git.oblomov.eu Git - linux-2.6/blob - fs/ocfs2/extent_map.c

   1 /* -*- mode: c; c-basic-offset: 8; -*-
   2  * vim: noexpandtab sw=8 ts=8 sts=0:
   3  *
   4  * extent_map.c
   5  *
   6  * In-memory extent map for OCFS2.  Man, this code was prettier in
   7  * the library.
   8  *
   9  * Copyright (C) 2004 Oracle.  All rights reserved.
  10  *
  11  * This program is free software; you can redistribute it and/or
  12  * modify it under the terms of the GNU General Public
  13  * License, version 2,  as published by the Free Software Foundation.
  14  *
  15  * This program is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU General Public
  21  * License along with this program; if not, write to the
  22  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  23  * Boston, MA 021110-1307, USA.
  24  */
  25
  26 #include <linux/fs.h>
  27 #include <linux/init.h>
  28 #include <linux/types.h>
  29 #include <linux/slab.h>
  30 #include <linux/rbtree.h>
  31
  32 #define MLOG_MASK_PREFIX ML_EXTENT_MAP
  33 #include <cluster/masklog.h>
  34
  35 #include "ocfs2.h"
  36
  37 #include "extent_map.h"
  38 #include "inode.h"
  39 #include "super.h"
  40
  41 #include "buffer_head_io.h"
  42
  43
  44 /*
  45  * SUCK SUCK SUCK
  46  * Our headers are so bad that struct ocfs2_extent_map is in ocfs.h
  47  */
  48
  49 struct ocfs2_extent_map_entry {
  50         struct rb_node e_node;
  51         int e_tree_depth;
  52         struct ocfs2_extent_rec e_rec;
  53 };
  54
  55 struct ocfs2_em_insert_context {
  56         int need_left;
  57         int need_right;
  58         struct ocfs2_extent_map_entry *new_ent;
  59         struct ocfs2_extent_map_entry *old_ent;
  60         struct ocfs2_extent_map_entry *left_ent;
  61         struct ocfs2_extent_map_entry *right_ent;
  62 };
  63
  64 static kmem_cache_t *ocfs2_em_ent_cachep = NULL;
  65
  66
  67 static struct ocfs2_extent_map_entry *
  68 ocfs2_extent_map_lookup(struct ocfs2_extent_map *em,
  69                         u32 cpos, u32 clusters,
  70                         struct rb_node ***ret_p,
  71                         struct rb_node **ret_parent);
  72 static int ocfs2_extent_map_insert(struct inode *inode,
  73                                    struct ocfs2_extent_rec *rec,
  74                                    int tree_depth);
  75 static int ocfs2_extent_map_insert_entry(struct ocfs2_extent_map *em,
  76                                          struct ocfs2_extent_map_entry *ent);
  77 static int ocfs2_extent_map_find_leaf(struct inode *inode,
  78                                       u32 cpos, u32 clusters,
  79                                       struct ocfs2_extent_list *el);
  80 static int ocfs2_extent_map_lookup_read(struct inode *inode,
  81                                         u32 cpos, u32 clusters,
  82                                         struct ocfs2_extent_map_entry **ret_ent);
  83 static int ocfs2_extent_map_try_insert(struct inode *inode,
  84                                        struct ocfs2_extent_rec *rec,
  85                                        int tree_depth,
  86                                        struct ocfs2_em_insert_context *ctxt);
  87
  88 /* returns 1 only if the rec contains all the given clusters -- that is that
  89  * rec's cpos is <= the cluster cpos and that the rec endpoint (cpos +
  90  * clusters) is >= the argument's endpoint */
  91 static int ocfs2_extent_rec_contains_clusters(struct ocfs2_extent_rec *rec,
  92                                               u32 cpos, u32 clusters)
  93 {
  94         if (le32_to_cpu(rec->e_cpos) > cpos)
  95                 return 0;
  96         if (cpos + clusters > le32_to_cpu(rec->e_cpos) +
  97                               le32_to_cpu(rec->e_clusters))
  98                 return 0;
  99         return 1;
 100 }
 101
 102
 103 /*
 104  * Find an entry in the tree that intersects the region passed in.
 105  * Note that this will find straddled intervals, it is up to the
 106  * callers to enforce any boundary conditions.
 107  *
 108  * Callers must hold ip_lock.  This lookup is not guaranteed to return
 109  * a tree_depth 0 match, and as such can race inserts if the lock
 110  * were not held.
 111  *
 112  * The rb_node garbage lets insertion share the search.  Trivial
 113  * callers pass NULL.
 114  */
 115 static struct ocfs2_extent_map_entry *
 116 ocfs2_extent_map_lookup(struct ocfs2_extent_map *em,
 117                         u32 cpos, u32 clusters,
 118                         struct rb_node ***ret_p,
 119                         struct rb_node **ret_parent)
 120 {
 121         struct rb_node **p = &em->em_extents.rb_node;
 122         struct rb_node *parent = NULL;
 123         struct ocfs2_extent_map_entry *ent = NULL;
 124
 125         while (*p)
 126         {
 127                 parent = *p;
 128                 ent = rb_entry(parent, struct ocfs2_extent_map_entry,
 129                                e_node);
 130                 if ((cpos + clusters) <= le32_to_cpu(ent->e_rec.e_cpos)) {
 131                         p = &(*p)->rb_left;
 132                         ent = NULL;
 133                 } else if (cpos >= (le32_to_cpu(ent->e_rec.e_cpos) +
 134                                     le32_to_cpu(ent->e_rec.e_clusters))) {
 135                         p = &(*p)->rb_right;
 136                         ent = NULL;
 137                 } else
 138                         break;
 139         }
 140
 141         if (ret_p != NULL)
 142                 *ret_p = p;
 143         if (ret_parent != NULL)
 144                 *ret_parent = parent;
 145         return ent;
 146 }
 147
 148 /*
 149  * Find the leaf containing the interval we want.  While we're on our
 150  * way down the tree, fill in every record we see at any depth, because
 151  * we might want it later.
 152  *
 153  * Note that this code is run without ip_lock.  That's because it
 154  * sleeps while reading.  If someone is also filling the extent list at
 155  * the same time we are, we might have to restart.
 156  */
 157 static int ocfs2_extent_map_find_leaf(struct inode *inode,
 158                                       u32 cpos, u32 clusters,
 159                                       struct ocfs2_extent_list *el)
 160 {
 161         int i, ret;
 162         struct buffer_head *eb_bh = NULL;
 163         u64 blkno;
 164         u32 rec_end;
 165         struct ocfs2_extent_block *eb;
 166         struct ocfs2_extent_rec *rec;
 167
 168         /*
 169          * The bh data containing the el cannot change here, because
 170          * we hold alloc_sem.  So we can do this without other
 171          * locks.
 172          */
 173         while (el->l_tree_depth)
 174         {
 175                 blkno = 0;
 176                 for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
 177                         rec = &el->l_recs[i];
 178                         rec_end = (le32_to_cpu(rec->e_cpos) +
 179                                    le32_to_cpu(rec->e_clusters));
 180
 181                         ret = -EBADR;
 182                         if (rec_end > OCFS2_I(inode)->ip_clusters) {
 183                                 mlog_errno(ret);
 184                                 goto out_free;
 185                         }
 186
 187                         if (rec_end <= cpos) {
 188                                 ret = ocfs2_extent_map_insert(inode, rec,
 189                                                 le16_to_cpu(el->l_tree_depth));
 190                                 if (ret && (ret != -EEXIST)) {
 191                                         mlog_errno(ret);
 192                                         goto out_free;
 193                                 }
 194                                 continue;
 195                         }
 196                         if ((cpos + clusters) <= le32_to_cpu(rec->e_cpos)) {
 197                                 ret = ocfs2_extent_map_insert(inode, rec,
 198                                                 le16_to_cpu(el->l_tree_depth));
 199                                 if (ret && (ret != -EEXIST)) {
 200                                         mlog_errno(ret);
 201                                         goto out_free;
 202                                 }
 203                                 continue;
 204                         }
 205
 206                         /*
 207                          * We've found a record that matches our
 208                          * interval.  We don't insert it because we're
 209                          * about to traverse it.
 210                          */
 211
 212                         /* Check to see if we're stradling */
 213                         ret = -ESRCH;
 214                         if (!ocfs2_extent_rec_contains_clusters(rec,
 215                                                                 cpos,
 216                                                                 clusters)) {
 217                                 mlog_errno(ret);
 218                                 goto out_free;
 219                         }
 220
 221                         /*
 222                          * If we've already found a record, the el has
 223                          * two records covering the same interval.
 224                          * EEEK!
 225                          */
 226                         ret = -EBADR;
 227                         if (blkno) {
 228                                 mlog_errno(ret);
 229                                 goto out_free;
 230                         }
 231
 232                         blkno = le64_to_cpu(rec->e_blkno);
 233                 }
 234
 235                 /*
 236                  * We don't support holes, and we're still up
 237                  * in the branches, so we'd better have found someone
 238                  */
 239                 ret = -EBADR;
 240                 if (!blkno) {
 241                         mlog_errno(ret);
 242                         goto out_free;
 243                 }
 244
 245                 if (eb_bh) {
 246                         brelse(eb_bh);
 247                         eb_bh = NULL;
 248                 }
 249                 ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
 250                                        blkno, &eb_bh, OCFS2_BH_CACHED,
 251                                        inode);
 252                 if (ret) {
 253                         mlog_errno(ret);
 254                         goto out_free;
 255                 }
 256                 eb = (struct ocfs2_extent_block *)eb_bh->b_data;
 257                 if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
 258                         OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
 259                         ret = -EIO;
 260                         goto out_free;
 261                 }
 262                 el = &eb->h_list;
 263         }
 264
 265         if (el->l_tree_depth)
 266                 BUG();
 267
 268         for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
 269                 rec = &el->l_recs[i];
 270                 ret = ocfs2_extent_map_insert(inode, rec,
 271                                               le16_to_cpu(el->l_tree_depth));
 272                 if (ret) {
 273                         mlog_errno(ret);
 274                         goto out_free;
 275                 }
 276         }
 277
 278         ret = 0;
 279
 280 out_free:
 281         if (eb_bh)
 282                 brelse(eb_bh);
 283
 284         return ret;
 285 }
 286
 287 /*
 288  * This lookup actually will read from disk.  It has one invariant:
 289  * It will never re-traverse blocks.  This means that all inserts should
 290  * be new regions or more granular regions (both allowed by insert).
 291  */
 292 static int ocfs2_extent_map_lookup_read(struct inode *inode,
 293                                         u32 cpos,
 294                                         u32 clusters,
 295                                         struct ocfs2_extent_map_entry **ret_ent)
 296 {
 297         int ret;
 298         u64 blkno;
 299         struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
 300         struct ocfs2_extent_map_entry *ent;
 301         struct buffer_head *bh = NULL;
 302         struct ocfs2_extent_block *eb;
 303         struct ocfs2_dinode *di;
 304         struct ocfs2_extent_list *el;
 305
 306         spin_lock(&OCFS2_I(inode)->ip_lock);
 307         ent = ocfs2_extent_map_lookup(em, cpos, clusters, NULL, NULL);
 308         if (ent) {
 309                 if (!ent->e_tree_depth) {
 310                         spin_unlock(&OCFS2_I(inode)->ip_lock);
 311                         *ret_ent = ent;
 312                         return 0;
 313                 }
 314                 blkno = le64_to_cpu(ent->e_rec.e_blkno);
 315                 spin_unlock(&OCFS2_I(inode)->ip_lock);
 316
 317                 ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), blkno, &bh,
 318                                        OCFS2_BH_CACHED, inode);
 319                 if (ret) {
 320                         mlog_errno(ret);
 321                         if (bh)
 322                                 brelse(bh);
 323                         return ret;
 324                 }
 325                 eb = (struct ocfs2_extent_block *)bh->b_data;
 326                 if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
 327                         OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
 328                         brelse(bh);
 329                         return -EIO;
 330                 }
 331                 el = &eb->h_list;
 332         } else {
 333                 spin_unlock(&OCFS2_I(inode)->ip_lock);
 334
 335                 ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
 336                                        OCFS2_I(inode)->ip_blkno, &bh,
 337                                        OCFS2_BH_CACHED, inode);
 338                 if (ret) {
 339                         mlog_errno(ret);
 340                         if (bh)
 341                                 brelse(bh);
 342                         return ret;
 343                 }
 344                 di = (struct ocfs2_dinode *)bh->b_data;
 345                 if (!OCFS2_IS_VALID_DINODE(di)) {
 346                         brelse(bh);
 347                         OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, di);
 348                         return -EIO;
 349                 }
 350                 el = &di->id2.i_list;
 351         }
 352
 353         ret = ocfs2_extent_map_find_leaf(inode, cpos, clusters, el);
 354         brelse(bh);
 355         if (ret) {
 356                 mlog_errno(ret);
 357                 return ret;
 358         }
 359
 360         ent = ocfs2_extent_map_lookup(em, cpos, clusters, NULL, NULL);
 361         if (!ent) {
 362                 ret = -ESRCH;
 363                 mlog_errno(ret);
 364                 return ret;
 365         }
 366
 367         if (ent->e_tree_depth)
 368                 BUG();  /* FIXME: Make sure this isn't a corruption */
 369
 370         *ret_ent = ent;
 371
 372         return 0;
 373 }
 374
 375 /*
 376  * Callers must hold ip_lock.  This can insert pieces of the tree,
 377  * thus racing lookup if the lock weren't held.
 378  */
 379 static int ocfs2_extent_map_insert_entry(struct ocfs2_extent_map *em,
 380                                          struct ocfs2_extent_map_entry *ent)
 381 {
 382         struct rb_node **p, *parent;
 383         struct ocfs2_extent_map_entry *old_ent;
 384
 385         old_ent = ocfs2_extent_map_lookup(em, le32_to_cpu(ent->e_rec.e_cpos),
 386                                           le32_to_cpu(ent->e_rec.e_clusters),
 387                                           &p, &parent);
 388         if (old_ent)
 389                 return -EEXIST;
 390
 391         rb_link_node(&ent->e_node, parent, p);
 392         rb_insert_color(&ent->e_node, &em->em_extents);
 393
 394         return 0;
 395 }
 396
 397
 398 /*
 399  * Simple rule: on any return code other than -EAGAIN, anything left
 400  * in the insert_context will be freed.
 401  */
 402 static int ocfs2_extent_map_try_insert(struct inode *inode,
 403                                        struct ocfs2_extent_rec *rec,
 404                                        int tree_depth,
 405                                        struct ocfs2_em_insert_context *ctxt)
 406 {
 407         int ret;
 408         struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
 409         struct ocfs2_extent_map_entry *old_ent;
 410
 411         ctxt->need_left = 0;
 412         ctxt->need_right = 0;
 413         ctxt->old_ent = NULL;
 414
 415         spin_lock(&OCFS2_I(inode)->ip_lock);
 416         ret = ocfs2_extent_map_insert_entry(em, ctxt->new_ent);
 417         if (!ret) {
 418                 ctxt->new_ent = NULL;
 419                 goto out_unlock;
 420         }
 421
 422         old_ent = ocfs2_extent_map_lookup(em, le32_to_cpu(rec->e_cpos),
 423                                           le32_to_cpu(rec->e_clusters), NULL,
 424                                           NULL);
 425
 426         if (!old_ent)
 427                 BUG();
 428
 429         ret = -EEXIST;
 430         if (old_ent->e_tree_depth < tree_depth)
 431                 goto out_unlock;
 432
 433         if (old_ent->e_tree_depth == tree_depth) {
 434                 if (!memcmp(rec, &old_ent->e_rec,
 435                             sizeof(struct ocfs2_extent_rec)))
 436                         ret = 0;
 437
 438                 /* FIXME: Should this be ESRCH/EBADR??? */
 439                 goto out_unlock;
 440         }
 441
 442         /*
 443          * We do it in this order specifically so that no actual tree
 444          * changes occur until we have all the pieces we need.  We
 445          * don't want malloc failures to leave an inconsistent tree.
 446          * Whenever we drop the lock, another process could be
 447          * inserting.  Also note that, if another process just beat us
 448          * to an insert, we might not need the same pieces we needed
 449          * the first go round.  In the end, the pieces we need will
 450          * be used, and the pieces we don't will be freed.
 451          */
 452         ctxt->need_left = !!(le32_to_cpu(rec->e_cpos) >
 453                              le32_to_cpu(old_ent->e_rec.e_cpos));
 454         ctxt->need_right = !!((le32_to_cpu(old_ent->e_rec.e_cpos) +
 455                                le32_to_cpu(old_ent->e_rec.e_clusters)) >
 456                               (le32_to_cpu(rec->e_cpos) + le32_to_cpu(rec->e_clusters)));
 457         ret = -EAGAIN;
 458         if (ctxt->need_left) {
 459                 if (!ctxt->left_ent)
 460                         goto out_unlock;
 461                 *(ctxt->left_ent) = *old_ent;
 462                 ctxt->left_ent->e_rec.e_clusters =
 463                         cpu_to_le32(le32_to_cpu(rec->e_cpos) -
 464                                     le32_to_cpu(ctxt->left_ent->e_rec.e_cpos));
 465         }
 466         if (ctxt->need_right) {
 467                 if (!ctxt->right_ent)
 468                         goto out_unlock;
 469                 *(ctxt->right_ent) = *old_ent;
 470                 ctxt->right_ent->e_rec.e_cpos =
 471                         cpu_to_le32(le32_to_cpu(rec->e_cpos) +
 472                                     le32_to_cpu(rec->e_clusters));
 473                 ctxt->right_ent->e_rec.e_clusters =
 474                         cpu_to_le32((le32_to_cpu(old_ent->e_rec.e_cpos) +
 475                                      le32_to_cpu(old_ent->e_rec.e_clusters)) -
 476                                     le32_to_cpu(ctxt->right_ent->e_rec.e_cpos));
 477         }
 478
 479         rb_erase(&old_ent->e_node, &em->em_extents);
 480         /* Now that he's erased, set him up for deletion */
 481         ctxt->old_ent = old_ent;
 482
 483         if (ctxt->need_left) {
 484                 ret = ocfs2_extent_map_insert_entry(em,
 485                                                     ctxt->left_ent);
 486                 if (ret)
 487                         goto out_unlock;
 488                 ctxt->left_ent = NULL;
 489         }
 490
 491         if (ctxt->need_right) {
 492                 ret = ocfs2_extent_map_insert_entry(em,
 493                                                     ctxt->right_ent);
 494                 if (ret)
 495                         goto out_unlock;
 496                 ctxt->right_ent = NULL;
 497         }
 498
 499         ret = ocfs2_extent_map_insert_entry(em, ctxt->new_ent);
 500
 501         if (!ret)
 502                 ctxt->new_ent = NULL;
 503
 504 out_unlock:
 505         spin_unlock(&OCFS2_I(inode)->ip_lock);
 506
 507         return ret;
 508 }
 509
 510
 511 static int ocfs2_extent_map_insert(struct inode *inode,
 512                                    struct ocfs2_extent_rec *rec,
 513                                    int tree_depth)
 514 {
 515         int ret;
 516         struct ocfs2_em_insert_context ctxt = {0, };
 517
 518         if ((le32_to_cpu(rec->e_cpos) + le32_to_cpu(rec->e_clusters)) >
 519             OCFS2_I(inode)->ip_map.em_clusters) {
 520                 ret = -EBADR;
 521                 mlog_errno(ret);
 522                 return ret;
 523         }
 524
 525         /* Zero e_clusters means a truncated tail record.  It better be EOF */
 526         if (!rec->e_clusters) {
 527                 if ((le32_to_cpu(rec->e_cpos) + le32_to_cpu(rec->e_clusters)) !=
 528                     OCFS2_I(inode)->ip_map.em_clusters) {
 529                         ret = -EBADR;
 530                         mlog_errno(ret);
 531                         return ret;
 532                 }
 533
 534                 /* Ignore the truncated tail */
 535                 return 0;
 536         }
 537
 538         ret = -ENOMEM;
 539         ctxt.new_ent = kmem_cache_alloc(ocfs2_em_ent_cachep,
 540                                         GFP_KERNEL);
 541         if (!ctxt.new_ent) {
 542                 mlog_errno(ret);
 543                 return ret;
 544         }
 545
 546         ctxt.new_ent->e_rec = *rec;
 547         ctxt.new_ent->e_tree_depth = tree_depth;
 548
 549         do {
 550                 ret = -ENOMEM;
 551                 if (ctxt.need_left && !ctxt.left_ent) {
 552                         ctxt.left_ent =
 553                                 kmem_cache_alloc(ocfs2_em_ent_cachep,
 554                                                  GFP_KERNEL);
 555                         if (!ctxt.left_ent)
 556                                 break;
 557                 }
 558                 if (ctxt.need_right && !ctxt.right_ent) {
 559                         ctxt.right_ent =
 560                                 kmem_cache_alloc(ocfs2_em_ent_cachep,
 561                                                  GFP_KERNEL);
 562                         if (!ctxt.right_ent)
 563                                 break;
 564                 }
 565
 566                 ret = ocfs2_extent_map_try_insert(inode, rec,
 567                                                   tree_depth, &ctxt);
 568         } while (ret == -EAGAIN);
 569
 570         if (ret < 0)
 571                 mlog_errno(ret);
 572
 573         if (ctxt.left_ent)
 574                 kmem_cache_free(ocfs2_em_ent_cachep, ctxt.left_ent);
 575         if (ctxt.right_ent)
 576                 kmem_cache_free(ocfs2_em_ent_cachep, ctxt.right_ent);
 577         if (ctxt.old_ent)
 578                 kmem_cache_free(ocfs2_em_ent_cachep, ctxt.old_ent);
 579         if (ctxt.new_ent)
 580                 kmem_cache_free(ocfs2_em_ent_cachep, ctxt.new_ent);
 581
 582         return ret;
 583 }
 584
 585 /*
 586  * Append this record to the tail of the extent map.  It must be
 587  * tree_depth 0.  The record might be an extension of an existing
 588  * record, and as such that needs to be handled.  eg:
 589  *
 590  * Existing record in the extent map:
 591  *
 592  *      cpos = 10, len = 10
 593  *      |---------|
 594  *
 595  * New Record:
 596  *
 597  *      cpos = 10, len = 20
 598  *      |------------------|
 599  *
 600  * The passed record is the new on-disk record.  The new_clusters value
 601  * is how many clusters were added to the file.  If the append is a
 602  * contiguous append, the new_clusters has been added to
 603  * rec->e_clusters.  If the append is an entirely new extent, then
 604  * rec->e_clusters is == new_clusters.
 605  */
 606 int ocfs2_extent_map_append(struct inode *inode,
 607                             struct ocfs2_extent_rec *rec,
 608                             u32 new_clusters)
 609 {
 610         int ret;
 611         struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
 612         struct ocfs2_extent_map_entry *ent;
 613         struct ocfs2_extent_rec *old;
 614
 615         BUG_ON(!new_clusters);
 616         BUG_ON(le32_to_cpu(rec->e_clusters) < new_clusters);
 617
 618         if (em->em_clusters < OCFS2_I(inode)->ip_clusters) {
 619                 /*
 620                  * Size changed underneath us on disk.  Drop any
 621                  * straddling records and update our idea of
 622                  * i_clusters
 623                  */
 624                 ocfs2_extent_map_drop(inode, em->em_clusters - 1);
 625                 em->em_clusters = OCFS2_I(inode)->ip_clusters;
 626         }
 627
 628         mlog_bug_on_msg((le32_to_cpu(rec->e_cpos) +
 629                          le32_to_cpu(rec->e_clusters)) !=
 630                         (em->em_clusters + new_clusters),
 631                         "Inode %"MLFu64":\n"
 632                         "rec->e_cpos = %u + rec->e_clusters = %u = %u\n"
 633                         "em->em_clusters = %u + new_clusters = %u = %u\n",
 634                         OCFS2_I(inode)->ip_blkno,
 635                         le32_to_cpu(rec->e_cpos), le32_to_cpu(rec->e_clusters),
 636                         le32_to_cpu(rec->e_cpos) + le32_to_cpu(rec->e_clusters),
 637                         em->em_clusters, new_clusters,
 638                         em->em_clusters + new_clusters);
 639
 640         em->em_clusters += new_clusters;
 641
 642         ret = -ENOENT;
 643         if (le32_to_cpu(rec->e_clusters) > new_clusters) {
 644                 /* This is a contiguous append */
 645                 ent = ocfs2_extent_map_lookup(em, le32_to_cpu(rec->e_cpos), 1,
 646                                               NULL, NULL);
 647                 if (ent) {
 648                         old = &ent->e_rec;
 649                         BUG_ON((le32_to_cpu(rec->e_cpos) +
 650                                 le32_to_cpu(rec->e_clusters)) !=
 651                                  (le32_to_cpu(old->e_cpos) +
 652                                   le32_to_cpu(old->e_clusters) +
 653                                   new_clusters));
 654                         if (ent->e_tree_depth == 0) {
 655                                 BUG_ON(le32_to_cpu(old->e_cpos) !=
 656                                        le32_to_cpu(rec->e_cpos));
 657                                 BUG_ON(le64_to_cpu(old->e_blkno) !=
 658                                        le64_to_cpu(rec->e_blkno));
 659                                 ret = 0;
 660                         }
 661                         /*
 662                          * Let non-leafs fall through as -ENOENT to
 663                          * force insertion of the new leaf.
 664                          */
 665                         le32_add_cpu(&old->e_clusters, new_clusters);
 666                 }
 667         }
 668
 669         if (ret == -ENOENT)
 670                 ret = ocfs2_extent_map_insert(inode, rec, 0);
 671         if (ret < 0)
 672                 mlog_errno(ret);
 673         return ret;
 674 }
 675
 676 #if 0
 677 /* Code here is included but defined out as it completes the extent
 678  * map api and may be used in the future. */
 679
 680 /*
 681  * Look up the record containing this cluster offset.  This record is
 682  * part of the extent map.  Do not free it.  Any changes you make to
 683  * it will reflect in the extent map.  So, if your last extent
 684  * is (cpos = 10, clusters = 10) and you truncate the file by 5
 685  * clusters, you can do:
 686  *
 687  * ret = ocfs2_extent_map_get_rec(em, orig_size - 5, &rec);
 688  * rec->e_clusters -= 5;
 689  *
 690  * The lookup does not read from disk.  If the map isn't filled in for
 691  * an entry, you won't find it.
 692  *
 693  * Also note that the returned record is valid until alloc_sem is
 694  * dropped.  After that, truncate and extend can happen.  Caveat Emptor.
 695  */
 696 int ocfs2_extent_map_get_rec(struct inode *inode, u32 cpos,
 697                              struct ocfs2_extent_rec **rec,
 698                              int *tree_depth)
 699 {
 700         int ret = -ENOENT;
 701         struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
 702         struct ocfs2_extent_map_entry *ent;
 703
 704         *rec = NULL;
 705
 706         if (cpos >= OCFS2_I(inode)->ip_clusters)
 707                 return -EINVAL;
 708
 709         if (cpos >= em->em_clusters) {
 710                 /*
 711                  * Size changed underneath us on disk.  Drop any
 712                  * straddling records and update our idea of
 713                  * i_clusters
 714                  */
 715                 ocfs2_extent_map_drop(inode, em->em_clusters - 1);
 716                 em->em_clusters = OCFS2_I(inode)->ip_clusters ;
 717         }
 718
 719         ent = ocfs2_extent_map_lookup(&OCFS2_I(inode)->ip_map, cpos, 1,
 720                                       NULL, NULL);
 721
 722         if (ent) {
 723                 *rec = &ent->e_rec;
 724                 if (tree_depth)
 725                         *tree_depth = ent->e_tree_depth;
 726                 ret = 0;
 727         }
 728
 729         return ret;
 730 }
 731
 732 int ocfs2_extent_map_get_clusters(struct inode *inode,
 733                                   u32 v_cpos, int count,
 734                                   u32 *p_cpos, int *ret_count)
 735 {
 736         int ret;
 737         u32 coff, ccount;
 738         struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
 739         struct ocfs2_extent_map_entry *ent = NULL;
 740
 741         *p_cpos = ccount = 0;
 742
 743         if ((v_cpos + count) > OCFS2_I(inode)->ip_clusters)
 744                 return -EINVAL;
 745
 746         if ((v_cpos + count) > em->em_clusters) {
 747                 /*
 748                  * Size changed underneath us on disk.  Drop any
 749                  * straddling records and update our idea of
 750                  * i_clusters
 751                  */
 752                 ocfs2_extent_map_drop(inode, em->em_clusters - 1);
 753                 em->em_clusters = OCFS2_I(inode)->ip_clusters;
 754         }
 755
 756
 757         ret = ocfs2_extent_map_lookup_read(inode, v_cpos, count, &ent);
 758         if (ret)
 759                 return ret;
 760
 761         if (ent) {
 762                 /* We should never find ourselves straddling an interval */
 763                 if (!ocfs2_extent_rec_contains_clusters(&ent->e_rec,
 764                                                         v_cpos,
 765                                                         count))
 766                         return -ESRCH;
 767
 768                 coff = v_cpos - le32_to_cpu(ent->e_rec.e_cpos);
 769                 *p_cpos = ocfs2_blocks_to_clusters(inode->i_sb,
 770                                 le64_to_cpu(ent->e_rec.e_blkno)) +
 771                           coff;
 772
 773                 if (ret_count)
 774                         *ret_count = le32_to_cpu(ent->e_rec.e_clusters) - coff;
 775
 776                 return 0;
 777         }
 778
 779
 780         return -ENOENT;
 781 }
 782
 783 #endif  /*  0  */
 784
 785 int ocfs2_extent_map_get_blocks(struct inode *inode,
 786                                 u64 v_blkno, int count,
 787                                 u64 *p_blkno, int *ret_count)
 788 {
 789         int ret;
 790         u64 boff;
 791         u32 cpos, clusters;
 792         int bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
 793         struct ocfs2_extent_map_entry *ent = NULL;
 794         struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
 795         struct ocfs2_extent_rec *rec;
 796
 797         *p_blkno = 0;
 798
 799         cpos = ocfs2_blocks_to_clusters(inode->i_sb, v_blkno);
 800         clusters = ocfs2_blocks_to_clusters(inode->i_sb,
 801                                             (u64)count + bpc - 1);
 802         if ((cpos + clusters) > OCFS2_I(inode)->ip_clusters) {
 803                 ret = -EINVAL;
 804                 mlog_errno(ret);
 805                 return ret;
 806         }
 807
 808         if ((cpos + clusters) > em->em_clusters) {
 809                 /*
 810                  * Size changed underneath us on disk.  Drop any
 811                  * straddling records and update our idea of
 812                  * i_clusters
 813                  */
 814                 ocfs2_extent_map_drop(inode, em->em_clusters - 1);
 815                 em->em_clusters = OCFS2_I(inode)->ip_clusters;
 816         }
 817
 818         ret = ocfs2_extent_map_lookup_read(inode, cpos, clusters, &ent);
 819         if (ret) {
 820                 mlog_errno(ret);
 821                 return ret;
 822         }
 823
 824         if (ent)
 825         {
 826                 rec = &ent->e_rec;
 827
 828                 /* We should never find ourselves straddling an interval */
 829                 if (!ocfs2_extent_rec_contains_clusters(rec, cpos, clusters)) {
 830                         ret = -ESRCH;
 831                         mlog_errno(ret);
 832                         return ret;
 833                 }
 834
 835                 boff = ocfs2_clusters_to_blocks(inode->i_sb, cpos -
 836                                                 le32_to_cpu(rec->e_cpos));
 837                 boff += (v_blkno & (u64)(bpc - 1));
 838                 *p_blkno = le64_to_cpu(rec->e_blkno) + boff;
 839
 840                 if (ret_count) {
 841                         *ret_count = ocfs2_clusters_to_blocks(inode->i_sb,
 842                                         le32_to_cpu(rec->e_clusters)) - boff;
 843                 }
 844
 845                 return 0;
 846         }
 847
 848         return -ENOENT;
 849 }
 850
 851 int ocfs2_extent_map_init(struct inode *inode)
 852 {
 853         struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
 854
 855         em->em_extents = RB_ROOT;
 856         em->em_clusters = 0;
 857
 858         return 0;
 859 }
 860
 861 /* Needs the lock */
 862 static void __ocfs2_extent_map_drop(struct inode *inode,
 863                                     u32 new_clusters,
 864                                     struct rb_node **free_head,
 865                                     struct ocfs2_extent_map_entry **tail_ent)
 866 {
 867         struct rb_node *node, *next;
 868         struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
 869         struct ocfs2_extent_map_entry *ent;
 870
 871         *free_head = NULL;
 872
 873         ent = NULL;
 874         node = rb_last(&em->em_extents);
 875         while (node)
 876         {
 877                 next = rb_prev(node);
 878
 879                 ent = rb_entry(node, struct ocfs2_extent_map_entry,
 880                                e_node);
 881                 if (le32_to_cpu(ent->e_rec.e_cpos) < new_clusters)
 882                         break;
 883
 884                 rb_erase(&ent->e_node, &em->em_extents);
 885
 886                 node->rb_right = *free_head;
 887                 *free_head = node;
 888
 889                 ent = NULL;
 890                 node = next;
 891         }
 892
 893         /* Do we have an entry straddling new_clusters? */
 894         if (tail_ent) {
 895                 if (ent &&
 896                     ((le32_to_cpu(ent->e_rec.e_cpos) +
 897                       le32_to_cpu(ent->e_rec.e_clusters)) > new_clusters))
 898                         *tail_ent = ent;
 899                 else
 900                         *tail_ent = NULL;
 901         }
 902 }
 903
 904 static void __ocfs2_extent_map_drop_cleanup(struct rb_node *free_head)
 905 {
 906         struct rb_node *node;
 907         struct ocfs2_extent_map_entry *ent;
 908
 909         while (free_head) {
 910                 node = free_head;
 911                 free_head = node->rb_right;
 912
 913                 ent = rb_entry(node, struct ocfs2_extent_map_entry,
 914                                e_node);
 915                 kmem_cache_free(ocfs2_em_ent_cachep, ent);
 916         }
 917 }
 918
 919 /*
 920  * Remove all entries past new_clusters, inclusive of an entry that
 921  * contains new_clusters.  This is effectively a cache forget.
 922  *
 923  * If you want to also clip the last extent by some number of clusters,
 924  * you need to call ocfs2_extent_map_trunc().
 925  * This code does not check or modify ip_clusters.
 926  */
 927 int ocfs2_extent_map_drop(struct inode *inode, u32 new_clusters)
 928 {
 929         struct rb_node *free_head = NULL;
 930         struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
 931         struct ocfs2_extent_map_entry *ent;
 932
 933         spin_lock(&OCFS2_I(inode)->ip_lock);
 934
 935         __ocfs2_extent_map_drop(inode, new_clusters, &free_head, &ent);
 936
 937         if (ent) {
 938                 rb_erase(&ent->e_node, &em->em_extents);
 939                 ent->e_node.rb_right = free_head;
 940                 free_head = &ent->e_node;
 941         }
 942
 943         spin_unlock(&OCFS2_I(inode)->ip_lock);
 944
 945         if (free_head)
 946                 __ocfs2_extent_map_drop_cleanup(free_head);
 947
 948         return 0;
 949 }
 950
 951 /*
 952  * Remove all entries past new_clusters and also clip any extent
 953  * straddling new_clusters, if there is one.  This does not check
 954  * or modify ip_clusters
 955  */
 956 int ocfs2_extent_map_trunc(struct inode *inode, u32 new_clusters)
 957 {
 958         struct rb_node *free_head = NULL;
 959         struct ocfs2_extent_map_entry *ent = NULL;
 960
 961         spin_lock(&OCFS2_I(inode)->ip_lock);
 962
 963         __ocfs2_extent_map_drop(inode, new_clusters, &free_head, &ent);
 964
 965         if (ent)
 966                 ent->e_rec.e_clusters = cpu_to_le32(new_clusters -
 967                                                le32_to_cpu(ent->e_rec.e_cpos));
 968
 969         OCFS2_I(inode)->ip_map.em_clusters = new_clusters;
 970
 971         spin_unlock(&OCFS2_I(inode)->ip_lock);
 972
 973         if (free_head)
 974                 __ocfs2_extent_map_drop_cleanup(free_head);
 975
 976         return 0;
 977 }
 978
 979 int __init init_ocfs2_extent_maps(void)
 980 {
 981         ocfs2_em_ent_cachep =
 982                 kmem_cache_create("ocfs2_em_ent",
 983                                   sizeof(struct ocfs2_extent_map_entry),
 984                                   0, SLAB_HWCACHE_ALIGN, NULL, NULL);
 985         if (!ocfs2_em_ent_cachep)
 986                 return -ENOMEM;
 987
 988         return 0;
 989 }
 990
 991 void __exit exit_ocfs2_extent_maps(void)
 992 {
 993         kmem_cache_destroy(ocfs2_em_ent_cachep);
 994 }