only) child block, then copy those bytes into
 782          * the root block's buffer and free the original child block.
 783          */
 784         child = be32_to_cpu(oldroot->btree[0].before);
 785         ASSERT(child != 0);
 786         error = xfs_da_read_buf(args->trans, args->dp, child, -1, &bp,
 787                                              args->whichfork);
 788         if (error)
 789                 return(error);
 790         ASSERT(bp != NULL);
 791         blkinfo = bp->data;
 792         if (be16_to_cpu(oldroot->hdr.level) == 1) {
 793                 ASSERT(be16_to_cpu(blkinfo->magic) == XFS_DIRX_LEAF_MAGIC(state->mp) ||
 794                        be16_to_cpu(blkinfo->magic) == XFS_ATTR_LEAF_MAGIC);
 795         } else {
 796                 ASSERT(be16_to_cpu(blkinfo->magic) == XFS_DA_NODE_MAGIC);
 797         }
 798         ASSERT(!blkinfo->forw);
 799         ASSERT(!blkinfo->back);
 800         memcpy(root_blk->bp->data, bp->data, state->blocksize);
 801         xfs_da_log_buf(args->trans, root_blk->bp, 0, state->blocksize - 1);
 802         error = xfs_da_shrink_inode(args, child, bp);
 803         return(error);
 804 }
 805
 806 /*
 807  * Check a node block and its neighbors to see if the block should be
 808  * collapsed into one or the other neighbor.  Always keep the block
 809  * with the smaller block number.
 810  * If the current block is over 50% full, don't try to join it, return 0.
 811  * If the block is empty, fill in the state structure and return 2.
 812  * If it can be collapsed, fill in the state structure and return 1.
 813  * If nothing can be done, return 0.
 814  */
 815 STATIC int
 816 xfs_da_node_toosmall(xfs_da_state_t *state, int *action)
 817 {
 818         xfs_da_intnode_t *node;
 819         xfs_da_state_blk_t *blk;
 820         xfs_da_blkinfo_t *info;
 821         int count, forward, error, retval, i;
 822         xfs_dablk_t blkno;
 823         xfs_dabuf_t *bp;
 824
 825         /*
 826          * Check for the degenerate case of the block being over 50% full.
 827          * If so, it's not worth even looking to see if we might be able
 828          * to coalesce with a sibling.
 829          */
 830         blk = &state->path.blk[ state->path.active-1 ];
 831         info = blk->bp->data;
 832         ASSERT(be16_to_cpu(info->magic) == XFS_DA_NODE_MAGIC);
 833         node = (xfs_da_intnode_t *)info;
 834         count = be16_to_cpu(node->hdr.count);
 835         if (count > (state->node_ents >> 1)) {
 836                 *action = 0;    /* blk over 50%, don't try to join */
 837                 return(0);      /* blk over 50%, don't try to join */
 838         }
 839
 840         /*
 841          * Check for the degenerate case of the block being empty.
 842          * If the block is empty, we'll simply delete it, no need to
 843          * coalesce it with a sibling block.  We choose (arbitrarily)
 844          * to merge with the forward block unless it is NULL.
 845          */
 846         if (count == 0) {
 847                 /*
 848                  * Make altpath point to the block we want to keep and
 849                  * path point to the block we want to drop (this one).
 850                  */
 851                 forward = (info->forw != 0);
 852                 memcpy(&state->altpath, &state->path, sizeof(state->path));
 853                 error = xfs_da_path_shift(state, &state->altpath, forward,
 854                                                  0, &retval);
 855                 if (error)
 856                         return(error);
 857                 if (retval) {
 858                         *action = 0;
 859                 } else {
 860                         *action = 2;
 861                 }
 862                 return(0);
 863         }
 864
 865         /*
 866          * Examine each sibling block to see if we can coalesce with
 867          * at least 25% free space to spare.  We need to figure out
 868          * whether to merge with the forward or the backward block.
 869          * We prefer coalescing with the lower numbered sibling so as
 870          * to shrink a directory over time.
 871          */
 872         /* start with smaller blk num */
 873         forward = (be32_to_cpu(info->forw) < be32_to_cpu(info->back));
 874         for (i = 0; i < 2; forward = !forward, i++) {
 875                 if (forward)
 876                         blkno = be32_to_cpu(info->forw);
 877                 else
 878                         blkno = be32_to_cpu(info->back);
 879                 if (blkno == 0)
 880                         continue;
 881                 error = xfs_da_read_buf(state->args->trans, state->args->dp,
 882                                         blkno, -1, &bp, state->args->whichfork);
 883                 if (error)
 884                         return(error);
 885                 ASSERT(bp != NULL);
 886
 887                 node = (xfs_da_intnode_t *)info;
 888                 count  = state->node_ents;
 889                 count -= state->node_ents >> 2;
 890                 count -= be16_to_cpu(node->hdr.count);
 891                 node = bp->data;
 892                 ASSERT(be16_to_cpu(node->hdr.info.magic) == XFS_DA_NODE_MAGIC);
 893                 count -= be16_to_cpu(node->hdr.count);
 894                 xfs_da_brelse(state->args->trans, bp);
 895                 if (count >= 0)
 896                         break;  /* fits with at least 25% to spare */
 897         }
 898         if (i >= 2) {
 899                 *action = 0;
 900                 return(0);
 901         }
 902
 903         /*
 904          * Make altpath point to the block we want to keep (the lower
 905          * numbered block) and path point to the block we want to drop.
 906          */
 907         memcpy(&state->altpath, &state->path, sizeof(state->path));
 908         if (blkno < blk->blkno) {
 909                 error = xfs_da_path_shift(state, &state->altpath, forward,
 910                                                  0, &retval);
 911                 if (error) {
 912                         return(error);
 913                 }
 914                 if (retval) {
 915                         *action = 0;
 916                         return(0);
 917                 }
 918         } else {
 919                 error = xfs_da_path_shift(state, &state->path, forward,
 920                                                  0, &retval);
 921                 if (error) {
 922                         return(error);
 923                 }
 924                 if (retval) {
 925                         *action = 0;
 926                         return(0);
 927                 }
 928         }
 929         *action = 1;
 930         return(0);
 931 }
 932
 933 /*
 934  * Walk back up the tree adjusting hash values as necessary,
 935  * when we stop making changes, return.
 936  */
 937 void
 938 xfs_da_fixhashpath(xfs_da_state_t *state, xfs_da_state_path_t *path)
 939 {
 940         xfs_da_state_blk_t *blk;
 941         xfs_da_intnode_t *node;
 942         xfs_da_node_entry_t *btree;
 943         xfs_dahash_t lasthash=0;
 944         int level, count;
 945
 946         level = path->active-1;
 947         blk = &path->blk[ level ];
 948         switch (blk->magic) {
 949         case XFS_ATTR_LEAF_MAGIC:
 950                 lasthash = xfs_attr_leaf_lasthash(blk->bp, &count);
 951                 if (count == 0)
 952                         return;
 953                 break;
 954         case XFS_DIR_LEAF_MAGIC:
 955                 ASSERT(XFS_DIR_IS_V1(state->mp));
 956                 lasthash = xfs_dir_leaf_lasthash(blk->bp, &count);
 957                 if (count == 0)
 958                         return;
 959                 break;
 960         case XFS_DIR2_LEAFN_MAGIC:
 961                 ASSERT(XFS_DIR_IS_V2(state->mp));
 962                 lasthash = xfs_dir2_leafn_lasthash(blk->bp, &count);
 963                 if (count == 0)
 964                         return;
 965                 break;
 966         case XFS_DA_NODE_MAGIC:
 967                 lasthash = xfs_da_node_lasthash(blk->bp, &count);
 968                 if (count == 0)
 969                         return;
 970                 break;
 971         }
 972         for (blk--, level--; level >= 0; blk--, level--) {
 973                 node = blk->bp->data;
 974                 ASSERT(be16_to_cpu(node->hdr.info.magic) == XFS_DA_NODE_MAGIC);
 975                 btree = &node->btree[ blk->index ];
 976                 if (be32_to_cpu(btree->hashval) == lasthash)
 977                         break;
 978                 blk->hashval = lasthash;
 979                 btree->hashval = cpu_to_be32(lasthash);
 980                 xfs_da_log_buf(state->args->trans, blk->bp,
 981                                   XFS_DA_LOGRANGE(node, btree, sizeof(*btree)));
 982
 983                 lasthash = be32_to_cpu(node->btree[be16_to_cpu(node->hdr.count)-1].hashval);
 984         }
 985 }
 986
 987 /*
 988  * Remove an entry from an intermediate node.
 989  */
 990 STATIC void
 991 xfs_da_node_remove(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk)
 992 {
 993         xfs_da_intnode_t *node;
 994         xfs_da_node_entry_t *btree;
 995         int tmp;
 996
 997         node = drop_blk->bp->data;
 998         ASSERT(drop_blk->index < be16_to_cpu(node->hdr.count));
 999         ASSERT(drop_blk->index >= 0);
1000
1001         /*
1002          * Copy over the offending entry, or just zero it out.
1003          */
1004         btree = &node->btree[drop_blk->index];
1005         if (drop_blk->index < (be16_to_cpu(node->hdr.count)-1)) {
1006                 tmp  = be16_to_cpu(node->hdr.count) - drop_blk->index - 1;
1007                 tmp *= (uint)sizeof(xfs_da_node_entry_t);
1008                 memmove(btree, btree + 1, tmp);
1009                 xfs_da_log_buf(state->args->trans, drop_blk->bp,
1010                     XFS_DA_LOGRANGE(node, btree, tmp));
1011                 btree = &node->btree[be16_to_cpu(node->hdr.count)-1];
1012         }
1013         memset((char *)btree, 0, sizeof(xfs_da_node_entry_t));
1014         xfs_da_log_buf(state->args->trans, drop_blk->bp,
1015             XFS_DA_LOGRANGE(node, btree, sizeof(*btree)));
1016         be16_add(&node->hdr.count, -1);
1017         xfs_da_log_buf(state->args->trans, drop_blk->bp,
1018             XFS_DA_LOGRANGE(node, &node->hdr, sizeof(node->hdr)));
1019
1020         /*
1021          * Copy the last hash value from the block to propagate upwards.
1022          */
1023         btree--;
1024         drop_blk->hashval = be32_to_cpu(btree->hashval);
1025 }
1026
1027 /*
1028  * Unbalance the btree elements between two intermediate nodes,
1029  * move all Btree elements from one node into another.
1030  */
1031 STATIC void
1032 xfs_da_node_unbalance(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
1033                                      xfs_da_state_blk_t *save_blk)
1034 {
1035         xfs_da_intnode_t *drop_node, *save_node;
1036         xfs_da_node_entry_t *btree;
1037         int tmp;
1038         xfs_trans_t *tp;
1039
1040         drop_node = drop_blk->bp->data;
1041         save_node = save_blk->bp->data;
1042         ASSERT(be16_to_cpu(drop_node->hdr.info.magic) == XFS_DA_NODE_MAGIC);
1043         ASSERT(be16_to_cpu(save_node->hdr.info.magic) == XFS_DA_NODE_MAGIC);
1044         tp = state->args->trans;
1045
1046         /*
1047          * If the dying block has lower hashvals, then move all the
1048          * elements in the remaining block up to make a hole.
1049          */
1050         if ((be32_to_cpu(drop_node->btree[0].hashval) < be32_to_cpu(save_node->btree[ 0 ].hashval)) ||
1051             (be32_to_cpu(drop_node->btree[be16_to_cpu(drop_node->hdr.count)-1].hashval) <
1052              be32_to_cpu(save_node->btree[be16_to_cpu(save_node->hdr.count)-1].hashval)))
1053         {
1054                 btree = &save_node->btree[be16_to_cpu(drop_node->hdr.count)];
1055                 tmp = be16_to_cpu(save_node->hdr.count) * (uint)sizeof(xfs_da_node_entry_t);
1056                 memmove(btree, &save_node->btree[0], tmp);
1057                 btree = &save_node->btree[0];
1058                 xfs_da_log_buf(tp, save_blk->bp,
1059                         XFS_DA_LOGRANGE(save_node, btree,
1060                                 (be16_to_cpu(save_node->hdr.count) + be16_to_cpu(drop_node->hdr.count)) *
1061                                 sizeof(xfs_da_node_entry_t)));
1062         } else {
1063                 btree = &save_node->btree[be16_to_cpu(save_node->hdr.count)];
1064                 xfs_da_log_buf(tp, save_blk->bp,
1065                         XFS_DA_LOGRANGE(save_node, btree,
1066                                 be16_to_cpu(drop_node->hdr.count) *
1067                                 sizeof(xfs_da_node_entry_t)));
1068         }
1069
1070         /*
1071          * Move all the B-tree elements from drop_blk to save_blk.
1072          */
1073         tmp = be16_to_cpu(drop_node->hdr.count) * (uint)sizeof(xfs_da_node_entry_t);
1074         memcpy(btree, &drop_node->btree[0], tmp);
1075         be16_add(&save_node->hdr.count, be16_to_cpu(drop_node->hdr.count));
1076
1077         xfs_da_log_buf(tp, save_blk->bp,
1078                 XFS_DA_LOGRANGE(save_node, &save_node->hdr,
1079                         sizeof(save_node->hdr)));
1080
1081         /*
1082          * Save the last hashval in the remaining block for upward propagation.
1083          */
1084         save_blk->hashval = be32_to_cpu(save_node->btree[be16_to_cpu(save_node->hdr.count)-1].hashval);
1085 }
1086
1087 /*========================================================================
1088  * Routines used for finding things in the Btree.
1089  *========================================================================*/
1090
1091 /*
1092  * Walk down the Btree looking for a particular filename, filling
1093  * in the state structure as we go.
1094  *
1095  * We will set the state structure to point to each of the elements
1096  * in each of the nodes where either the hashval is or should be.
1097  *
1098  * We support duplicate hashval's so for each entry in the current
1099  * node that could contain the desired hashval, descend.  This is a
1100  * pruned depth-first tree search.
1101  */
1102 int                                                     /* error */
1103 xfs_da_node_lookup_int(xfs_da_state_t *state, int *result)
1104 {
1105         xfs_da_state_blk_t *blk;
1106         xfs_da_blkinfo_t *curr;
1107         xfs_da_intnode_t *node;
1108         xfs_da_node_entry_t *btree;
1109         xfs_dablk_t blkno;
1110         int probe, span, max, error, retval;
1111         xfs_dahash_t hashval;
1112         xfs_da_args_t *args;
1113
1114         args = state->args;
1115
1116         /*
1117          * Descend thru the B-tree searching each level for the right
1118          * node to use, until the right hashval is found.
1119          */
1120         if (args->whichfork == XFS_DATA_FORK && XFS_DIR_IS_V2(state->mp))
1121                 blkno = state->mp->m_dirleafblk;
1122         else
1123                 blkno = 0;
1124         for (blk = &state->path.blk[0], state->path.active = 1;
1125                          state->path.active <= XFS_DA_NODE_MAXDEPTH;
1126                          blk++, state->path.active++) {
1127                 /*
1128                  * Read the next node down in the tree.
1129                  */
1130                 blk->blkno = blkno;
1131                 error = xfs_da_read_buf(args->trans, args->dp, blkno,
1132                                         -1, &blk->bp, args->whichfork);
1133                 if (error) {
1134                         blk->blkno = 0;
1135                         state->path.active--;
1136                         return(error);
1137                 }
1138                 curr = blk->bp->data;
1139                 ASSERT(be16_to_cpu(curr->magic) == XFS_DA_NODE_MAGIC ||
1140                        be16_to_cpu(curr->magic) == XFS_DIRX_LEAF_MAGIC(state->mp) ||
1141                        be16_to_cpu(curr->magic) == XFS_ATTR_LEAF_MAGIC);
1142
1143                 /*
1144                  * Search an intermediate node for a match.
1145                  */
1146                 blk->magic = be16_to_cpu(curr->magic);
1147                 if (blk->magic == XFS_DA_NODE_MAGIC) {
1148                         node = blk->bp->data;
1149                         blk->hashval = be32_to_cpu(node->btree[be16_to_cpu(node->hdr.count)-1].hashval);
1150
1151                         /*
1152                          * Binary search.  (note: small blocks will skip loop)
1153                          */
1154                         max = be16_to_cpu(node->hdr.count);
1155                         probe = span = max / 2;
1156                         hashval = args->hashval;
1157                         for (btree = &node->btree[probe]; span > 4;
1158                                    btree = &node->btree[probe]) {
1159                                 span /= 2;
1160                                 if (be32_to_cpu(btree->hashval) < hashval)
1161                                         probe += span;
1162                                 else if (be32_to_cpu(btree->hashval) > hashval)
1163                                         probe -= span;
1164                                 else
1165                                         break;
1166                         }
1167                         ASSERT((probe >= 0) && (probe < max));
1168                         ASSERT((span <= 4) || (be32_to_cpu(btree->hashval) == hashval));
1169
1170                         /*
1171                          * Since we may have duplicate hashval's, find the first
1172                          * matching hashval in the node.
1173                          */
1174                         while ((probe > 0) && (be32_to_cpu(btree->hashval) >= hashval)) {
1175                                 btree--;
1176                                 probe--;
1177                         }
1178                         while ((probe < max) && (be32_to_cpu(btree->hashval) < hashval)) {
1179                                 btree++;
1180                                 probe++;
1181                         }
1182
1183                         /*
1184                          * Pick the right block to descend on.
1185                          */
1186                         if (probe == max) {
1187                                 blk->index = max-1;
1188                                 blkno = be32_to_cpu(node->btree[max-1].before);
1189                         } else {
1190                                 blk->index = probe;
1191                                 blkno = be32_to_cpu(btree->before);
1192                         }
1193                 }
1194                 else if (be16_to_cpu(curr->magic) == XFS_ATTR_LEAF_MAGIC) {
1195                         blk->hashval = xfs_attr_leaf_lasthash(blk->bp, NULL);
1196                         break;
1197                 }
1198                 else if (be16_to_cpu(curr->magic) == XFS_DIR_LEAF_MAGIC) {
1199                         blk->hashval = xfs_dir_leaf_lasthash(blk->bp, NULL);
1200                         break;
1201                 }
1202                 else if (be16_to_cpu(curr->magic) == XFS_DIR2_LEAFN_MAGIC) {
1203                         blk->hashval = xfs_dir2_leafn_lasthash(blk->bp, NULL);
1204                         break;
1205                 }
1206         }
1207
1208         /*
1209          * A leaf block that ends in the hashval that we are interested in
1210          * (final hashval == search hashval) means that the next block may
1211          * contain more entries with the same hashval, shift upward to the
1212          * next leaf and keep searching.
1213          */
1214         for (;;) {
1215                 if (blk->magic == XFS_DIR_LEAF_MAGIC) {
1216                         ASSERT(XFS_DIR_IS_V1(state->mp));
1217                         retval = xfs_dir_leaf_lookup_int(blk->bp, args,
1218                                                                   &blk->index);
1219                 } else if (blk->magic == XFS_DIR2_LEAFN_MAGIC) {
1220                         ASSERT(XFS_DIR_IS_V2(state->mp));
1221                         retval = xfs_dir2_leafn_lookup_int(blk->bp, args,
1222                                                         &blk->index, state);
1223                 }
1224                 else if (blk->magic == XFS_ATTR_LEAF_MAGIC) {
1225                         retval = xfs_attr_leaf_lookup_int(blk->bp, args);
1226                         blk->index = args->index;
1227                         args->blkno = blk->blkno;
1228                 }
1229                 if (((retval == ENOENT) || (retval == ENOATTR)) &&
1230                     (blk->hashval == args->hashval)) {
1231                         error = xfs_da_path_shift(state, &state->path, 1, 1,
1232                                                          &retval);
1233                         if (error)
1234                                 return(error);
1235                         if (retval == 0) {
1236                                 continue;
1237                         }
1238                         else if (blk->magic == XFS_ATTR_LEAF_MAGIC) {
1239                                 /* path_shift() gives ENOENT */
1240                                 retval = XFS_ERROR(ENOATTR);
1241                         }
1242                 }
1243                 break;
1244         }
1245         *result = retval;
1246         return(0);
1247 }
1248
1249 /*========================================================================
1250  * Utility routines.
1251  *========================================================================*/
1252
1253 /*
1254  * Link a new block into a doubly linked list of blocks (of whatever type).
1255  */
1256 int                                                     /* error */
1257 xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
1258                                xfs_da_state_blk_t *new_blk)
1259 {
1260         xfs_da_blkinfo_t *old_info, *new_info, *tmp_info;
1261         xfs_da_args_t *args;
1262         int before=0, error;
1263         xfs_dabuf_t *bp;
1264
1265         /*
1266          * Set up environment.
1267          */
1268         args = state->args;
1269         ASSERT(args != NULL);
1270         old_info = old_blk->bp->data;
1271         new_info = new_blk->bp->data;
1272         ASSERT(old_blk->magic == XFS_DA_NODE_MAGIC ||
1273                old_blk->magic == XFS_DIRX_LEAF_MAGIC(state->mp) ||
1274                old_blk->magic == XFS_ATTR_LEAF_MAGIC);
1275         ASSERT(old_blk->magic == be16_to_cpu(old_info->magic));
1276         ASSERT(new_blk->magic == be16_to_cpu(new_info->magic));
1277         ASSERT(old_blk->magic == new_blk->magic);
1278
1279         switch (old_blk->magic) {
1280         case XFS_ATTR_LEAF_MAGIC:
1281                 before = xfs_attr_leaf_order(old_blk->bp, new_blk->bp);
1282                 break;
1283         case XFS_DIR_LEAF_MAGIC:
1284                 ASSERT(XFS_DIR_IS_V1(state->mp));
1285                 before = xfs_dir_leaf_order(old_blk->bp, new_blk->bp);
1286                 break;
1287         case XFS_DIR2_LEAFN_MAGIC:
1288                 ASSERT(XFS_DIR_IS_V2(state->mp));
1289                 before = xfs_dir2_leafn_order(old_blk->bp, new_blk->bp);
1290                 break;
1291         case XFS_DA_NODE_MAGIC:
1292                 before = xfs_da_node_order(old_blk->bp, new_blk->bp);
1293                 break;
1294         }
1295
1296         /*
1297          * Link blocks in appropriate order.
1298          */
1299         if (before) {
1300                 /*
1301                  * Link new block in before existing block.
1302                  */
1303                 new_info->forw = cpu_to_be32(old_blk->blkno);
1304                 new_info->back = old_info->back;
1305                 if (old_info->back) {
1306                         error = xfs_da_read_buf(args->trans, args->dp,
1307                                                 be32_to_cpu(old_info->back),
1308                                                 -1, &bp, args->whichfork);
1309                         if (error)
1310                                 return(error);
1311                         ASSERT(bp != NULL);
1312                         tmp_info = bp->data;
1313                         ASSERT(be16_to_cpu(tmp_info->magic) == be16_to_cpu(old_info->magic));
1314                         ASSERT(be32_to_cpu(tmp_info->forw) == old_blk->blkno);
1315                         tmp_info->forw = cpu_to_be32(new_blk->blkno);
1316                         xfs_da_log_buf(args->trans, bp, 0, sizeof(*tmp_info)-1);
1317                         xfs_da_buf_done(bp);
1318                 }
1319                 old_info->back = cpu_to_be32(new_blk->blkno);
1320         } else {
1321                 /*
1322                  * Link new block in after existing block.
1323                  */
1324                 new_info->forw = old_info->forw;
1325                 new_info->back = cpu_to_be32(old_blk->blkno);
1326                 if (old_info->forw) {
1327                         error = xfs_da_read_buf(args->trans, args->dp,
1328                                                 be32_to_cpu(old_info->forw),
1329                                                 -1, &bp, args->whichfork);
1330                         if (error)
1331                                 return(error);
1332                         ASSERT(bp != NULL);
1333                         tmp_info = bp->data;
1334                         ASSERT(tmp_info->magic == old_info->magic);
1335                         ASSERT(be32_to_cpu(tmp_info->back) == old_blk->blkno);
1336                         tmp_info->back = cpu_to_be32(new_blk->blkno);
1337                         xfs_da_log_buf(args->trans, bp, 0, sizeof(*tmp_info)-1);
1338                         xfs_da_buf_done(bp);
1339                 }
1340                 old_info->forw = cpu_to_be32(new_blk->blkno);
1341         }
1342
1343         xfs_da_log_buf(args->trans, old_blk->bp, 0, sizeof(*tmp_info) - 1);
1344         xfs_da_log_buf(args->trans, new_blk->bp, 0, sizeof(*tmp_info) - 1);
1345         return(0);
1346 }
1347
1348 /*
1349  * Compare two intermediate nodes for "order".
1350  */
1351 STATIC int
1352 xfs_da_node_order(xfs_dabuf_t *node1_bp, xfs_dabuf_t *node2_bp)
1353 {
1354         xfs_da_intnode_t *node1, *node2;
1355
1356         node1 = node1_bp->data;
1357         node2 = node2_bp->data;
1358         ASSERT((be16_to_cpu(node1->hdr.info.magic) == XFS_DA_NODE_MAGIC) &&
1359                (be16_to_cpu(node2->hdr.info.magic) == XFS_DA_NODE_MAGIC));
1360         if ((be16_to_cpu(node1->hdr.count) > 0) && (be16_to_cpu(node2->hdr.count) > 0) &&
1361             ((be32_to_cpu(node2->btree[0].hashval) <
1362               be32_to_cpu(node1->btree[0].hashval)) ||
1363              (be32_to_cpu(node2->btree[be16_to_cpu(node2->hdr.count)-1].hashval) <
1364               be32_to_cpu(node1->btree[be16_to_cpu(node1->hdr.count)-1].hashval)))) {
1365                 return(1);
1366         }
1367         return(0);
1368 }
1369
1370 /*
1371  * Pick up the last hashvalue from an intermediate node.
1372  */
1373 STATIC uint
1374 xfs_da_node_lasthash(xfs_dabuf_t *bp, int *count)
1375 {
1376         xfs_da_intnode_t *node;
1377
1378         node = bp->data;
1379         ASSERT(be16_to_cpu(node->hdr.info.magic) == XFS_DA_NODE_MAGIC);
1380         if (count)
1381                 *count = be16_to_cpu(node->hdr.count);
1382         if (!node->hdr.count)
1383                 return(0);
1384         return be32_to_cpu(node->btree[be16_to_cpu(node->hdr.count)-1].hashval);
1385 }
1386
1387 /*
1388  * Unlink a block from a doubly linked list of blocks.
1389  */
1390 STATIC int                                              /* error */
1391 xfs_da_blk_unlink(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
1392                                  xfs_da_state_blk_t *save_blk)
1393 {
1394         xfs_da_blkinfo_t *drop_info, *save_info, *tmp_info;
1395         xfs_da_args_t *args;
1396         xfs_dabuf_t *bp;
1397         int error;
1398
1399         /*
1400          * Set up environment.
1401          */
1402         args = state->args;
1403         ASSERT(args != NULL);
1404         save_info = save_blk->bp->data;
1405         drop_info = drop_blk->bp->data;
1406         ASSERT(save_blk->magic == XFS_DA_NODE_MAGIC ||
1407                save_blk->magic == XFS_DIRX_LEAF_MAGIC(state->mp) ||
1408                save_blk->magic == XFS_ATTR_LEAF_MAGIC);
1409         ASSERT(save_blk->magic == be16_to_cpu(save_info->magic));
1410         ASSERT(drop_blk->magic == be16_to_cpu(drop_info->magic));
1411         ASSERT(save_blk->magic == drop_blk->magic);
1412         ASSERT((be32_to_cpu(save_info->forw) == drop_blk->blkno) ||
1413                (be32_to_cpu(save_info->back) == drop_blk->blkno));
1414         ASSERT((be32_to_cpu(drop_info->forw) == save_blk->blkno) ||
1415                (be32_to_cpu(drop_info->back) == save_blk->blkno));
1416
1417         /*
1418          * Unlink the leaf block from the doubly linked chain of leaves.
1419          */
1420         if (be32_to_cpu(save_info->back) == drop_blk->blkno) {
1421                 save_info->back = drop_info->back;
1422                 if (drop_info->back) {
1423                         error = xfs_da_read_buf(args->trans, args->dp,
1424                                                 be32_to_cpu(drop_info->back),
1425                                                 -1, &bp, args->whichfork);
1426                         if (error)
1427                                 return(error);
1428                         ASSERT(bp != NULL);
1429                         tmp_info = bp->data;
1430                         ASSERT(tmp_info->magic == save_info->magic);
1431                         ASSERT(be32_to_cpu(tmp_info->forw) == drop_blk->blkno);
1432                         tmp_info->forw = cpu_to_be32(save_blk->blkno);
1433                         xfs_da_log_buf(args->trans, bp, 0,
1434                                                     sizeof(*tmp_info) - 1);
1435                         xfs_da_buf_done(bp);
1436                 }
1437         } else {
1438                 save_info->forw = drop_info->forw;
1439                 if (drop_info->forw) {
1440                         error = xfs_da_read_buf(args->trans, args->dp,
1441                                                 be32_to_cpu(drop_info->forw),
1442                                                 -1, &bp, args->whichfork);
1443                         if (error)
1444                                 return(error);
1445                         ASSERT(bp != NULL);
1446                         tmp_info = bp->data;
1447                         ASSERT(tmp_info->magic == save_info->magic);
1448                         ASSERT(be32_to_cpu(tmp_info->back) == drop_blk->blkno);
1449                         tmp_info->back = cpu_to_be32(save_blk->blkno);
1450                         xfs_da_log_buf(args->trans, bp, 0,
1451                                                     sizeof(*tmp_info) - 1);
1452                         xfs_da_buf_done(bp);
1453                 }
1454         }
1455
1456         xfs_da_log_buf(args->trans, save_blk->bp, 0, sizeof(*save_info) - 1);
1457         return(0);
1458 }
1459
1460 /*
1461  * Move a path "forward" or "!forward" one block at the current level.
1462  *
1463  * This routine will adjust a "path" to point to the next block
1464  * "forward" (higher hashvalues) or "!forward" (lower hashvals) in the
1465  * Btree, including updating pointers to the intermediate nodes between
1466  * the new bottom and the root.
1467  */
1468 int                                                     /* error */
1469 xfs_da_path_shift(xfs_da_state_t *state, xfs_da_state_path_t *path,
1470                                  int forward, int release, int *result)
1471 {
1472         xfs_da_state_blk_t *blk;
1473         xfs_da_blkinfo_t *info;
1474         xfs_da_intnode_t *node;
1475         xfs_da_args_t *args;
1476         xfs_dablk_t blkno=0;
1477         int level, error;
1478
1479         /*
1480          * Roll up the Btree looking for the first block where our
1481          * current index is not at the edge of the block.  Note that
1482          * we skip the bottom layer because we want the sibling block.
1483          */
1484         args = state->args;
1485         ASSERT(args != NULL);
1486         ASSERT(path != NULL);
1487         ASSERT((path->active > 0) && (path->active < XFS_DA_NODE_MAXDEPTH));
1488         level = (path->active-1) - 1;   /* skip bottom layer in path */
1489         for (blk = &path->blk[level]; level >= 0; blk--, level--) {
1490                 ASSERT(blk->bp != NULL);
1491                 node = blk->bp->data;
1492                 ASSERT(be16_to_cpu(node->hdr.info.magic) == XFS_DA_NODE_MAGIC);
1493                 if (forward && (blk->index < be16_to_cpu(node->hdr.count)-1)) {
1494                         blk->index++;
1495                         blkno = be32_to_cpu(node->btree[blk->index].before);
1496                         break;
1497                 } else if (!forward && (blk->index > 0)) {
1498                         blk->index--;
1499                         blkno = be32_to_cpu(node->btree[blk->index].before);
1500                         break;
1501                 }
1502         }
1503         if (level < 0) {
1504                 *result = XFS_ERROR(ENOENT);    /* we're out of our tree */
1505                 ASSERT(args->oknoent);
1506                 return(0);
1507         }
1508
1509         /*
1510          * Roll down the edge of the subtree until we reach the
1511          * same depth we were at originally.
1512          */
1513         for (blk++, level++; level < path->active; blk++, level++) {
1514                 /*
1515                  * Release the old block.
1516                  * (if it's dirty, trans won't actually let go)
1517                  */
1518                 if (release)
1519                         xfs_da_brelse(args->trans, blk->bp);
1520
1521                 /*
1522                  * Read the next child block.
1523                  */
1524                 blk->blkno = blkno;
1525                 error = xfs_da_read_buf(args->trans, args->dp, blkno, -1,
1526                                                      &blk->bp, args->whichfork);
1527                 if (error)
1528                         return(error);
1529                 ASSERT(blk->bp != NULL);
1530                 info = blk->bp->data;
1531                 ASSERT(be16_to_cpu(info->magic) == XFS_DA_NODE_MAGIC ||
1532                        be16_to_cpu(info->magic) == XFS_DIRX_LEAF_MAGIC(state->mp) ||
1533                        be16_to_cpu(info->magic) == XFS_ATTR_LEAF_MAGIC);
1534                 blk->magic = be16_to_cpu(info->magic);
1535                 if (blk->magic == XFS_DA_NODE_MAGIC) {
1536                         node = (xfs_da_intnode_t *)info;
1537                         blk->hashval = be32_to_cpu(node->btree[be16_to_cpu(node->hdr.count)-1].hashval);
1538                         if (forward)
1539                                 blk->index = 0;
1540                         else
1541                                 blk->index = be16_to_cpu(node->hdr.count)-1;
1542                         blkno = be32_to_cpu(node->btree[blk->index].before);
1543                 } else {
1544                         ASSERT(level == path->active-1);
1545                         blk->index = 0;
1546                         switch(blk->magic) {
1547                         case XFS_ATTR_LEAF_MAGIC:
1548                                 blk->hashval = xfs_attr_leaf_lasthash(blk->bp,
1549                                                                       NULL);
1550                                 break;
1551                         case XFS_DIR_LEAF_MAGIC:
1552                                 ASSERT(XFS_DIR_IS_V1(state->mp));
1553                                 blk->hashval = xfs_dir_leaf_lasthash(blk->bp,
1554                                                                      NULL);
1555                                 break;
1556                         case XFS_DIR2_LEAFN_MAGIC:
1557                                 ASSERT(XFS_DIR_IS_V2(state->mp));
1558                                 blk->hashval = xfs_dir2_leafn_lasthash(blk->bp,
1559                                                                        NULL);
1560                                 break;
1561                         default:
1562                                 ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC ||
1563                                        blk->magic ==
1564                                        XFS_DIRX_LEAF_MAGIC(state->mp));
1565                                 break;
1566                         }
1567                 }
1568         }
1569         *result = 0;
1570         return(0);
1571 }
1572
1573
1574 /*========================================================================
1575  * Utility routines.
1576  *========================================================================*/
1577
1578 /*
1579  * Implement a simple hash on a character string.
1580  * Rotate the hash value by 7 bits, then XOR each character in.
1581  * This is implemented with some source-level loop unrolling.
1582  */
1583 xfs_dahash_t
1584 xfs_da_hashname(const uchar_t *name, int namelen)
1585 {
1586         xfs_dahash_t hash;
1587
1588         /*
1589          * Do four characters at a time as long as we can.
1590          */
1591         for (hash = 0; namelen >= 4; namelen -= 4, name += 4)
1592                 hash = (name[0] << 21) ^ (name[1] << 14) ^ (name[2] << 7) ^
1593                        (name[3] << 0) ^ rol32(hash, 7 * 4);
1594
1595         /*
1596          * Now do the rest of the characters.
1597          */
1598         switch (namelen) {
1599         case 3:
1600                 return (name[0] << 14) ^ (name[1] << 7) ^ (name[2] << 0) ^
1601                        rol32(hash, 7 * 3);
1602         case 2:
1603                 return (name[0] << 7) ^ (name[1] << 0) ^ rol32(hash, 7 * 2);
1604         case 1:
1605                 return (name[0] << 0) ^ rol32(hash, 7 * 1);
1606         default: /* case 0: */
1607                 return hash;
1608         }
1609 }
1610
1611 /*
1612  * Add a block to the btree ahead of the file.
1613  * Return the new block number to the caller.
1614  */
1615 int
1616 xfs_da_grow_inode(xfs_da_args_t *args, xfs_dablk_t *new_blkno)
1617 {
1618         xfs_fileoff_t bno, b;
1619         xfs_bmbt_irec_t map;
1620         xfs_bmbt_irec_t *mapp;
1621         xfs_inode_t *dp;
1622         int nmap, error, w, count, c, got, i, mapi;
1623         xfs_fsize_t size;
1624         xfs_trans_t *tp;
1625         xfs_mount_t *mp;
1626
1627         dp = args->dp;
1628         mp = dp->i_mount;
1629         w = args->whichfork;
1630         tp = args->trans;
1631         /*
1632          * For new directories adjust the file offset and block count.
1633          */
1634         if (w == XFS_DATA_FORK && XFS_DIR_IS_V2(mp)) {
1635                 bno = mp->m_dirleafblk;
1636                 count = mp->m_dirblkfsbs;
1637         } else {
1638                 bno = 0;
1639                 count = 1;
1640         }
1641         /*
1642          * Find a spot in the file space to put the new block.
1643          */
1644         if ((error = xfs_bmap_first_unused(tp, dp, count, &bno, w))) {
1645                 return error;
1646         }
1647         if (w == XFS_DATA_FORK && XFS_DIR_IS_V2(mp))
1648                 ASSERT(bno >= mp->m_dirleafblk && bno < mp->m_dirfreeblk);
1649         /*
1650          * Try mapping it in one filesystem block.
1651          */
1652         nmap = 1;
1653         ASSERT(args->firstblock != NULL);
1654         if ((error = xfs_bmapi(tp, dp, bno, count,
1655                         XFS_BMAPI_AFLAG(w)|XFS_BMAPI_WRITE|XFS_BMAPI_METADATA|
1656                         XFS_BMAPI_CONTIG,
1657                         args->firstblock, args->total, &map, &nmap,
1658                         args->flist, NULL))) {
1659                 return error;
1660         }
1661         ASSERT(nmap <= 1);
1662         if (nmap == 1) {
1663                 mapp = &map;
1664                 mapi = 1;
1665         }
1666         /*
1667          * If we didn't get it and the block might work if fragmented,
1668          * try without the CONTIG flag.  Loop until we get it all.
1669          */
1670         else if (nmap == 0 && count > 1) {
1671                 mapp = kmem_alloc(sizeof(*mapp) * count, KM_SLEEP);
1672                 for (b = bno, mapi = 0; b < bno + count; ) {
1673                         nmap = MIN(XFS_BMAP_MAX_NMAP, count);
1674                         c = (int)(bno + count - b);
1675                         if ((error = xfs_bmapi(tp, dp, b, c,
1676                                         XFS_BMAPI_AFLAG(w)|XFS_BMAPI_WRITE|
1677                                         XFS_BMAPI_METADATA,
1678                                         args->firstblock, args->total,
1679                                         &mapp[mapi], &nmap, args->flist,
1680                                         NULL))) {
1681                                 kmem_free(mapp, sizeof(*mapp) * count);
1682                                 return error;
1683                         }
1684                         if (nmap < 1)
1685                                 break;
1686                         mapi += nmap;
1687                         b = mapp[mapi - 1].br_startoff +
1688                             mapp[mapi - 1].br_blockcount;
1689                 }
1690         } else {
1691                 mapi = 0;
1692                 mapp = NULL;
1693         }
1694         /*
1695          * Count the blocks we got, make sure it matches the total.
1696          */
1697         for (i = 0, got = 0; i < mapi; i++)
1698                 got += mapp[i].br_blockcount;
1699         if (got != count || mapp[0].br_startoff != bno ||
1700             mapp[mapi - 1].br_startoff + mapp[mapi - 1].br_blockcount !=
1701             bno + count) {
1702                 if (mapp != &map)
1703                         kmem_free(mapp, sizeof(*mapp) * count);
1704                 return XFS_ERROR(ENOSPC);
1705         }
1706         if (mapp != &map)
1707                 kmem_free(mapp, sizeof(*mapp) * count);
1708         *new_blkno = (xfs_dablk_t)bno;
1709         /*
1710          * For version 1 directories, adjust the file size if it changed.
1711          */
1712         if (w == XFS_DATA_FORK && XFS_DIR_IS_V1(mp)) {
1713                 ASSERT(mapi == 1);
1714                 if ((error = xfs_bmap_last_offset(tp, dp, &bno, w)))
1715                         return error;
1716                 size = XFS_FSB_TO_B(mp, bno);
1717                 if (size != dp->i_d.di_size) {
1718                         dp->i_d.di_size = size;
1719                         xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
1720                 }
1721         }
1722         return 0;
1723 }
1724
1725 /*
1726  * Ick.  We need to always be able to remove a btree block, even
1727  * if there's no space reservation because the filesystem is full.
1728  * This is called if xfs_bunmapi on a btree block fails due to ENOSPC.
1729  * It swaps the target block with the last block in the file.  The
1730  * last block in the file can always be removed since it can't cause
1731  * a bmap btree split to do that.
1732  */
1733 STATIC int
1734 xfs_da_swap_lastblock(xfs_da_args_t *args, xfs_dablk_t *dead_blknop,
1735                       xfs_dabuf_t **dead_bufp)
1736 {
1737         xfs_dablk_t dead_blkno, last_blkno, sib_blkno, par_blkno;
1738         xfs_dabuf_t *dead_buf, *last_buf, *sib_buf, *par_buf;
1739         xfs_fileoff_t lastoff;
1740         xfs_inode_t *ip;
1741         xfs_trans_t *tp;
1742         xfs_mount_t *mp;
1743         int error, w, entno, level, dead_level;
1744         xfs_da_blkinfo_t *dead_info, *sib_info;
1745         xfs_da_intnode_t *par_node, *dead_node;
1746         xfs_dir_leafblock_t *dead_leaf;
1747         xfs_dir2_leaf_t *dead_leaf2;
1748         xfs_dahash_t dead_hash;
1749
1750         dead_buf = *dead_bufp;
1751         dead_blkno = *dead_blknop;
1752         tp = args->trans;
1753         ip = args->dp;
1754         w = args->whichfork;
1755         ASSERT(w == XFS_DATA_FORK);
1756         mp = ip->i_mount;
1757         if (XFS_DIR_IS_V2(mp)) {
1758                 lastoff = mp->m_dirfreeblk;
1759                 error = xfs_bmap_last_before(tp, ip, &lastoff, w);
1760         } else
1761                 error = xfs_bmap_last_offset(tp, ip, &lastoff, w);
1762         if (error)
1763                 return error;
1764         if (unlikely(lastoff == 0)) {
1765                 XFS_ERROR_REPORT("xfs_da_swap_lastblock(1)", XFS_ERRLEVEL_LOW,
1766                                  mp);
1767                 return XFS_ERROR(EFSCORRUPTED);
1768         }
1769         /*
1770          * Read the last block in the btree space.
1771          */
1772         last_blkno = (xfs_dablk_t)lastoff - mp->m_dirblkfsbs;
1773         if ((error = xfs_da_read_buf(tp, ip, last_blkno, -1, &last_buf, w)))
1774                 return error;
1775         /*
1776          * Copy the last block into the dead buffer and log it.
1777          */
1778         memcpy(dead_buf->data, last_buf->data, mp->m_dirblksize);
1779         xfs_da_log_buf(tp, dead_buf, 0, mp->m_dirblksize - 1);
1780         dead_info = dead_buf->data;
1781         /*
1782          * Get values from the moved block.
1783          */
1784         if (be16_to_cpu(dead_info->magic) == XFS_DIR_LEAF_MAGIC) {
1785                 ASSERT(XFS_DIR_IS_V1(mp));
1786                 dead_leaf = (xfs_dir_leafblock_t *)dead_info;
1787                 dead_level = 0;
1788                 dead_hash =
1789                         INT_GET(dead_leaf->entries[INT_GET(dead_leaf->hdr.count, ARCH_CONVERT) - 1].hashval, ARCH_CONVERT);
1790         } else if (be16_to_cpu(dead_info->magic) == XFS_DIR2_LEAFN_MAGIC) {
1791                 ASSERT(XFS_DIR_IS_V2(mp));
1792                 dead_leaf2 = (xfs_dir2_leaf_t *)dead_info;
1793                 dead_level = 0;
1794                 dead_hash = be32_to_cpu(dead_leaf2->ents[be16_to_cpu(dead_leaf2->hdr.count) - 1].hashval);
1795         } else {
1796                 ASSERT(be16_to_cpu(dead_info->magic) == XFS_DA_NODE_MAGIC);
1797                 dead_node = (xfs_da_intnode_t *)dead_info;
1798                 dead_level = be16_to_cpu(dead_node->hdr.level);
1799                 dead_hash = be32_to_cpu(dead_node->btree[be16_to_cpu(dead_node->hdr.count) - 1].hashval);
1800         }
1801         sib_buf = par_buf = NULL;
1802         /*
1803          * If the moved block has a left sibling, fix up the pointers.
1804          */
1805         if ((sib_blkno = be32_to_cpu(dead_info->back))) {
1806                 if ((error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w)))
1807                         goto done;
1808                 sib_info = sib_buf->data;
1809                 if (unlikely(
1810                     be32_to_cpu(sib_info->forw) != last_blkno ||
1811                     sib_info->magic != dead_info->magic)) {
1812                         XFS_ERROR_REPORT("xfs_da_swap_lastblock(2)",
1813                                          XFS_ERRLEVEL_LOW, mp);
1814                         error = XFS_ERROR(EFSCORRUPTED);
1815                         goto done;
1816                 }
1817                 sib_info->forw = cpu_to_be32(dead_blkno);
1818                 xfs_da_log_buf(tp, sib_buf,
1819                         XFS_DA_LOGRANGE(sib_info, &sib_info->forw,
1820                                         sizeof(sib_info->forw)));
1821                 xfs_da_buf_done(sib_buf);
1822                 sib_buf = NULL;
1823         }
1824         /*
1825          * If the moved block has a right sibling, fix up the pointers.
1826          */
1827         if ((sib_blkno = be32_to_cpu(dead_info->forw))) {
1828                 if ((error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w)))
1829                         goto done;
1830                 sib_info = sib_buf->data;
1831                 if (unlikely(
1832                        be32_to_cpu(sib_info->back) != last_blkno ||
1833                        sib_info->magic != dead_info->magic)) {
1834                         XFS_ERROR_REPORT("xfs_da_swap_lastblock(3)",
1835                                          XFS_ERRLEVEL_LOW, mp);
1836                         error = XFS_ERROR(EFSCORRUPTED);
1837                         goto done;
1838                 }
1839                 sib_info->back = cpu_to_be32(dead_blkno);
1840                 xfs_da_log_buf(tp, sib_buf,
1841                         XFS_DA_LOGRANGE(sib_info, &sib_info->back,
1842                                         sizeof(sib_info->back)));
1843                 xfs_da_buf_done(sib_buf);
1844                 sib_buf = NULL;
1845         }
1846         par_blkno = XFS_DIR_IS_V1(mp) ? 0 : mp->m_dirleafblk;
1847         level = -1;
1848         /*
1849          * Walk down the tree looking for the parent of the moved block.
1850          */
1851         for (;;) {
1852                 if ((error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w)))
1853                         goto done;
1854                 par_node = par_buf->data;
1855                 if (unlikely(
1856                     be16_to_cpu(par_node->hdr.info.magic) != XFS_DA_NODE_MAGIC ||
1857                     (level >= 0 && level != be16_to_cpu(par_node->hdr.level) + 1))) {
1858                         XFS_ERROR_REPORT("xfs_da_swap_lastblock(4)",
1859                                          XFS_ERRLEVEL_LOW, mp);
1860                         error = XFS_ERROR(EFSCORRUPTED);
1861                         goto done;
1862                 }
1863                 level = be16_to_cpu(par_node->hdr.level);
1864                 for (entno = 0;
1865                      entno < be16_to_cpu(par_node->hdr.count) &&
1866                      be32_to_cpu(par_node->btree[entno].hashval) < dead_hash;
1867                      entno++)
1868                         continue;
1869                 if (unlikely(entno == be16_to_cpu(par_node->hdr.count))) {
1870                         XFS_ERROR_REPORT("xfs_da_swap_lastblock(5)",
1871                                          XFS_ERRLEVEL_LOW, mp);
1872                         error = XFS_ERROR(EFSCORRUPTED);
1873                         goto done;
1874                 }
1875                 par_blkno = be32_to_cpu(par_node->btree[entno].before);
1876                 if (level == dead_level + 1)
1877                         break;
1878                 xfs_da_brelse(tp, par_buf);
1879                 par_buf = NULL;
1880         }
1881         /*
1882          * We're in the right parent block.
1883          * Look for the right entry.
1884          */
1885         for (;;) {
1886                 for (;
1887                      entno < be16_to_cpu(par_node->hdr.count) &&
1888                      be32_to_cpu(par_node->btree[entno].before) != last_blkno;
1889                      entno++)
1890                         continue;
1891                 if (entno < be16_to_cpu(par_node->hdr.count))
1892                         break;
1893                 par_blkno = be32_to_cpu(par_node->hdr.info.forw);
1894                 xfs_da_brelse(tp, par_buf);
1895                 par_buf = NULL;
1896                 if (unlikely(par_blkno == 0)) {
1897                         XFS_ERROR_REPORT("xfs_da_swap_lastblock(6)",
1898                                          XFS_ERRLEVEL_LOW, mp);
1899                         error = XFS_ERROR(EFSCORRUPTED);
1900                         goto done;
1901                 }
1902                 if ((error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w)))
1903                         goto done;
1904                 par_node = par_buf->data;
1905                 if (unlikely(
1906                     be16_to_cpu(par_node->hdr.level) != level ||
1907                     be16_to_cpu(par_node->hdr.info.magic) != XFS_DA_NODE_MAGIC)) {
1908                         XFS_ERROR_REPORT("xfs_da_swap_lastblock(7)",
1909                                          XFS_ERRLEVEL_LOW, mp);
1910                         error = XFS_ERROR(EFSCORRUPTED);
1911                         goto done;
1912                 }
1913                 entno = 0;
1914         }
1915         /*
1916          * Update the parent entry pointing to the moved block.
1917          */
1918         par_node->btree[entno].before = cpu_to_be32(dead_blkno);
1919         xfs_da_log_buf(tp, par_buf,
1920                 XFS_DA_LOGRANGE(par_node, &par_node->btree[entno].before,
1921                                 sizeof(par_node->btree[entno].before)));
1922         xfs_da_buf_done(par_buf);
1923         xfs_da_buf_done(dead_buf);
1924         *dead_blknop = last_blkno;
1925         *dead_bufp = last_buf;
1926         return 0;
1927 done:
1928         if (par_buf)
1929                 xfs_da_brelse(tp, par_buf);
1930         if (sib_buf)
1931                 xfs_da_brelse(tp, sib_buf);
1932         xfs_da_brelse(tp, last_buf);
1933         return error;
1934 }
1935
1936 /*
1937  * Remove a btree block from a directory or attribute.
1938  */
1939 int
1940 xfs_da_shrink_inode(xfs_da_args_t *args, xfs_dablk_t dead_blkno,
1941                     xfs_dabuf_t *dead_buf)
1942 {
1943         xfs_inode_t *dp;
1944         int done, error, w, count;
1945         xfs_fileoff_t bno;
1946         xfs_fsize_t size;
1947         xfs_trans_t *tp;
1948         xfs_mount_t *mp;
1949
1950         dp = args->dp;
1951         w = args->whichfork;
1952         tp = args->trans;
1953         mp = dp->i_mount;
1954         if (w == XFS_DATA_FORK && XFS_DIR_IS_V2(mp))
1955                 count = mp->m_dirblkfsbs;
1956         else
1957                 count = 1;
1958         for (;;) {
1959                 /*
1960                  * Remove extents.  If we get ENOSPC for a dir we have to move
1961                  * the last block to the place we want to kill.
1962                  */
1963                 if ((error = xfs_bunmapi(tp, dp, dead_blkno, count,
1964                                 XFS_BMAPI_AFLAG(w)|XFS_BMAPI_METADATA,
1965                                 0, args->firstblock, args->flist, NULL,
1966                                 &done)) == ENOSPC) {
1967                         if (w != XFS_DATA_FORK)
1968                                 goto done;
1969                         if ((error = xfs_da_swap_lastblock(args, &dead_blkno,
1970                                         &dead_buf)))
1971                                 goto done;
1972                 } else if (error)
1973                         goto done;
1974                 else
1975                         break;
1976         }
1977         ASSERT(done);
1978         xfs_da_binval(tp, dead_buf);
1979         /*
1980          * Adjust the directory size for version 1.
1981          */
1982         if (w == XFS_DATA_FORK && XFS_DIR_IS_V1(mp)) {
1983                 if ((error = xfs_bmap_last_offset(tp, dp, &bno, w)))
1984                         return error;
1985                 size = XFS_FSB_TO_B(dp->i_mount, bno);
1986                 if (size != dp->i_d.di_size) {
1987                         dp->i_d.di_size = size;
1988                         xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
1989                 }
1990         }
1991         return 0;
1992 done:
1993         xfs_da_binval(tp, dead_buf);
1994         return error;
1995 }
1996
1997 /*
1998  * See if the mapping(s) for this btree block are valid, i.e.
1999  * don't contain holes, are logically contiguous, and cover the whole range.
2000  */
2001 STATIC int
2002 xfs_da_map_covers_blocks(
2003         int             nmap,
2004         xfs_bmbt_irec_t *mapp,
2005         xfs_dablk_t     bno,
2006         int             count)
2007 {
2008         int             i;
2009         xfs_fileoff_t   off;
2010
2011         for (i = 0, off = bno; i < nmap; i++) {
2012                 if (mapp[i].br_startblock == HOLESTARTBLOCK ||
2013                     mapp[i].br_startblock == DELAYSTARTBLOCK) {
2014                         return 0;
2015                 }
2016                 if (off != mapp[i].br_startoff) {
2017                         return 0;
2018                 }
2019                 off += mapp[i].br_blockcount;
2020         }
2021         return off == bno + count;
2022 }
2023
2024 /*
2025  * Make a dabuf.
2026  * Used for get_buf, read_buf, read_bufr, and reada_buf.
2027  */
2028 STATIC int
2029 xfs_da_do_buf(
2030         xfs_trans_t     *trans,
2031         xfs_inode_t     *dp,
2032         xfs_dablk_t     bno,
2033         xfs_daddr_t     *mappedbnop,
2034         xfs_dabuf_t     **bpp,
2035         int             whichfork,
2036         int             caller,
2037         inst_t          *ra)
2038 {
2039         xfs_buf_t       *bp = NULL;
2040         xfs_buf_t       **bplist;
2041         int             error=0;
2042         int             i;
2043         xfs_bmbt_irec_t map;
2044         xfs_bmbt_irec_t *mapp;
2045         xfs_daddr_t     mappedbno;
2046         xfs_mount_t     *mp;
2047         int             nbplist=0;
2048         int             nfsb;
2049         int             nmap;
2050         xfs_dabuf_t     *rbp;
2051
2052         mp = dp->i_mount;
2053         if (whichfork == XFS_DATA_FORK && XFS_DIR_IS_V2(mp))
2054                 nfsb = mp->m_dirblkfsbs;
2055         else
2056                 nfsb = 1;
2057         mappedbno = *mappedbnop;
2058         /*
2059          * Caller doesn't have a mapping.  -2 means don't complain
2060          * if we land in a hole.
2061          */
2062         if (mappedbno == -1 || mappedbno == -2) {
2063                 /*
2064                  * Optimize the one-block case.
2065                  */
2066                 if (nfsb == 1) {
2067                         xfs_fsblock_t   fsb;
2068
2069                         if ((error =
2070                             xfs_bmapi_single(trans, dp, whichfork, &fsb,
2071                                     (xfs_fileoff_t)bno))) {
2072                                 return error;
2073                         }
2074                         mapp = &map;
2075                         if (fsb == NULLFSBLOCK) {
2076                                 nmap = 0;
2077                         } else {
2078                                 map.br_startblock = fsb;
2079                                 map.br_startoff = (xfs_fileoff_t)bno;
2080                                 map.br_blockcount = 1;
2081                                 nmap = 1;
2082                         }
2083                 } else {
2084                         mapp = kmem_alloc(sizeof(*mapp) * nfsb, KM_SLEEP);
2085                         nmap = nfsb;
2086                         if ((error = xfs_bmapi(trans, dp, (xfs_fileoff_t)bno,
2087                                         nfsb,
2088                                         XFS_BMAPI_METADATA |
2089                                                 XFS_BMAPI_AFLAG(whichfork),
2090                                         NULL, 0, mapp, &nmap, NULL, NULL)))
2091                                 goto exit0;
2092                 }
2093         } else {
2094                 map.br_startblock = XFS_DADDR_TO_FSB(mp, mappedbno);
2095                 map.br_startoff = (xfs_fileoff_t)bno;
2096                 map.br_blockcount = nfsb;
2097                 mapp = &map;
2098                 nmap = 1;
2099         }
2100         if (!xfs_da_map_covers_blocks(nmap, mapp, bno, nfsb)) {
2101                 error = mappedbno == -2 ? 0 : XFS_ERROR(EFSCORRUPTED);
2102                 if (unlikely(error == EFSCORRUPTED)) {
2103                         if (xfs_error_level >= XFS_ERRLEVEL_LOW) {
2104                                 int     i;
2105                                 cmn_err(CE_ALERT, "xfs_da_do_buf: bno %lld\n",
2106                                         (long long)bno);
2107                                 cmn_err(CE_ALERT, "dir: inode %lld\n",
2108                                         (long long)dp->i_ino);
2109                                 for (i = 0; i < nmap; i++) {
2110                                         cmn_err(CE_ALERT,
2111                                                 "[%02d] br_startoff %lld br_startblock %lld br_blockcount %lld br_state %d\n",
2112                                                 i,
2113                                                 (long long)mapp[i].br_startoff,
2114                                                 (long long)mapp[i].br_startblock,
2115                                                 (long long)mapp[i].br_blockcount,
2116                                                 mapp[i].br_state);
2117                                 }
2118                         }
2119                         XFS_ERROR_REPORT("xfs_da_do_buf(1)",
2120                                          XFS_ERRLEVEL_LOW, mp);
2121                 }
2122                 goto exit0;
2123         }
2124         if (caller != 3 && nmap > 1) {
2125                 bplist = kmem_alloc(sizeof(*bplist) * nmap, KM_SLEEP);
2126                 nbplist = 0;
2127         } else
2128                 bplist = NULL;
2129         /*
2130          * Turn the mapping(s) into buffer(s).
2131          */
2132         for (i = 0; i < nmap; i++) {
2133                 int     nmapped;
2134
2135                 mappedbno = XFS_FSB_TO_DADDR(mp, mapp[i].br_startblock);
2136                 if (i == 0)
2137                         *mappedbnop = mappedbno;
2138                 nmapped = (int)XFS_FSB_TO_BB(mp, mapp[i].br_blockcount);
2139                 switch (caller) {
2140                 case 0:
2141                         bp = xfs_trans_get_buf(trans, mp->m_ddev_targp,
2142                                 mappedbno, nmapped, 0);
2143                         error = bp ? XFS_BUF_GETERROR(bp) : XFS_ERROR(EIO);
2144                         break;
2145                 case 1:
2146                 case 2:
2147                         bp = NULL;
2148                         error = xfs_trans_read_buf(mp, trans, mp->m_ddev_targp,
2149                                 mappedbno, nmapped, 0, &bp);
2150                         break;
2151                 case 3:
2152                         xfs_baread(mp->m_ddev_targp, mappedbno, nmapped);
2153                         error = 0;
2154                         bp = NULL;
2155                         break;
2156                 }
2157                 if (error) {
2158                         if (bp)
2159                                 xfs_trans_brelse(trans, bp);
2160                         goto exit1;
2161                 }
2162                 if (!bp)
2163                         continue;
2164                 if (caller == 1) {
2165                         if (whichfork == XFS_ATTR_FORK) {
2166                                 XFS_BUF_SET_VTYPE_REF(bp, B_FS_ATTR_BTREE,
2167                                                 XFS_ATTR_BTREE_REF);
2168                         } else {
2169                                 XFS_BUF_SET_VTYPE_REF(bp, B_FS_DIR_BTREE,
2170                                                 XFS_DIR_BTREE_REF);
2171                         }
2172                 }
2173                 if (bplist) {
2174                         bplist[nbplist++] = bp;
2175                 }
2176         }
2177         /*
2178          * Build a dabuf structure.
2179          */
2180         if (bplist) {
2181                 rbp = xfs_da_buf_make(nbplist, bplist, ra);
2182         } else if (bp)
2183                 rbp = xfs_da_buf_make(1, &bp, ra);
2184         else
2185                 rbp = NULL;
2186         /*
2187          * For read_buf, check the magic number.
2188          */
2189         if (caller == 1) {
2190                 xfs_dir2_data_t         *data;
2191                 xfs_dir2_free_t         *free;
2192                 xfs_da_blkinfo_t        *info;
2193                 uint                    magic, magic1;
2194
2195                 info = rbp->data;
2196                 data = rbp->data;
2197                 free = rbp->data;
2198                 magic = be16_to_cpu(info->magic);
2199                 magic1 = be32_to_cpu(data->hdr.magic);
2200                 if (unlikely(
2201                     XFS_TEST_ERROR((magic != XFS_DA_NODE_MAGIC) &&
2202                                    (magic != XFS_DIR_LEAF_MAGIC) &&
2203                                    (magic != XFS_ATTR_LEAF_MAGIC) &&
2204                                    (magic != XFS_DIR2_LEAF1_MAGIC) &&
2205                                    (magic != XFS_DIR2_LEAFN_MAGIC) &&
2206                                    (magic1 != XFS_DIR2_BLOCK_MAGIC) &&
2207                                    (magic1 != XFS_DIR2_DATA_MAGIC) &&
2208                                    (be32_to_cpu(free->hdr.magic) != XFS_DIR2_FREE_MAGIC),
2209                                 mp, XFS_ERRTAG_DA_READ_BUF,
2210                                 XFS_RANDOM_DA_READ_BUF))) {
2211                         xfs_buftrace("DA READ ERROR", rbp->bps[0]);
2212                         XFS_CORRUPTION_ERROR("xfs_da_do_buf(2)",
2213                                              XFS_ERRLEVEL_LOW, mp, info);
2214                         error = XFS_ERROR(EFSCORRUPTED);
2215                         xfs_da_brelse(trans, rbp);
2216                         nbplist = 0;
2217                         goto exit1;
2218                 }
2219         }
2220         if (bplist) {
2221                 kmem_free(bplist, sizeof(*bplist) * nmap);
2222         }
2223         if (mapp != &map) {
2224                 kmem_free(mapp, sizeof(*mapp) * nfsb);
2225         }
2226         if (bpp)
2227                 *bpp = rbp;
2228         return 0;
2229 exit1:
2230         if (bplist) {
2231                 for (i = 0; i < nbplist; i++)
2232                         xfs_trans_brelse(trans, bplist[i]);
2233                 kmem_free(bplist, sizeof(*bplist) * nmap);
2234         }
2235 exit0:
2236         if (mapp != &map)
2237                 kmem_free(mapp, sizeof(*mapp) * nfsb);
2238         if (bpp)
2239                 *bpp = NULL;
2240         return error;
2241 }
2242
2243 /*
2244  * Get a buffer for the dir/attr block.
2245  */
2246 int
2247 xfs_da_get_buf(
2248         xfs_trans_t     *trans,
2249         xfs_inode_t     *dp,
2250         xfs_dablk_t     bno,
2251         xfs_daddr_t             mappedbno,
2252         xfs_dabuf_t     **bpp,
2253         int             whichfork)
2254 {
2255         return xfs_da_do_buf(trans, dp, bno, &mappedbno, bpp, whichfork, 0,
2256                                                  (inst_t *)__return_address);
2257 }
2258
2259 /*
2260  * Get a buffer for the dir/attr block, fill in the contents.
2261  */
2262 int
2263 xfs_da_read_buf(
2264         xfs_trans_t     *trans,
2265         xfs_inode_t     *dp,
2266         xfs_dablk_t     bno,
2267         xfs_daddr_t             mappedbno,
2268         xfs_dabuf_t     **bpp,
2269         int             whichfork)
2270 {
2271         return xfs_da_do_buf(trans, dp, bno, &mappedbno, bpp, whichfork, 1,
2272                 (inst_t *)__return_address);
2273 }
2274
2275 /*
2276  * Readahead the dir/attr block.
2277  */
2278 xfs_daddr_t
2279 xfs_da_reada_buf(
2280         xfs_trans_t     *trans,
2281         xfs_inode_t     *dp,
2282         xfs_dablk_t     bno,
2283         int             whichfork)
2284 {
2285         xfs_daddr_t             rval;
2286
2287         rval = -1;
2288         if (xfs_da_do_buf(trans, dp, bno, &rval, NULL, whichfork, 3,
2289                         (inst_t *)__return_address))
2290                 return -1;
2291         else
2292                 return rval;
2293 }
2294
2295 /*
2296  * Calculate the number of bits needed to hold i different values.
2297  */
2298 uint
2299 xfs_da_log2_roundup(uint i)
2300 {
2301         uint rval;
2302
2303         for (rval = 0; rval < NBBY * sizeof(i); rval++) {
2304                 if ((1 << rval) >= i)
2305                         break;
2306         }
2307         return(rval);
2308 }
2309
2310 kmem_zone_t *xfs_da_state_zone; /* anchor for state struct zone */
2311 kmem_zone_t *xfs_dabuf_zone;            /* dabuf zone */
2312
2313 /*
2314  * Allocate a dir-state structure.
2315  * We don't put them on the stack since they're large.
2316  */
2317 xfs_da_state_t *
2318 xfs_da_state_alloc(void)
2319 {
2320         return kmem_zone_zalloc(xfs_da_state_zone, KM_SLEEP);
2321 }
2322
2323 /*
2324  * Kill the altpath contents of a da-state structure.
2325  */
2326 STATIC void
2327 xfs_da_state_kill_altpath(xfs_da_state_t *state)
2328 {
2329         int     i;
2330
2331         for (i = 0; i < state->altpath.active; i++) {
2332                 if (state->altpath.blk[i].bp) {
2333                         if (state->altpath.blk[i].bp != state->path.blk[i].bp)
2334                                 xfs_da_buf_done(state->altpath.blk[i].bp);
2335                         state->altpath.blk[i].bp = NULL;
2336                 }
2337         }
2338         state->altpath.active = 0;
2339 }
2340
2341 /*
2342  * Free a da-state structure.
2343  */
2344 void
2345 xfs_da_state_free(xfs_da_state_t *state)
2346 {
2347         int     i;
2348
2349         xfs_da_state_kill_altpath(state);
2350         for (i = 0; i < state->path.active; i++) {
2351                 if (state->path.blk[i].bp)
2352                         xfs_da_buf_done(state->path.blk[i].bp);
2353         }
2354         if (state->extravalid && state->extrablk.bp)
2355                 xfs_da_buf_done(state->extrablk.bp);
2356 #ifdef DEBUG
2357         memset((char *)state, 0, sizeof(*state));
2358 #endif /* DEBUG */
2359         kmem_zone_free(xfs_da_state_zone, state);
2360 }
2361
2362 #ifdef XFS_DABUF_DEBUG
2363 xfs_dabuf_t     *xfs_dabuf_global_list;
2364 lock_t          xfs_dabuf_global_lock;
2365 #endif
2366
2367 /*
2368  * Create a dabuf.
2369  */
2370 /* ARGSUSED */
2371 STATIC xfs_dabuf_t *
2372 xfs_da_buf_make(int nbuf, xfs_buf_t **bps, inst_t *ra)
2373 {
2374         xfs_buf_t       *bp;
2375         xfs_dabuf_t     *dabuf;
2376         int             i;
2377         int             off;
2378
2379         if (nbuf == 1)
2380                 dabuf = kmem_zone_alloc(xfs_dabuf_zone, KM_SLEEP);
2381         else
2382                 dabuf = kmem_alloc(XFS_DA_BUF_SIZE(nbuf), KM_SLEEP);
2383         dabuf->dirty = 0;
2384 #ifdef XFS_DABUF_DEBUG
2385         dabuf->ra = ra;
2386         dabuf->target = XFS_BUF_TARGET(bps[0]);
2387         dabuf->blkno = XFS_BUF_ADDR(bps[0]);
2388 #endif
2389         if (nbuf == 1) {
2390                 dabuf->nbuf = 1;
2391                 bp = bps[0];
2392                 dabuf->bbcount = (short)BTOBB(XFS_BUF_COUNT(bp));
2393                 dabuf->data = XFS_BUF_PTR(bp);
2394                 dabuf->bps[0] = bp;
2395         } else {
2396                 dabuf->nbuf = nbuf;
2397                 for (i = 0, dabuf->bbcount = 0; i < nbuf; i++) {
2398                         dabuf->bps[i] = bp = bps[i];
2399                         dabuf->bbcount += BTOBB(XFS_BUF_COUNT(bp));
2400                 }
2401                 dabuf->data = kmem_alloc(BBTOB(dabuf->bbcount), KM_SLEEP);
2402                 for (i = off = 0; i < nbuf; i++, off += XFS_BUF_COUNT(bp)) {
2403                         bp = bps[i];
2404                         memcpy((char *)dabuf->data + off, XFS_BUF_PTR(bp),
2405                                 XFS_BUF_COUNT(bp));
2406                 }
2407         }
2408 #ifdef XFS_DABUF_DEBUG
2409         {
2410                 SPLDECL(s);
2411                 xfs_dabuf_t     *p;
2412
2413                 s = mutex_spinlock(&xfs_dabuf_global_lock);
2414                 for (p = xfs_dabuf_global_list; p; p = p->next) {
2415                         ASSERT(p->blkno != dabuf->blkno ||
2416                                p->target != dabuf->target);
2417                 }
2418                 dabuf->prev = NULL;
2419                 if (xfs_dabuf_global_list)
2420                         xfs_dabuf_global_list->prev = dabuf;
2421                 dabuf->next = xfs_dabuf_global_list;
2422                 xfs_dabuf_global_list = dabuf;
2423                 mutex_spinunlock(&xfs_dabuf_global_lock, s);
2424         }
2425 #endif
2426         return dabuf;
2427 }
2428
2429 /*
2430  * Un-dirty a dabuf.
2431  */
2432 STATIC void
2433 xfs_da_buf_clean(xfs_dabuf_t *dabuf)
2434 {
2435         xfs_buf_t       *bp;
2436         int             i;
2437         int             off;
2438
2439         if (dabuf->dirty) {
2440                 ASSERT(dabuf->nbuf > 1);
2441                 dabuf->dirty = 0;
2442                 for (i = off = 0; i < dabuf->nbuf;
2443                                 i++, off += XFS_BUF_COUNT(bp)) {
2444                         bp = dabuf->bps[i];
2445                         memcpy(XFS_BUF_PTR(bp), (char *)dabuf->data + off,
2446                                 XFS_BUF_COUNT(bp));
2447                 }
2448         }
2449 }
2450
2451 /*
2452  * Release a dabuf.
2453  */
2454 void
2455 xfs_da_buf_done(xfs_dabuf_t *dabuf)
2456 {
2457         ASSERT(dabuf);
2458         ASSERT(dabuf->nbuf && dabuf->data && dabuf->bbcount && dabuf->bps[0]);
2459         if (dabuf->dirty)
2460                 xfs_da_buf_clean(dabuf);
2461         if (dabuf->nbuf > 1)
2462                 kmem_free(dabuf->data, BBTOB(dabuf->bbcount));
2463 #ifdef XFS_DABUF_DEBUG
2464         {
2465                 SPLDECL(s);
2466
2467                 s = mutex_spinlock(&xfs_dabuf_global_lock);
2468                 if (dabuf->prev)
2469                         dabuf->prev->next = dabuf->next;
2470                 else
2471                         xfs_dabuf_global_list = dabuf->next;
2472                 if (dabuf->next)
2473                         dabuf->next->prev = dabuf->prev;
2474                 mutex_spinunlock(&xfs_dabuf_global_lock, s);
2475         }
2476         memset(dabuf, 0, XFS_DA_BUF_SIZE(dabuf->nbuf));
2477 #endif
2478         if (dabuf->nbuf == 1)
2479                 kmem_zone_free(xfs_dabuf_zone, dabuf);
2480         else
2481                 kmem_free(dabuf, XFS_DA_BUF_SIZE(dabuf->nbuf));
2482 }
2483
2484 /*
2485  * Log transaction from a dabuf.
2486  */
2487 void
2488 xfs_da_log_buf(xfs_trans_t *tp, xfs_dabuf_t *dabuf, uint first, uint last)
2489 {
2490         xfs_buf_t       *bp;
2491         uint            f;
2492         int             i;
2493         uint            l;
2494         int             off;
2495
2496         ASSERT(dabuf->nbuf && dabuf->data && dabuf->bbcount && dabuf->bps[0]);
2497         if (dabuf->nbuf == 1) {
2498                 ASSERT(dabuf->data == (void *)XFS_BUF_PTR(dabuf->bps[0]));
2499                 xfs_trans_log_buf(tp, dabuf->bps[0], first, last);
2500                 return;
2501         }
2502         dabuf->dirty = 1;
2503         ASSERT(first <= last);
2504         for (i = off = 0; i < dabuf->nbuf; i++, off += XFS_BUF_COUNT(bp)) {
2505                 bp = dabuf->bps[i];
2506                 f = off;
2507                 l = f + XFS_BUF_COUNT(bp) - 1;
2508                 if (f < first)
2509                         f = first;
2510                 if (l > last)
2511                         l = last;
2512                 if (f <= l)
2513                         xfs_trans_log_buf(tp, bp, f - off, l - off);
2514                 /*
2515                  * B_DONE is set by xfs_trans_log buf.
2516                  * If we don't set it on a new buffer (get not read)
2517                  * then if we don't put anything in the buffer it won't
2518                  * be set, and at commit it it released into the cache,
2519                  * and then a read will fail.
2520                  */
2521                 else if (!(XFS_BUF_ISDONE(bp)))
2522                   XFS_BUF_DONE(bp);
2523         }
2524         ASSERT(last < off);
2525 }
2526
2527 /*
2528  * Release dabuf from a transaction.
2529  * Have to free up the dabuf before the buffers are released,
2530  * since the synchronization on the dabuf is really the lock on the buffer.
2531  */
2532 void
2533 xfs_da_brelse(xfs_trans_t *tp, xfs_dabuf_t *dabuf)
2534 {
2535         xfs_buf_t       *bp;
2536         xfs_buf_t       **bplist;
2537         int             i;
2538         int             nbuf;
2539
2540         ASSERT(dabuf->nbuf && dabuf->data && dabuf->bbcount && dabuf->bps[0]);
2541         if ((nbuf = dabuf->nbuf) == 1) {
2542                 bplist = &bp;
2543                 bp = dabuf->bps[0];
2544         } else {
2545                 bplist = kmem_alloc(nbuf * sizeof(*bplist), KM_SLEEP);
2546                 memcpy(bplist, dabuf->bps, nbuf * sizeof(*bplist));
2547         }
2548         xfs_da_buf_done(dabuf);
2549         for (i = 0; i < nbuf; i++)
2550                 xfs_trans_brelse(tp, bplist[i]);
2551         if (bplist != &bp)
2552                 kmem_free(bplist, nbuf * sizeof(*bplist));
2553 }
2554
2555 /*
2556  * Invalidate dabuf from a transaction.
2557  */
2558 void
2559 xfs_da_binval(xfs_trans_t *tp, xfs_dabuf_t *dabuf)
2560 {
2561         xfs_buf_t       *bp;
2562         xfs_buf_t       **bplist;
2563         int             i;
2564         int             nbuf;
2565
2566         ASSERT(dabuf->nbuf && dabuf->data && dabuf->bbcount && dabuf->bps[0]);
2567         if ((nbuf = dabuf->nbuf) == 1) {
2568                 bplist = &bp;
2569                 bp = dabuf->bps[0];
2570         } else {
2571                 bplist = kmem_alloc(nbuf * sizeof(*bplist), KM_SLEEP);
2572                 memcpy(bplist, dabuf->bps, nbuf * sizeof(*bplist));
2573         }
2574         xfs_da_buf_done(dabuf);
2575         for (i = 0; i < nbuf; i++)
2576                 xfs_trans_binval(tp, bplist[i]);
2577         if (bplist != &bp)
2578                 kmem_free(bplist, nbuf * sizeof(*bplist));
2579 }
2580
2581 /*
2582  * Get the first daddr from a dabuf.
2583  */
2584 xfs_daddr_t
2585 xfs_da_blkno(xfs_dabuf_t *dabuf)
2586 {
2587         ASSERT(dabuf->nbuf);
2588         ASSERT(dabuf->data);
2589         return XFS_BUF_ADDR(dabuf->bps[0]);
2590 }