linux/fs/ocfs2/alloc.c
<<
>>
Prefs
   1/* -*- mode: c; c-basic-offset: 8; -*-
   2 * vim: noexpandtab sw=8 ts=8 sts=0:
   3 *
   4 * alloc.c
   5 *
   6 * Extent allocs and frees
   7 *
   8 * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
   9 *
  10 * This program is free software; you can redistribute it and/or
  11 * modify it under the terms of the GNU General Public
  12 * License as published by the Free Software Foundation; either
  13 * version 2 of the License, or (at your option) any later version.
  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/types.h>
  28#include <linux/slab.h>
  29#include <linux/highmem.h>
  30#include <linux/swap.h>
  31
  32#define MLOG_MASK_PREFIX ML_DISK_ALLOC
  33#include <cluster/masklog.h>
  34
  35#include "ocfs2.h"
  36
  37#include "alloc.h"
  38#include "aops.h"
  39#include "dlmglue.h"
  40#include "extent_map.h"
  41#include "inode.h"
  42#include "journal.h"
  43#include "localalloc.h"
  44#include "suballoc.h"
  45#include "sysfile.h"
  46#include "file.h"
  47#include "super.h"
  48#include "uptodate.h"
  49
  50#include "buffer_head_io.h"
  51
  52static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc);
  53static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt,
  54                                         struct ocfs2_extent_block *eb);
  55
  56/*
  57 * Structures which describe a path through a btree, and functions to
  58 * manipulate them.
  59 *
  60 * The idea here is to be as generic as possible with the tree
  61 * manipulation code.
  62 */
  63struct ocfs2_path_item {
  64        struct buffer_head              *bh;
  65        struct ocfs2_extent_list        *el;
  66};
  67
  68#define OCFS2_MAX_PATH_DEPTH    5
  69
  70struct ocfs2_path {
  71        int                     p_tree_depth;
  72        struct ocfs2_path_item  p_node[OCFS2_MAX_PATH_DEPTH];
  73};
  74
  75#define path_root_bh(_path) ((_path)->p_node[0].bh)
  76#define path_root_el(_path) ((_path)->p_node[0].el)
  77#define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh)
  78#define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el)
  79#define path_num_items(_path) ((_path)->p_tree_depth + 1)
  80
  81/*
  82 * Reset the actual path elements so that we can re-use the structure
  83 * to build another path. Generally, this involves freeing the buffer
  84 * heads.
  85 */
  86static void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root)
  87{
  88        int i, start = 0, depth = 0;
  89        struct ocfs2_path_item *node;
  90
  91        if (keep_root)
  92                start = 1;
  93
  94        for(i = start; i < path_num_items(path); i++) {
  95                node = &path->p_node[i];
  96
  97                brelse(node->bh);
  98                node->bh = NULL;
  99                node->el = NULL;
 100        }
 101
 102        /*
 103         * Tree depth may change during truncate, or insert. If we're
 104         * keeping the root extent list, then make sure that our path
 105         * structure reflects the proper depth.
 106         */
 107        if (keep_root)
 108                depth = le16_to_cpu(path_root_el(path)->l_tree_depth);
 109
 110        path->p_tree_depth = depth;
 111}
 112
 113static void ocfs2_free_path(struct ocfs2_path *path)
 114{
 115        if (path) {
 116                ocfs2_reinit_path(path, 0);
 117                kfree(path);
 118        }
 119}
 120
 121/*
 122 * All the elements of src into dest. After this call, src could be freed
 123 * without affecting dest.
 124 *
 125 * Both paths should have the same root. Any non-root elements of dest
 126 * will be freed.
 127 */
 128static void ocfs2_cp_path(struct ocfs2_path *dest, struct ocfs2_path *src)
 129{
 130        int i;
 131
 132        BUG_ON(path_root_bh(dest) != path_root_bh(src));
 133        BUG_ON(path_root_el(dest) != path_root_el(src));
 134
 135        ocfs2_reinit_path(dest, 1);
 136
 137        for(i = 1; i < OCFS2_MAX_PATH_DEPTH; i++) {
 138                dest->p_node[i].bh = src->p_node[i].bh;
 139                dest->p_node[i].el = src->p_node[i].el;
 140
 141                if (dest->p_node[i].bh)
 142                        get_bh(dest->p_node[i].bh);
 143        }
 144}
 145
 146/*
 147 * Make the *dest path the same as src and re-initialize src path to
 148 * have a root only.
 149 */
 150static void ocfs2_mv_path(struct ocfs2_path *dest, struct ocfs2_path *src)
 151{
 152        int i;
 153
 154        BUG_ON(path_root_bh(dest) != path_root_bh(src));
 155
 156        for(i = 1; i < OCFS2_MAX_PATH_DEPTH; i++) {
 157                brelse(dest->p_node[i].bh);
 158
 159                dest->p_node[i].bh = src->p_node[i].bh;
 160                dest->p_node[i].el = src->p_node[i].el;
 161
 162                src->p_node[i].bh = NULL;
 163                src->p_node[i].el = NULL;
 164        }
 165}
 166
 167/*
 168 * Insert an extent block at given index.
 169 *
 170 * This will not take an additional reference on eb_bh.
 171 */
 172static inline void ocfs2_path_insert_eb(struct ocfs2_path *path, int index,
 173                                        struct buffer_head *eb_bh)
 174{
 175        struct ocfs2_extent_block *eb = (struct ocfs2_extent_block *)eb_bh->b_data;
 176
 177        /*
 178         * Right now, no root bh is an extent block, so this helps
 179         * catch code errors with dinode trees. The assertion can be
 180         * safely removed if we ever need to insert extent block
 181         * structures at the root.
 182         */
 183        BUG_ON(index == 0);
 184
 185        path->p_node[index].bh = eb_bh;
 186        path->p_node[index].el = &eb->h_list;
 187}
 188
 189static struct ocfs2_path *ocfs2_new_path(struct buffer_head *root_bh,
 190                                         struct ocfs2_extent_list *root_el)
 191{
 192        struct ocfs2_path *path;
 193
 194        BUG_ON(le16_to_cpu(root_el->l_tree_depth) >= OCFS2_MAX_PATH_DEPTH);
 195
 196        path = kzalloc(sizeof(*path), GFP_NOFS);
 197        if (path) {
 198                path->p_tree_depth = le16_to_cpu(root_el->l_tree_depth);
 199                get_bh(root_bh);
 200                path_root_bh(path) = root_bh;
 201                path_root_el(path) = root_el;
 202        }
 203
 204        return path;
 205}
 206
 207/*
 208 * Allocate and initialize a new path based on a disk inode tree.
 209 */
 210static struct ocfs2_path *ocfs2_new_inode_path(struct buffer_head *di_bh)
 211{
 212        struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
 213        struct ocfs2_extent_list *el = &di->id2.i_list;
 214
 215        return ocfs2_new_path(di_bh, el);
 216}
 217
 218/*
 219 * Convenience function to journal all components in a path.
 220 */
 221static int ocfs2_journal_access_path(struct inode *inode, handle_t *handle,
 222                                     struct ocfs2_path *path)
 223{
 224        int i, ret = 0;
 225
 226        if (!path)
 227                goto out;
 228
 229        for(i = 0; i < path_num_items(path); i++) {
 230                ret = ocfs2_journal_access(handle, inode, path->p_node[i].bh,
 231                                           OCFS2_JOURNAL_ACCESS_WRITE);
 232                if (ret < 0) {
 233                        mlog_errno(ret);
 234                        goto out;
 235                }
 236        }
 237
 238out:
 239        return ret;
 240}
 241
 242/*
 243 * Return the index of the extent record which contains cluster #v_cluster.
 244 * -1 is returned if it was not found.
 245 *
 246 * Should work fine on interior and exterior nodes.
 247 */
 248int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster)
 249{
 250        int ret = -1;
 251        int i;
 252        struct ocfs2_extent_rec *rec;
 253        u32 rec_end, rec_start, clusters;
 254
 255        for(i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
 256                rec = &el->l_recs[i];
 257
 258                rec_start = le32_to_cpu(rec->e_cpos);
 259                clusters = ocfs2_rec_clusters(el, rec);
 260
 261                rec_end = rec_start + clusters;
 262
 263                if (v_cluster >= rec_start && v_cluster < rec_end) {
 264                        ret = i;
 265                        break;
 266                }
 267        }
 268
 269        return ret;
 270}
 271
 272enum ocfs2_contig_type {
 273        CONTIG_NONE = 0,
 274        CONTIG_LEFT,
 275        CONTIG_RIGHT,
 276        CONTIG_LEFTRIGHT,
 277};
 278
 279
 280/*
 281 * NOTE: ocfs2_block_extent_contig(), ocfs2_extents_adjacent() and
 282 * ocfs2_extent_contig only work properly against leaf nodes!
 283 */
 284static int ocfs2_block_extent_contig(struct super_block *sb,
 285                                     struct ocfs2_extent_rec *ext,
 286                                     u64 blkno)
 287{
 288        u64 blk_end = le64_to_cpu(ext->e_blkno);
 289
 290        blk_end += ocfs2_clusters_to_blocks(sb,
 291                                    le16_to_cpu(ext->e_leaf_clusters));
 292
 293        return blkno == blk_end;
 294}
 295
 296static int ocfs2_extents_adjacent(struct ocfs2_extent_rec *left,
 297                                  struct ocfs2_extent_rec *right)
 298{
 299        u32 left_range;
 300
 301        left_range = le32_to_cpu(left->e_cpos) +
 302                le16_to_cpu(left->e_leaf_clusters);
 303
 304        return (left_range == le32_to_cpu(right->e_cpos));
 305}
 306
 307static enum ocfs2_contig_type
 308        ocfs2_extent_contig(struct inode *inode,
 309                            struct ocfs2_extent_rec *ext,
 310                            struct ocfs2_extent_rec *insert_rec)
 311{
 312        u64 blkno = le64_to_cpu(insert_rec->e_blkno);
 313
 314        /*
 315         * Refuse to coalesce extent records with different flag
 316         * fields - we don't want to mix unwritten extents with user
 317         * data.
 318         */
 319        if (ext->e_flags != insert_rec->e_flags)
 320                return CONTIG_NONE;
 321
 322        if (ocfs2_extents_adjacent(ext, insert_rec) &&
 323            ocfs2_block_extent_contig(inode->i_sb, ext, blkno))
 324                        return CONTIG_RIGHT;
 325
 326        blkno = le64_to_cpu(ext->e_blkno);
 327        if (ocfs2_extents_adjacent(insert_rec, ext) &&
 328            ocfs2_block_extent_contig(inode->i_sb, insert_rec, blkno))
 329                return CONTIG_LEFT;
 330
 331        return CONTIG_NONE;
 332}
 333
 334/*
 335 * NOTE: We can have pretty much any combination of contiguousness and
 336 * appending.
 337 *
 338 * The usefulness of APPEND_TAIL is more in that it lets us know that
 339 * we'll have to update the path to that leaf.
 340 */
 341enum ocfs2_append_type {
 342        APPEND_NONE = 0,
 343        APPEND_TAIL,
 344};
 345
 346enum ocfs2_split_type {
 347        SPLIT_NONE = 0,
 348        SPLIT_LEFT,
 349        SPLIT_RIGHT,
 350};
 351
 352struct ocfs2_insert_type {
 353        enum ocfs2_split_type   ins_split;
 354        enum ocfs2_append_type  ins_appending;
 355        enum ocfs2_contig_type  ins_contig;
 356        int                     ins_contig_index;
 357        int                     ins_tree_depth;
 358};
 359
 360struct ocfs2_merge_ctxt {
 361        enum ocfs2_contig_type  c_contig_type;
 362        int                     c_has_empty_extent;
 363        int                     c_split_covers_rec;
 364};
 365
 366/*
 367 * How many free extents have we got before we need more meta data?
 368 */
 369int ocfs2_num_free_extents(struct ocfs2_super *osb,
 370                           struct inode *inode,
 371                           struct ocfs2_dinode *fe)
 372{
 373        int retval;
 374        struct ocfs2_extent_list *el;
 375        struct ocfs2_extent_block *eb;
 376        struct buffer_head *eb_bh = NULL;
 377
 378        mlog_entry_void();
 379
 380        if (!OCFS2_IS_VALID_DINODE(fe)) {
 381                OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
 382                retval = -EIO;
 383                goto bail;
 384        }
 385
 386        if (fe->i_last_eb_blk) {
 387                retval = ocfs2_read_block(osb, le64_to_cpu(fe->i_last_eb_blk),
 388                                          &eb_bh, OCFS2_BH_CACHED, inode);
 389                if (retval < 0) {
 390                        mlog_errno(retval);
 391                        goto bail;
 392                }
 393                eb = (struct ocfs2_extent_block *) eb_bh->b_data;
 394                el = &eb->h_list;
 395        } else
 396                el = &fe->id2.i_list;
 397
 398        BUG_ON(el->l_tree_depth != 0);
 399
 400        retval = le16_to_cpu(el->l_count) - le16_to_cpu(el->l_next_free_rec);
 401bail:
 402        if (eb_bh)
 403                brelse(eb_bh);
 404
 405        mlog_exit(retval);
 406        return retval;
 407}
 408
 409/* expects array to already be allocated
 410 *
 411 * sets h_signature, h_blkno, h_suballoc_bit, h_suballoc_slot, and
 412 * l_count for you
 413 */
 414static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb,
 415                                     handle_t *handle,
 416                                     struct inode *inode,
 417                                     int wanted,
 418                                     struct ocfs2_alloc_context *meta_ac,
 419                                     struct buffer_head *bhs[])
 420{
 421        int count, status, i;
 422        u16 suballoc_bit_start;
 423        u32 num_got;
 424        u64 first_blkno;
 425        struct ocfs2_extent_block *eb;
 426
 427        mlog_entry_void();
 428
 429        count = 0;
 430        while (count < wanted) {
 431                status = ocfs2_claim_metadata(osb,
 432                                              handle,
 433                                              meta_ac,
 434                                              wanted - count,
 435                                              &suballoc_bit_start,
 436                                              &num_got,
 437                                              &first_blkno);
 438                if (status < 0) {
 439                        mlog_errno(status);
 440                        goto bail;
 441                }
 442
 443                for(i = count;  i < (num_got + count); i++) {
 444                        bhs[i] = sb_getblk(osb->sb, first_blkno);
 445                        if (bhs[i] == NULL) {
 446                                status = -EIO;
 447                                mlog_errno(status);
 448                                goto bail;
 449                        }
 450                        ocfs2_set_new_buffer_uptodate(inode, bhs[i]);
 451
 452                        status = ocfs2_journal_access(handle, inode, bhs[i],
 453                                                      OCFS2_JOURNAL_ACCESS_CREATE);
 454                        if (status < 0) {
 455                                mlog_errno(status);
 456                                goto bail;
 457                        }
 458
 459                        memset(bhs[i]->b_data, 0, osb->sb->s_blocksize);
 460                        eb = (struct ocfs2_extent_block *) bhs[i]->b_data;
 461                        /* Ok, setup the minimal stuff here. */
 462                        strcpy(eb->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE);
 463                        eb->h_blkno = cpu_to_le64(first_blkno);
 464                        eb->h_fs_generation = cpu_to_le32(osb->fs_generation);
 465                        eb->h_suballoc_slot = cpu_to_le16(osb->slot_num);
 466                        eb->h_suballoc_bit = cpu_to_le16(suballoc_bit_start);
 467                        eb->h_list.l_count =
 468                                cpu_to_le16(ocfs2_extent_recs_per_eb(osb->sb));
 469
 470                        suballoc_bit_start++;
 471                        first_blkno++;
 472
 473                        /* We'll also be dirtied by the caller, so
 474                         * this isn't absolutely necessary. */
 475                        status = ocfs2_journal_dirty(handle, bhs[i]);
 476                        if (status < 0) {
 477                                mlog_errno(status);
 478                                goto bail;
 479                        }
 480                }
 481
 482                count += num_got;
 483        }
 484
 485        status = 0;
 486bail:
 487        if (status < 0) {
 488                for(i = 0; i < wanted; i++) {
 489                        if (bhs[i])
 490                                brelse(bhs[i]);
 491                        bhs[i] = NULL;
 492                }
 493        }
 494        mlog_exit(status);
 495        return status;
 496}
 497
 498/*
 499 * Helper function for ocfs2_add_branch() and ocfs2_shift_tree_depth().
 500 *
 501 * Returns the sum of the rightmost extent rec logical offset and
 502 * cluster count.
 503 *
 504 * ocfs2_add_branch() uses this to determine what logical cluster
 505 * value should be populated into the leftmost new branch records.
 506 *
 507 * ocfs2_shift_tree_depth() uses this to determine the # clusters
 508 * value for the new topmost tree record.
 509 */
 510static inline u32 ocfs2_sum_rightmost_rec(struct ocfs2_extent_list  *el)
 511{
 512        int i;
 513
 514        i = le16_to_cpu(el->l_next_free_rec) - 1;
 515
 516        return le32_to_cpu(el->l_recs[i].e_cpos) +
 517                ocfs2_rec_clusters(el, &el->l_recs[i]);
 518}
 519
 520/*
 521 * Add an entire tree branch to our inode. eb_bh is the extent block
 522 * to start at, if we don't want to start the branch at the dinode
 523 * structure.
 524 *
 525 * last_eb_bh is required as we have to update it's next_leaf pointer
 526 * for the new last extent block.
 527 *
 528 * the new branch will be 'empty' in the sense that every block will
 529 * contain a single record with cluster count == 0.
 530 */
 531static int ocfs2_add_branch(struct ocfs2_super *osb,
 532                            handle_t *handle,
 533                            struct inode *inode,
 534                            struct buffer_head *fe_bh,
 535                            struct buffer_head *eb_bh,
 536                            struct buffer_head **last_eb_bh,
 537                            struct ocfs2_alloc_context *meta_ac)
 538{
 539        int status, new_blocks, i;
 540        u64 next_blkno, new_last_eb_blk;
 541        struct buffer_head *bh;
 542        struct buffer_head **new_eb_bhs = NULL;
 543        struct ocfs2_dinode *fe;
 544        struct ocfs2_extent_block *eb;
 545        struct ocfs2_extent_list  *eb_el;
 546        struct ocfs2_extent_list  *el;
 547        u32 new_cpos;
 548
 549        mlog_entry_void();
 550
 551        BUG_ON(!last_eb_bh || !*last_eb_bh);
 552
 553        fe = (struct ocfs2_dinode *) fe_bh->b_data;
 554
 555        if (eb_bh) {
 556                eb = (struct ocfs2_extent_block *) eb_bh->b_data;
 557                el = &eb->h_list;
 558        } else
 559                el = &fe->id2.i_list;
 560
 561        /* we never add a branch to a leaf. */
 562        BUG_ON(!el->l_tree_depth);
 563
 564        new_blocks = le16_to_cpu(el->l_tree_depth);
 565
 566        /* allocate the number of new eb blocks we need */
 567        new_eb_bhs = kcalloc(new_blocks, sizeof(struct buffer_head *),
 568                             GFP_KERNEL);
 569        if (!new_eb_bhs) {
 570                status = -ENOMEM;
 571                mlog_errno(status);
 572                goto bail;
 573        }
 574
 575        status = ocfs2_create_new_meta_bhs(osb, handle, inode, new_blocks,
 576                                           meta_ac, new_eb_bhs);
 577        if (status < 0) {
 578                mlog_errno(status);
 579                goto bail;
 580        }
 581
 582        eb = (struct ocfs2_extent_block *)(*last_eb_bh)->b_data;
 583        new_cpos = ocfs2_sum_rightmost_rec(&eb->h_list);
 584
 585        /* Note: new_eb_bhs[new_blocks - 1] is the guy which will be
 586         * linked with the rest of the tree.
 587         * conversly, new_eb_bhs[0] is the new bottommost leaf.
 588         *
 589         * when we leave the loop, new_last_eb_blk will point to the
 590         * newest leaf, and next_blkno will point to the topmost extent
 591         * block. */
 592        next_blkno = new_last_eb_blk = 0;
 593        for(i = 0; i < new_blocks; i++) {
 594                bh = new_eb_bhs[i];
 595                eb = (struct ocfs2_extent_block *) bh->b_data;
 596                if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
 597                        OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
 598                        status = -EIO;
 599                        goto bail;
 600                }
 601                eb_el = &eb->h_list;
 602
 603                status = ocfs2_journal_access(handle, inode, bh,
 604                                              OCFS2_JOURNAL_ACCESS_CREATE);
 605                if (status < 0) {
 606                        mlog_errno(status);
 607                        goto bail;
 608                }
 609
 610                eb->h_next_leaf_blk = 0;
 611                eb_el->l_tree_depth = cpu_to_le16(i);
 612                eb_el->l_next_free_rec = cpu_to_le16(1);
 613                /*
 614                 * This actually counts as an empty extent as
 615                 * c_clusters == 0
 616                 */
 617                eb_el->l_recs[0].e_cpos = cpu_to_le32(new_cpos);
 618                eb_el->l_recs[0].e_blkno = cpu_to_le64(next_blkno);
 619                /*
 620                 * eb_el isn't always an interior node, but even leaf
 621                 * nodes want a zero'd flags and reserved field so
 622                 * this gets the whole 32 bits regardless of use.
 623                 */
 624                eb_el->l_recs[0].e_int_clusters = cpu_to_le32(0);
 625                if (!eb_el->l_tree_depth)
 626                        new_last_eb_blk = le64_to_cpu(eb->h_blkno);
 627
 628                status = ocfs2_journal_dirty(handle, bh);
 629                if (status < 0) {
 630                        mlog_errno(status);
 631                        goto bail;
 632                }
 633
 634                next_blkno = le64_to_cpu(eb->h_blkno);
 635        }
 636
 637        /* This is a bit hairy. We want to update up to three blocks
 638         * here without leaving any of them in an inconsistent state
 639         * in case of error. We don't have to worry about
 640         * journal_dirty erroring as it won't unless we've aborted the
 641         * handle (in which case we would never be here) so reserving
 642         * the write with journal_access is all we need to do. */
 643        status = ocfs2_journal_access(handle, inode, *last_eb_bh,
 644                                      OCFS2_JOURNAL_ACCESS_WRITE);
 645        if (status < 0) {
 646                mlog_errno(status);
 647                goto bail;
 648        }
 649        status = ocfs2_journal_access(handle, inode, fe_bh,
 650                                      OCFS2_JOURNAL_ACCESS_WRITE);
 651        if (status < 0) {
 652                mlog_errno(status);
 653                goto bail;
 654        }
 655        if (eb_bh) {
 656                status = ocfs2_journal_access(handle, inode, eb_bh,
 657                                              OCFS2_JOURNAL_ACCESS_WRITE);
 658                if (status < 0) {
 659                        mlog_errno(status);
 660                        goto bail;
 661                }
 662        }
 663
 664        /* Link the new branch into the rest of the tree (el will
 665         * either be on the fe, or the extent block passed in. */
 666        i = le16_to_cpu(el->l_next_free_rec);
 667        el->l_recs[i].e_blkno = cpu_to_le64(next_blkno);
 668        el->l_recs[i].e_cpos = cpu_to_le32(new_cpos);
 669        el->l_recs[i].e_int_clusters = 0;
 670        le16_add_cpu(&el->l_next_free_rec, 1);
 671
 672        /* fe needs a new last extent block pointer, as does the
 673         * next_leaf on the previously last-extent-block. */
 674        fe->i_last_eb_blk = cpu_to_le64(new_last_eb_blk);
 675
 676        eb = (struct ocfs2_extent_block *) (*last_eb_bh)->b_data;
 677        eb->h_next_leaf_blk = cpu_to_le64(new_last_eb_blk);
 678
 679        status = ocfs2_journal_dirty(handle, *last_eb_bh);
 680        if (status < 0)
 681                mlog_errno(status);
 682        status = ocfs2_journal_dirty(handle, fe_bh);
 683        if (status < 0)
 684                mlog_errno(status);
 685        if (eb_bh) {
 686                status = ocfs2_journal_dirty(handle, eb_bh);
 687                if (status < 0)
 688                        mlog_errno(status);
 689        }
 690
 691        /*
 692         * Some callers want to track the rightmost leaf so pass it
 693         * back here.
 694         */
 695        brelse(*last_eb_bh);
 696        get_bh(new_eb_bhs[0]);
 697        *last_eb_bh = new_eb_bhs[0];
 698
 699        status = 0;
 700bail:
 701        if (new_eb_bhs) {
 702                for (i = 0; i < new_blocks; i++)
 703                        if (new_eb_bhs[i])
 704                                brelse(new_eb_bhs[i]);
 705                kfree(new_eb_bhs);
 706        }
 707
 708        mlog_exit(status);
 709        return status;
 710}
 711
 712/*
 713 * adds another level to the allocation tree.
 714 * returns back the new extent block so you can add a branch to it
 715 * after this call.
 716 */
 717static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
 718                                  handle_t *handle,
 719                                  struct inode *inode,
 720                                  struct buffer_head *fe_bh,
 721                                  struct ocfs2_alloc_context *meta_ac,
 722                                  struct buffer_head **ret_new_eb_bh)
 723{
 724        int status, i;
 725        u32 new_clusters;
 726        struct buffer_head *new_eb_bh = NULL;
 727        struct ocfs2_dinode *fe;
 728        struct ocfs2_extent_block *eb;
 729        struct ocfs2_extent_list  *fe_el;
 730        struct ocfs2_extent_list  *eb_el;
 731
 732        mlog_entry_void();
 733
 734        status = ocfs2_create_new_meta_bhs(osb, handle, inode, 1, meta_ac,
 735                                           &new_eb_bh);
 736        if (status < 0) {
 737                mlog_errno(status);
 738                goto bail;
 739        }
 740
 741        eb = (struct ocfs2_extent_block *) new_eb_bh->b_data;
 742        if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
 743                OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
 744                status = -EIO;
 745                goto bail;
 746        }
 747
 748        eb_el = &eb->h_list;
 749        fe = (struct ocfs2_dinode *) fe_bh->b_data;
 750        fe_el = &fe->id2.i_list;
 751
 752        status = ocfs2_journal_access(handle, inode, new_eb_bh,
 753                                      OCFS2_JOURNAL_ACCESS_CREATE);
 754        if (status < 0) {
 755                mlog_errno(status);
 756                goto bail;
 757        }
 758
 759        /* copy the fe data into the new extent block */
 760        eb_el->l_tree_depth = fe_el->l_tree_depth;
 761        eb_el->l_next_free_rec = fe_el->l_next_free_rec;
 762        for(i = 0; i < le16_to_cpu(fe_el->l_next_free_rec); i++)
 763                eb_el->l_recs[i] = fe_el->l_recs[i];
 764
 765        status = ocfs2_journal_dirty(handle, new_eb_bh);
 766        if (status < 0) {
 767                mlog_errno(status);
 768                goto bail;
 769        }
 770
 771        status = ocfs2_journal_access(handle, inode, fe_bh,
 772                                      OCFS2_JOURNAL_ACCESS_WRITE);
 773        if (status < 0) {
 774                mlog_errno(status);
 775                goto bail;
 776        }
 777
 778        new_clusters = ocfs2_sum_rightmost_rec(eb_el);
 779
 780        /* update fe now */
 781        le16_add_cpu(&fe_el->l_tree_depth, 1);
 782        fe_el->l_recs[0].e_cpos = 0;
 783        fe_el->l_recs[0].e_blkno = eb->h_blkno;
 784        fe_el->l_recs[0].e_int_clusters = cpu_to_le32(new_clusters);
 785        for(i = 1; i < le16_to_cpu(fe_el->l_next_free_rec); i++)
 786                memset(&fe_el->l_recs[i], 0, sizeof(struct ocfs2_extent_rec));
 787        fe_el->l_next_free_rec = cpu_to_le16(1);
 788
 789        /* If this is our 1st tree depth shift, then last_eb_blk
 790         * becomes the allocated extent block */
 791        if (fe_el->l_tree_depth == cpu_to_le16(1))
 792                fe->i_last_eb_blk = eb->h_blkno;
 793
 794        status = ocfs2_journal_dirty(handle, fe_bh);
 795        if (status < 0) {
 796                mlog_errno(status);
 797                goto bail;
 798        }
 799
 800        *ret_new_eb_bh = new_eb_bh;
 801        new_eb_bh = NULL;
 802        status = 0;
 803bail:
 804        if (new_eb_bh)
 805                brelse(new_eb_bh);
 806
 807        mlog_exit(status);
 808        return status;
 809}
 810
 811/*
 812 * Should only be called when there is no space left in any of the
 813 * leaf nodes. What we want to do is find the lowest tree depth
 814 * non-leaf extent block with room for new records. There are three
 815 * valid results of this search:
 816 *
 817 * 1) a lowest extent block is found, then we pass it back in
 818 *    *lowest_eb_bh and return '0'
 819 *
 820 * 2) the search fails to find anything, but the dinode has room. We
 821 *    pass NULL back in *lowest_eb_bh, but still return '0'
 822 *
 823 * 3) the search fails to find anything AND the dinode is full, in
 824 *    which case we return > 0
 825 *
 826 * return status < 0 indicates an error.
 827 */
 828static int ocfs2_find_branch_target(struct ocfs2_super *osb,
 829                                    struct inode *inode,
 830                                    struct buffer_head *fe_bh,
 831                                    struct buffer_head **target_bh)
 832{
 833        int status = 0, i;
 834        u64 blkno;
 835        struct ocfs2_dinode *fe;
 836        struct ocfs2_extent_block *eb;
 837        struct ocfs2_extent_list  *el;
 838        struct buffer_head *bh = NULL;
 839        struct buffer_head *lowest_bh = NULL;
 840
 841        mlog_entry_void();
 842
 843        *target_bh = NULL;
 844
 845        fe = (struct ocfs2_dinode *) fe_bh->b_data;
 846        el = &fe->id2.i_list;
 847
 848        while(le16_to_cpu(el->l_tree_depth) > 1) {
 849                if (le16_to_cpu(el->l_next_free_rec) == 0) {
 850                        ocfs2_error(inode->i_sb, "Dinode %llu has empty "
 851                                    "extent list (next_free_rec == 0)",
 852                                    (unsigned long long)OCFS2_I(inode)->ip_blkno);
 853                        status = -EIO;
 854                        goto bail;
 855                }
 856                i = le16_to_cpu(el->l_next_free_rec) - 1;
 857                blkno = le64_to_cpu(el->l_recs[i].e_blkno);
 858                if (!blkno) {
 859                        ocfs2_error(inode->i_sb, "Dinode %llu has extent "
 860                                    "list where extent # %d has no physical "
 861                                    "block start",
 862                                    (unsigned long long)OCFS2_I(inode)->ip_blkno, i);
 863                        status = -EIO;
 864                        goto bail;
 865                }
 866
 867                if (bh) {
 868                        brelse(bh);
 869                        bh = NULL;
 870                }
 871
 872                status = ocfs2_read_block(osb, blkno, &bh, OCFS2_BH_CACHED,
 873                                          inode);
 874                if (status < 0) {
 875                        mlog_errno(status);
 876                        goto bail;
 877                }
 878
 879                eb = (struct ocfs2_extent_block *) bh->b_data;
 880                if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
 881                        OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
 882                        status = -EIO;
 883                        goto bail;
 884                }
 885                el = &eb->h_list;
 886
 887                if (le16_to_cpu(el->l_next_free_rec) <
 888                    le16_to_cpu(el->l_count)) {
 889                        if (lowest_bh)
 890                                brelse(lowest_bh);
 891                        lowest_bh = bh;
 892                        get_bh(lowest_bh);
 893                }
 894        }
 895
 896        /* If we didn't find one and the fe doesn't have any room,
 897         * then return '1' */
 898        if (!lowest_bh
 899            && (fe->id2.i_list.l_next_free_rec == fe->id2.i_list.l_count))
 900                status = 1;
 901
 902        *target_bh = lowest_bh;
 903bail:
 904        if (bh)
 905                brelse(bh);
 906
 907        mlog_exit(status);
 908        return status;
 909}
 910
 911/*
 912 * Grow a b-tree so that it has more records.
 913 *
 914 * We might shift the tree depth in which case existing paths should
 915 * be considered invalid.
 916 *
 917 * Tree depth after the grow is returned via *final_depth.
 918 *
 919 * *last_eb_bh will be updated by ocfs2_add_branch().
 920 */
 921static int ocfs2_grow_tree(struct inode *inode, handle_t *handle,
 922                           struct buffer_head *di_bh, int *final_depth,
 923                           struct buffer_head **last_eb_bh,
 924                           struct ocfs2_alloc_context *meta_ac)
 925{
 926        int ret, shift;
 927        struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
 928        int depth = le16_to_cpu(di->id2.i_list.l_tree_depth);
 929        struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 930        struct buffer_head *bh = NULL;
 931
 932        BUG_ON(meta_ac == NULL);
 933
 934        shift = ocfs2_find_branch_target(osb, inode, di_bh, &bh);
 935        if (shift < 0) {
 936                ret = shift;
 937                mlog_errno(ret);
 938                goto out;
 939        }
 940
 941        /* We traveled all the way to the bottom of the allocation tree
 942         * and didn't find room for any more extents - we need to add
 943         * another tree level */
 944        if (shift) {
 945                BUG_ON(bh);
 946                mlog(0, "need to shift tree depth (current = %d)\n", depth);
 947
 948                /* ocfs2_shift_tree_depth will return us a buffer with
 949                 * the new extent block (so we can pass that to
 950                 * ocfs2_add_branch). */
 951                ret = ocfs2_shift_tree_depth(osb, handle, inode, di_bh,
 952                                             meta_ac, &bh);
 953                if (ret < 0) {
 954                        mlog_errno(ret);
 955                        goto out;
 956                }
 957                depth++;
 958                if (depth == 1) {
 959                        /*
 960                         * Special case: we have room now if we shifted from
 961                         * tree_depth 0, so no more work needs to be done.
 962                         *
 963                         * We won't be calling add_branch, so pass
 964                         * back *last_eb_bh as the new leaf. At depth
 965                         * zero, it should always be null so there's
 966                         * no reason to brelse.
 967                         */
 968                        BUG_ON(*last_eb_bh);
 969                        get_bh(bh);
 970                        *last_eb_bh = bh;
 971                        goto out;
 972                }
 973        }
 974
 975        /* call ocfs2_add_branch to add the final part of the tree with
 976         * the new data. */
 977        mlog(0, "add branch. bh = %p\n", bh);
 978        ret = ocfs2_add_branch(osb, handle, inode, di_bh, bh, last_eb_bh,
 979                               meta_ac);
 980        if (ret < 0) {
 981                mlog_errno(ret);
 982                goto out;
 983        }
 984
 985out:
 986        if (final_depth)
 987                *final_depth = depth;
 988        brelse(bh);
 989        return ret;
 990}
 991
 992/*
 993 * This is only valid for leaf nodes, which are the only ones that can
 994 * have empty extents anyway.
 995 */
 996static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec)
 997{
 998        return !rec->e_leaf_clusters;
 999}
1000
1001/*
1002 * This function will discard the rightmost extent record.
1003 */
1004static void ocfs2_shift_records_right(struct ocfs2_extent_list *el)
1005{
1006        int next_free = le16_to_cpu(el->l_next_free_rec);
1007        int count = le16_to_cpu(el->l_count);
1008        unsigned int num_bytes;
1009
1010        BUG_ON(!next_free);
1011        /* This will cause us to go off the end of our extent list. */
1012        BUG_ON(next_free >= count);
1013
1014        num_bytes = sizeof(struct ocfs2_extent_rec) * next_free;
1015
1016        memmove(&el->l_recs[1], &el->l_recs[0], num_bytes);
1017}
1018
1019static void ocfs2_rotate_leaf(struct ocfs2_extent_list *el,
1020                              struct ocfs2_extent_rec *insert_rec)
1021{
1022        int i, insert_index, next_free, has_empty, num_bytes;
1023        u32 insert_cpos = le32_to_cpu(insert_rec->e_cpos);
1024        struct ocfs2_extent_rec *rec;
1025
1026        next_free = le16_to_cpu(el->l_next_free_rec);
1027        has_empty = ocfs2_is_empty_extent(&el->l_recs[0]);
1028
1029        BUG_ON(!next_free);
1030
1031        /* The tree code before us didn't allow enough room in the leaf. */
1032        if (el->l_next_free_rec == el->l_count && !has_empty)
1033                BUG();
1034
1035        /*
1036         * The easiest way to approach this is to just remove the
1037         * empty extent and temporarily decrement next_free.
1038         */
1039        if (has_empty) {
1040                /*
1041                 * If next_free was 1 (only an empty extent), this
1042                 * loop won't execute, which is fine. We still want
1043                 * the decrement above to happen.
1044                 */
1045                for(i = 0; i < (next_free - 1); i++)
1046                        el->l_recs[i] = el->l_recs[i+1];
1047
1048                next_free--;
1049        }
1050
1051        /*
1052         * Figure out what the new record index should be.
1053         */
1054        for(i = 0; i < next_free; i++) {
1055                rec = &el->l_recs[i];
1056
1057                if (insert_cpos < le32_to_cpu(rec->e_cpos))
1058                        break;
1059        }
1060        insert_index = i;
1061
1062        mlog(0, "ins %u: index %d, has_empty %d, next_free %d, count %d\n",
1063             insert_cpos, insert_index, has_empty, next_free, le16_to_cpu(el->l_count));
1064
1065        BUG_ON(insert_index < 0);
1066        BUG_ON(insert_index >= le16_to_cpu(el->l_count));
1067        BUG_ON(insert_index > next_free);
1068
1069        /*
1070         * No need to memmove if we're just adding to the tail.
1071         */
1072        if (insert_index != next_free) {
1073                BUG_ON(next_free >= le16_to_cpu(el->l_count));
1074
1075                num_bytes = next_free - insert_index;
1076                num_bytes *= sizeof(struct ocfs2_extent_rec);
1077                memmove(&el->l_recs[insert_index + 1],
1078                        &el->l_recs[insert_index],
1079                        num_bytes);
1080        }
1081
1082        /*
1083         * Either we had an empty extent, and need to re-increment or
1084         * there was no empty extent on a non full rightmost leaf node,
1085         * in which case we still need to increment.
1086         */
1087        next_free++;
1088        el->l_next_free_rec = cpu_to_le16(next_free);
1089        /*
1090         * Make sure none of the math above just messed up our tree.
1091         */
1092        BUG_ON(le16_to_cpu(el->l_next_free_rec) > le16_to_cpu(el->l_count));
1093
1094        el->l_recs[insert_index] = *insert_rec;
1095
1096}
1097
1098static void ocfs2_remove_empty_extent(struct ocfs2_extent_list *el)
1099{
1100        int size, num_recs = le16_to_cpu(el->l_next_free_rec);
1101
1102        BUG_ON(num_recs == 0);
1103
1104        if (ocfs2_is_empty_extent(&el->l_recs[0])) {
1105                num_recs--;
1106                size = num_recs * sizeof(struct ocfs2_extent_rec);
1107                memmove(&el->l_recs[0], &el->l_recs[1], size);
1108                memset(&el->l_recs[num_recs], 0,
1109                       sizeof(struct ocfs2_extent_rec));
1110                el->l_next_free_rec = cpu_to_le16(num_recs);
1111        }
1112}
1113
1114/*
1115 * Create an empty extent record .
1116 *
1117 * l_next_free_rec may be updated.
1118 *
1119 * If an empty extent already exists do nothing.
1120 */
1121static void ocfs2_create_empty_extent(struct ocfs2_extent_list *el)
1122{
1123        int next_free = le16_to_cpu(el->l_next_free_rec);
1124
1125        BUG_ON(le16_to_cpu(el->l_tree_depth) != 0);
1126
1127        if (next_free == 0)
1128                goto set_and_inc;
1129
1130        if (ocfs2_is_empty_extent(&el->l_recs[0]))
1131                return;
1132
1133        mlog_bug_on_msg(el->l_count == el->l_next_free_rec,
1134                        "Asked to create an empty extent in a full list:\n"
1135                        "count = %u, tree depth = %u",
1136                        le16_to_cpu(el->l_count),
1137                        le16_to_cpu(el->l_tree_depth));
1138
1139        ocfs2_shift_records_right(el);
1140
1141set_and_inc:
1142        le16_add_cpu(&el->l_next_free_rec, 1);
1143        memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
1144}
1145
1146/*
1147 * For a rotation which involves two leaf nodes, the "root node" is
1148 * the lowest level tree node which contains a path to both leafs. This
1149 * resulting set of information can be used to form a complete "subtree"
1150 *
1151 * This function is passed two full paths from the dinode down to a
1152 * pair of adjacent leaves. It's task is to figure out which path
1153 * index contains the subtree root - this can be the root index itself
1154 * in a worst-case rotation.
1155 *
1156 * The array index of the subtree root is passed back.
1157 */
1158static int ocfs2_find_subtree_root(struct inode *inode,
1159                                   struct ocfs2_path *left,
1160                                   struct ocfs2_path *right)
1161{
1162        int i = 0;
1163
1164        /*
1165         * Check that the caller passed in two paths from the same tree.
1166         */
1167        BUG_ON(path_root_bh(left) != path_root_bh(right));
1168
1169        do {
1170                i++;
1171
1172                /*
1173                 * The caller didn't pass two adjacent paths.
1174                 */
1175                mlog_bug_on_msg(i > left->p_tree_depth,
1176                                "Inode %lu, left depth %u, right depth %u\n"
1177                                "left leaf blk %llu, right leaf blk %llu\n",
1178                                inode->i_ino, left->p_tree_depth,
1179                                right->p_tree_depth,
1180                                (unsigned long long)path_leaf_bh(left)->b_blocknr,
1181                                (unsigned long long)path_leaf_bh(right)->b_blocknr);
1182        } while (left->p_node[i].bh->b_blocknr ==
1183                 right->p_node[i].bh->b_blocknr);
1184
1185        return i - 1;
1186}
1187
1188typedef void (path_insert_t)(void *, struct buffer_head *);
1189
1190/*
1191 * Traverse a btree path in search of cpos, starting at root_el.
1192 *
1193 * This code can be called with a cpos larger than the tree, in which
1194 * case it will return the rightmost path.
1195 */
1196static int __ocfs2_find_path(struct inode *inode,
1197                             struct ocfs2_extent_list *root_el, u32 cpos,
1198                             path_insert_t *func, void *data)
1199{
1200        int i, ret = 0;
1201        u32 range;
1202        u64 blkno;
1203        struct buffer_head *bh = NULL;
1204        struct ocfs2_extent_block *eb;
1205        struct ocfs2_extent_list *el;
1206        struct ocfs2_extent_rec *rec;
1207        struct ocfs2_inode_info *oi = OCFS2_I(inode);
1208
1209        el = root_el;
1210        while (el->l_tree_depth) {
1211                if (le16_to_cpu(el->l_next_free_rec) == 0) {
1212                        ocfs2_error(inode->i_sb,
1213                                    "Inode %llu has empty extent list at "
1214                                    "depth %u\n",
1215                                    (unsigned long long)oi->ip_blkno,
1216                                    le16_to_cpu(el->l_tree_depth));
1217                        ret = -EROFS;
1218                        goto out;
1219
1220                }
1221
1222                for(i = 0; i < le16_to_cpu(el->l_next_free_rec) - 1; i++) {
1223                        rec = &el->l_recs[i];
1224
1225                        /*
1226                         * In the case that cpos is off the allocation
1227                         * tree, this should just wind up returning the
1228                         * rightmost record.
1229                         */
1230                        range = le32_to_cpu(rec->e_cpos) +
1231                                ocfs2_rec_clusters(el, rec);
1232                        if (cpos >= le32_to_cpu(rec->e_cpos) && cpos < range)
1233                            break;
1234                }
1235
1236                blkno = le64_to_cpu(el->l_recs[i].e_blkno);
1237                if (blkno == 0) {
1238                        ocfs2_error(inode->i_sb,
1239                                    "Inode %llu has bad blkno in extent list "
1240                                    "at depth %u (index %d)\n",
1241                                    (unsigned long long)oi->ip_blkno,
1242                                    le16_to_cpu(el->l_tree_depth), i);
1243                        ret = -EROFS;
1244                        goto out;
1245                }
1246
1247                brelse(bh);
1248                bh = NULL;
1249                ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), blkno,
1250                                       &bh, OCFS2_BH_CACHED, inode);
1251                if (ret) {
1252                        mlog_errno(ret);
1253                        goto out;
1254                }
1255
1256                eb = (struct ocfs2_extent_block *) bh->b_data;
1257                el = &eb->h_list;
1258                if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
1259                        OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
1260                        ret = -EIO;
1261                        goto out;
1262                }
1263
1264                if (le16_to_cpu(el->l_next_free_rec) >
1265                    le16_to_cpu(el->l_count)) {
1266                        ocfs2_error(inode->i_sb,
1267                                    "Inode %llu has bad count in extent list "
1268                                    "at block %llu (next free=%u, count=%u)\n",
1269                                    (unsigned long long)oi->ip_blkno,
1270                                    (unsigned long long)bh->b_blocknr,
1271                                    le16_to_cpu(el->l_next_free_rec),
1272                                    le16_to_cpu(el->l_count));
1273                        ret = -EROFS;
1274                        goto out;
1275                }
1276
1277                if (func)
1278                        func(data, bh);
1279        }
1280
1281out:
1282        /*
1283         * Catch any trailing bh that the loop didn't handle.
1284         */
1285        brelse(bh);
1286
1287        return ret;
1288}
1289
1290/*
1291 * Given an initialized path (that is, it has a valid root extent
1292 * list), this function will traverse the btree in search of the path
1293 * which would contain cpos.
1294 *
1295 * The path traveled is recorded in the path structure.
1296 *
1297 * Note that this will not do any comparisons on leaf node extent
1298 * records, so it will work fine in the case that we just added a tree
1299 * branch.
1300 */
1301struct find_path_data {
1302        int index;
1303        struct ocfs2_path *path;
1304};
1305static void find_path_ins(void *data, struct buffer_head *bh)
1306{
1307        struct find_path_data *fp = data;
1308
1309        get_bh(bh);
1310        ocfs2_path_insert_eb(fp->path, fp->index, bh);
1311        fp->index++;
1312}
1313static int ocfs2_find_path(struct inode *inode, struct ocfs2_path *path,
1314                           u32 cpos)
1315{
1316        struct find_path_data data;
1317
1318        data.index = 1;
1319        data.path = path;
1320        return __ocfs2_find_path(inode, path_root_el(path), cpos,
1321                                 find_path_ins, &data);
1322}
1323
1324static void find_leaf_ins(void *data, struct buffer_head *bh)
1325{
1326        struct ocfs2_extent_block *eb =(struct ocfs2_extent_block *)bh->b_data;
1327        struct ocfs2_extent_list *el = &eb->h_list;
1328        struct buffer_head **ret = data;
1329
1330        /* We want to retain only the leaf block. */
1331        if (le16_to_cpu(el->l_tree_depth) == 0) {
1332                get_bh(bh);
1333                *ret = bh;
1334        }
1335}
1336/*
1337 * Find the leaf block in the tree which would contain cpos. No
1338 * checking of the actual leaf is done.
1339 *
1340 * Some paths want to call this instead of allocating a path structure
1341 * and calling ocfs2_find_path().
1342 *
1343 * This function doesn't handle non btree extent lists.
1344 */
1345int ocfs2_find_leaf(struct inode *inode, struct ocfs2_extent_list *root_el,
1346                    u32 cpos, struct buffer_head **leaf_bh)
1347{
1348        int ret;
1349        struct buffer_head *bh = NULL;
1350
1351        ret = __ocfs2_find_path(inode, root_el, cpos, find_leaf_ins, &bh);
1352        if (ret) {
1353                mlog_errno(ret);
1354                goto out;
1355        }
1356
1357        *leaf_bh = bh;
1358out:
1359        return ret;
1360}
1361
1362/*
1363 * Adjust the adjacent records (left_rec, right_rec) involved in a rotation.
1364 *
1365 * Basically, we've moved stuff around at the bottom of the tree and
1366 * we need to fix up the extent records above the changes to reflect
1367 * the new changes.
1368 *
1369 * left_rec: the record on the left.
1370 * left_child_el: is the child list pointed to by left_rec
1371 * right_rec: the record to the right of left_rec
1372 * right_child_el: is the child list pointed to by right_rec
1373 *
1374 * By definition, this only works on interior nodes.
1375 */
1376static void ocfs2_adjust_adjacent_records(struct ocfs2_extent_rec *left_rec,
1377                                  struct ocfs2_extent_list *left_child_el,
1378                                  struct ocfs2_extent_rec *right_rec,
1379                                  struct ocfs2_extent_list *right_child_el)
1380{
1381        u32 left_clusters, right_end;
1382
1383        /*
1384         * Interior nodes never have holes. Their cpos is the cpos of
1385         * the leftmost record in their child list. Their cluster
1386         * count covers the full theoretical range of their child list
1387         * - the range between their cpos and the cpos of the record
1388         * immediately to their right.
1389         */
1390        left_clusters = le32_to_cpu(right_child_el->l_recs[0].e_cpos);
1391        if (ocfs2_is_empty_extent(&right_child_el->l_recs[0])) {
1392                BUG_ON(le16_to_cpu(right_child_el->l_next_free_rec) <= 1);
1393                left_clusters = le32_to_cpu(right_child_el->l_recs[1].e_cpos);
1394        }
1395        left_clusters -= le32_to_cpu(left_rec->e_cpos);
1396        left_rec->e_int_clusters = cpu_to_le32(left_clusters);
1397
1398        /*
1399         * Calculate the rightmost cluster count boundary before
1400         * moving cpos - we will need to adjust clusters after
1401         * updating e_cpos to keep the same highest cluster count.
1402         */
1403        right_end = le32_to_cpu(right_rec->e_cpos);
1404        right_end += le32_to_cpu(right_rec->e_int_clusters);
1405
1406        right_rec->e_cpos = left_rec->e_cpos;
1407        le32_add_cpu(&right_rec->e_cpos, left_clusters);
1408
1409        right_end -= le32_to_cpu(right_rec->e_cpos);
1410        right_rec->e_int_clusters = cpu_to_le32(right_end);
1411}
1412
1413/*
1414 * Adjust the adjacent root node records involved in a
1415 * rotation. left_el_blkno is passed in as a key so that we can easily
1416 * find it's index in the root list.
1417 */
1418static void ocfs2_adjust_root_records(struct ocfs2_extent_list *root_el,
1419                                      struct ocfs2_extent_list *left_el,
1420                                      struct ocfs2_extent_list *right_el,
1421                                      u64 left_el_blkno)
1422{
1423        int i;
1424
1425        BUG_ON(le16_to_cpu(root_el->l_tree_depth) <=
1426               le16_to_cpu(left_el->l_tree_depth));
1427
1428        for(i = 0; i < le16_to_cpu(root_el->l_next_free_rec) - 1; i++) {
1429                if (le64_to_cpu(root_el->l_recs[i].e_blkno) == left_el_blkno)
1430                        break;
1431        }
1432
1433        /*
1434         * The path walking code should have never returned a root and
1435         * two paths which are not adjacent.
1436         */
1437        BUG_ON(i >= (le16_to_cpu(root_el->l_next_free_rec) - 1));
1438
1439        ocfs2_adjust_adjacent_records(&root_el->l_recs[i], left_el,
1440                                      &root_el->l_recs[i + 1], right_el);
1441}
1442
1443/*
1444 * We've changed a leaf block (in right_path) and need to reflect that
1445 * change back up the subtree.
1446 *
1447 * This happens in multiple places:
1448 *   - When we've moved an extent record from the left path leaf to the right
1449 *     path leaf to make room for an empty extent in the left path leaf.
1450 *   - When our insert into the right path leaf is at the leftmost edge
1451 *     and requires an update of the path immediately to it's left. This
1452 *     can occur at the end of some types of rotation and appending inserts.
1453 */
1454static void ocfs2_complete_edge_insert(struct inode *inode, handle_t *handle,
1455                                       struct ocfs2_path *left_path,
1456                                       struct ocfs2_path *right_path,
1457                                       int subtree_index)
1458{
1459        int ret, i, idx;
1460        struct ocfs2_extent_list *el, *left_el, *right_el;
1461        struct ocfs2_extent_rec *left_rec, *right_rec;
1462        struct buffer_head *root_bh = left_path->p_node[subtree_index].bh;
1463
1464        /*
1465         * Update the counts and position values within all the
1466         * interior nodes to reflect the leaf rotation we just did.
1467         *
1468         * The root node is handled below the loop.
1469         *
1470         * We begin the loop with right_el and left_el pointing to the
1471         * leaf lists and work our way up.
1472         *
1473         * NOTE: within this loop, left_el and right_el always refer
1474         * to the *child* lists.
1475         */
1476        left_el = path_leaf_el(left_path);
1477        right_el = path_leaf_el(right_path);
1478        for(i = left_path->p_tree_depth - 1; i > subtree_index; i--) {
1479                mlog(0, "Adjust records at index %u\n", i);
1480
1481                /*
1482                 * One nice property of knowing that all of these
1483                 * nodes are below the root is that we only deal with
1484                 * the leftmost right node record and the rightmost
1485                 * left node record.
1486                 */
1487                el = left_path->p_node[i].el;
1488                idx = le16_to_cpu(left_el->l_next_free_rec) - 1;
1489                left_rec = &el->l_recs[idx];
1490
1491                el = right_path->p_node[i].el;
1492                right_rec = &el->l_recs[0];
1493
1494                ocfs2_adjust_adjacent_records(left_rec, left_el, right_rec,
1495                                              right_el);
1496
1497                ret = ocfs2_journal_dirty(handle, left_path->p_node[i].bh);
1498                if (ret)
1499                        mlog_errno(ret);
1500
1501                ret = ocfs2_journal_dirty(handle, right_path->p_node[i].bh);
1502                if (ret)
1503                        mlog_errno(ret);
1504
1505                /*
1506                 * Setup our list pointers now so that the current
1507                 * parents become children in the next iteration.
1508                 */
1509                left_el = left_path->p_node[i].el;
1510                right_el = right_path->p_node[i].el;
1511        }
1512
1513        /*
1514         * At the root node, adjust the two adjacent records which
1515         * begin our path to the leaves.
1516         */
1517
1518        el = left_path->p_node[subtree_index].el;
1519        left_el = left_path->p_node[subtree_index + 1].el;
1520        right_el = right_path->p_node[subtree_index + 1].el;
1521
1522        ocfs2_adjust_root_records(el, left_el, right_el,
1523                                  left_path->p_node[subtree_index + 1].bh->b_blocknr);
1524
1525        root_bh = left_path->p_node[subtree_index].bh;
1526
1527        ret = ocfs2_journal_dirty(handle, root_bh);
1528        if (ret)
1529                mlog_errno(ret);
1530}
1531
1532static int ocfs2_rotate_subtree_right(struct inode *inode,
1533                                      handle_t *handle,
1534                                      struct ocfs2_path *left_path,
1535                                      struct ocfs2_path *right_path,
1536                                      int subtree_index)
1537{
1538        int ret, i;
1539        struct buffer_head *right_leaf_bh;
1540        struct buffer_head *left_leaf_bh = NULL;
1541        struct buffer_head *root_bh;
1542        struct ocfs2_extent_list *right_el, *left_el;
1543        struct ocfs2_extent_rec move_rec;
1544
1545        left_leaf_bh = path_leaf_bh(left_path);
1546        left_el = path_leaf_el(left_path);
1547
1548        if (left_el->l_next_free_rec != left_el->l_count) {
1549                ocfs2_error(inode->i_sb,
1550                            "Inode %llu has non-full interior leaf node %llu"
1551                            "(next free = %u)",
1552                            (unsigned long long)OCFS2_I(inode)->ip_blkno,
1553                            (unsigned long long)left_leaf_bh->b_blocknr,
1554                            le16_to_cpu(left_el->l_next_free_rec));
1555                return -EROFS;
1556        }
1557
1558        /*
1559         * This extent block may already have an empty record, so we
1560         * return early if so.
1561         */
1562        if (ocfs2_is_empty_extent(&left_el->l_recs[0]))
1563                return 0;
1564
1565        root_bh = left_path->p_node[subtree_index].bh;
1566        BUG_ON(root_bh != right_path->p_node[subtree_index].bh);
1567
1568        ret = ocfs2_journal_access(handle, inode, root_bh,
1569                                   OCFS2_JOURNAL_ACCESS_WRITE);
1570        if (ret) {
1571                mlog_errno(ret);
1572                goto out;
1573        }
1574
1575        for(i = subtree_index + 1; i < path_num_items(right_path); i++) {
1576                ret = ocfs2_journal_access(handle, inode,
1577                                           right_path->p_node[i].bh,
1578                                           OCFS2_JOURNAL_ACCESS_WRITE);
1579                if (ret) {
1580                        mlog_errno(ret);
1581                        goto out;
1582                }
1583
1584                ret = ocfs2_journal_access(handle, inode,
1585                                           left_path->p_node[i].bh,
1586                                           OCFS2_JOURNAL_ACCESS_WRITE);
1587                if (ret) {
1588                        mlog_errno(ret);
1589                        goto out;
1590                }
1591        }
1592
1593        right_leaf_bh = path_leaf_bh(right_path);
1594        right_el = path_leaf_el(right_path);
1595
1596        /* This is a code error, not a disk corruption. */
1597        mlog_bug_on_msg(!right_el->l_next_free_rec, "Inode %llu: Rotate fails "
1598                        "because rightmost leaf block %llu is empty\n",
1599                        (unsigned long long)OCFS2_I(inode)->ip_blkno,
1600                        (unsigned long long)right_leaf_bh->b_blocknr);
1601
1602        ocfs2_create_empty_extent(right_el);
1603
1604        ret = ocfs2_journal_dirty(handle, right_leaf_bh);
1605        if (ret) {
1606                mlog_errno(ret);
1607                goto out;
1608        }
1609
1610        /* Do the copy now. */
1611        i = le16_to_cpu(left_el->l_next_free_rec) - 1;
1612        move_rec = left_el->l_recs[i];
1613        right_el->l_recs[0] = move_rec;
1614
1615        /*
1616         * Clear out the record we just copied and shift everything
1617         * over, leaving an empty extent in the left leaf.
1618         *
1619         * We temporarily subtract from next_free_rec so that the
1620         * shift will lose the tail record (which is now defunct).
1621         */
1622        le16_add_cpu(&left_el->l_next_free_rec, -1);
1623        ocfs2_shift_records_right(left_el);
1624        memset(&left_el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
1625        le16_add_cpu(&left_el->l_next_free_rec, 1);
1626
1627        ret = ocfs2_journal_dirty(handle, left_leaf_bh);
1628        if (ret) {
1629                mlog_errno(ret);
1630                goto out;
1631        }
1632
1633        ocfs2_complete_edge_insert(inode, handle, left_path, right_path,
1634                                subtree_index);
1635
1636out:
1637        return ret;
1638}
1639
1640/*
1641 * Given a full path, determine what cpos value would return us a path
1642 * containing the leaf immediately to the left of the current one.
1643 *
1644 * Will return zero if the path passed in is already the leftmost path.
1645 */
1646static int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
1647                                         struct ocfs2_path *path, u32 *cpos)
1648{
1649        int i, j, ret = 0;
1650        u64 blkno;
1651        struct ocfs2_extent_list *el;
1652
1653        BUG_ON(path->p_tree_depth == 0);
1654
1655        *cpos = 0;
1656
1657        blkno = path_leaf_bh(path)->b_blocknr;
1658
1659        /* Start at the tree node just above the leaf and work our way up. */
1660        i = path->p_tree_depth - 1;
1661        while (i >= 0) {
1662                el = path->p_node[i].el;
1663
1664                /*
1665                 * Find the extent record just before the one in our
1666                 * path.
1667                 */
1668                for(j = 0; j < le16_to_cpu(el->l_next_free_rec); j++) {
1669                        if (le64_to_cpu(el->l_recs[j].e_blkno) == blkno) {
1670                                if (j == 0) {
1671                                        if (i == 0) {
1672                                                /*
1673                                                 * We've determined that the
1674                                                 * path specified is already
1675                                                 * the leftmost one - return a
1676                                                 * cpos of zero.
1677                                                 */
1678                                                goto out;
1679                                        }
1680                                        /*
1681                                         * The leftmost record points to our
1682                                         * leaf - we need to travel up the
1683                                         * tree one level.
1684                                         */
1685                                        goto next_node;
1686                                }
1687
1688                                *cpos = le32_to_cpu(el->l_recs[j - 1].e_cpos);
1689                                *cpos = *cpos + ocfs2_rec_clusters(el,
1690                                                           &el->l_recs[j - 1]);
1691                                *cpos = *cpos - 1;
1692                                goto out;
1693                        }
1694                }
1695
1696                /*
1697                 * If we got here, we never found a valid node where
1698                 * the tree indicated one should be.
1699                 */
1700                ocfs2_error(sb,
1701                            "Invalid extent tree at extent block %llu\n",
1702                            (unsigned long long)blkno);
1703                ret = -EROFS;
1704                goto out;
1705
1706next_node:
1707                blkno = path->p_node[i].bh->b_blocknr;
1708                i--;
1709        }
1710
1711out:
1712        return ret;
1713}
1714
1715/*
1716 * Extend the transaction by enough credits to complete the rotation,
1717 * and still leave at least the original number of credits allocated
1718 * to this transaction.
1719 */
1720static int ocfs2_extend_rotate_transaction(handle_t *handle, int subtree_depth,
1721                                           int op_credits,
1722                                           struct ocfs2_path *path)
1723{
1724        int credits = (path->p_tree_depth - subtree_depth) * 2 + 1 + op_credits;
1725
1726        if (handle->h_buffer_credits < credits)
1727                return ocfs2_extend_trans(handle, credits);
1728
1729        return 0;
1730}
1731
1732/*
1733 * Trap the case where we're inserting into the theoretical range past
1734 * the _actual_ left leaf range. Otherwise, we'll rotate a record
1735 * whose cpos is less than ours into the right leaf.
1736 *
1737 * It's only necessary to look at the rightmost record of the left
1738 * leaf because the logic that calls us should ensure that the
1739 * theoretical ranges in the path components above the leaves are
1740 * correct.
1741 */
1742static int ocfs2_rotate_requires_path_adjustment(struct ocfs2_path *left_path,
1743                                                 u32 insert_cpos)
1744{
1745        struct ocfs2_extent_list *left_el;
1746        struct ocfs2_extent_rec *rec;
1747        int next_free;
1748
1749        left_el = path_leaf_el(left_path);
1750        next_free = le16_to_cpu(left_el->l_next_free_rec);
1751        rec = &left_el->l_recs[next_free - 1];
1752
1753        if (insert_cpos > le32_to_cpu(rec->e_cpos))
1754                return 1;
1755        return 0;
1756}
1757
1758static int ocfs2_leftmost_rec_contains(struct ocfs2_extent_list *el, u32 cpos)
1759{
1760        int next_free = le16_to_cpu(el->l_next_free_rec);
1761        unsigned int range;
1762        struct ocfs2_extent_rec *rec;
1763
1764        if (next_free == 0)
1765                return 0;
1766
1767        rec = &el->l_recs[0];
1768        if (ocfs2_is_empty_extent(rec)) {
1769                /* Empty list. */
1770                if (next_free == 1)
1771                        return 0;
1772                rec = &el->l_recs[1];
1773        }
1774
1775        range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
1776        if (cpos >= le32_to_cpu(rec->e_cpos) && cpos < range)
1777                return 1;
1778        return 0;
1779}
1780
1781/*
1782 * Rotate all the records in a btree right one record, starting at insert_cpos.
1783 *
1784 * The path to the rightmost leaf should be passed in.
1785 *
1786 * The array is assumed to be large enough to hold an entire path (tree depth).
1787 *
1788 * Upon succesful return from this function:
1789 *
1790 * - The 'right_path' array will contain a path to the leaf block
1791 *   whose range contains e_cpos.
1792 * - That leaf block will have a single empty extent in list index 0.
1793 * - In the case that the rotation requires a post-insert update,
1794 *   *ret_left_path will contain a valid path which can be passed to
1795 *   ocfs2_insert_path().
1796 */
1797static int ocfs2_rotate_tree_right(struct inode *inode,
1798                                   handle_t *handle,
1799                                   enum ocfs2_split_type split,
1800                                   u32 insert_cpos,
1801                                   struct ocfs2_path *right_path,
1802                                   struct ocfs2_path **ret_left_path)
1803{
1804        int ret, start, orig_credits = handle->h_buffer_credits;
1805        u32 cpos;
1806        struct ocfs2_path *left_path = NULL;
1807
1808        *ret_left_path = NULL;
1809
1810        left_path = ocfs2_new_path(path_root_bh(right_path),
1811                                   path_root_el(right_path));
1812        if (!left_path) {
1813                ret = -ENOMEM;
1814                mlog_errno(ret);
1815                goto out;
1816        }
1817
1818        ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, right_path, &cpos);
1819        if (ret) {
1820                mlog_errno(ret);
1821                goto out;
1822        }
1823
1824        mlog(0, "Insert: %u, first left path cpos: %u\n", insert_cpos, cpos);
1825
1826        /*
1827         * What we want to do here is:
1828         *
1829         * 1) Start with the rightmost path.
1830         *
1831         * 2) Determine a path to the leaf block directly to the left
1832         *    of that leaf.
1833         *
1834         * 3) Determine the 'subtree root' - the lowest level tree node
1835         *    which contains a path to both leaves.
1836         *
1837         * 4) Rotate the subtree.
1838         *
1839         * 5) Find the next subtree by considering the left path to be
1840         *    the new right path.
1841         *
1842         * The check at the top of this while loop also accepts
1843         * insert_cpos == cpos because cpos is only a _theoretical_
1844         * value to get us the left path - insert_cpos might very well
1845         * be filling that hole.
1846         *
1847         * Stop at a cpos of '0' because we either started at the
1848         * leftmost branch (i.e., a tree with one branch and a
1849         * rotation inside of it), or we've gone as far as we can in
1850         * rotating subtrees.
1851         */
1852        while (cpos && insert_cpos <= cpos) {
1853                mlog(0, "Rotating a tree: ins. cpos: %u, left path cpos: %u\n",
1854                     insert_cpos, cpos);
1855
1856                ret = ocfs2_find_path(inode, left_path, cpos);
1857                if (ret) {
1858                        mlog_errno(ret);
1859                        goto out;
1860                }
1861
1862                mlog_bug_on_msg(path_leaf_bh(left_path) ==
1863                                path_leaf_bh(right_path),
1864                                "Inode %lu: error during insert of %u "
1865                                "(left path cpos %u) results in two identical "
1866                                "paths ending at %llu\n",
1867                                inode->i_ino, insert_cpos, cpos,
1868                                (unsigned long long)
1869                                path_leaf_bh(left_path)->b_blocknr);
1870
1871                if (split == SPLIT_NONE &&
1872                    ocfs2_rotate_requires_path_adjustment(left_path,
1873                                                          insert_cpos)) {
1874
1875                        /*
1876                         * We've rotated the tree as much as we
1877                         * should. The rest is up to
1878                         * ocfs2_insert_path() to complete, after the
1879                         * record insertion. We indicate this
1880                         * situation by returning the left path.
1881                         *
1882                         * The reason we don't adjust the records here
1883                         * before the record insert is that an error
1884                         * later might break the rule where a parent
1885                         * record e_cpos will reflect the actual
1886                         * e_cpos of the 1st nonempty record of the
1887                         * child list.
1888                         */
1889                        *ret_left_path = left_path;
1890                        goto out_ret_path;
1891                }
1892
1893                start = ocfs2_find_subtree_root(inode, left_path, right_path);
1894
1895                mlog(0, "Subtree root at index %d (blk %llu, depth %d)\n",
1896                     start,
1897                     (unsigned long long) right_path->p_node[start].bh->b_blocknr,
1898                     right_path->p_tree_depth);
1899
1900                ret = ocfs2_extend_rotate_transaction(handle, start,
1901                                                      orig_credits, right_path);
1902                if (ret) {
1903                        mlog_errno(ret);
1904                        goto out;
1905                }
1906
1907                ret = ocfs2_rotate_subtree_right(inode, handle, left_path,
1908                                                 right_path, start);
1909                if (ret) {
1910                        mlog_errno(ret);
1911                        goto out;
1912                }
1913
1914                if (split != SPLIT_NONE &&
1915                    ocfs2_leftmost_rec_contains(path_leaf_el(right_path),
1916                                                insert_cpos)) {
1917                        /*
1918                         * A rotate moves the rightmost left leaf
1919                         * record over to the leftmost right leaf
1920                         * slot. If we're doing an extent split
1921                         * instead of a real insert, then we have to
1922                         * check that the extent to be split wasn't
1923                         * just moved over. If it was, then we can
1924                         * exit here, passing left_path back -
1925                         * ocfs2_split_extent() is smart enough to
1926                         * search both leaves.
1927                         */
1928                        *ret_left_path = left_path;
1929                        goto out_ret_path;
1930                }
1931
1932                /*
1933                 * There is no need to re-read the next right path
1934                 * as we know that it'll be our current left
1935                 * path. Optimize by copying values instead.
1936                 */
1937                ocfs2_mv_path(right_path, left_path);
1938
1939                ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, right_path,
1940                                                    &cpos);
1941                if (ret) {
1942                        mlog_errno(ret);
1943                        goto out;
1944                }
1945        }
1946
1947out:
1948        ocfs2_free_path(left_path);
1949
1950out_ret_path:
1951        return ret;
1952}
1953
1954static void ocfs2_update_edge_lengths(struct inode *inode, handle_t *handle,
1955                                      struct ocfs2_path *path)
1956{
1957        int i, idx;
1958        struct ocfs2_extent_rec *rec;
1959        struct ocfs2_extent_list *el;
1960        struct ocfs2_extent_block *eb;
1961        u32 range;
1962
1963        /* Path should always be rightmost. */
1964        eb = (struct ocfs2_extent_block *)path_leaf_bh(path)->b_data;
1965        BUG_ON(eb->h_next_leaf_blk != 0ULL);
1966
1967        el = &eb->h_list;
1968        BUG_ON(le16_to_cpu(el->l_next_free_rec) == 0);
1969        idx = le16_to_cpu(el->l_next_free_rec) - 1;
1970        rec = &el->l_recs[idx];
1971        range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
1972
1973        for (i = 0; i < path->p_tree_depth; i++) {
1974                el = path->p_node[i].el;
1975                idx = le16_to_cpu(el->l_next_free_rec) - 1;
1976                rec = &el->l_recs[idx];
1977
1978                rec->e_int_clusters = cpu_to_le32(range);
1979                le32_add_cpu(&rec->e_int_clusters, -le32_to_cpu(rec->e_cpos));
1980
1981                ocfs2_journal_dirty(handle, path->p_node[i].bh);
1982        }
1983}
1984
1985static void ocfs2_unlink_path(struct inode *inode, handle_t *handle,
1986                              struct ocfs2_cached_dealloc_ctxt *dealloc,
1987                              struct ocfs2_path *path, int unlink_start)
1988{
1989        int ret, i;
1990        struct ocfs2_extent_block *eb;
1991        struct ocfs2_extent_list *el;
1992        struct buffer_head *bh;
1993
1994        for(i = unlink_start; i < path_num_items(path); i++) {
1995                bh = path->p_node[i].bh;
1996
1997                eb = (struct ocfs2_extent_block *)bh->b_data;
1998                /*
1999                 * Not all nodes might have had their final count
2000                 * decremented by the caller - handle this here.
2001                 */
2002                el = &eb->h_list;
2003                if (le16_to_cpu(el->l_next_free_rec) > 1) {
2004                        mlog(ML_ERROR,
2005                             "Inode %llu, attempted to remove extent block "
2006                             "%llu with %u records\n",
2007                             (unsigned long long)OCFS2_I(inode)->ip_blkno,
2008                             (unsigned long long)le64_to_cpu(eb->h_blkno),
2009                             le16_to_cpu(el->l_next_free_rec));
2010
2011                        ocfs2_journal_dirty(handle, bh);
2012                        ocfs2_remove_from_cache(inode, bh);
2013                        continue;
2014                }
2015
2016                el->l_next_free_rec = 0;
2017                memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
2018
2019                ocfs2_journal_dirty(handle, bh);
2020
2021                ret = ocfs2_cache_extent_block_free(dealloc, eb);
2022                if (ret)
2023                        mlog_errno(ret);
2024
2025                ocfs2_remove_from_cache(inode, bh);
2026        }
2027}
2028
2029static void ocfs2_unlink_subtree(struct inode *inode, handle_t *handle,
2030                                 struct ocfs2_path *left_path,
2031                                 struct ocfs2_path *right_path,
2032                                 int subtree_index,
2033                                 struct ocfs2_cached_dealloc_ctxt *dealloc)
2034{
2035        int i;
2036        struct buffer_head *root_bh = left_path->p_node[subtree_index].bh;
2037        struct ocfs2_extent_list *root_el = left_path->p_node[subtree_index].el;
2038        struct ocfs2_extent_list *el;
2039        struct ocfs2_extent_block *eb;
2040
2041        el = path_leaf_el(left_path);
2042
2043        eb = (struct ocfs2_extent_block *)right_path->p_node[subtree_index + 1].bh->b_data;
2044
2045        for(i = 1; i < le16_to_cpu(root_el->l_next_free_rec); i++)
2046                if (root_el->l_recs[i].e_blkno == eb->h_blkno)
2047                        break;
2048
2049        BUG_ON(i >= le16_to_cpu(root_el->l_next_free_rec));
2050
2051        memset(&root_el->l_recs[i], 0, sizeof(struct ocfs2_extent_rec));
2052        le16_add_cpu(&root_el->l_next_free_rec, -1);
2053
2054        eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data;
2055        eb->h_next_leaf_blk = 0;
2056
2057        ocfs2_journal_dirty(handle, root_bh);
2058        ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
2059
2060        ocfs2_unlink_path(inode, handle, dealloc, right_path,
2061                          subtree_index + 1);
2062}
2063
2064static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle,
2065                                     struct ocfs2_path *left_path,
2066                                     struct ocfs2_path *right_path,
2067                                     int subtree_index,
2068                                     struct ocfs2_cached_dealloc_ctxt *dealloc,
2069                                     int *deleted)
2070{
2071        int ret, i, del_right_subtree = 0, right_has_empty = 0;
2072        struct buffer_head *root_bh, *di_bh = path_root_bh(right_path);
2073        struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
2074        struct ocfs2_extent_list *right_leaf_el, *left_leaf_el;
2075        struct ocfs2_extent_block *eb;
2076
2077        *deleted = 0;
2078
2079        right_leaf_el = path_leaf_el(right_path);
2080        left_leaf_el = path_leaf_el(left_path);
2081        root_bh = left_path->p_node[subtree_index].bh;
2082        BUG_ON(root_bh != right_path->p_node[subtree_index].bh);
2083
2084        if (!ocfs2_is_empty_extent(&left_leaf_el->l_recs[0]))
2085                return 0;
2086
2087        eb = (struct ocfs2_extent_block *)path_leaf_bh(right_path)->b_data;
2088        if (ocfs2_is_empty_extent(&right_leaf_el->l_recs[0])) {
2089                /*
2090                 * It's legal for us to proceed if the right leaf is
2091                 * the rightmost one and it has an empty extent. There
2092                 * are two cases to handle - whether the leaf will be
2093                 * empty after removal or not. If the leaf isn't empty
2094                 * then just remove the empty extent up front. The
2095                 * next block will handle empty leaves by flagging
2096                 * them for unlink.
2097                 *
2098                 * Non rightmost leaves will throw -EAGAIN and the
2099                 * caller can manually move the subtree and retry.
2100                 */
2101
2102                if (eb->h_next_leaf_blk != 0ULL)
2103                        return -EAGAIN;
2104
2105                if (le16_to_cpu(right_leaf_el->l_next_free_rec) > 1) {
2106                        ret = ocfs2_journal_access(handle, inode,
2107                                                   path_leaf_bh(right_path),
2108                                                   OCFS2_JOURNAL_ACCESS_WRITE);
2109                        if (ret) {
2110                                mlog_errno(ret);
2111                                goto out;
2112                        }
2113
2114                        ocfs2_remove_empty_extent(right_leaf_el);
2115                } else
2116                        right_has_empty = 1;
2117        }
2118
2119        if (eb->h_next_leaf_blk == 0ULL &&
2120            le16_to_cpu(right_leaf_el->l_next_free_rec) == 1) {
2121                /*
2122                 * We have to update i_last_eb_blk during the meta
2123                 * data delete.
2124                 */
2125                ret = ocfs2_journal_access(handle, inode, di_bh,
2126                                           OCFS2_JOURNAL_ACCESS_WRITE);
2127                if (ret) {
2128                        mlog_errno(ret);
2129                        goto out;
2130                }
2131
2132                del_right_subtree = 1;
2133        }
2134
2135        /*
2136         * Getting here with an empty extent in the right path implies
2137         * that it's the rightmost path and will be deleted.
2138         */
2139        BUG_ON(right_has_empty && !del_right_subtree);
2140
2141        ret = ocfs2_journal_access(handle, inode, root_bh,
2142                                   OCFS2_JOURNAL_ACCESS_WRITE);
2143        if (ret) {
2144                mlog_errno(ret);
2145                goto out;
2146        }
2147
2148        for(i = subtree_index + 1; i < path_num_items(right_path); i++) {
2149                ret = ocfs2_journal_access(handle, inode,
2150                                           right_path->p_node[i].bh,
2151                                           OCFS2_JOURNAL_ACCESS_WRITE);
2152                if (ret) {
2153                        mlog_errno(ret);
2154                        goto out;
2155                }
2156
2157                ret = ocfs2_journal_access(handle, inode,
2158                                           left_path->p_node[i].bh,
2159                                           OCFS2_JOURNAL_ACCESS_WRITE);
2160                if (ret) {
2161                        mlog_errno(ret);
2162                        goto out;
2163                }
2164        }
2165
2166        if (!right_has_empty) {
2167                /*
2168                 * Only do this if we're moving a real
2169                 * record. Otherwise, the action is delayed until
2170                 * after removal of the right path in which case we
2171                 * can do a simple shift to remove the empty extent.
2172                 */
2173                ocfs2_rotate_leaf(left_leaf_el, &right_leaf_el->l_recs[0]);
2174                memset(&right_leaf_el->l_recs[0], 0,
2175                       sizeof(struct ocfs2_extent_rec));
2176        }
2177        if (eb->h_next_leaf_blk == 0ULL) {
2178                /*
2179                 * Move recs over to get rid of empty extent, decrease
2180                 * next_free. This is allowed to remove the last
2181                 * extent in our leaf (setting l_next_free_rec to
2182                 * zero) - the delete code below won't care.
2183                 */
2184                ocfs2_remove_empty_extent(right_leaf_el);
2185        }
2186
2187        ret = ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
2188        if (ret)
2189                mlog_errno(ret);
2190        ret = ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
2191        if (ret)
2192                mlog_errno(ret);
2193
2194        if (del_right_subtree) {
2195                ocfs2_unlink_subtree(inode, handle, left_path, right_path,
2196                                     subtree_index, dealloc);
2197                ocfs2_update_edge_lengths(inode, handle, left_path);
2198
2199                eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data;
2200                di->i_last_eb_blk = eb->h_blkno;
2201
2202                /*
2203                 * Removal of the extent in the left leaf was skipped
2204                 * above so we could delete the right path
2205                 * 1st.
2206                 */
2207                if (right_has_empty)
2208                        ocfs2_remove_empty_extent(left_leaf_el);
2209
2210                ret = ocfs2_journal_dirty(handle, di_bh);
2211                if (ret)
2212                        mlog_errno(ret);
2213
2214                *deleted = 1;
2215        } else
2216                ocfs2_complete_edge_insert(inode, handle, left_path, right_path,
2217                                           subtree_index);
2218
2219out:
2220        return ret;
2221}
2222
2223/*
2224 * Given a full path, determine what cpos value would return us a path
2225 * containing the leaf immediately to the right of the current one.
2226 *
2227 * Will return zero if the path passed in is already the rightmost path.
2228 *
2229 * This looks similar, but is subtly different to
2230 * ocfs2_find_cpos_for_left_leaf().
2231 */
2232static int ocfs2_find_cpos_for_right_leaf(struct super_block *sb,
2233                                          struct ocfs2_path *path, u32 *cpos)
2234{
2235        int i, j, ret = 0;
2236        u64 blkno;
2237        struct ocfs2_extent_list *el;
2238
2239        *cpos = 0;
2240
2241        if (path->p_tree_depth == 0)
2242                return 0;
2243
2244        blkno = path_leaf_bh(path)->b_blocknr;
2245
2246        /* Start at the tree node just above the leaf and work our way up. */
2247        i = path->p_tree_depth - 1;
2248        while (i >= 0) {
2249                int next_free;
2250
2251                el = path->p_node[i].el;
2252
2253                /*
2254                 * Find the extent record just after the one in our
2255                 * path.
2256                 */
2257                next_free = le16_to_cpu(el->l_next_free_rec);
2258                for(j = 0; j < le16_to_cpu(el->l_next_free_rec); j++) {
2259                        if (le64_to_cpu(el->l_recs[j].e_blkno) == blkno) {
2260                                if (j == (next_free - 1)) {
2261                                        if (i == 0) {
2262                                                /*
2263                                                 * We've determined that the
2264                                                 * path specified is already
2265                                                 * the rightmost one - return a
2266                                                 * cpos of zero.
2267                                                 */
2268                                                goto out;
2269                                        }
2270                                        /*
2271                                         * The rightmost record points to our
2272                                         * leaf - we need to travel up the
2273                                         * tree one level.
2274                                         */
2275                                        goto next_node;
2276                                }
2277
2278                                *cpos = le32_to_cpu(el->l_recs[j + 1].e_cpos);
2279                                goto out;
2280                        }
2281                }
2282
2283                /*
2284                 * If we got here, we never found a valid node where
2285                 * the tree indicated one should be.
2286                 */
2287                ocfs2_error(sb,
2288                            "Invalid extent tree at extent block %llu\n",
2289                            (unsigned long long)blkno);
2290                ret = -EROFS;
2291                goto out;
2292
2293next_node:
2294                blkno = path->p_node[i].bh->b_blocknr;
2295                i--;
2296        }
2297
2298out:
2299        return ret;
2300}
2301
2302static int ocfs2_rotate_rightmost_leaf_left(struct inode *inode,
2303                                            handle_t *handle,
2304                                            struct buffer_head *bh,
2305                                            struct ocfs2_extent_list *el)
2306{
2307        int ret;
2308
2309        if (!ocfs2_is_empty_extent(&el->l_recs[0]))
2310                return 0;
2311
2312        ret = ocfs2_journal_access(handle, inode, bh,
2313                                   OCFS2_JOURNAL_ACCESS_WRITE);
2314        if (ret) {
2315                mlog_errno(ret);
2316                goto out;
2317        }
2318
2319        ocfs2_remove_empty_extent(el);
2320
2321        ret = ocfs2_journal_dirty(handle, bh);
2322        if (ret)
2323                mlog_errno(ret);
2324
2325out:
2326        return ret;
2327}
2328
2329static int __ocfs2_rotate_tree_left(struct inode *inode,
2330                                    handle_t *handle, int orig_credits,
2331                                    struct ocfs2_path *path,
2332                                    struct ocfs2_cached_dealloc_ctxt *dealloc,
2333                                    struct ocfs2_path **empty_extent_path)
2334{
2335        int ret, subtree_root, deleted;
2336        u32 right_cpos;
2337        struct ocfs2_path *left_path = NULL;
2338        struct ocfs2_path *right_path = NULL;
2339
2340        BUG_ON(!ocfs2_is_empty_extent(&(path_leaf_el(path)->l_recs[0])));
2341
2342        *empty_extent_path = NULL;
2343
2344        ret = ocfs2_find_cpos_for_right_leaf(inode->i_sb, path,
2345                                             &right_cpos);
2346        if (ret) {
2347                mlog_errno(ret);
2348                goto out;
2349        }
2350
2351        left_path = ocfs2_new_path(path_root_bh(path),
2352                                   path_root_el(path));
2353        if (!left_path) {
2354                ret = -ENOMEM;
2355                mlog_errno(ret);
2356                goto out;
2357        }
2358
2359        ocfs2_cp_path(left_path, path);
2360
2361        right_path = ocfs2_new_path(path_root_bh(path),
2362                                    path_root_el(path));
2363        if (!right_path) {
2364                ret = -ENOMEM;
2365                mlog_errno(ret);
2366                goto out;
2367        }
2368
2369        while (right_cpos) {
2370                ret = ocfs2_find_path(inode, right_path, right_cpos);
2371                if (ret) {
2372                        mlog_errno(ret);
2373                        goto out;
2374                }
2375
2376                subtree_root = ocfs2_find_subtree_root(inode, left_path,
2377                                                       right_path);
2378
2379                mlog(0, "Subtree root at index %d (blk %llu, depth %d)\n",
2380                     subtree_root,
2381                     (unsigned long long)
2382                     right_path->p_node[subtree_root].bh->b_blocknr,
2383                     right_path->p_tree_depth);
2384
2385                ret = ocfs2_extend_rotate_transaction(handle, subtree_root,
2386                                                      orig_credits, left_path);
2387                if (ret) {
2388                        mlog_errno(ret);
2389                        goto out;
2390                }
2391
2392                /*
2393                 * Caller might still want to make changes to the
2394                 * tree root, so re-add it to the journal here.
2395                 */
2396                ret = ocfs2_journal_access(handle, inode,
2397                                           path_root_bh(left_path),
2398                                           OCFS2_JOURNAL_ACCESS_WRITE);
2399                if (ret) {
2400                        mlog_errno(ret);
2401                        goto out;
2402                }
2403
2404                ret = ocfs2_rotate_subtree_left(inode, handle, left_path,
2405                                                right_path, subtree_root,
2406                                                dealloc, &deleted);
2407                if (ret == -EAGAIN) {
2408                        /*
2409                         * The rotation has to temporarily stop due to
2410                         * the right subtree having an empty
2411                         * extent. Pass it back to the caller for a
2412                         * fixup.
2413                         */
2414                        *empty_extent_path = right_path;
2415                        right_path = NULL;
2416                        goto out;
2417                }
2418                if (ret) {
2419                        mlog_errno(ret);
2420                        goto out;
2421                }
2422
2423                /*
2424                 * The subtree rotate might have removed records on
2425                 * the rightmost edge. If so, then rotation is
2426                 * complete.
2427                 */
2428                if (deleted)
2429                        break;
2430
2431                ocfs2_mv_path(left_path, right_path);
2432
2433                ret = ocfs2_find_cpos_for_right_leaf(inode->i_sb, left_path,
2434                                                     &right_cpos);
2435                if (ret) {
2436                        mlog_errno(ret);
2437                        goto out;
2438                }
2439        }
2440
2441out:
2442        ocfs2_free_path(right_path);
2443        ocfs2_free_path(left_path);
2444
2445        return ret;
2446}
2447
2448static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle,
2449                                       struct ocfs2_path *path,
2450                                       struct ocfs2_cached_dealloc_ctxt *dealloc)
2451{
2452        int ret, subtree_index;
2453        u32 cpos;
2454        struct ocfs2_path *left_path = NULL;
2455        struct ocfs2_dinode *di;
2456        struct ocfs2_extent_block *eb;
2457        struct ocfs2_extent_list *el;
2458
2459        /*
2460         * XXX: This code assumes that the root is an inode, which is
2461         * true for now but may change as tree code gets generic.
2462         */
2463        di = (struct ocfs2_dinode *)path_root_bh(path)->b_data;
2464        if (!OCFS2_IS_VALID_DINODE(di)) {
2465                ret = -EIO;
2466                ocfs2_error(inode->i_sb,
2467                            "Inode %llu has invalid path root",
2468                            (unsigned long long)OCFS2_I(inode)->ip_blkno);
2469                goto out;
2470        }
2471
2472        /*
2473         * There's two ways we handle this depending on
2474         * whether path is the only existing one.
2475         */
2476        ret = ocfs2_extend_rotate_transaction(handle, 0,
2477                                              handle->h_buffer_credits,
2478                                              path);
2479        if (ret) {
2480                mlog_errno(ret);
2481                goto out;
2482        }
2483
2484        ret = ocfs2_journal_access_path(inode, handle, path);
2485        if (ret) {
2486                mlog_errno(ret);
2487                goto out;
2488        }
2489
2490        ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path, &cpos);
2491        if (ret) {
2492                mlog_errno(ret);
2493                goto out;
2494        }
2495
2496        if (cpos) {
2497                /*
2498                 * We have a path to the left of this one - it needs
2499                 * an update too.
2500                 */
2501                left_path = ocfs2_new_path(path_root_bh(path),
2502                                           path_root_el(path));
2503                if (!left_path) {
2504                        ret = -ENOMEM;
2505                        mlog_errno(ret);
2506                        goto out;
2507                }
2508
2509                ret = ocfs2_find_path(inode, left_path, cpos);
2510                if (ret) {
2511                        mlog_errno(ret);
2512                        goto out;
2513                }
2514
2515                ret = ocfs2_journal_access_path(inode, handle, left_path);
2516                if (ret) {
2517                        mlog_errno(ret);
2518                        goto out;
2519                }
2520
2521                subtree_index = ocfs2_find_subtree_root(inode, left_path, path);
2522
2523                ocfs2_unlink_subtree(inode, handle, left_path, path,
2524                                     subtree_index, dealloc);
2525                ocfs2_update_edge_lengths(inode, handle, left_path);
2526
2527                eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data;
2528                di->i_last_eb_blk = eb->h_blkno;
2529        } else {
2530                /*
2531                 * 'path' is also the leftmost path which
2532                 * means it must be the only one. This gets
2533                 * handled differently because we want to
2534                 * revert the inode back to having extents
2535                 * in-line.
2536                 */
2537                ocfs2_unlink_path(inode, handle, dealloc, path, 1);
2538
2539                el = &di->id2.i_list;
2540                el->l_tree_depth = 0;
2541                el->l_next_free_rec = 0;
2542                memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
2543
2544                di->i_last_eb_blk = 0;
2545        }
2546
2547        ocfs2_journal_dirty(handle, path_root_bh(path));
2548
2549out:
2550        ocfs2_free_path(left_path);
2551        return ret;
2552}
2553
2554/*
2555 * Left rotation of btree records.
2556 *
2557 * In many ways, this is (unsurprisingly) the opposite of right
2558 * rotation. We start at some non-rightmost path containing an empty
2559 * extent in the leaf block. The code works its way to the rightmost
2560 * path by rotating records to the left in every subtree.
2561 *
2562 * This is used by any code which reduces the number of extent records
2563 * in a leaf. After removal, an empty record should be placed in the
2564 * leftmost list position.
2565 *
2566 * This won't handle a length update of the rightmost path records if
2567 * the rightmost tree leaf record is removed so the caller is
2568 * responsible for detecting and correcting that.
2569 */
2570static int ocfs2_rotate_tree_left(struct inode *inode, handle_t *handle,
2571                                  struct ocfs2_path *path,
2572                                  struct ocfs2_cached_dealloc_ctxt *dealloc)
2573{
2574        int ret, orig_credits = handle->h_buffer_credits;
2575        struct ocfs2_path *tmp_path = NULL, *restart_path = NULL;
2576        struct ocfs2_extent_block *eb;
2577        struct ocfs2_extent_list *el;
2578
2579        el = path_leaf_el(path);
2580        if (!ocfs2_is_empty_extent(&el->l_recs[0]))
2581                return 0;
2582
2583        if (path->p_tree_depth == 0) {
2584rightmost_no_delete:
2585                /*
2586                 * In-inode extents. This is trivially handled, so do
2587                 * it up front.
2588                 */
2589                ret = ocfs2_rotate_rightmost_leaf_left(inode, handle,
2590                                                       path_leaf_bh(path),
2591                                                       path_leaf_el(path));
2592                if (ret)
2593                        mlog_errno(ret);
2594                goto out;
2595        }
2596
2597        /*
2598         * Handle rightmost branch now. There's several cases:
2599         *  1) simple rotation leaving records in there. That's trivial.
2600         *  2) rotation requiring a branch delete - there's no more
2601         *     records left. Two cases of this:
2602         *     a) There are branches to the left.
2603         *     b) This is also the leftmost (the only) branch.
2604         *
2605         *  1) is handled via ocfs2_rotate_rightmost_leaf_left()
2606         *  2a) we need the left branch so that we can update it with the unlink
2607         *  2b) we need to bring the inode back to inline extents.
2608         */
2609
2610        eb = (struct ocfs2_extent_block *)path_leaf_bh(path)->b_data;
2611        el = &eb->h_list;
2612        if (eb->h_next_leaf_blk == 0) {
2613                /*
2614                 * This gets a bit tricky if we're going to delete the
2615                 * rightmost path. Get the other cases out of the way
2616                 * 1st.
2617                 */
2618                if (le16_to_cpu(el->l_next_free_rec) > 1)
2619                        goto rightmost_no_delete;
2620
2621                if (le16_to_cpu(el->l_next_free_rec) == 0) {
2622                        ret = -EIO;
2623                        ocfs2_error(inode->i_sb,
2624                                    "Inode %llu has empty extent block at %llu",
2625                                    (unsigned long long)OCFS2_I(inode)->ip_blkno,
2626                                    (unsigned long long)le64_to_cpu(eb->h_blkno));
2627                        goto out;
2628                }
2629
2630                /*
2631                 * XXX: The caller can not trust "path" any more after
2632                 * this as it will have been deleted. What do we do?
2633                 *
2634                 * In theory the rotate-for-merge code will never get
2635                 * here because it'll always ask for a rotate in a
2636                 * nonempty list.
2637                 */
2638
2639                ret = ocfs2_remove_rightmost_path(inode, handle, path,
2640                                                  dealloc);
2641                if (ret)
2642                        mlog_errno(ret);
2643                goto out;
2644        }
2645
2646        /*
2647         * Now we can loop, remembering the path we get from -EAGAIN
2648         * and restarting from there.
2649         */
2650try_rotate:
2651        ret = __ocfs2_rotate_tree_left(inode, handle, orig_credits, path,
2652                                       dealloc, &restart_path);
2653        if (ret && ret != -EAGAIN) {
2654                mlog_errno(ret);
2655                goto out;
2656        }
2657
2658        while (ret == -EAGAIN) {
2659                tmp_path = restart_path;
2660                restart_path = NULL;
2661
2662                ret = __ocfs2_rotate_tree_left(inode, handle, orig_credits,
2663                                               tmp_path, dealloc,
2664                                               &restart_path);
2665                if (ret && ret != -EAGAIN) {
2666                        mlog_errno(ret);
2667                        goto out;
2668                }
2669
2670                ocfs2_free_path(tmp_path);
2671                tmp_path = NULL;
2672
2673                if (ret == 0)
2674                        goto try_rotate;
2675        }
2676
2677out:
2678        ocfs2_free_path(tmp_path);
2679        ocfs2_free_path(restart_path);
2680        return ret;
2681}
2682
2683static void ocfs2_cleanup_merge(struct ocfs2_extent_list *el,
2684                                int index)
2685{
2686        struct ocfs2_extent_rec *rec = &el->l_recs[index];
2687        unsigned int size;
2688
2689        if (rec->e_leaf_clusters == 0) {
2690                /*
2691                 * We consumed all of the merged-from record. An empty
2692                 * extent cannot exist anywhere but the 1st array
2693                 * position, so move things over if the merged-from
2694                 * record doesn't occupy that position.
2695                 *
2696                 * This creates a new empty extent so the caller
2697                 * should be smart enough to have removed any existing
2698                 * ones.
2699                 */
2700                if (index > 0) {
2701                        BUG_ON(ocfs2_is_empty_extent(&el->l_recs[0]));
2702                        size = index * sizeof(struct ocfs2_extent_rec);
2703                        memmove(&el->l_recs[1], &el->l_recs[0], size);
2704                }
2705
2706                /*
2707                 * Always memset - the caller doesn't check whether it
2708                 * created an empty extent, so there could be junk in
2709                 * the other fields.
2710                 */
2711                memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
2712        }
2713}
2714
2715/*
2716 * Remove split_rec clusters from the record at index and merge them
2717 * onto the beginning of the record at index + 1.
2718 */
2719static int ocfs2_merge_rec_right(struct inode *inode, struct buffer_head *bh,
2720                                handle_t *handle,
2721                                struct ocfs2_extent_rec *split_rec,
2722                                struct ocfs2_extent_list *el, int index)
2723{
2724        int ret;
2725        unsigned int split_clusters = le16_to_cpu(split_rec->e_leaf_clusters);
2726        struct ocfs2_extent_rec *left_rec;
2727        struct ocfs2_extent_rec *right_rec;
2728
2729        BUG_ON(index >= le16_to_cpu(el->l_next_free_rec));
2730
2731        left_rec = &el->l_recs[index];
2732        right_rec = &el->l_recs[index + 1];
2733
2734        ret = ocfs2_journal_access(handle, inode, bh,
2735                                   OCFS2_JOURNAL_ACCESS_WRITE);
2736        if (ret) {
2737                mlog_errno(ret);
2738                goto out;
2739        }
2740
2741        le16_add_cpu(&left_rec->e_leaf_clusters, -split_clusters);
2742
2743        le32_add_cpu(&right_rec->e_cpos, -split_clusters);
2744        le64_add_cpu(&right_rec->e_blkno,
2745                     -ocfs2_clusters_to_blocks(inode->i_sb, split_clusters));
2746        le16_add_cpu(&right_rec->e_leaf_clusters, split_clusters);
2747
2748        ocfs2_cleanup_merge(el, index);
2749
2750        ret = ocfs2_journal_dirty(handle, bh);
2751        if (ret)
2752                mlog_errno(ret);
2753
2754out:
2755        return ret;
2756}
2757
2758/*
2759 * Remove split_rec clusters from the record at index and merge them
2760 * onto the tail of the record at index - 1.
2761 */
2762static int ocfs2_merge_rec_left(struct inode *inode, struct buffer_head *bh,
2763                                handle_t *handle,
2764                                struct ocfs2_extent_rec *split_rec,
2765                                struct ocfs2_extent_list *el, int index)
2766{
2767        int ret, has_empty_extent = 0;
2768        unsigned int split_clusters = le16_to_cpu(split_rec->e_leaf_clusters);
2769        struct ocfs2_extent_rec *left_rec;
2770        struct ocfs2_extent_rec *right_rec;
2771
2772        BUG_ON(index <= 0);
2773
2774        left_rec = &el->l_recs[index - 1];
2775        right_rec = &el->l_recs[index];
2776        if (ocfs2_is_empty_extent(&el->l_recs[0]))
2777                has_empty_extent = 1;
2778
2779        ret = ocfs2_journal_access(handle, inode, bh,
2780                                   OCFS2_JOURNAL_ACCESS_WRITE);
2781        if (ret) {
2782                mlog_errno(ret);
2783                goto out;
2784        }
2785
2786        if (has_empty_extent && index == 1) {
2787                /*
2788                 * The easy case - we can just plop the record right in.
2789                 */
2790                *left_rec = *split_rec;
2791
2792                has_empty_extent = 0;
2793        } else {
2794                le16_add_cpu(&left_rec->e_leaf_clusters, split_clusters);
2795        }
2796
2797        le32_add_cpu(&right_rec->e_cpos, split_clusters);
2798        le64_add_cpu(&right_rec->e_blkno,
2799                     ocfs2_clusters_to_blocks(inode->i_sb, split_clusters));
2800        le16_add_cpu(&right_rec->e_leaf_clusters, -split_clusters);
2801
2802        ocfs2_cleanup_merge(el, index);
2803
2804        ret = ocfs2_journal_dirty(handle, bh);
2805        if (ret)
2806                mlog_errno(ret);
2807
2808out:
2809        return ret;
2810}
2811
2812static int ocfs2_try_to_merge_extent(struct inode *inode,
2813                                     handle_t *handle,
2814                                     struct ocfs2_path *left_path,
2815                                     int split_index,
2816                                     struct ocfs2_extent_rec *split_rec,
2817                                     struct ocfs2_cached_dealloc_ctxt *dealloc,
2818                                     struct ocfs2_merge_ctxt *ctxt)
2819
2820{
2821        int ret = 0;
2822        struct ocfs2_extent_list *el = path_leaf_el(left_path);
2823        struct ocfs2_extent_rec *rec = &el->l_recs[split_index];
2824
2825        BUG_ON(ctxt->c_contig_type == CONTIG_NONE);
2826
2827        if (ctxt->c_split_covers_rec && ctxt->c_has_empty_extent) {
2828                /*
2829                 * The merge code will need to create an empty
2830                 * extent to take the place of the newly
2831                 * emptied slot. Remove any pre-existing empty
2832                 * extents - having more than one in a leaf is
2833                 * illegal.
2834                 */
2835                ret = ocfs2_rotate_tree_left(inode, handle, left_path,
2836                                             dealloc);
2837                if (ret) {
2838                        mlog_errno(ret);
2839                        goto out;
2840                }
2841                split_index--;
2842                rec = &el->l_recs[split_index];
2843        }
2844
2845        if (ctxt->c_contig_type == CONTIG_LEFTRIGHT) {
2846                /*
2847                 * Left-right contig implies this.
2848                 */
2849                BUG_ON(!ctxt->c_split_covers_rec);
2850                BUG_ON(split_index == 0);
2851
2852                /*
2853                 * Since the leftright insert always covers the entire
2854                 * extent, this call will delete the insert record
2855                 * entirely, resulting in an empty extent record added to
2856                 * the extent block.
2857                 *
2858                 * Since the adding of an empty extent shifts
2859                 * everything back to the right, there's no need to
2860                 * update split_index here.
2861                 */
2862                ret = ocfs2_merge_rec_left(inode, path_leaf_bh(left_path),
2863                                           handle, split_rec, el, split_index);
2864                if (ret) {
2865                        mlog_errno(ret);
2866                        goto out;
2867                }
2868
2869                /*
2870                 * We can only get this from logic error above.
2871                 */
2872                BUG_ON(!ocfs2_is_empty_extent(&el->l_recs[0]));
2873
2874                /*
2875                 * The left merge left us with an empty extent, remove
2876                 * it.
2877                 */
2878                ret = ocfs2_rotate_tree_left(inode, handle, left_path, dealloc);
2879                if (ret) {
2880                        mlog_errno(ret);
2881                        goto out;
2882                }
2883                split_index--;
2884                rec = &el->l_recs[split_index];
2885
2886                /*
2887                 * Note that we don't pass split_rec here on purpose -
2888                 * we've merged it into the left side.
2889                 */
2890                ret = ocfs2_merge_rec_right(inode, path_leaf_bh(left_path),
2891                                            handle, rec, el, split_index);
2892                if (ret) {
2893                        mlog_errno(ret);
2894                        goto out;
2895                }
2896
2897                BUG_ON(!ocfs2_is_empty_extent(&el->l_recs[0]));
2898
2899                ret = ocfs2_rotate_tree_left(inode, handle, left_path,
2900                                             dealloc);
2901                /*
2902                 * Error from this last rotate is not critical, so
2903                 * print but don't bubble it up.
2904                 */
2905                if (ret)
2906                        mlog_errno(ret);
2907                ret = 0;
2908        } else {
2909                /*
2910                 * Merge a record to the left or right.
2911                 *
2912                 * 'contig_type' is relative to the existing record,
2913                 * so for example, if we're "right contig", it's to
2914                 * the record on the left (hence the left merge).
2915                 */
2916                if (ctxt->c_contig_type == CONTIG_RIGHT) {
2917                        ret = ocfs2_merge_rec_left(inode,
2918                                                   path_leaf_bh(left_path),
2919                                                   handle, split_rec, el,
2920                                                   split_index);
2921                        if (ret) {
2922                                mlog_errno(ret);
2923                                goto out;
2924                        }
2925                } else {
2926                        ret = ocfs2_merge_rec_right(inode,
2927                                                    path_leaf_bh(left_path),
2928                                                    handle, split_rec, el,
2929                                                    split_index);
2930                        if (ret) {
2931                                mlog_errno(ret);
2932                                goto out;
2933                        }
2934                }
2935
2936                if (ctxt->c_split_covers_rec) {
2937                        /*
2938                         * The merge may have left an empty extent in
2939                         * our leaf. Try to rotate it away.
2940                         */
2941                        ret = ocfs2_rotate_tree_left(inode, handle, left_path,
2942                                                     dealloc);
2943                        if (ret)
2944                                mlog_errno(ret);
2945                        ret = 0;
2946                }
2947        }
2948
2949out:
2950        return ret;
2951}
2952
2953static void ocfs2_subtract_from_rec(struct super_block *sb,
2954                                    enum ocfs2_split_type split,
2955                                    struct ocfs2_extent_rec *rec,
2956                                    struct ocfs2_extent_rec *split_rec)
2957{
2958        u64 len_blocks;
2959
2960        len_blocks = ocfs2_clusters_to_blocks(sb,
2961                                le16_to_cpu(split_rec->e_leaf_clusters));
2962
2963        if (split == SPLIT_LEFT) {
2964                /*
2965                 * Region is on the left edge of the existing
2966                 * record.
2967                 */
2968                le32_add_cpu(&rec->e_cpos,
2969                             le16_to_cpu(split_rec->e_leaf_clusters));
2970                le64_add_cpu(&rec->e_blkno, len_blocks);
2971                le16_add_cpu(&rec->e_leaf_clusters,
2972                             -le16_to_cpu(split_rec->e_leaf_clusters));
2973        } else {
2974                /*
2975                 * Region is on the right edge of the existing
2976                 * record.
2977                 */
2978                le16_add_cpu(&rec->e_leaf_clusters,
2979                             -le16_to_cpu(split_rec->e_leaf_clusters));
2980        }
2981}
2982
2983/*
2984 * Do the final bits of extent record insertion at the target leaf
2985 * list. If this leaf is part of an allocation tree, it is assumed
2986 * that the tree above has been prepared.
2987 */
2988static void ocfs2_insert_at_leaf(struct ocfs2_extent_rec *insert_rec,
2989                                 struct ocfs2_extent_list *el,
2990                                 struct ocfs2_insert_type *insert,
2991                                 struct inode *inode)
2992{
2993        int i = insert->ins_contig_index;
2994        unsigned int range;
2995        struct ocfs2_extent_rec *rec;
2996
2997        BUG_ON(le16_to_cpu(el->l_tree_depth) != 0);
2998
2999        if (insert->ins_split != SPLIT_NONE) {
3000                i = ocfs2_search_extent_list(el, le32_to_cpu(insert_rec->e_cpos));
3001                BUG_ON(i == -1);
3002                rec = &el->l_recs[i];
3003                ocfs2_subtract_from_rec(inode->i_sb, insert->ins_split, rec,
3004                                        insert_rec);
3005                goto rotate;
3006        }
3007
3008        /*
3009         * Contiguous insert - either left or right.
3010         */
3011        if (insert->ins_contig != CONTIG_NONE) {
3012                rec = &el->l_recs[i];
3013                if (insert->ins_contig == CONTIG_LEFT) {
3014                        rec->e_blkno = insert_rec->e_blkno;
3015                        rec->e_cpos = insert_rec->e_cpos;
3016                }
3017                le16_add_cpu(&rec->e_leaf_clusters,
3018                             le16_to_cpu(insert_rec->e_leaf_clusters));
3019                return;
3020        }
3021
3022        /*
3023         * Handle insert into an empty leaf.
3024         */
3025        if (le16_to_cpu(el->l_next_free_rec) == 0 ||
3026            ((le16_to_cpu(el->l_next_free_rec) == 1) &&
3027             ocfs2_is_empty_extent(&el->l_recs[0]))) {
3028                el->l_recs[0] = *insert_rec;
3029                el->l_next_free_rec = cpu_to_le16(1);
3030                return;
3031        }
3032
3033        /*
3034         * Appending insert.
3035         */
3036        if (insert->ins_appending == APPEND_TAIL) {
3037                i = le16_to_cpu(el->l_next_free_rec) - 1;
3038                rec = &el->l_recs[i];
3039                range = le32_to_cpu(rec->e_cpos)
3040                        + le16_to_cpu(rec->e_leaf_clusters);
3041                BUG_ON(le32_to_cpu(insert_rec->e_cpos) < range);
3042
3043                mlog_bug_on_msg(le16_to_cpu(el->l_next_free_rec) >=
3044                                le16_to_cpu(el->l_count),
3045                                "inode %lu, depth %u, count %u, next free %u, "
3046                                "rec.cpos %u, rec.clusters %u, "
3047                                "insert.cpos %u, insert.clusters %u\n",
3048                                inode->i_ino,
3049                                le16_to_cpu(el->l_tree_depth),
3050                                le16_to_cpu(el->l_count),
3051                                le16_to_cpu(el->l_next_free_rec),
3052                                le32_to_cpu(el->l_recs[i].e_cpos),
3053                                le16_to_cpu(el->l_recs[i].e_leaf_clusters),
3054                                le32_to_cpu(insert_rec->e_cpos),
3055                                le16_to_cpu(insert_rec->e_leaf_clusters));
3056                i++;
3057                el->l_recs[i] = *insert_rec;
3058                le16_add_cpu(&el->l_next_free_rec, 1);
3059                return;
3060        }
3061
3062rotate:
3063        /*
3064         * Ok, we have to rotate.
3065         *
3066         * At this point, it is safe to assume that inserting into an
3067         * empty leaf and appending to a leaf have both been handled
3068         * above.
3069         *
3070         * This leaf needs to have space, either by the empty 1st
3071         * extent record, or by virtue of an l_next_rec < l_count.
3072         */
3073        ocfs2_rotate_leaf(el, insert_rec);
3074}
3075
3076static inline void ocfs2_update_dinode_clusters(struct inode *inode,
3077                                                struct ocfs2_dinode *di,
3078                                                u32 clusters)
3079{
3080        le32_add_cpu(&di->i_clusters, clusters);
3081        spin_lock(&OCFS2_I(inode)->ip_lock);
3082        OCFS2_I(inode)->ip_clusters = le32_to_cpu(di->i_clusters);
3083        spin_unlock(&OCFS2_I(inode)->ip_lock);
3084}
3085
3086static void ocfs2_adjust_rightmost_records(struct inode *inode,
3087                                           handle_t *handle,
3088                                           struct ocfs2_path *path,
3089                                           struct ocfs2_extent_rec *insert_rec)
3090{
3091        int ret, i, next_free;
3092        struct buffer_head *bh;
3093        struct ocfs2_extent_list *el;
3094        struct ocfs2_extent_rec *rec;
3095
3096        /*
3097         * Update everything except the leaf block.
3098         */
3099        for (i = 0; i < path->p_tree_depth; i++) {
3100                bh = path->p_node[i].bh;
3101                el = path->p_node[i].el;
3102
3103                next_free = le16_to_cpu(el->l_next_free_rec);
3104                if (next_free == 0) {
3105                        ocfs2_error(inode->i_sb,
3106                                    "Dinode %llu has a bad extent list",
3107                                    (unsigned long long)OCFS2_I(inode)->ip_blkno);
3108                        ret = -EIO;
3109                        return;
3110                }
3111
3112                rec = &el->l_recs[next_free - 1];
3113
3114                rec->e_int_clusters = insert_rec->e_cpos;
3115                le32_add_cpu(&rec->e_int_clusters,
3116                             le16_to_cpu(insert_rec->e_leaf_clusters));
3117                le32_add_cpu(&rec->e_int_clusters,
3118                             -le32_to_cpu(rec->e_cpos));
3119
3120                ret = ocfs2_journal_dirty(handle, bh);
3121                if (ret)
3122                        mlog_errno(ret);
3123
3124        }
3125}
3126
3127static int ocfs2_append_rec_to_path(struct inode *inode, handle_t *handle,
3128                                    struct ocfs2_extent_rec *insert_rec,
3129                                    struct ocfs2_path *right_path,
3130                                    struct ocfs2_path **ret_left_path)
3131{
3132        int ret, next_free;
3133        struct ocfs2_extent_list *el;
3134        struct ocfs2_path *left_path = NULL;
3135
3136        *ret_left_path = NULL;
3137
3138        /*
3139         * This shouldn't happen for non-trees. The extent rec cluster
3140         * count manipulation below only works for interior nodes.
3141         */
3142        BUG_ON(right_path->p_tree_depth == 0);
3143
3144        /*
3145         * If our appending insert is at the leftmost edge of a leaf,
3146         * then we might need to update the rightmost records of the
3147         * neighboring path.
3148         */
3149        el = path_leaf_el(right_path);
3150        next_free = le16_to_cpu(el->l_next_free_rec);
3151        if (next_free == 0 ||
3152            (next_free == 1 && ocfs2_is_empty_extent(&el->l_recs[0]))) {
3153                u32 left_cpos;
3154
3155                ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, right_path,
3156                                                    &left_cpos);
3157                if (ret) {
3158                        mlog_errno(ret);
3159                        goto out;
3160                }
3161
3162                mlog(0, "Append may need a left path update. cpos: %u, "
3163                     "left_cpos: %u\n", le32_to_cpu(insert_rec->e_cpos),
3164                     left_cpos);
3165
3166                /*
3167                 * No need to worry if the append is already in the
3168                 * leftmost leaf.
3169                 */
3170                if (left_cpos) {
3171                        left_path = ocfs2_new_path(path_root_bh(right_path),
3172                                                   path_root_el(right_path));
3173                        if (!left_path) {
3174                                ret = -ENOMEM;
3175                                mlog_errno(ret);
3176                                goto out;
3177                        }
3178
3179                        ret = ocfs2_find_path(inode, left_path, left_cpos);
3180                        if (ret) {
3181                                mlog_errno(ret);
3182                                goto out;
3183                        }
3184
3185                        /*
3186                         * ocfs2_insert_path() will pass the left_path to the
3187                         * journal for us.
3188                         */
3189                }
3190        }
3191
3192        ret = ocfs2_journal_access_path(inode, handle, right_path);
3193        if (ret) {
3194                mlog_errno(ret);
3195                goto out;
3196        }
3197
3198        ocfs2_adjust_rightmost_records(inode, handle, right_path, insert_rec);
3199
3200        *ret_left_path = left_path;
3201        ret = 0;
3202out:
3203        if (ret != 0)
3204                ocfs2_free_path(left_path);
3205
3206        return ret;
3207}
3208
3209static void ocfs2_split_record(struct inode *inode,
3210                               struct ocfs2_path *left_path,
3211                               struct ocfs2_path *right_path,
3212                               struct ocfs2_extent_rec *split_rec,
3213                               enum ocfs2_split_type split)
3214{
3215        int index;
3216        u32 cpos = le32_to_cpu(split_rec->e_cpos);
3217        struct ocfs2_extent_list *left_el = NULL, *right_el, *insert_el, *el;
3218        struct ocfs2_extent_rec *rec, *tmprec;
3219
3220        right_el = path_leaf_el(right_path);;
3221        if (left_path)
3222                left_el = path_leaf_el(left_path);
3223
3224        el = right_el;
3225        insert_el = right_el;
3226        index = ocfs2_search_extent_list(el, cpos);
3227        if (index != -1) {
3228                if (index == 0 && left_path) {
3229                        BUG_ON(ocfs2_is_empty_extent(&el->l_recs[0]));
3230
3231                        /*
3232                         * This typically means that the record
3233                         * started in the left path but moved to the
3234                         * right as a result of rotation. We either
3235                         * move the existing record to the left, or we
3236                         * do the later insert there.
3237                         *
3238                         * In this case, the left path should always
3239                         * exist as the rotate code will have passed
3240                         * it back for a post-insert update.
3241                         */
3242
3243                        if (split == SPLIT_LEFT) {
3244                                /*
3245                                 * It's a left split. Since we know
3246                                 * that the rotate code gave us an
3247                                 * empty extent in the left path, we
3248                                 * can just do the insert there.
3249                                 */
3250                                insert_el = left_el;
3251                        } else {
3252                                /*
3253                                 * Right split - we have to move the
3254                                 * existing record over to the left
3255                                 * leaf. The insert will be into the
3256                                 * newly created empty extent in the
3257                                 * right leaf.
3258                                 */
3259                                tmprec = &right_el->l_recs[index];
3260                                ocfs2_rotate_leaf(left_el, tmprec);
3261                                el = left_el;
3262
3263                                memset(tmprec, 0, sizeof(*tmprec));
3264                                index = ocfs2_search_extent_list(left_el, cpos);
3265                                BUG_ON(index == -1);
3266                        }
3267                }
3268        } else {
3269                BUG_ON(!left_path);
3270                BUG_ON(!ocfs2_is_empty_extent(&left_el->l_recs[0]));
3271                /*
3272                 * Left path is easy - we can just allow the insert to
3273                 * happen.
3274                 */
3275                el = left_el;
3276                insert_el = left_el;
3277                index = ocfs2_search_extent_list(el, cpos);
3278                BUG_ON(index == -1);
3279        }
3280
3281        rec = &el->l_recs[index];
3282        ocfs2_subtract_from_rec(inode->i_sb, split, rec, split_rec);
3283        ocfs2_rotate_leaf(insert_el, split_rec);
3284}
3285
3286/*
3287 * This function only does inserts on an allocation b-tree. For dinode
3288 * lists, ocfs2_insert_at_leaf() is called directly.
3289 *
3290 * right_path is the path we want to do the actual insert
3291 * in. left_path should only be passed in if we need to update that
3292 * portion of the tree after an edge insert.
3293 */
3294static int ocfs2_insert_path(struct inode *inode,
3295                             handle_t *handle,
3296                             struct ocfs2_path *left_path,
3297                             struct ocfs2_path *right_path,
3298                             struct ocfs2_extent_rec *insert_rec,
3299                             struct ocfs2_insert_type *insert)
3300{
3301        int ret, subtree_index;
3302        struct buffer_head *leaf_bh = path_leaf_bh(right_path);
3303
3304        if (left_path) {
3305                int credits = handle->h_buffer_credits;
3306
3307                /*
3308                 * There's a chance that left_path got passed back to
3309                 * us without being accounted for in the
3310                 * journal. Extend our transaction here to be sure we
3311                 * can change those blocks.
3312                 */
3313                credits += left_path->p_tree_depth;
3314
3315                ret = ocfs2_extend_trans(handle, credits);
3316                if (ret < 0) {
3317                        mlog_errno(ret);
3318                        goto out;
3319                }
3320
3321                ret = ocfs2_journal_access_path(inode, handle, left_path);
3322                if (ret < 0) {
3323                        mlog_errno(ret);
3324                        goto out;
3325                }
3326        }
3327
3328        /*
3329         * Pass both paths to the journal. The majority of inserts
3330         * will be touching all components anyway.
3331         */
3332        ret = ocfs2_journal_access_path(inode, handle, right_path);
3333        if (ret < 0) {
3334                mlog_errno(ret);
3335                goto out;
3336        }
3337
3338        if (insert->ins_split != SPLIT_NONE) {
3339                /*
3340                 * We could call ocfs2_insert_at_leaf() for some types
3341                 * of splits, but it's easier to just let one seperate
3342                 * function sort it all out.
3343                 */
3344                ocfs2_split_record(inode, left_path, right_path,
3345                                   insert_rec, insert->ins_split);
3346
3347                /*
3348                 * Split might have modified either leaf and we don't
3349                 * have a guarantee that the later edge insert will
3350                 * dirty this for us.
3351                 */
3352                if (left_path)
3353                        ret = ocfs2_journal_dirty(handle,
3354                                                  path_leaf_bh(left_path));
3355                        if (ret)
3356                                mlog_errno(ret);
3357        } else
3358                ocfs2_insert_at_leaf(insert_rec, path_leaf_el(right_path),
3359                                     insert, inode);
3360
3361        ret = ocfs2_journal_dirty(handle, leaf_bh);
3362        if (ret)
3363                mlog_errno(ret);
3364
3365        if (left_path) {
3366                /*
3367                 * The rotate code has indicated that we need to fix
3368                 * up portions of the tree after the insert.
3369                 *
3370                 * XXX: Should we extend the transaction here?
3371                 */
3372                subtree_index = ocfs2_find_subtree_root(inode, left_path,
3373                                                        right_path);
3374                ocfs2_complete_edge_insert(inode, handle, left_path,
3375                                           right_path, subtree_index);
3376        }
3377
3378        ret = 0;
3379out:
3380        return ret;
3381}
3382
3383static int ocfs2_do_insert_extent(struct inode *inode,
3384                                  handle_t *handle,
3385                                  struct buffer_head *di_bh,
3386                                  struct ocfs2_extent_rec *insert_rec,
3387                                  struct ocfs2_insert_type *type)
3388{
3389        int ret, rotate = 0;
3390        u32 cpos;
3391        struct ocfs2_path *right_path = NULL;
3392        struct ocfs2_path *left_path = NULL;
3393        struct ocfs2_dinode *di;
3394        struct ocfs2_extent_list *el;
3395
3396        di = (struct ocfs2_dinode *) di_bh->b_data;
3397        el = &di->id2.i_list;
3398
3399        ret = ocfs2_journal_access(handle, inode, di_bh,
3400                                   OCFS2_JOURNAL_ACCESS_WRITE);
3401        if (ret) {
3402                mlog_errno(ret);
3403                goto out;
3404        }
3405
3406        if (le16_to_cpu(el->l_tree_depth) == 0) {
3407                ocfs2_insert_at_leaf(insert_rec, el, type, inode);
3408                goto out_update_clusters;
3409        }
3410
3411        right_path = ocfs2_new_inode_path(di_bh);
3412        if (!right_path) {
3413                ret = -ENOMEM;
3414                mlog_errno(ret);
3415                goto out;
3416        }
3417
3418        /*
3419         * Determine the path to start with. Rotations need the
3420         * rightmost path, everything else can go directly to the
3421         * target leaf.
3422         */
3423        cpos = le32_to_cpu(insert_rec->e_cpos);
3424        if (type->ins_appending == APPEND_NONE &&
3425            type->ins_contig == CONTIG_NONE) {
3426                rotate = 1;
3427                cpos = UINT_MAX;
3428        }
3429
3430        ret = ocfs2_find_path(inode, right_path, cpos);
3431        if (ret) {
3432                mlog_errno(ret);
3433                goto out;
3434        }
3435
3436        /*
3437         * Rotations and appends need special treatment - they modify
3438         * parts of the tree's above them.
3439         *
3440         * Both might pass back a path immediate to the left of the
3441         * one being inserted to. This will be cause
3442         * ocfs2_insert_path() to modify the rightmost records of
3443         * left_path to account for an edge insert.
3444         *
3445         * XXX: When modifying this code, keep in mind that an insert
3446         * can wind up skipping both of these two special cases...
3447         */
3448        if (rotate) {
3449                ret = ocfs2_rotate_tree_right(inode, handle, type->ins_split,
3450                                              le32_to_cpu(insert_rec->e_cpos),
3451                                              right_path, &left_path);
3452                if (ret) {
3453                        mlog_errno(ret);
3454                        goto out;
3455                }
3456
3457                /*
3458                 * ocfs2_rotate_tree_right() might have extended the
3459                 * transaction without re-journaling our tree root.
3460                 */
3461                ret = ocfs2_journal_access(handle, inode, di_bh,
3462                                           OCFS2_JOURNAL_ACCESS_WRITE);
3463                if (ret) {
3464                        mlog_errno(ret);
3465                        goto out;
3466                }
3467        } else if (type->ins_appending == APPEND_TAIL
3468                   && type->ins_contig != CONTIG_LEFT) {
3469                ret = ocfs2_append_rec_to_path(inode, handle, insert_rec,
3470                                               right_path, &left_path);
3471                if (ret) {
3472                        mlog_errno(ret);
3473                        goto out;
3474                }
3475        }
3476
3477        ret = ocfs2_insert_path(inode, handle, left_path, right_path,
3478                                insert_rec, type);
3479        if (ret) {
3480                mlog_errno(ret);
3481                goto out;
3482        }
3483
3484out_update_clusters:
3485        if (type->ins_split == SPLIT_NONE)
3486                ocfs2_update_dinode_clusters(inode, di,
3487                                             le16_to_cpu(insert_rec->e_leaf_clusters));
3488
3489        ret = ocfs2_journal_dirty(handle, di_bh);
3490        if (ret)
3491                mlog_errno(ret);
3492
3493out:
3494        ocfs2_free_path(left_path);
3495        ocfs2_free_path(right_path);
3496
3497        return ret;
3498}
3499
3500static enum ocfs2_contig_type
3501ocfs2_figure_merge_contig_type(struct inode *inode,
3502                               struct ocfs2_extent_list *el, int index,
3503                               struct ocfs2_extent_rec *split_rec)
3504{
3505        struct ocfs2_extent_rec *rec;
3506        enum ocfs2_contig_type ret = CONTIG_NONE;
3507
3508        /*
3509         * We're careful to check for an empty extent record here -
3510         * the merge code will know what to do if it sees one.
3511         */
3512
3513        if (index > 0) {
3514                rec = &el->l_recs[index - 1];
3515                if (index == 1 && ocfs2_is_empty_extent(rec)) {
3516                        if (split_rec->e_cpos == el->l_recs[index].e_cpos)
3517                                ret = CONTIG_RIGHT;
3518                } else {
3519                        ret = ocfs2_extent_contig(inode, rec, split_rec);
3520                }
3521        }
3522
3523        if (index < (le16_to_cpu(el->l_next_free_rec) - 1)) {
3524                enum ocfs2_contig_type contig_type;
3525
3526                rec = &el->l_recs[index + 1];
3527                contig_type = ocfs2_extent_contig(inode, rec, split_rec);
3528
3529                if (contig_type == CONTIG_LEFT && ret == CONTIG_RIGHT)
3530                        ret = CONTIG_LEFTRIGHT;
3531                else if (ret == CONTIG_NONE)
3532                        ret = contig_type;
3533        }
3534
3535        return ret;
3536}
3537
3538static void ocfs2_figure_contig_type(struct inode *inode,
3539                                     struct ocfs2_insert_type *insert,
3540                                     struct ocfs2_extent_list *el,
3541                                     struct ocfs2_extent_rec *insert_rec)
3542{
3543        int i;
3544        enum ocfs2_contig_type contig_type = CONTIG_NONE;
3545
3546        BUG_ON(le16_to_cpu(el->l_tree_depth) != 0);
3547
3548        for(i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
3549                contig_type = ocfs2_extent_contig(inode, &el->l_recs[i],
3550                                                  insert_rec);
3551                if (contig_type != CONTIG_NONE) {
3552                        insert->ins_contig_index = i;
3553                        break;
3554                }
3555        }
3556        insert->ins_contig = contig_type;
3557}
3558
3559/*
3560 * This should only be called against the righmost leaf extent list.
3561 *
3562 * ocfs2_figure_appending_type() will figure out whether we'll have to
3563 * insert at the tail of the rightmost leaf.
3564 *
3565 * This should also work against the dinode list for tree's with 0
3566 * depth. If we consider the dinode list to be the rightmost leaf node
3567 * then the logic here makes sense.
3568 */
3569static void ocfs2_figure_appending_type(struct ocfs2_insert_type *insert,
3570                                        struct ocfs2_extent_list *el,
3571                                        struct ocfs2_extent_rec *insert_rec)
3572{
3573        int i;
3574        u32 cpos = le32_to_cpu(insert_rec->e_cpos);
3575        struct ocfs2_extent_rec *rec;
3576
3577        insert->ins_appending = APPEND_NONE;
3578
3579        BUG_ON(le16_to_cpu(el->l_tree_depth) != 0);
3580
3581        if (!el->l_next_free_rec)
3582                goto set_tail_append;
3583
3584        if (ocfs2_is_empty_extent(&el->l_recs[0])) {
3585                /* Were all records empty? */
3586                if (le16_to_cpu(el->l_next_free_rec) == 1)
3587                        goto set_tail_append;
3588        }
3589
3590        i = le16_to_cpu(el->l_next_free_rec) - 1;
3591        rec = &el->l_recs[i];
3592
3593        if (cpos >=
3594            (le32_to_cpu(rec->e_cpos) + le16_to_cpu(rec->e_leaf_clusters)))
3595                goto set_tail_append;
3596
3597        return;
3598
3599set_tail_append:
3600        insert->ins_appending = APPEND_TAIL;
3601}
3602
3603/*
3604 * Helper function called at the begining of an insert.
3605 *
3606 * This computes a few things that are commonly used in the process of
3607 * inserting into the btree:
3608 *   - Whether the new extent is contiguous with an existing one.
3609 *   - The current tree depth.
3610 *   - Whether the insert is an appending one.
3611 *   - The total # of free records in the tree.
3612 *
3613 * All of the information is stored on the ocfs2_insert_type
3614 * structure.
3615 */
3616static int ocfs2_figure_insert_type(struct inode *inode,
3617                                    struct buffer_head *di_bh,
3618                                    struct buffer_head **last_eb_bh,
3619                                    struct ocfs2_extent_rec *insert_rec,
3620                                    int *free_records,
3621                                    struct ocfs2_insert_type *insert)
3622{
3623        int ret;
3624        struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
3625        struct ocfs2_extent_block *eb;
3626        struct ocfs2_extent_list *el;
3627        struct ocfs2_path *path = NULL;
3628        struct buffer_head *bh = NULL;
3629
3630        insert->ins_split = SPLIT_NONE;
3631
3632        el = &di->id2.i_list;
3633        insert->ins_tree_depth = le16_to_cpu(el->l_tree_depth);
3634
3635        if (el->l_tree_depth) {
3636                /*
3637                 * If we have tree depth, we read in the
3638                 * rightmost extent block ahead of time as
3639                 * ocfs2_figure_insert_type() and ocfs2_add_branch()
3640                 * may want it later.
3641                 */
3642                ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
3643                                       le64_to_cpu(di->i_last_eb_blk), &bh,
3644                                       OCFS2_BH_CACHED, inode);
3645                if (ret) {
3646                        mlog_exit(ret);
3647                        goto out;
3648                }
3649                eb = (struct ocfs2_extent_block *) bh->b_data;
3650                el = &eb->h_list;
3651        }
3652
3653        /*
3654         * Unless we have a contiguous insert, we'll need to know if
3655         * there is room left in our allocation tree for another
3656         * extent record.
3657         *
3658         * XXX: This test is simplistic, we can search for empty
3659         * extent records too.
3660         */
3661        *free_records = le16_to_cpu(el->l_count) -
3662                le16_to_cpu(el->l_next_free_rec);
3663
3664        if (!insert->ins_tree_depth) {
3665                ocfs2_figure_contig_type(inode, insert, el, insert_rec);
3666                ocfs2_figure_appending_type(insert, el, insert_rec);
3667                return 0;
3668        }
3669
3670        path = ocfs2_new_inode_path(di_bh);
3671        if (!path) {
3672                ret = -ENOMEM;
3673                mlog_errno(ret);
3674                goto out;
3675        }
3676
3677        /*
3678         * In the case that we're inserting past what the tree
3679         * currently accounts for, ocfs2_find_path() will return for
3680         * us the rightmost tree path. This is accounted for below in
3681         * the appending code.
3682         */
3683        ret = ocfs2_find_path(inode, path, le32_to_cpu(insert_rec->e_cpos));
3684        if (ret) {
3685                mlog_errno(ret);
3686                goto out;
3687        }
3688
3689        el = path_leaf_el(path);
3690
3691        /*
3692         * Now that we have the path, there's two things we want to determine:
3693         * 1) Contiguousness (also set contig_index if this is so)
3694         *
3695         * 2) Are we doing an append? We can trivially break this up
3696         *     into two types of appends: simple record append, or a
3697         *     rotate inside the tail leaf.
3698         */
3699        ocfs2_figure_contig_type(inode, insert, el, insert_rec);
3700
3701        /*
3702         * The insert code isn't quite ready to deal with all cases of
3703         * left contiguousness. Specifically, if it's an insert into
3704         * the 1st record in a leaf, it will require the adjustment of
3705         * cluster count on the last record of the path directly to it's
3706         * left. For now, just catch that case and fool the layers
3707         * above us. This works just fine for tree_depth == 0, which
3708         * is why we allow that above.
3709         */
3710        if (insert->ins_contig == CONTIG_LEFT &&
3711            insert->ins_contig_index == 0)
3712                insert->ins_contig = CONTIG_NONE;
3713
3714        /*
3715         * Ok, so we can simply compare against last_eb to figure out
3716         * whether the path doesn't exist. This will only happen in
3717         * the case that we're doing a tail append, so maybe we can
3718         * take advantage of that information somehow.
3719         */
3720        if (le64_to_cpu(di->i_last_eb_blk) == path_leaf_bh(path)->b_blocknr) {
3721                /*
3722                 * Ok, ocfs2_find_path() returned us the rightmost
3723                 * tree path. This might be an appending insert. There are
3724                 * two cases:
3725                 *    1) We're doing a true append at the tail:
3726                 *      -This might even be off the end of the leaf
3727                 *    2) We're "appending" by rotating in the tail
3728                 */
3729                ocfs2_figure_appending_type(insert, el, insert_rec);
3730        }
3731
3732out:
3733        ocfs2_free_path(path);
3734
3735        if (ret == 0)
3736                *last_eb_bh = bh;
3737        else
3738                brelse(bh);
3739        return ret;
3740}
3741
3742/*
3743 * Insert an extent into an inode btree.
3744 *
3745 * The caller needs to update fe->i_clusters
3746 */
3747int ocfs2_insert_extent(struct ocfs2_super *osb,
3748                        handle_t *handle,
3749                        struct inode *inode,
3750                        struct buffer_head *fe_bh,
3751                        u32 cpos,
3752                        u64 start_blk,
3753                        u32 new_clusters,
3754                        u8 flags,
3755                        struct ocfs2_alloc_context *meta_ac)
3756{
3757        int status;
3758        int uninitialized_var(free_records);
3759        struct buffer_head *last_eb_bh = NULL;
3760        struct ocfs2_insert_type insert = {0, };
3761        struct ocfs2_extent_rec rec;
3762
3763        BUG_ON(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL);
3764
3765        mlog(0, "add %u clusters at position %u to inode %llu\n",
3766             new_clusters, cpos, (unsigned long long)OCFS2_I(inode)->ip_blkno);
3767
3768        mlog_bug_on_msg(!ocfs2_sparse_alloc(osb) &&
3769                        (OCFS2_I(inode)->ip_clusters != cpos),
3770                        "Device %s, asking for sparse allocation: inode %llu, "
3771                        "cpos %u, clusters %u\n",
3772                        osb->dev_str,
3773                        (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos,
3774                        OCFS2_I(inode)->ip_clusters);
3775
3776        memset(&rec, 0, sizeof(rec));
3777        rec.e_cpos = cpu_to_le32(cpos);
3778        rec.e_blkno = cpu_to_le64(start_blk);
3779        rec.e_leaf_clusters = cpu_to_le16(new_clusters);
3780        rec.e_flags = flags;
3781
3782        status = ocfs2_figure_insert_type(inode, fe_bh, &last_eb_bh, &rec,
3783                                          &free_records, &insert);
3784        if (status < 0) {
3785                mlog_errno(status);
3786                goto bail;
3787        }
3788
3789        mlog(0, "Insert.appending: %u, Insert.Contig: %u, "
3790             "Insert.contig_index: %d, Insert.free_records: %d, "
3791             "Insert.tree_depth: %d\n",
3792             insert.ins_appending, insert.ins_contig, insert.ins_contig_index,
3793             free_records, insert.ins_tree_depth);
3794
3795        if (insert.ins_contig == CONTIG_NONE && free_records == 0) {
3796                status = ocfs2_grow_tree(inode, handle, fe_bh,
3797                                         &insert.ins_tree_depth, &last_eb_bh,
3798                                         meta_ac);
3799                if (status) {
3800                        mlog_errno(status);
3801                        goto bail;
3802                }
3803        }
3804
3805        /* Finally, we can add clusters. This might rotate the tree for us. */
3806        status = ocfs2_do_insert_extent(inode, handle, fe_bh, &rec, &insert);
3807        if (status < 0)
3808                mlog_errno(status);
3809        else
3810                ocfs2_extent_map_insert_rec(inode, &rec);
3811
3812bail:
3813        if (last_eb_bh)
3814                brelse(last_eb_bh);
3815
3816        mlog_exit(status);
3817        return status;
3818}
3819
3820static void ocfs2_make_right_split_rec(struct super_block *sb,
3821                                       struct ocfs2_extent_rec *split_rec,
3822                                       u32 cpos,
3823                                       struct ocfs2_extent_rec *rec)
3824{
3825        u32 rec_cpos = le32_to_cpu(rec->e_cpos);
3826        u32 rec_range = rec_cpos + le16_to_cpu(rec->e_leaf_clusters);
3827
3828        memset(split_rec, 0, sizeof(struct ocfs2_extent_rec));
3829
3830        split_rec->e_cpos = cpu_to_le32(cpos);
3831        split_rec->e_leaf_clusters = cpu_to_le16(rec_range - cpos);
3832
3833        split_rec->e_blkno = rec->e_blkno;
3834        le64_add_cpu(&split_rec->e_blkno,
3835                     ocfs2_clusters_to_blocks(sb, cpos - rec_cpos));
3836
3837        split_rec->e_flags = rec->e_flags;
3838}
3839
3840static int ocfs2_split_and_insert(struct inode *inode,
3841                                  handle_t *handle,
3842                                  struct ocfs2_path *path,
3843                                  struct buffer_head *di_bh,
3844                                  struct buffer_head **last_eb_bh,
3845                                  int split_index,
3846                                  struct ocfs2_extent_rec *orig_split_rec,
3847                                  struct ocfs2_alloc_context *meta_ac)
3848{
3849        int ret = 0, depth;
3850        unsigned int insert_range, rec_range, do_leftright = 0;
3851        struct ocfs2_extent_rec tmprec;
3852        struct ocfs2_extent_list *rightmost_el;
3853        struct ocfs2_extent_rec rec;
3854        struct ocfs2_extent_rec split_rec = *orig_split_rec;
3855        struct ocfs2_insert_type insert;
3856        struct ocfs2_extent_block *eb;
3857        struct ocfs2_dinode *di;
3858
3859leftright:
3860        /*
3861         * Store a copy of the record on the stack - it might move
3862         * around as the tree is manipulated below.
3863         */
3864        rec = path_leaf_el(path)->l_recs[split_index];
3865
3866        di = (struct ocfs2_dinode *)di_bh->b_data;
3867        rightmost_el = &di->id2.i_list;
3868
3869        depth = le16_to_cpu(rightmost_el->l_tree_depth);
3870        if (depth) {
3871                BUG_ON(!(*last_eb_bh));
3872                eb = (struct ocfs2_extent_block *) (*last_eb_bh)->b_data;
3873                rightmost_el = &eb->h_list;
3874        }
3875
3876        if (le16_to_cpu(rightmost_el->l_next_free_rec) ==
3877            le16_to_cpu(rightmost_el->l_count)) {
3878                ret = ocfs2_grow_tree(inode, handle, di_bh, &depth, last_eb_bh,
3879                                      meta_ac);
3880                if (ret) {
3881                        mlog_errno(ret);
3882                        goto out;
3883                }
3884        }
3885
3886        memset(&insert, 0, sizeof(struct ocfs2_insert_type));
3887        insert.ins_appending = APPEND_NONE;
3888        insert.ins_contig = CONTIG_NONE;
3889        insert.ins_tree_depth = depth;
3890
3891        insert_range = le32_to_cpu(split_rec.e_cpos) +
3892                le16_to_cpu(split_rec.e_leaf_clusters);
3893        rec_range = le32_to_cpu(rec.e_cpos) +
3894                le16_to_cpu(rec.e_leaf_clusters);
3895
3896        if (split_rec.e_cpos == rec.e_cpos) {
3897                insert.ins_split = SPLIT_LEFT;
3898        } else if (insert_range == rec_range) {
3899                insert.ins_split = SPLIT_RIGHT;
3900        } else {
3901                /*
3902                 * Left/right split. We fake this as a right split
3903                 * first and then make a second pass as a left split.
3904                 */
3905                insert.ins_split = SPLIT_RIGHT;
3906
3907                ocfs2_make_right_split_rec(inode->i_sb, &tmprec, insert_range,
3908                                           &rec);
3909
3910                split_rec = tmprec;
3911
3912                BUG_ON(do_leftright);
3913                do_leftright = 1;
3914        }
3915
3916        ret = ocfs2_do_insert_extent(inode, handle, di_bh, &split_rec,
3917                                     &insert);
3918        if (ret) {
3919                mlog_errno(ret);
3920                goto out;
3921        }
3922
3923        if (do_leftright == 1) {
3924                u32 cpos;
3925                struct ocfs2_extent_list *el;
3926
3927                do_leftright++;
3928                split_rec = *orig_split_rec;
3929
3930                ocfs2_reinit_path(path, 1);
3931
3932                cpos = le32_to_cpu(split_rec.e_cpos);
3933                ret = ocfs2_find_path(inode, path, cpos);
3934                if (ret) {
3935                        mlog_errno(ret);
3936                        goto out;
3937                }
3938
3939                el = path_leaf_el(path);
3940                split_index = ocfs2_search_extent_list(el, cpos);
3941                goto leftright;
3942        }
3943out:
3944
3945        return ret;
3946}
3947
3948/*
3949 * Mark part or all of the extent record at split_index in the leaf
3950 * pointed to by path as written. This removes the unwritten
3951 * extent flag.
3952 *
3953 * Care is taken to handle contiguousness so as to not grow the tree.
3954 *
3955 * meta_ac is not strictly necessary - we only truly need it if growth
3956 * of the tree is required. All other cases will degrade into a less
3957 * optimal tree layout.
3958 *
3959 * last_eb_bh should be the rightmost leaf block for any inode with a
3960 * btree. Since a split may grow the tree or a merge might shrink it, the caller cannot trust the contents of that buffer after this call.
3961 *
3962 * This code is optimized for readability - several passes might be
3963 * made over certain portions of the tree. All of those blocks will
3964 * have been brought into cache (and pinned via the journal), so the
3965 * extra overhead is not expressed in terms of disk reads.
3966 */
3967static int __ocfs2_mark_extent_written(struct inode *inode,
3968                                       struct buffer_head *di_bh,
3969                                       handle_t *handle,
3970                                       struct ocfs2_path *path,
3971                                       int split_index,
3972                                       struct ocfs2_extent_rec *split_rec,
3973                                       struct ocfs2_alloc_context *meta_ac,
3974                                       struct ocfs2_cached_dealloc_ctxt *dealloc)
3975{
3976        int ret = 0;
3977        struct ocfs2_extent_list *el = path_leaf_el(path);
3978        struct buffer_head *last_eb_bh = NULL;
3979        struct ocfs2_extent_rec *rec = &el->l_recs[split_index];
3980        struct ocfs2_merge_ctxt ctxt;
3981        struct ocfs2_extent_list *rightmost_el;
3982
3983        if (!(rec->e_flags & OCFS2_EXT_UNWRITTEN)) {
3984                ret = -EIO;
3985                mlog_errno(ret);
3986                goto out;
3987        }
3988
3989        if (le32_to_cpu(rec->e_cpos) > le32_to_cpu(split_rec->e_cpos) ||
3990            ((le32_to_cpu(rec->e_cpos) + le16_to_cpu(rec->e_leaf_clusters)) <
3991             (le32_to_cpu(split_rec->e_cpos) + le16_to_cpu(split_rec->e_leaf_clusters)))) {
3992                ret = -EIO;
3993                mlog_errno(ret);
3994                goto out;
3995        }
3996
3997        ctxt.c_contig_type = ocfs2_figure_merge_contig_type(inode, el,
3998                                                            split_index,
3999                                                            split_rec);
4000
4001        /*
4002         * The core merge / split code wants to know how much room is
4003         * left in this inodes allocation tree, so we pass the
4004         * rightmost extent list.
4005         */
4006        if (path->p_tree_depth) {
4007                struct ocfs2_extent_block *eb;
4008                struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
4009
4010                ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
4011                                       le64_to_cpu(di->i_last_eb_blk),
4012                                       &last_eb_bh, OCFS2_BH_CACHED, inode);
4013                if (ret) {
4014                        mlog_exit(ret);
4015                        goto out;
4016                }
4017
4018                eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
4019                if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
4020                        OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
4021                        ret = -EROFS;
4022                        goto out;
4023                }
4024
4025                rightmost_el = &eb->h_list;
4026        } else
4027                rightmost_el = path_root_el(path);
4028
4029        if (rec->e_cpos == split_rec->e_cpos &&
4030            rec->e_leaf_clusters == split_rec->e_leaf_clusters)
4031                ctxt.c_split_covers_rec = 1;
4032        else
4033                ctxt.c_split_covers_rec = 0;
4034
4035        ctxt.c_has_empty_extent = ocfs2_is_empty_extent(&el->l_recs[0]);
4036
4037        mlog(0, "index: %d, contig: %u, has_empty: %u, split_covers: %u\n",
4038             split_index, ctxt.c_contig_type, ctxt.c_has_empty_extent,
4039             ctxt.c_split_covers_rec);
4040
4041        if (ctxt.c_contig_type == CONTIG_NONE) {
4042                if (ctxt.c_split_covers_rec)
4043                        el->l_recs[split_index] = *split_rec;
4044                else
4045                        ret = ocfs2_split_and_insert(inode, handle, path, di_bh,
4046                                                     &last_eb_bh, split_index,
4047                                                     split_rec, meta_ac);
4048                if (ret)
4049                        mlog_errno(ret);
4050        } else {
4051                ret = ocfs2_try_to_merge_extent(inode, handle, path,
4052                                                split_index, split_rec,
4053                                                dealloc, &ctxt);
4054                if (ret)
4055                        mlog_errno(ret);
4056        }
4057
4058out:
4059        brelse(last_eb_bh);
4060        return ret;
4061}
4062
4063/*
4064 * Mark the already-existing extent at cpos as written for len clusters.
4065 *
4066 * If the existing extent is larger than the request, initiate a
4067 * split. An attempt will be made at merging with adjacent extents.
4068 *
4069 * The caller is responsible for passing down meta_ac if we'll need it.
4070 */
4071int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *di_bh,
4072                              handle_t *handle, u32 cpos, u32 len, u32 phys,
4073                              struct ocfs2_alloc_context *meta_ac,
4074                              struct ocfs2_cached_dealloc_ctxt *dealloc)
4075{
4076        int ret, index;
4077        u64 start_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys);
4078        struct ocfs2_extent_rec split_rec;
4079        struct ocfs2_path *left_path = NULL;
4080        struct ocfs2_extent_list *el;
4081
4082        mlog(0, "Inode %lu cpos %u, len %u, phys %u (%llu)\n",
4083             inode->i_ino, cpos, len, phys, (unsigned long long)start_blkno);
4084
4085        if (!ocfs2_writes_unwritten_extents(OCFS2_SB(inode->i_sb))) {
4086                ocfs2_error(inode->i_sb, "Inode %llu has unwritten extents "
4087                            "that are being written to, but the feature bit "
4088                            "is not set in the super block.",
4089                            (unsigned long long)OCFS2_I(inode)->ip_blkno);
4090                ret = -EROFS;
4091                goto out;
4092        }
4093
4094        /*
4095         * XXX: This should be fixed up so that we just re-insert the
4096         * next extent records.
4097         */
4098        ocfs2_extent_map_trunc(inode, 0);
4099
4100        left_path = ocfs2_new_inode_path(di_bh);
4101        if (!left_path) {
4102                ret = -ENOMEM;
4103                mlog_errno(ret);
4104                goto out;
4105        }
4106
4107        ret = ocfs2_find_path(inode, left_path, cpos);
4108        if (ret) {
4109                mlog_errno(ret);
4110                goto out;
4111        }
4112        el = path_leaf_el(left_path);
4113
4114        index = ocfs2_search_extent_list(el, cpos);
4115        if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) {
4116                ocfs2_error(inode->i_sb,
4117                            "Inode %llu has an extent at cpos %u which can no "
4118                            "longer be found.\n",
4119                            (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos);
4120                ret = -EROFS;
4121                goto out;
4122        }
4123
4124        memset(&split_rec, 0, sizeof(struct ocfs2_extent_rec));
4125        split_rec.e_cpos = cpu_to_le32(cpos);
4126        split_rec.e_leaf_clusters = cpu_to_le16(len);
4127        split_rec.e_blkno = cpu_to_le64(start_blkno);
4128        split_rec.e_flags = path_leaf_el(left_path)->l_recs[index].e_flags;
4129        split_rec.e_flags &= ~OCFS2_EXT_UNWRITTEN;
4130
4131        ret = __ocfs2_mark_extent_written(inode, di_bh, handle, left_path,
4132                                          index, &split_rec, meta_ac, dealloc);
4133        if (ret)
4134                mlog_errno(ret);
4135
4136out:
4137        ocfs2_free_path(left_path);
4138        return ret;
4139}
4140
4141static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh,
4142                            handle_t *handle, struct ocfs2_path *path,
4143                            int index, u32 new_range,
4144                            struct ocfs2_alloc_context *meta_ac)
4145{
4146        int ret, depth, credits = handle->h_buffer_credits;
4147        struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
4148        struct buffer_head *last_eb_bh = NULL;
4149        struct ocfs2_extent_block *eb;
4150        struct ocfs2_extent_list *rightmost_el, *el;
4151        struct ocfs2_extent_rec split_rec;
4152        struct ocfs2_extent_rec *rec;
4153        struct ocfs2_insert_type insert;
4154
4155        /*
4156         * Setup the record to split before we grow the tree.
4157         */
4158        el = path_leaf_el(path);
4159        rec = &el->l_recs[index];
4160        ocfs2_make_right_split_rec(inode->i_sb, &split_rec, new_range, rec);
4161
4162        depth = path->p_tree_depth;
4163        if (depth > 0) {
4164                ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
4165                                       le64_to_cpu(di->i_last_eb_blk),
4166                                       &last_eb_bh, OCFS2_BH_CACHED, inode);
4167                if (ret < 0) {
4168                        mlog_errno(ret);
4169                        goto out;
4170                }
4171
4172                eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
4173                rightmost_el = &eb->h_list;
4174        } else
4175                rightmost_el = path_leaf_el(path);
4176
4177        credits += path->p_tree_depth + ocfs2_extend_meta_needed(di);
4178        ret = ocfs2_extend_trans(handle, credits);
4179        if (ret) {
4180                mlog_errno(ret);
4181                goto out;
4182        }
4183
4184        if (le16_to_cpu(rightmost_el->l_next_free_rec) ==
4185            le16_to_cpu(rightmost_el->l_count)) {
4186                ret = ocfs2_grow_tree(inode, handle, di_bh, &depth, &last_eb_bh,
4187                                      meta_ac);
4188                if (ret) {
4189                        mlog_errno(ret);
4190                        goto out;
4191                }
4192        }
4193
4194        memset(&insert, 0, sizeof(struct ocfs2_insert_type));
4195        insert.ins_appending = APPEND_NONE;
4196        insert.ins_contig = CONTIG_NONE;
4197        insert.ins_split = SPLIT_RIGHT;
4198        insert.ins_tree_depth = depth;
4199
4200        ret = ocfs2_do_insert_extent(inode, handle, di_bh, &split_rec, &insert);
4201        if (ret)
4202                mlog_errno(ret);
4203
4204out:
4205        brelse(last_eb_bh);
4206        return ret;
4207}
4208
4209static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle,
4210                              struct ocfs2_path *path, int index,
4211                              struct ocfs2_cached_dealloc_ctxt *dealloc,
4212                              u32 cpos, u32 len)
4213{
4214        int ret;
4215        u32 left_cpos, rec_range, trunc_range;
4216        int wants_rotate = 0, is_rightmost_tree_rec = 0;
4217        struct super_block *sb = inode->i_sb;
4218        struct ocfs2_path *left_path = NULL;
4219        struct ocfs2_extent_list *el = path_leaf_el(path);
4220        struct ocfs2_extent_rec *rec;
4221        struct ocfs2_extent_block *eb;
4222
4223        if (ocfs2_is_empty_extent(&el->l_recs[0]) && index > 0) {
4224                ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc);
4225                if (ret) {
4226                        mlog_errno(ret);
4227                        goto out;
4228                }
4229
4230                index--;
4231        }
4232
4233        if (index == (le16_to_cpu(el->l_next_free_rec) - 1) &&
4234            path->p_tree_depth) {
4235                /*
4236                 * Check whether this is the rightmost tree record. If
4237                 * we remove all of this record or part of its right
4238                 * edge then an update of the record lengths above it
4239                 * will be required.
4240                 */
4241                eb = (struct ocfs2_extent_block *)path_leaf_bh(path)->b_data;
4242                if (eb->h_next_leaf_blk == 0)
4243                        is_rightmost_tree_rec = 1;
4244        }
4245
4246        rec = &el->l_recs[index];
4247        if (index == 0 && path->p_tree_depth &&
4248            le32_to_cpu(rec->e_cpos) == cpos) {
4249                /*
4250                 * Changing the leftmost offset (via partial or whole
4251                 * record truncate) of an interior (or rightmost) path
4252                 * means we have to update the subtree that is formed
4253                 * by this leaf and the one to it's left.
4254                 *
4255                 * There are two cases we can skip:
4256                 *   1) Path is the leftmost one in our inode tree.
4257                 *   2) The leaf is rightmost and will be empty after
4258                 *      we remove the extent record - the rotate code
4259                 *      knows how to update the newly formed edge.
4260                 */
4261
4262                ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path,
4263                                                    &left_cpos);
4264                if (ret) {
4265                        mlog_errno(ret);
4266                        goto out;
4267                }
4268
4269                if (left_cpos && le16_to_cpu(el->l_next_free_rec) > 1) {
4270                        left_path = ocfs2_new_path(path_root_bh(path),
4271                                                   path_root_el(path));
4272                        if (!left_path) {
4273                                ret = -ENOMEM;
4274                                mlog_errno(ret);
4275                                goto out;
4276                        }
4277
4278                        ret = ocfs2_find_path(inode, left_path, left_cpos);
4279                        if (ret) {
4280                                mlog_errno(ret);
4281                                goto out;
4282                        }
4283                }
4284        }
4285
4286        ret = ocfs2_extend_rotate_transaction(handle, 0,
4287                                              handle->h_buffer_credits,
4288                                              path);
4289        if (ret) {
4290                mlog_errno(ret);
4291                goto out;
4292        }
4293
4294        ret = ocfs2_journal_access_path(inode, handle, path);
4295        if (ret) {
4296                mlog_errno(ret);
4297                goto out;
4298        }
4299
4300        ret = ocfs2_journal_access_path(inode, handle, left_path);
4301        if (ret) {
4302                mlog_errno(ret);
4303                goto out;
4304        }
4305
4306        rec_range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
4307        trunc_range = cpos + len;
4308
4309        if (le32_to_cpu(rec->e_cpos) == cpos && rec_range == trunc_range) {
4310                int next_free;
4311
4312                memset(rec, 0, sizeof(*rec));
4313                ocfs2_cleanup_merge(el, index);
4314                wants_rotate = 1;
4315
4316                next_free = le16_to_cpu(el->l_next_free_rec);
4317                if (is_rightmost_tree_rec && next_free > 1) {
4318                        /*
4319                         * We skip the edge update if this path will
4320                         * be deleted by the rotate code.
4321                         */
4322                        rec = &el->l_recs[next_free - 1];
4323                        ocfs2_adjust_rightmost_records(inode, handle, path,
4324                                                       rec);
4325                }
4326        } else if (le32_to_cpu(rec->e_cpos) == cpos) {
4327                /* Remove leftmost portion of the record. */
4328                le32_add_cpu(&rec->e_cpos, len);
4329                le64_add_cpu(&rec->e_blkno, ocfs2_clusters_to_blocks(sb, len));
4330                le16_add_cpu(&rec->e_leaf_clusters, -len);
4331        } else if (rec_range == trunc_range) {
4332                /* Remove rightmost portion of the record */
4333                le16_add_cpu(&rec->e_leaf_clusters, -len);
4334                if (is_rightmost_tree_rec)
4335                        ocfs2_adjust_rightmost_records(inode, handle, path, rec);
4336        } else {
4337                /* Caller should have trapped this. */
4338                mlog(ML_ERROR, "Inode %llu: Invalid record truncate: (%u, %u) "
4339                     "(%u, %u)\n", (unsigned long long)OCFS2_I(inode)->ip_blkno,
4340                     le32_to_cpu(rec->e_cpos),
4341                     le16_to_cpu(rec->e_leaf_clusters), cpos, len);
4342                BUG();
4343        }
4344
4345        if (left_path) {
4346                int subtree_index;
4347
4348                subtree_index = ocfs2_find_subtree_root(inode, left_path, path);
4349                ocfs2_complete_edge_insert(inode, handle, left_path, path,
4350                                           subtree_index);
4351        }
4352
4353        ocfs2_journal_dirty(handle, path_leaf_bh(path));
4354
4355        ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc);
4356        if (ret) {
4357                mlog_errno(ret);
4358                goto out;
4359        }
4360
4361out:
4362        ocfs2_free_path(left_path);
4363        return ret;
4364}
4365
4366int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh,
4367                        u32 cpos, u32 len, handle_t *handle,
4368                        struct ocfs2_alloc_context *meta_ac,
4369                        struct ocfs2_cached_dealloc_ctxt *dealloc)
4370{
4371        int ret, index;
4372        u32 rec_range, trunc_range;
4373        struct ocfs2_extent_rec *rec;
4374        struct ocfs2_extent_list *el;
4375        struct ocfs2_path *path;
4376
4377        ocfs2_extent_map_trunc(inode, 0);
4378
4379        path = ocfs2_new_inode_path(di_bh);
4380        if (!path) {
4381                ret = -ENOMEM;
4382                mlog_errno(ret);
4383                goto out;
4384        }
4385
4386        ret = ocfs2_find_path(inode, path, cpos);
4387        if (ret) {
4388                mlog_errno(ret);
4389                goto out;
4390        }
4391
4392        el = path_leaf_el(path);
4393        index = ocfs2_search_extent_list(el, cpos);
4394        if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) {
4395                ocfs2_error(inode->i_sb,
4396                            "Inode %llu has an extent at cpos %u which can no "
4397                            "longer be found.\n",
4398                            (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos);
4399                ret = -EROFS;
4400                goto out;
4401        }
4402
4403        /*
4404         * We have 3 cases of extent removal:
4405         *   1) Range covers the entire extent rec
4406         *   2) Range begins or ends on one edge of the extent rec
4407         *   3) Range is in the middle of the extent rec (no shared edges)
4408         *
4409         * For case 1 we remove the extent rec and left rotate to
4410         * fill the hole.
4411         *
4412         * For case 2 we just shrink the existing extent rec, with a
4413         * tree update if the shrinking edge is also the edge of an
4414         * extent block.
4415         *
4416         * For case 3 we do a right split to turn the extent rec into
4417         * something case 2 can handle.
4418         */
4419        rec = &el->l_recs[index];
4420        rec_range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
4421        trunc_range = cpos + len;
4422
4423        BUG_ON(cpos < le32_to_cpu(rec->e_cpos) || trunc_range > rec_range);
4424
4425        mlog(0, "Inode %llu, remove (cpos %u, len %u). Existing index %d "
4426             "(cpos %u, len %u)\n",
4427             (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos, len, index,
4428             le32_to_cpu(rec->e_cpos), ocfs2_rec_clusters(el, rec));
4429
4430        if (le32_to_cpu(rec->e_cpos) == cpos || rec_range == trunc_range) {
4431                ret = ocfs2_truncate_rec(inode, handle, path, index, dealloc,
4432                                         cpos, len);
4433                if (ret) {
4434                        mlog_errno(ret);
4435                        goto out;
4436                }
4437        } else {
4438                ret = ocfs2_split_tree(inode, di_bh, handle, path, index,
4439                                       trunc_range, meta_ac);
4440                if (ret) {
4441                        mlog_errno(ret);
4442                        goto out;
4443                }
4444
4445                /*
4446                 * The split could have manipulated the tree enough to
4447                 * move the record location, so we have to look for it again.
4448                 */
4449                ocfs2_reinit_path(path, 1);
4450
4451                ret = ocfs2_find_path(inode, path, cpos);
4452                if (ret) {
4453                        mlog_errno(ret);
4454                        goto out;
4455                }
4456
4457                el = path_leaf_el(path);
4458                index = ocfs2_search_extent_list(el, cpos);
4459                if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) {
4460                        ocfs2_error(inode->i_sb,
4461                                    "Inode %llu: split at cpos %u lost record.",
4462                                    (unsigned long long)OCFS2_I(inode)->ip_blkno,
4463                                    cpos);
4464                        ret = -EROFS;
4465                        goto out;
4466                }
4467
4468                /*
4469                 * Double check our values here. If anything is fishy,
4470                 * it's easier to catch it at the top level.
4471                 */
4472                rec = &el->l_recs[index];
4473                rec_range = le32_to_cpu(rec->e_cpos) +
4474                        ocfs2_rec_clusters(el, rec);
4475                if (rec_range != trunc_range) {
4476                        ocfs2_error(inode->i_sb,
4477                                    "Inode %llu: error after split at cpos %u"
4478                                    "trunc len %u, existing record is (%u,%u)",
4479                                    (unsigned long long)OCFS2_I(inode)->ip_blkno,
4480                                    cpos, len, le32_to_cpu(rec->e_cpos),
4481                                    ocfs2_rec_clusters(el, rec));
4482                        ret = -EROFS;
4483                        goto out;
4484                }
4485
4486                ret = ocfs2_truncate_rec(inode, handle, path, index, dealloc,
4487                                         cpos, len);
4488                if (ret) {
4489                        mlog_errno(ret);
4490                        goto out;
4491                }
4492        }
4493
4494out:
4495        ocfs2_free_path(path);
4496        return ret;
4497}
4498
4499int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb)
4500{
4501        struct buffer_head *tl_bh = osb->osb_tl_bh;
4502        struct ocfs2_dinode *di;
4503        struct ocfs2_truncate_log *tl;
4504
4505        di = (struct ocfs2_dinode *) tl_bh->b_data;
4506        tl = &di->id2.i_dealloc;
4507
4508        mlog_bug_on_msg(le16_to_cpu(tl->tl_used) > le16_to_cpu(tl->tl_count),
4509                        "slot %d, invalid truncate log parameters: used = "
4510                        "%u, count = %u\n", osb->slot_num,
4511                        le16_to_cpu(tl->tl_used), le16_to_cpu(tl->tl_count));
4512        return le16_to_cpu(tl->tl_used) == le16_to_cpu(tl->tl_count);
4513}
4514
4515static int ocfs2_truncate_log_can_coalesce(struct ocfs2_truncate_log *tl,
4516                                           unsigned int new_start)
4517{
4518        unsigned int tail_index;
4519        unsigned int current_tail;
4520
4521        /* No records, nothing to coalesce */
4522        if (!le16_to_cpu(tl->tl_used))
4523                return 0;
4524
4525        tail_index = le16_to_cpu(tl->tl_used) - 1;
4526        current_tail = le32_to_cpu(tl->tl_recs[tail_index].t_start);
4527        current_tail += le32_to_cpu(tl->tl_recs[tail_index].t_clusters);
4528
4529        return current_tail == new_start;
4530}
4531
4532int ocfs2_truncate_log_append(struct ocfs2_super *osb,
4533                              handle_t *handle,
4534                              u64 start_blk,
4535                              unsigned int num_clusters)
4536{
4537        int status, index;
4538        unsigned int start_cluster, tl_count;
4539        struct inode *tl_inode = osb->osb_tl_inode;
4540        struct buffer_head *tl_bh = osb->osb_tl_bh;
4541        struct ocfs2_dinode *di;
4542        struct ocfs2_truncate_log *tl;
4543
4544        mlog_entry("start_blk = %llu, num_clusters = %u\n",
4545                   (unsigned long long)start_blk, num_clusters);
4546
4547        BUG_ON(mutex_trylock(&tl_inode->i_mutex));
4548
4549        start_cluster = ocfs2_blocks_to_clusters(osb->sb, start_blk);
4550
4551        di = (struct ocfs2_dinode *) tl_bh->b_data;
4552        tl = &di->id2.i_dealloc;
4553        if (!OCFS2_IS_VALID_DINODE(di)) {
4554                OCFS2_RO_ON_INVALID_DINODE(osb->sb, di);
4555                status = -EIO;
4556                goto bail;
4557        }
4558
4559        tl_count = le16_to_cpu(tl->tl_count);
4560        mlog_bug_on_msg(tl_count > ocfs2_truncate_recs_per_inode(osb->sb) ||
4561                        tl_count == 0,
4562                        "Truncate record count on #%llu invalid "
4563                        "wanted %u, actual %u\n",
4564                        (unsigned long long)OCFS2_I(tl_inode)->ip_blkno,
4565                        ocfs2_truncate_recs_per_inode(osb->sb),
4566                        le16_to_cpu(tl->tl_count));
4567
4568        /* Caller should have known to flush before calling us. */
4569        index = le16_to_cpu(tl->tl_used);
4570        if (index >= tl_count) {
4571                status = -ENOSPC;
4572                mlog_errno(status);
4573                goto bail;
4574        }
4575
4576        status = ocfs2_journal_access(handle, tl_inode, tl_bh,
4577                                      OCFS2_JOURNAL_ACCESS_WRITE);
4578        if (status < 0) {
4579                mlog_errno(status);
4580                goto bail;
4581        }
4582
4583        mlog(0, "Log truncate of %u clusters starting at cluster %u to "
4584             "%llu (index = %d)\n", num_clusters, start_cluster,
4585             (unsigned long long)OCFS2_I(tl_inode)->ip_blkno, index);
4586
4587        if (ocfs2_truncate_log_can_coalesce(tl, start_cluster)) {
4588                /*
4589                 * Move index back to the record we are coalescing with.
4590                 * ocfs2_truncate_log_can_coalesce() guarantees nonzero
4591                 */
4592                index--;
4593
4594                num_clusters += le32_to_cpu(tl->tl_recs[index].t_clusters);
4595                mlog(0, "Coalesce with index %u (start = %u, clusters = %u)\n",
4596                     index, le32_to_cpu(tl->tl_recs[index].t_start),
4597                     num_clusters);
4598        } else {
4599                tl->tl_recs[index].t_start = cpu_to_le32(start_cluster);
4600                tl->tl_used = cpu_to_le16(index + 1);
4601        }
4602        tl->tl_recs[index].t_clusters = cpu_to_le32(num_clusters);
4603
4604        status = ocfs2_journal_dirty(handle, tl_bh);
4605        if (status < 0) {
4606                mlog_errno(status);
4607                goto bail;
4608        }
4609
4610bail:
4611        mlog_exit(status);
4612        return status;
4613}
4614
4615static int ocfs2_replay_truncate_records(struct ocfs2_super *osb,
4616                                         handle_t *handle,
4617                                         struct inode *data_alloc_inode,
4618                                         struct buffer_head *data_alloc_bh)
4619{
4620        int status = 0;
4621        int i;
4622        unsigned int num_clusters;
4623        u64 start_blk;
4624        struct ocfs2_truncate_rec rec;
4625        struct ocfs2_dinode *di;
4626        struct ocfs2_truncate_log *tl;
4627        struct inode *tl_inode = osb->osb_tl_inode;
4628        struct buffer_head *tl_bh = osb->osb_tl_bh;
4629
4630        mlog_entry_void();
4631
4632        di = (struct ocfs2_dinode *) tl_bh->b_data;
4633        tl = &di->id2.i_dealloc;
4634        i = le16_to_cpu(tl->tl_used) - 1;
4635        while (i >= 0) {
4636                /* Caller has given us at least enough credits to
4637                 * update the truncate log dinode */
4638                status = ocfs2_journal_access(handle, tl_inode, tl_bh,
4639                                              OCFS2_JOURNAL_ACCESS_WRITE);
4640                if (status < 0) {
4641                        mlog_errno(status);
4642                        goto bail;
4643                }
4644
4645                tl->tl_used = cpu_to_le16(i);
4646
4647                status = ocfs2_journal_dirty(handle, tl_bh);
4648                if (status < 0) {
4649                        mlog_errno(status);
4650                        goto bail;
4651                }
4652
4653                /* TODO: Perhaps we can calculate the bulk of the
4654                 * credits up front rather than extending like
4655                 * this. */
4656                status = ocfs2_extend_trans(handle,
4657                                            OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC);
4658                if (status < 0) {
4659                        mlog_errno(status);
4660                        goto bail;
4661                }
4662
4663                rec = tl->tl_recs[i];
4664                start_blk = ocfs2_clusters_to_blocks(data_alloc_inode->i_sb,
4665                                                    le32_to_cpu(rec.t_start));
4666                num_clusters = le32_to_cpu(rec.t_clusters);
4667
4668                /* if start_blk is not set, we ignore the record as
4669                 * invalid. */
4670                if (start_blk) {
4671                        mlog(0, "free record %d, start = %u, clusters = %u\n",
4672                             i, le32_to_cpu(rec.t_start), num_clusters);
4673
4674                        status = ocfs2_free_clusters(handle, data_alloc_inode,
4675                                                     data_alloc_bh, start_blk,
4676                                                     num_clusters);
4677                        if (status < 0) {
4678                                mlog_errno(status);
4679                                goto bail;
4680                        }
4681                }
4682                i--;
4683        }
4684
4685bail:
4686        mlog_exit(status);
4687        return status;
4688}
4689
4690/* Expects you to already be holding tl_inode->i_mutex */
4691int __ocfs2_flush_truncate_log(struct ocfs2_super *osb)
4692{
4693        int status;
4694        unsigned int num_to_flush;
4695        handle_t *handle;
4696        struct inode *tl_inode = osb->osb_tl_inode;
4697        struct inode *data_alloc_inode = NULL;
4698        struct buffer_head *tl_bh = osb->osb_tl_bh;
4699        struct buffer_head *data_alloc_bh = NULL;
4700        struct ocfs2_dinode *di;
4701        struct ocfs2_truncate_log *tl;
4702
4703        mlog_entry_void();
4704
4705        BUG_ON(mutex_trylock(&tl_inode->i_mutex));
4706
4707        di = (struct ocfs2_dinode *) tl_bh->b_data;
4708        tl = &di->id2.i_dealloc;
4709        if (!OCFS2_IS_VALID_DINODE(di)) {
4710                OCFS2_RO_ON_INVALID_DINODE(osb->sb, di);
4711                status = -EIO;
4712                goto out;
4713        }
4714
4715        num_to_flush = le16_to_cpu(tl->tl_used);
4716        mlog(0, "Flush %u records from truncate log #%llu\n",
4717             num_to_flush, (unsigned long long)OCFS2_I(tl_inode)->ip_blkno);
4718        if (!num_to_flush) {
4719                status = 0;
4720                goto out;
4721        }
4722
4723        data_alloc_inode = ocfs2_get_system_file_inode(osb,
4724                                                       GLOBAL_BITMAP_SYSTEM_INODE,
4725                                                       OCFS2_INVALID_SLOT);
4726        if (!data_alloc_inode) {
4727                status = -EINVAL;
4728                mlog(ML_ERROR, "Could not get bitmap inode!\n");
4729                goto out;
4730        }
4731
4732        mutex_lock(&data_alloc_inode->i_mutex);
4733
4734        status = ocfs2_meta_lock(data_alloc_inode, &data_alloc_bh, 1);
4735        if (status < 0) {
4736                mlog_errno(status);
4737                goto out_mutex;
4738        }
4739
4740        handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE);
4741        if (IS_ERR(handle)) {
4742                status = PTR_ERR(handle);
4743                mlog_errno(status);
4744                goto out_unlock;
4745        }
4746
4747        status = ocfs2_replay_truncate_records(osb, handle, data_alloc_inode,
4748                                               data_alloc_bh);
4749        if (status < 0)
4750                mlog_errno(status);
4751
4752        ocfs2_commit_trans(osb, handle);
4753
4754out_unlock:
4755        brelse(data_alloc_bh);
4756        ocfs2_meta_unlock(data_alloc_inode, 1);
4757
4758out_mutex:
4759        mutex_unlock(&data_alloc_inode->i_mutex);
4760        iput(data_alloc_inode);
4761
4762out:
4763        mlog_exit(status);
4764        return status;
4765}
4766
4767int ocfs2_flush_truncate_log(struct ocfs2_super *osb)
4768{
4769        int status;
4770        struct inode *tl_inode = osb->osb_tl_inode;
4771
4772        mutex_lock(&tl_inode->i_mutex);
4773        status = __ocfs2_flush_truncate_log(osb);
4774        mutex_unlock(&tl_inode->i_mutex);
4775
4776        return status;
4777}
4778
4779static void ocfs2_truncate_log_worker(struct work_struct *work)
4780{
4781        int status;
4782        struct ocfs2_super *osb =
4783                container_of(work, struct ocfs2_super,
4784                             osb_truncate_log_wq.work);
4785
4786        mlog_entry_void();
4787
4788        status = ocfs2_flush_truncate_log(osb);
4789        if (status < 0)
4790                mlog_errno(status);
4791
4792        mlog_exit(status);
4793}
4794
4795#define OCFS2_TRUNCATE_LOG_FLUSH_INTERVAL (2 * HZ)
4796void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
4797                                       int cancel)
4798{
4799        if (osb->osb_tl_inode) {
4800                /* We want to push off log flushes while truncates are
4801                 * still running. */
4802                if (cancel)
4803                        cancel_delayed_work(&osb->osb_truncate_log_wq);
4804
4805                queue_delayed_work(ocfs2_wq, &osb->osb_truncate_log_wq,
4806                                   OCFS2_TRUNCATE_LOG_FLUSH_INTERVAL);
4807        }
4808}
4809
4810static int ocfs2_get_truncate_log_info(struct ocfs2_super *osb,
4811                                       int slot_num,
4812                                       struct inode **tl_inode,
4813                                       struct buffer_head **tl_bh)
4814{
4815        int status;
4816        struct inode *inode = NULL;
4817        struct buffer_head *bh = NULL;
4818
4819        inode = ocfs2_get_system_file_inode(osb,
4820                                           TRUNCATE_LOG_SYSTEM_INODE,
4821                                           slot_num);
4822        if (!inode) {
4823                status = -EINVAL;
4824                mlog(ML_ERROR, "Could not get load truncate log inode!\n");
4825                goto bail;
4826        }
4827
4828        status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno, &bh,
4829                                  OCFS2_BH_CACHED, inode);
4830        if (status < 0) {
4831                iput(inode);
4832                mlog_errno(status);
4833                goto bail;
4834        }
4835
4836        *tl_inode = inode;
4837        *tl_bh    = bh;
4838bail:
4839        mlog_exit(status);
4840        return status;
4841}
4842
4843/* called during the 1st stage of node recovery. we stamp a clean
4844 * truncate log and pass back a copy for processing later. if the
4845 * truncate log does not require processing, a *tl_copy is set to
4846 * NULL. */
4847int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
4848                                      int slot_num,
4849                                      struct ocfs2_dinode **tl_copy)
4850{
4851        int status;
4852        struct inode *tl_inode = NULL;
4853        struct buffer_head *tl_bh = NULL;
4854        struct ocfs2_dinode *di;
4855        struct ocfs2_truncate_log *tl;
4856
4857        *tl_copy = NULL;
4858
4859        mlog(0, "recover truncate log from slot %d\n", slot_num);
4860
4861        status = ocfs2_get_truncate_log_info(osb, slot_num, &tl_inode, &tl_bh);
4862        if (status < 0) {
4863                mlog_errno(status);
4864                goto bail;
4865        }
4866
4867        di = (struct ocfs2_dinode *) tl_bh->b_data;
4868        tl = &di->id2.i_dealloc;
4869        if (!OCFS2_IS_VALID_DINODE(di)) {
4870                OCFS2_RO_ON_INVALID_DINODE(tl_inode->i_sb, di);
4871                status = -EIO;
4872                goto bail;
4873        }
4874
4875        if (le16_to_cpu(tl->tl_used)) {
4876                mlog(0, "We'll have %u logs to recover\n",
4877                     le16_to_cpu(tl->tl_used));
4878
4879                *tl_copy = kmalloc(tl_bh->b_size, GFP_KERNEL);
4880                if (!(*tl_copy)) {
4881                        status = -ENOMEM;
4882                        mlog_errno(status);
4883                        goto bail;
4884                }
4885
4886                /* Assuming the write-out below goes well, this copy
4887                 * will be passed back to recovery for processing. */
4888                memcpy(*tl_copy, tl_bh->b_data, tl_bh->b_size);
4889
4890                /* All we need to do to clear the truncate log is set
4891                 * tl_used. */
4892                tl->tl_used = 0;
4893
4894                status = ocfs2_write_block(osb, tl_bh, tl_inode);
4895                if (status < 0) {
4896                        mlog_errno(status);
4897                        goto bail;
4898                }
4899        }
4900
4901bail:
4902        if (tl_inode)
4903                iput(tl_inode);
4904        if (tl_bh)
4905                brelse(tl_bh);
4906
4907        if (status < 0 && (*tl_copy)) {
4908                kfree(*tl_copy);
4909                *tl_copy = NULL;
4910        }
4911
4912        mlog_exit(status);
4913        return status;
4914}
4915
4916int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
4917                                         struct ocfs2_dinode *tl_copy)
4918{
4919        int status = 0;
4920        int i;
4921        unsigned int clusters, num_recs, start_cluster;
4922        u64 start_blk;
4923        handle_t *handle;
4924        struct inode *tl_inode = osb->osb_tl_inode;
4925        struct ocfs2_truncate_log *tl;
4926
4927        mlog_entry_void();
4928
4929        if (OCFS2_I(tl_inode)->ip_blkno == le64_to_cpu(tl_copy->i_blkno)) {
4930                mlog(ML_ERROR, "Asked to recover my own truncate log!\n");
4931                return -EINVAL;
4932        }
4933
4934        tl = &tl_copy->id2.i_dealloc;
4935        num_recs = le16_to_cpu(tl->tl_used);
4936        mlog(0, "cleanup %u records from %llu\n", num_recs,
4937             (unsigned long long)le64_to_cpu(tl_copy->i_blkno));
4938
4939        mutex_lock(&tl_inode->i_mutex);
4940        for(i = 0; i < num_recs; i++) {
4941                if (ocfs2_truncate_log_needs_flush(osb)) {
4942                        status = __ocfs2_flush_truncate_log(osb);
4943                        if (status < 0) {
4944                                mlog_errno(status);
4945                                goto bail_up;
4946                        }
4947                }
4948
4949                handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE);
4950                if (IS_ERR(handle)) {
4951                        status = PTR_ERR(handle);
4952                        mlog_errno(status);
4953                        goto bail_up;
4954                }
4955
4956                clusters = le32_to_cpu(tl->tl_recs[i].t_clusters);
4957                start_cluster = le32_to_cpu(tl->tl_recs[i].t_start);
4958                start_blk = ocfs2_clusters_to_blocks(osb->sb, start_cluster);
4959
4960                status = ocfs2_truncate_log_append(osb, handle,
4961                                                   start_blk, clusters);
4962                ocfs2_commit_trans(osb, handle);
4963                if (status < 0) {
4964                        mlog_errno(status);
4965                        goto bail_up;
4966                }
4967        }
4968
4969bail_up:
4970        mutex_unlock(&tl_inode->i_mutex);
4971
4972        mlog_exit(status);
4973        return status;
4974}
4975
4976void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb)
4977{
4978        int status;
4979        struct inode *tl_inode = osb->osb_tl_inode;
4980
4981        mlog_entry_void();
4982
4983        if (tl_inode) {
4984                cancel_delayed_work(&osb->osb_truncate_log_wq);
4985                flush_workqueue(ocfs2_wq);
4986
4987                status = ocfs2_flush_truncate_log(osb);
4988                if (status < 0)
4989                        mlog_errno(status);
4990
4991                brelse(osb->osb_tl_bh);
4992                iput(osb->osb_tl_inode);
4993        }
4994
4995        mlog_exit_void();
4996}
4997
4998int ocfs2_truncate_log_init(struct ocfs2_super *osb)
4999{
5000        int status;
5001        struct inode *tl_inode = NULL;
5002        struct buffer_head *tl_bh = NULL;
5003
5004        mlog_entry_void();
5005
5006        status = ocfs2_get_truncate_log_info(osb,
5007                                             osb->slot_num,
5008                                             &tl_inode,
5009                                             &tl_bh);
5010        if (status < 0)
5011                mlog_errno(status);
5012
5013        /* ocfs2_truncate_log_shutdown keys on the existence of
5014         * osb->osb_tl_inode so we don't set any of the osb variables
5015         * until we're sure all is well. */
5016        INIT_DELAYED_WORK(&osb->osb_truncate_log_wq,
5017                          ocfs2_truncate_log_worker);
5018        osb->osb_tl_bh    = tl_bh;
5019        osb->osb_tl_inode = tl_inode;
5020
5021        mlog_exit(status);
5022        return status;
5023}
5024
5025/*
5026 * Delayed de-allocation of suballocator blocks.
5027 *
5028 * Some sets of block de-allocations might involve multiple suballocator inodes.
5029 *
5030 * The locking for this can get extremely complicated, especially when
5031 * the suballocator inodes to delete from aren't known until deep
5032 * within an unrelated codepath.
5033 *
5034 * ocfs2_extent_block structures are a good example of this - an inode
5035 * btree could have been grown by any number of nodes each allocating
5036 * out of their own suballoc inode.
5037 *
5038 * These structures allow the delay of block de-allocation until a
5039 * later time, when locking of multiple cluster inodes won't cause
5040 * deadlock.
5041 */
5042
5043/*
5044 * Describes a single block free from a suballocator
5045 */
5046struct ocfs2_cached_block_free {
5047        struct ocfs2_cached_block_free          *free_next;
5048        u64                                     free_blk;
5049        unsigned int                            free_bit;
5050};
5051
5052struct ocfs2_per_slot_free_list {
5053        struct ocfs2_per_slot_free_list         *f_next_suballocator;
5054        int                                     f_inode_type;
5055        int                                     f_slot;
5056        struct ocfs2_cached_block_free          *f_first;
5057};
5058
5059static int ocfs2_free_cached_items(struct ocfs2_super *osb,
5060                                   int sysfile_type,
5061                                   int slot,
5062                                   struct ocfs2_cached_block_free *head)
5063{
5064        int ret;
5065        u64 bg_blkno;
5066        handle_t *handle;
5067        struct inode *inode;
5068        struct buffer_head *di_bh = NULL;
5069        struct ocfs2_cached_block_free *tmp;
5070
5071        inode = ocfs2_get_system_file_inode(osb, sysfile_type, slot);
5072        if (!inode) {
5073                ret = -EINVAL;
5074                mlog_errno(ret);
5075                goto out;
5076        }
5077
5078        mutex_lock(&inode->i_mutex);
5079
5080        ret = ocfs2_meta_lock(inode, &di_bh, 1);
5081        if (ret) {
5082                mlog_errno(ret);
5083                goto out_mutex;
5084        }
5085
5086        handle = ocfs2_start_trans(osb, OCFS2_SUBALLOC_FREE);
5087        if (IS_ERR(handle)) {
5088                ret = PTR_ERR(handle);
5089                mlog_errno(ret);
5090                goto out_unlock;
5091        }
5092
5093        while (head) {
5094                bg_blkno = ocfs2_which_suballoc_group(head->free_blk,
5095                                                      head->free_bit);
5096                mlog(0, "Free bit: (bit %u, blkno %llu)\n",
5097                     head->free_bit, (unsigned long long)head->free_blk);
5098
5099                ret = ocfs2_free_suballoc_bits(handle, inode, di_bh,
5100                                               head->free_bit, bg_blkno, 1);
5101                if (ret) {
5102                        mlog_errno(ret);
5103                        goto out_journal;
5104                }
5105
5106                ret = ocfs2_extend_trans(handle, OCFS2_SUBALLOC_FREE);
5107                if (ret) {
5108                        mlog_errno(ret);
5109                        goto out_journal;
5110                }
5111
5112                tmp = head;
5113                head = head->free_next;
5114                kfree(tmp);
5115        }
5116
5117out_journal:
5118        ocfs2_commit_trans(osb, handle);
5119
5120out_unlock:
5121        ocfs2_meta_unlock(inode, 1);
5122        brelse(di_bh);
5123out_mutex:
5124        mutex_unlock(&inode->i_mutex);
5125        iput(inode);
5126out:
5127        while(head) {
5128                /* Premature exit may have left some dangling items. */
5129                tmp = head;
5130                head = head->free_next;
5131                kfree(tmp);
5132        }
5133
5134        return ret;
5135}
5136
5137int ocfs2_run_deallocs(struct ocfs2_super *osb,
5138                       struct ocfs2_cached_dealloc_ctxt *ctxt)
5139{
5140        int ret = 0, ret2;
5141        struct ocfs2_per_slot_free_list *fl;
5142
5143        if (!ctxt)
5144                return 0;
5145
5146        while (ctxt->c_first_suballocator) {
5147                fl = ctxt->c_first_suballocator;
5148
5149                if (fl->f_first) {
5150                        mlog(0, "Free items: (type %u, slot %d)\n",
5151                             fl->f_inode_type, fl->f_slot);
5152                        ret2 = ocfs2_free_cached_items(osb, fl->f_inode_type,
5153                                                       fl->f_slot, fl->f_first);
5154                        if (ret2)
5155                                mlog_errno(ret2);
5156                        if (!ret)
5157                                ret = ret2;
5158                }
5159
5160                ctxt->c_first_suballocator = fl->f_next_suballocator;
5161                kfree(fl);
5162        }
5163
5164        return ret;
5165}
5166
5167static struct ocfs2_per_slot_free_list *
5168ocfs2_find_per_slot_free_list(int type,
5169                              int slot,
5170                              struct ocfs2_cached_dealloc_ctxt *ctxt)
5171{
5172        struct ocfs2_per_slot_free_list *fl = ctxt->c_first_suballocator;
5173
5174        while (fl) {
5175                if (fl->f_inode_type == type && fl->f_slot == slot)
5176                        return fl;
5177
5178                fl = fl->f_next_suballocator;
5179        }
5180
5181        fl = kmalloc(sizeof(*fl), GFP_NOFS);
5182        if (fl) {
5183                fl->f_inode_type = type;
5184                fl->f_slot = slot;
5185                fl->f_first = NULL;
5186                fl->f_next_suballocator = ctxt->c_first_suballocator;
5187
5188                ctxt->c_first_suballocator = fl;
5189        }
5190        return fl;
5191}
5192
5193static int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
5194                                     int type, int slot, u64 blkno,
5195                                     unsigned int bit)
5196{
5197        int ret;
5198        struct ocfs2_per_slot_free_list *fl;
5199        struct ocfs2_cached_block_free *item;
5200
5201        fl = ocfs2_find_per_slot_free_list(type, slot, ctxt);
5202        if (fl == NULL) {
5203                ret = -ENOMEM;
5204                mlog_errno(ret);
5205                goto out;
5206        }
5207
5208        item = kmalloc(sizeof(*item), GFP_NOFS);
5209        if (item == NULL) {
5210                ret = -ENOMEM;
5211                mlog_errno(ret);
5212                goto out;
5213        }
5214
5215        mlog(0, "Insert: (type %d, slot %u, bit %u, blk %llu)\n",
5216             type, slot, bit, (unsigned long long)blkno);
5217
5218        item->free_blk = blkno;
5219        item->free_bit = bit;
5220        item->free_next = fl->f_first;
5221
5222        fl->f_first = item;
5223
5224        ret = 0;
5225out:
5226        return ret;
5227}
5228
5229static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt,
5230                                         struct ocfs2_extent_block *eb)
5231{
5232        return ocfs2_cache_block_dealloc(ctxt, EXTENT_ALLOC_SYSTEM_INODE,
5233                                         le16_to_cpu(eb->h_suballoc_slot),
5234                                         le64_to_cpu(eb->h_blkno),
5235                                         le16_to_cpu(eb->h_suballoc_bit));
5236}
5237
5238/* This function will figure out whether the currently last extent
5239 * block will be deleted, and if it will, what the new last extent
5240 * block will be so we can update his h_next_leaf_blk field, as well
5241 * as the dinodes i_last_eb_blk */
5242static int ocfs2_find_new_last_ext_blk(struct inode *inode,
5243                                       unsigned int clusters_to_del,
5244                                       struct ocfs2_path *path,
5245                                       struct buffer_head **new_last_eb)
5246{
5247        int next_free, ret = 0;
5248        u32 cpos;
5249        struct ocfs2_extent_rec *rec;
5250        struct ocfs2_extent_block *eb;
5251        struct ocfs2_extent_list *el;
5252        struct buffer_head *bh = NULL;
5253
5254        *new_last_eb = NULL;
5255
5256        /* we have no tree, so of course, no last_eb. */
5257        if (!path->p_tree_depth)
5258                goto out;
5259
5260        /* trunc to zero special case - this makes tree_depth = 0
5261         * regardless of what it is.  */
5262        if (OCFS2_I(inode)->ip_clusters == clusters_to_del)
5263                goto out;
5264
5265        el = path_leaf_el(path);
5266        BUG_ON(!el->l_next_free_rec);
5267
5268        /*
5269         * Make sure that this extent list will actually be empty
5270         * after we clear away the data. We can shortcut out if
5271         * there's more than one non-empty extent in the
5272         * list. Otherwise, a check of the remaining extent is
5273         * necessary.
5274         */
5275        next_free = le16_to_cpu(el->l_next_free_rec);
5276        rec = NULL;
5277        if (ocfs2_is_empty_extent(&el->l_recs[0])) {
5278                if (next_free > 2)
5279                        goto out;
5280
5281                /* We may have a valid extent in index 1, check it. */
5282                if (next_free == 2)
5283                        rec = &el->l_recs[1];
5284
5285                /*
5286                 * Fall through - no more nonempty extents, so we want
5287                 * to delete this leaf.
5288                 */
5289        } else {
5290                if (next_free > 1)
5291                        goto out;
5292
5293                rec = &el->l_recs[0];
5294        }
5295
5296        if (rec) {
5297                /*
5298                 * Check it we'll only be trimming off the end of this
5299                 * cluster.
5300                 */
5301                if (le16_to_cpu(rec->e_leaf_clusters) > clusters_to_del)
5302                        goto out;
5303        }
5304
5305        ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path, &cpos);
5306        if (ret) {
5307                mlog_errno(ret);
5308                goto out;
5309        }
5310
5311        ret = ocfs2_find_leaf(inode, path_root_el(path), cpos, &bh);
5312        if (ret) {
5313                mlog_errno(ret);
5314                goto out;
5315        }
5316
5317        eb = (struct ocfs2_extent_block *) bh->b_data;
5318        el = &eb->h_list;
5319        if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
5320                OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
5321                ret = -EROFS;
5322                goto out;
5323        }
5324
5325        *new_last_eb = bh;
5326        get_bh(*new_last_eb);
5327        mlog(0, "returning block %llu, (cpos: %u)\n",
5328             (unsigned long long)le64_to_cpu(eb->h_blkno), cpos);
5329out:
5330        brelse(bh);
5331
5332        return ret;
5333}
5334
5335/*
5336 * Trim some clusters off the rightmost edge of a tree. Only called
5337 * during truncate.
5338 *
5339 * The caller needs to:
5340 *   - start journaling of each path component.
5341 *   - compute and fully set up any new last ext block
5342 */
5343static int ocfs2_trim_tree(struct inode *inode, struct ocfs2_path *path,
5344                           handle_t *handle, struct ocfs2_truncate_context *tc,
5345                           u32 clusters_to_del, u64 *delete_start)
5346{
5347        int ret, i, index = path->p_tree_depth;
5348        u32 new_edge = 0;
5349        u64 deleted_eb = 0;
5350        struct buffer_head *bh;
5351        struct ocfs2_extent_list *el;
5352        struct ocfs2_extent_rec *rec;
5353
5354        *delete_start = 0;
5355
5356        while (index >= 0) {
5357                bh = path->p_node[index].bh;
5358                el = path->p_node[index].el;
5359
5360                mlog(0, "traveling tree (index = %d, block = %llu)\n",
5361                     index,  (unsigned long long)bh->b_blocknr);
5362
5363                BUG_ON(le16_to_cpu(el->l_next_free_rec) == 0);
5364
5365                if (index !=
5366                    (path->p_tree_depth - le16_to_cpu(el->l_tree_depth))) {
5367                        ocfs2_error(inode->i_sb,
5368                                    "Inode %lu has invalid ext. block %llu",
5369                                    inode->i_ino,
5370                                    (unsigned long long)bh->b_blocknr);
5371                        ret = -EROFS;
5372                        goto out;
5373                }
5374
5375find_tail_record:
5376                i = le16_to_cpu(el->l_next_free_rec) - 1;
5377                rec = &el->l_recs[i];
5378
5379                mlog(0, "Extent list before: record %d: (%u, %u, %llu), "
5380                     "next = %u\n", i, le32_to_cpu(rec->e_cpos),
5381                     ocfs2_rec_clusters(el, rec),
5382                     (unsigned long long)le64_to_cpu(rec->e_blkno),
5383                     le16_to_cpu(el->l_next_free_rec));
5384
5385                BUG_ON(ocfs2_rec_clusters(el, rec) < clusters_to_del);
5386
5387                if (le16_to_cpu(el->l_tree_depth) == 0) {
5388                        /*
5389                         * If the leaf block contains a single empty
5390                         * extent and no records, we can just remove
5391                         * the block.
5392                         */
5393                        if (i == 0 && ocfs2_is_empty_extent(rec)) {
5394                                memset(rec, 0,
5395                                       sizeof(struct ocfs2_extent_rec));
5396                                el->l_next_free_rec = cpu_to_le16(0);
5397
5398                                goto delete;
5399                        }
5400
5401                        /*
5402                         * Remove any empty extents by shifting things
5403                         * left. That should make life much easier on
5404                         * the code below. This condition is rare
5405                         * enough that we shouldn't see a performance
5406                         * hit.
5407                         */
5408                        if (ocfs2_is_empty_extent(&el->l_recs[0])) {
5409                                le16_add_cpu(&el->l_next_free_rec, -1);
5410
5411                                for(i = 0;
5412                                    i < le16_to_cpu(el->l_next_free_rec); i++)
5413                                        el->l_recs[i] = el->l_recs[i + 1];
5414
5415                                memset(&el->l_recs[i], 0,
5416                                       sizeof(struct ocfs2_extent_rec));
5417
5418                                /*
5419                                 * We've modified our extent list. The
5420                                 * simplest way to handle this change
5421                                 * is to being the search from the
5422                                 * start again.
5423                                 */
5424                                goto find_tail_record;
5425                        }
5426
5427                        le16_add_cpu(&rec->e_leaf_clusters, -clusters_to_del);
5428
5429                        /*
5430                         * We'll use "new_edge" on our way back up the
5431                         * tree to know what our rightmost cpos is.
5432                         */
5433                        new_edge = le16_to_cpu(rec->e_leaf_clusters);
5434                        new_edge += le32_to_cpu(rec->e_cpos);
5435
5436                        /*
5437                         * The caller will use this to delete data blocks.
5438                         */
5439                        *delete_start = le64_to_cpu(rec->e_blkno)
5440                                + ocfs2_clusters_to_blocks(inode->i_sb,
5441                                        le16_to_cpu(rec->e_leaf_clusters));
5442
5443                        /*
5444                         * If it's now empty, remove this record.
5445                         */
5446                        if (le16_to_cpu(rec->e_leaf_clusters) == 0) {
5447                                memset(rec, 0,
5448                                       sizeof(struct ocfs2_extent_rec));
5449                                le16_add_cpu(&el->l_next_free_rec, -1);
5450                        }
5451                } else {
5452                        if (le64_to_cpu(rec->e_blkno) == deleted_eb) {
5453                                memset(rec, 0,
5454                                       sizeof(struct ocfs2_extent_rec));
5455                                le16_add_cpu(&el->l_next_free_rec, -1);
5456
5457                                goto delete;
5458                        }
5459
5460                        /* Can this actually happen? */
5461                        if (le16_to_cpu(el->l_next_free_rec) == 0)
5462                                goto delete;
5463
5464                        /*
5465                         * We never actually deleted any clusters
5466                         * because our leaf was empty. There's no
5467                         * reason to adjust the rightmost edge then.
5468                         */
5469                        if (new_edge == 0)
5470                                goto delete;
5471
5472                        rec->e_int_clusters = cpu_to_le32(new_edge);
5473                        le32_add_cpu(&rec->e_int_clusters,
5474                                     -le32_to_cpu(rec->e_cpos));
5475
5476                         /*
5477                          * A deleted child record should have been
5478                          * caught above.
5479                          */
5480                         BUG_ON(le32_to_cpu(rec->e_int_clusters) == 0);
5481                }
5482
5483delete:
5484                ret = ocfs2_journal_dirty(handle, bh);
5485                if (ret) {
5486                        mlog_errno(ret);
5487                        goto out;
5488                }
5489
5490                mlog(0, "extent list container %llu, after: record %d: "
5491                     "(%u, %u, %llu), next = %u.\n",
5492                     (unsigned long long)bh->b_blocknr, i,
5493                     le32_to_cpu(rec->e_cpos), ocfs2_rec_clusters(el, rec),
5494                     (unsigned long long)le64_to_cpu(rec->e_blkno),
5495                     le16_to_cpu(el->l_next_free_rec));
5496
5497                /*
5498                 * We must be careful to only attempt delete of an
5499                 * extent block (and not the root inode block).
5500                 */
5501                if (index > 0 && le16_to_cpu(el->l_next_free_rec) == 0) {
5502                        struct ocfs2_extent_block *eb =
5503                                (struct ocfs2_extent_block *)bh->b_data;
5504
5505                        /*
5506                         * Save this for use when processing the
5507                         * parent block.
5508                         */
5509                        deleted_eb = le64_to_cpu(eb->h_blkno);
5510
5511                        mlog(0, "deleting this extent block.\n");
5512
5513                        ocfs2_remove_from_cache(inode, bh);
5514
5515                        BUG_ON(ocfs2_rec_clusters(el, &el->l_recs[0]));
5516                        BUG_ON(le32_to_cpu(el->l_recs[0].e_cpos));
5517                        BUG_ON(le64_to_cpu(el->l_recs[0].e_blkno));
5518
5519                        ret = ocfs2_cache_extent_block_free(&tc->tc_dealloc, eb);
5520                        /* An error here is not fatal. */
5521                        if (ret < 0)
5522                                mlog_errno(ret);
5523                } else {
5524                        deleted_eb = 0;
5525                }
5526
5527                index--;
5528        }
5529
5530        ret = 0;
5531out:
5532        return ret;
5533}
5534
5535static int ocfs2_do_truncate(struct ocfs2_super *osb,
5536                             unsigned int clusters_to_del,
5537                             struct inode *inode,
5538                             struct buffer_head *fe_bh,
5539                             handle_t *handle,
5540                             struct ocfs2_truncate_context *tc,
5541                             struct ocfs2_path *path)
5542{
5543        int status;
5544        struct ocfs2_dinode *fe;
5545        struct ocfs2_extent_block *last_eb = NULL;
5546        struct ocfs2_extent_list *el;
5547        struct buffer_head *last_eb_bh = NULL;
5548        u64 delete_blk = 0;
5549
5550        fe = (struct ocfs2_dinode *) fe_bh->b_data;
5551
5552        status = ocfs2_find_new_last_ext_blk(inode, clusters_to_del,
5553                                             path, &last_eb_bh);
5554        if (status < 0) {
5555                mlog_errno(status);
5556                goto bail;
5557        }
5558
5559        /*
5560         * Each component will be touched, so we might as well journal
5561         * here to avoid having to handle errors later.
5562         */
5563        status = ocfs2_journal_access_path(inode, handle, path);
5564        if (status < 0) {
5565                mlog_errno(status);
5566                goto bail;
5567        }
5568
5569        if (last_eb_bh) {
5570                status = ocfs2_journal_access(handle, inode, last_eb_bh,
5571                                              OCFS2_JOURNAL_ACCESS_WRITE);
5572                if (status < 0) {
5573                        mlog_errno(status);
5574                        goto bail;
5575                }
5576
5577                last_eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
5578        }
5579
5580        el = &(fe->id2.i_list);
5581
5582        /*
5583         * Lower levels depend on this never happening, but it's best
5584         * to check it up here before changing the tree.
5585         */
5586        if (el->l_tree_depth && el->l_recs[0].e_int_clusters == 0) {
5587                ocfs2_error(inode->i_sb,
5588                            "Inode %lu has an empty extent record, depth %u\n",
5589                            inode->i_ino, le16_to_cpu(el->l_tree_depth));
5590                status = -EROFS;
5591                goto bail;
5592        }
5593
5594        spin_lock(&OCFS2_I(inode)->ip_lock);
5595        OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters) -
5596                                      clusters_to_del;
5597        spin_unlock(&OCFS2_I(inode)->ip_lock);
5598        le32_add_cpu(&fe->i_clusters, -clusters_to_del);
5599        inode->i_blocks = ocfs2_inode_sector_count(inode);
5600
5601        status = ocfs2_trim_tree(inode, path, handle, tc,
5602                                 clusters_to_del, &delete_blk);
5603        if (status) {
5604                mlog_errno(status);
5605                goto bail;
5606        }
5607
5608        if (le32_to_cpu(fe->i_clusters) == 0) {
5609                /* trunc to zero is a special case. */
5610                el->l_tree_depth = 0;
5611                fe->i_last_eb_blk = 0;
5612        } else if (last_eb)
5613                fe->i_last_eb_blk = last_eb->h_blkno;
5614
5615        status = ocfs2_journal_dirty(handle, fe_bh);
5616        if (status < 0) {
5617                mlog_errno(status);
5618                goto bail;
5619        }
5620
5621        if (last_eb) {
5622                /* If there will be a new last extent block, then by
5623                 * definition, there cannot be any leaves to the right of
5624                 * him. */
5625                last_eb->h_next_leaf_blk = 0;
5626                status = ocfs2_journal_dirty(handle, last_eb_bh);
5627                if (status < 0) {
5628                        mlog_errno(status);
5629                        goto bail;
5630                }
5631        }
5632
5633        if (delete_blk) {
5634                status = ocfs2_truncate_log_append(osb, handle, delete_blk,
5635                                                   clusters_to_del);
5636                if (status < 0) {
5637                        mlog_errno(status);
5638                        goto bail;
5639                }
5640        }
5641        status = 0;
5642bail:
5643
5644        mlog_exit(status);
5645        return status;
5646}
5647
5648static int ocfs2_writeback_zero_func(handle_t *handle, struct buffer_head *bh)
5649{
5650        set_buffer_uptodate(bh);
5651        mark_buffer_dirty(bh);
5652        return 0;
5653}
5654
5655static int ocfs2_ordered_zero_func(handle_t *handle, struct buffer_head *bh)
5656{
5657        set_buffer_uptodate(bh);
5658        mark_buffer_dirty(bh);
5659        return ocfs2_journal_dirty_data(handle, bh);
5660}
5661
5662static void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle,
5663                                     unsigned int from, unsigned int to,
5664                                     struct page *page, int zero, u64 *phys)
5665{
5666        int ret, partial = 0;
5667
5668        ret = ocfs2_map_page_blocks(page, phys, inode, from, to, 0);
5669        if (ret)
5670                mlog_errno(ret);
5671
5672        if (zero)
5673                zero_user_page(page, from, to - from, KM_USER0);
5674
5675        /*
5676         * Need to set the buffers we zero'd into uptodate
5677         * here if they aren't - ocfs2_map_page_blocks()
5678         * might've skipped some
5679         */
5680        if (ocfs2_should_order_data(inode)) {
5681                ret = walk_page_buffers(handle,
5682                                        page_buffers(page),
5683                                        from, to, &partial,
5684                                        ocfs2_ordered_zero_func);
5685                if (ret < 0)
5686                        mlog_errno(ret);
5687        } else {
5688                ret = walk_page_buffers(handle, page_buffers(page),
5689                                        from, to, &partial,
5690                                        ocfs2_writeback_zero_func);
5691                if (ret < 0)
5692                        mlog_errno(ret);
5693        }
5694
5695        if (!partial)
5696                SetPageUptodate(page);
5697
5698        flush_dcache_page(page);
5699}
5700
5701static void ocfs2_zero_cluster_pages(struct inode *inode, loff_t start,
5702                                     loff_t end, struct page **pages,
5703                                     int numpages, u64 phys, handle_t *handle)
5704{
5705        int i;
5706        struct page *page;
5707        unsigned int from, to = PAGE_CACHE_SIZE;
5708        struct super_block *sb = inode->i_sb;
5709
5710        BUG_ON(!ocfs2_sparse_alloc(OCFS2_SB(sb)));
5711
5712        if (numpages == 0)
5713                goto out;
5714
5715        to = PAGE_CACHE_SIZE;
5716        for(i = 0; i < numpages; i++) {
5717                page = pages[i];
5718
5719                from = start & (PAGE_CACHE_SIZE - 1);
5720                if ((end >> PAGE_CACHE_SHIFT) == page->index)
5721                        to = end & (PAGE_CACHE_SIZE - 1);
5722
5723                BUG_ON(from > PAGE_CACHE_SIZE);
5724                BUG_ON(to > PAGE_CACHE_SIZE);
5725
5726                ocfs2_map_and_dirty_page(inode, handle, from, to, page, 1,
5727                                         &phys);
5728
5729                start = (page->index + 1) << PAGE_CACHE_SHIFT;
5730        }
5731out:
5732        if (pages)
5733                ocfs2_unlock_and_free_pages(pages, numpages);
5734}
5735
5736static int ocfs2_grab_eof_pages(struct inode *inode, loff_t start, loff_t end,
5737                                struct page **pages, int *num)
5738{
5739        int numpages, ret = 0;
5740        struct super_block *sb = inode->i_sb;
5741        struct address_space *mapping = inode->i_mapping;
5742        unsigned long index;
5743        loff_t last_page_bytes;
5744
5745        BUG_ON(start > end);
5746
5747        BUG_ON(start >> OCFS2_SB(sb)->s_clustersize_bits !=
5748               (end - 1) >> OCFS2_SB(sb)->s_clustersize_bits);
5749
5750        numpages = 0;
5751        last_page_bytes = PAGE_ALIGN(end);
5752        index = start >> PAGE_CACHE_SHIFT;
5753        do {
5754                pages[numpages] = grab_cache_page(mapping, index);
5755                if (!pages[numpages]) {
5756                        ret = -ENOMEM;
5757                        mlog_errno(ret);
5758                        goto out;
5759                }
5760
5761                numpages++;
5762                index++;
5763        } while (index < (last_page_bytes >> PAGE_CACHE_SHIFT));
5764
5765out:
5766        if (ret != 0) {
5767                if (pages)
5768                        ocfs2_unlock_and_free_pages(pages, numpages);
5769                numpages = 0;
5770        }
5771
5772        *num = numpages;
5773
5774        return ret;
5775}
5776
5777/*
5778 * Zero the area past i_size but still within an allocated
5779 * cluster. This avoids exposing nonzero data on subsequent file
5780 * extends.
5781 *
5782 * We need to call this before i_size is updated on the inode because
5783 * otherwise block_write_full_page() will skip writeout of pages past
5784 * i_size. The new_i_size parameter is passed for this reason.
5785 */
5786int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle,
5787                                  u64 range_start, u64 range_end)
5788{
5789        int ret = 0, numpages;
5790        struct page **pages = NULL;
5791        u64 phys;
5792        unsigned int ext_flags;
5793        struct super_block *sb = inode->i_sb;
5794
5795        /*
5796         * File systems which don't support sparse files zero on every
5797         * extend.
5798         */
5799        if (!ocfs2_sparse_alloc(OCFS2_SB(sb)))
5800                return 0;
5801
5802        pages = kcalloc(ocfs2_pages_per_cluster(sb),
5803                        sizeof(struct page *), GFP_NOFS);
5804        if (pages == NULL) {
5805                ret = -ENOMEM;
5806                mlog_errno(ret);
5807                goto out;
5808        }
5809
5810        if (range_start == range_end)
5811                goto out;
5812
5813        ret = ocfs2_extent_map_get_blocks(inode,
5814                                          range_start >> sb->s_blocksize_bits,
5815                                          &phys, NULL, &ext_flags);
5816        if (ret) {
5817                mlog_errno(ret);
5818                goto out;
5819        }
5820
5821        /*
5822         * Tail is a hole, or is marked unwritten. In either case, we
5823         * can count on read and write to return/push zero's.
5824         */
5825        if (phys == 0 || ext_flags & OCFS2_EXT_UNWRITTEN)
5826                goto out;
5827
5828        ret = ocfs2_grab_eof_pages(inode, range_start, range_end, pages,
5829                                   &numpages);
5830        if (ret) {
5831                mlog_errno(ret);
5832                goto out;
5833        }
5834
5835        ocfs2_zero_cluster_pages(inode, range_start, range_end, pages,
5836                                 numpages, phys, handle);
5837
5838        /*
5839         * Initiate writeout of the pages we zero'd here. We don't
5840         * wait on them - the truncate_inode_pages() call later will
5841         * do that for us.
5842         */
5843        ret = do_sync_mapping_range(inode->i_mapping, range_start,
5844                                    range_end - 1, SYNC_FILE_RANGE_WRITE);
5845        if (ret)
5846                mlog_errno(ret);
5847
5848out:
5849        if (pages)
5850                kfree(pages);
5851
5852        return ret;
5853}
5854
5855static void ocfs2_zero_dinode_id2(struct inode *inode, struct ocfs2_dinode *di)
5856{
5857        unsigned int blocksize = 1 << inode->i_sb->s_blocksize_bits;
5858
5859        memset(&di->id2, 0, blocksize - offsetof(struct ocfs2_dinode, id2));
5860}
5861
5862void ocfs2_dinode_new_extent_list(struct inode *inode,
5863                                  struct ocfs2_dinode *di)
5864{
5865        ocfs2_zero_dinode_id2(inode, di);
5866        di->id2.i_list.l_tree_depth = 0;
5867        di->id2.i_list.l_next_free_rec = 0;
5868        di->id2.i_list.l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(inode->i_sb));
5869}
5870
5871void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di)
5872{
5873        struct ocfs2_inode_info *oi = OCFS2_I(inode);
5874        struct ocfs2_inline_data *idata = &di->id2.i_data;
5875
5876        spin_lock(&oi->ip_lock);
5877        oi->ip_dyn_features |= OCFS2_INLINE_DATA_FL;
5878        di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
5879        spin_unlock(&oi->ip_lock);
5880
5881        /*
5882         * We clear the entire i_data structure here so that all
5883         * fields can be properly initialized.
5884         */
5885        ocfs2_zero_dinode_id2(inode, di);
5886
5887        idata->id_count = cpu_to_le16(ocfs2_max_inline_data(inode->i_sb));
5888}
5889
5890int ocfs2_convert_inline_data_to_extents(struct inode *inode,
5891                                         struct buffer_head *di_bh)
5892{
5893        int ret, i, has_data, num_pages = 0;
5894        handle_t *handle;
5895        u64 uninitialized_var(block);
5896        struct ocfs2_inode_info *oi = OCFS2_I(inode);
5897        struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
5898        struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
5899        struct ocfs2_alloc_context *data_ac = NULL;
5900        struct page **pages = NULL;
5901        loff_t end = osb->s_clustersize;
5902
5903        has_data = i_size_read(inode) ? 1 : 0;
5904
5905        if (has_data) {
5906                pages = kcalloc(ocfs2_pages_per_cluster(osb->sb),
5907                                sizeof(struct page *), GFP_NOFS);
5908                if (pages == NULL) {
5909                        ret = -ENOMEM;
5910                        mlog_errno(ret);
5911                        goto out;
5912                }
5913
5914                ret = ocfs2_reserve_clusters(osb, 1, &data_ac);
5915                if (ret) {
5916                        mlog_errno(ret);
5917                        goto out;
5918                }
5919        }
5920
5921        handle = ocfs2_start_trans(osb, OCFS2_INLINE_TO_EXTENTS_CREDITS);
5922        if (IS_ERR(handle)) {
5923                ret = PTR_ERR(handle);
5924                mlog_errno(ret);
5925                goto out_unlock;
5926        }
5927
5928        ret = ocfs2_journal_access(handle, inode, di_bh,
5929                                   OCFS2_JOURNAL_ACCESS_WRITE);
5930        if (ret) {
5931                mlog_errno(ret);
5932                goto out_commit;
5933        }
5934
5935        if (has_data) {
5936                u32 bit_off, num;
5937                unsigned int page_end;
5938                u64 phys;
5939
5940                ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off,
5941                                           &num);
5942                if (ret) {
5943                        mlog_errno(ret);
5944                        goto out_commit;
5945                }
5946
5947                /*
5948                 * Save two copies, one for insert, and one that can
5949                 * be changed by ocfs2_map_and_dirty_page() below.
5950                 */
5951                block = phys = ocfs2_clusters_to_blocks(inode->i_sb, bit_off);
5952
5953                /*
5954                 * Non sparse file systems zero on extend, so no need
5955                 * to do that now.
5956                 */
5957                if (!ocfs2_sparse_alloc(osb) &&
5958                    PAGE_CACHE_SIZE < osb->s_clustersize)
5959                        end = PAGE_CACHE_SIZE;
5960
5961                ret = ocfs2_grab_eof_pages(inode, 0, end, pages, &num_pages);
5962                if (ret) {
5963                        mlog_errno(ret);
5964                        goto out_commit;
5965                }
5966
5967                /*
5968                 * This should populate the 1st page for us and mark
5969                 * it up to date.
5970                 */
5971                ret = ocfs2_read_inline_data(inode, pages[0], di_bh);
5972                if (ret) {
5973                        mlog_errno(ret);
5974                        goto out_commit;
5975                }
5976
5977                page_end = PAGE_CACHE_SIZE;
5978                if (PAGE_CACHE_SIZE > osb->s_clustersize)
5979                        page_end = osb->s_clustersize;
5980
5981                for (i = 0; i < num_pages; i++)
5982                        ocfs2_map_and_dirty_page(inode, handle, 0, page_end,
5983                                                 pages[i], i > 0, &phys);
5984        }
5985
5986        spin_lock(&oi->ip_lock);
5987        oi->ip_dyn_features &= ~OCFS2_INLINE_DATA_FL;
5988        di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
5989        spin_unlock(&oi->ip_lock);
5990
5991        ocfs2_dinode_new_extent_list(inode, di);
5992
5993        ocfs2_journal_dirty(handle, di_bh);
5994
5995        if (has_data) {
5996                /*
5997                 * An error at this point should be extremely rare. If
5998                 * this proves to be false, we could always re-build
5999                 * the in-inode data from our pages.
6000                 */
6001                ret = ocfs2_insert_extent(osb, handle, inode, di_bh,
6002                                          0, block, 1, 0, NULL);
6003                if (ret) {
6004                        mlog_errno(ret);
6005                        goto out_commit;
6006                }
6007
6008                inode->i_blocks = ocfs2_inode_sector_count(inode);
6009        }
6010
6011out_commit:
6012        ocfs2_commit_trans(osb, handle);
6013
6014out_unlock:
6015        if (data_ac)
6016                ocfs2_free_alloc_context(data_ac);
6017
6018out:
6019        if (pages) {
6020                ocfs2_unlock_and_free_pages(pages, num_pages);
6021                kfree(pages);
6022        }
6023
6024        return ret;
6025}
6026
6027/*
6028 * It is expected, that by the time you call this function,
6029 * inode->i_size and fe->i_size have been adjusted.
6030 *
6031 * WARNING: This will kfree the truncate context
6032 */
6033int ocfs2_commit_truncate(struct ocfs2_super *osb,
6034                          struct inode *inode,
6035                          struct buffer_head *fe_bh,
6036                          struct ocfs2_truncate_context *tc)
6037{
6038        int status, i, credits, tl_sem = 0;
6039        u32 clusters_to_del, new_highest_cpos, range;
6040        struct ocfs2_extent_list *el;
6041        handle_t *handle = NULL;
6042        struct inode *tl_inode = osb->osb_tl_inode;
6043        struct ocfs2_path *path = NULL;
6044
6045        mlog_entry_void();
6046
6047        new_highest_cpos = ocfs2_clusters_for_bytes(osb->sb,
6048                                                     i_size_read(inode));
6049
6050        path = ocfs2_new_inode_path(fe_bh);
6051        if (!path) {
6052                status = -ENOMEM;
6053                mlog_errno(status);
6054                goto bail;
6055        }
6056
6057        ocfs2_extent_map_trunc(inode, new_highest_cpos);
6058
6059start:
6060        /*
6061         * Check that we still have allocation to delete.
6062         */
6063        if (OCFS2_I(inode)->ip_clusters == 0) {
6064                status = 0;
6065                goto bail;
6066        }
6067
6068        /*
6069         * Truncate always works against the rightmost tree branch.
6070         */
6071        status = ocfs2_find_path(inode, path, UINT_MAX);
6072        if (status) {
6073                mlog_errno(status);
6074                goto bail;
6075        }
6076
6077        mlog(0, "inode->ip_clusters = %u, tree_depth = %u\n",
6078             OCFS2_I(inode)->ip_clusters, path->p_tree_depth);
6079
6080        /*
6081         * By now, el will point to the extent list on the bottom most
6082         * portion of this tree. Only the tail record is considered in
6083         * each pass.
6084         *
6085         * We handle the following cases, in order:
6086         * - empty extent: delete the remaining branch
6087         * - remove the entire record
6088         * - remove a partial record
6089         * - no record needs to be removed (truncate has completed)
6090         */
6091        el = path_leaf_el(path);
6092        if (le16_to_cpu(el->l_next_free_rec) == 0) {
6093                ocfs2_error(inode->i_sb,
6094                            "Inode %llu has empty extent block at %llu\n",
6095                            (unsigned long long)OCFS2_I(inode)->ip_blkno,
6096                            (unsigned long long)path_leaf_bh(path)->b_blocknr);
6097                status = -EROFS;
6098                goto bail;
6099        }
6100
6101        i = le16_to_cpu(el->l_next_free_rec) - 1;
6102        range = le32_to_cpu(el->l_recs[i].e_cpos) +
6103                ocfs2_rec_clusters(el, &el->l_recs[i]);
6104        if (i == 0 && ocfs2_is_empty_extent(&el->l_recs[i])) {
6105                clusters_to_del = 0;
6106        } else if (le32_to_cpu(el->l_recs[i].e_cpos) >= new_highest_cpos) {
6107                clusters_to_del = ocfs2_rec_clusters(el, &el->l_recs[i]);
6108        } else if (range > new_highest_cpos) {
6109                clusters_to_del = (ocfs2_rec_clusters(el, &el->l_recs[i]) +
6110                                   le32_to_cpu(el->l_recs[i].e_cpos)) -
6111                                  new_highest_cpos;
6112        } else {
6113                status = 0;
6114                goto bail;
6115        }
6116
6117        mlog(0, "clusters_to_del = %u in this pass, tail blk=%llu\n",
6118             clusters_to_del, (unsigned long long)path_leaf_bh(path)->b_blocknr);
6119
6120        mutex_lock(&tl_inode->i_mutex);
6121        tl_sem = 1;
6122        /* ocfs2_truncate_log_needs_flush guarantees us at least one
6123         * record is free for use. If there isn't any, we flush to get
6124         * an empty truncate log.  */
6125        if (ocfs2_truncate_log_needs_flush(osb)) {
6126                status = __ocfs2_flush_truncate_log(osb);
6127                if (status < 0) {
6128                        mlog_errno(status);
6129                        goto bail;
6130                }
6131        }
6132
6133        credits = ocfs2_calc_tree_trunc_credits(osb->sb, clusters_to_del,
6134                                                (struct ocfs2_dinode *)fe_bh->b_data,
6135                                                el);
6136        handle = ocfs2_start_trans(osb, credits);
6137        if (IS_ERR(handle)) {
6138                status = PTR_ERR(handle);
6139                handle = NULL;
6140                mlog_errno(status);
6141                goto bail;
6142        }
6143
6144        status = ocfs2_do_truncate(osb, clusters_to_del, inode, fe_bh, handle,
6145                                   tc, path);
6146        if (status < 0) {
6147                mlog_errno(status);
6148                goto bail;
6149        }
6150
6151        mutex_unlock(&tl_inode->i_mutex);
6152        tl_sem = 0;
6153
6154        ocfs2_commit_trans(osb, handle);
6155        handle = NULL;
6156
6157        ocfs2_reinit_path(path, 1);
6158
6159        /*
6160         * The check above will catch the case where we've truncated
6161         * away all allocation.
6162         */
6163        goto start;
6164
6165bail:
6166
6167        ocfs2_schedule_truncate_log_flush(osb, 1);
6168
6169        if (tl_sem)
6170                mutex_unlock(&tl_inode->i_mutex);
6171
6172        if (handle)
6173                ocfs2_commit_trans(osb, handle);
6174
6175        ocfs2_run_deallocs(osb, &tc->tc_dealloc);
6176
6177        ocfs2_free_path(path);
6178
6179        /* This will drop the ext_alloc cluster lock for us */
6180        ocfs2_free_truncate_context(tc);
6181
6182        mlog_exit(status);
6183        return status;
6184}
6185
6186/*
6187 * Expects the inode to already be locked.
6188 */
6189int ocfs2_prepare_truncate(struct ocfs2_super *osb,
6190                           struct inode *inode,
6191                           struct buffer_head *fe_bh,
6192                           struct ocfs2_truncate_context **tc)
6193{
6194        int status;
6195        unsigned int new_i_clusters;
6196        struct ocfs2_dinode *fe;
6197        struct ocfs2_extent_block *eb;
6198        struct buffer_head *last_eb_bh = NULL;
6199
6200        mlog_entry_void();
6201
6202        *tc = NULL;
6203
6204        new_i_clusters = ocfs2_clusters_for_bytes(osb->sb,
6205                                                  i_size_read(inode));
6206        fe = (struct ocfs2_dinode *) fe_bh->b_data;
6207
6208        mlog(0, "fe->i_clusters = %u, new_i_clusters = %u, fe->i_size ="
6209             "%llu\n", le32_to_cpu(fe->i_clusters), new_i_clusters,
6210             (unsigned long long)le64_to_cpu(fe->i_size));
6211
6212        *tc = kzalloc(sizeof(struct ocfs2_truncate_context), GFP_KERNEL);
6213        if (!(*tc)) {
6214                status = -ENOMEM;
6215                mlog_errno(status);
6216                goto bail;
6217        }
6218        ocfs2_init_dealloc_ctxt(&(*tc)->tc_dealloc);
6219
6220        if (fe->id2.i_list.l_tree_depth) {
6221                status = ocfs2_read_block(osb, le64_to_cpu(fe->i_last_eb_blk),
6222                                          &last_eb_bh, OCFS2_BH_CACHED, inode);
6223                if (status < 0) {
6224                        mlog_errno(status);
6225                        goto bail;
6226                }
6227                eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
6228                if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
6229                        OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
6230
6231                        brelse(last_eb_bh);
6232                        status = -EIO;
6233                        goto bail;
6234                }
6235        }
6236
6237        (*tc)->tc_last_eb_bh = last_eb_bh;
6238
6239        status = 0;
6240bail:
6241        if (status < 0) {
6242                if (*tc)
6243                        ocfs2_free_truncate_context(*tc);
6244                *tc = NULL;
6245        }
6246        mlog_exit_void();
6247        return status;
6248}
6249
6250/*
6251 * 'start' is inclusive, 'end' is not.
6252 */
6253int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
6254                          unsigned int start, unsigned int end, int trunc)
6255{
6256        int ret;
6257        unsigned int numbytes;
6258        handle_t *handle;
6259        struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
6260        struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
6261        struct ocfs2_inline_data *idata = &di->id2.i_data;
6262
6263        if (end > i_size_read(inode))
6264                end = i_size_read(inode);
6265
6266        BUG_ON(start >= end);
6267
6268        if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
6269            !(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL) ||
6270            !ocfs2_supports_inline_data(osb)) {
6271                ocfs2_error(inode->i_sb,
6272                            "Inline data flags for inode %llu don't agree! "
6273                            "Disk: 0x%x, Memory: 0x%x, Superblock: 0x%x\n",
6274                            (unsigned long long)OCFS2_I(inode)->ip_blkno,
6275                            le16_to_cpu(di->i_dyn_features),
6276                            OCFS2_I(inode)->ip_dyn_features,
6277                            osb->s_feature_incompat);
6278                ret = -EROFS;
6279                goto out;
6280        }
6281
6282        handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
6283        if (IS_ERR(handle)) {
6284                ret = PTR_ERR(handle);
6285                mlog_errno(ret);
6286                goto out;
6287        }
6288
6289        ret = ocfs2_journal_access(handle, inode, di_bh,
6290                                   OCFS2_JOURNAL_ACCESS_WRITE);
6291        if (ret) {
6292                mlog_errno(ret);
6293                goto out_commit;
6294        }
6295
6296        numbytes = end - start;
6297        memset(idata->id_data + start, 0, numbytes);
6298
6299        /*
6300         * No need to worry about the data page here - it's been
6301         * truncated already and inline data doesn't need it for
6302         * pushing zero's to disk, so we'll let readpage pick it up
6303         * later.
6304         */
6305        if (trunc) {
6306                i_size_write(inode, start);
6307                di->i_size = cpu_to_le64(start);
6308        }
6309
6310        inode->i_blocks = ocfs2_inode_sector_count(inode);
6311        inode->i_ctime = inode->i_mtime = CURRENT_TIME;
6312
6313        di->i_ctime = di->i_mtime = cpu_to_le64(inode->i_ctime.tv_sec);
6314        di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
6315
6316        ocfs2_journal_dirty(handle, di_bh);
6317
6318out_commit:
6319        ocfs2_commit_trans(osb, handle);
6320
6321out:
6322        return ret;
6323}
6324
6325static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc)
6326{
6327        /*
6328         * The caller is responsible for completing deallocation
6329         * before freeing the context.
6330         */
6331        if (tc->tc_dealloc.c_first_suballocator != NULL)
6332                mlog(ML_NOTICE,
6333                     "Truncate completion has non-empty dealloc context\n");
6334
6335        if (tc->tc_last_eb_bh)
6336                brelse(tc->tc_last_eb_bh);
6337
6338        kfree(tc);
6339}
6340