fs/ext4/extents.c

   1 /*
   2  * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
   3  * Written by Alex Tomas <alex@clusterfs.com>
   4  *
   5  * Architecture independence:
   6  *   Copyright (c) 2005, Bull S.A.
   7  *   Written by Pierre Peiffer <pierre.peiffer@bull.net>
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of the GNU General Public License version 2 as
  11  * published by the Free Software Foundation.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public Licens
  19  * along with this program; if not, write to the Free Software
  20  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-
  21  */
  22
  23 /*
  24  * Extents support for EXT4
  25  *
  26  * TODO:
  27  *   - ext4*_error() should be used in some situations
  28  *   - analyze all BUG()/BUG_ON(), use -EIO where appropriate
  29  *   - smart tree reduction
  30  */
  31
  32 #include <linux/fs.h>
  33 #include <linux/time.h>
  34 #include <linux/jbd2.h>
  35 #include <linux/highuid.h>
  36 #include <linux/pagemap.h>
  37 #include <linux/quotaops.h>
  38 #include <linux/string.h>
  39 #include <linux/slab.h>
  40 #include <asm/uaccess.h>
  41 #include <linux/fiemap.h>
  42 #include "ext4_jbd2.h"
  43 #include "ext4_extents.h"
  44 #include "xattr.h"
  45
  46 #include <trace/events/ext4.h>
  47
  48 /*
  49  * used by extent splitting.
  50  */
  51 #define EXT4_EXT_MAY_ZEROOUT    0x1  /* safe to zeroout if split fails \
  52                                         due to ENOSPC */
  53 #define EXT4_EXT_MARK_UNWRIT1   0x2  /* mark first half unwritten */
  54 #define EXT4_EXT_MARK_UNWRIT2   0x4  /* mark second half unwritten */
  55
  56 #define EXT4_EXT_DATA_VALID1    0x8  /* first half contains valid data */
  57 #define EXT4_EXT_DATA_VALID2    0x10 /* second half contains valid data */
  58
  59 static __le32 ext4_extent_block_csum(struct inode *inode,
  60                                      struct ext4_extent_header *eh)
  61 {
  62         struct ext4_inode_info *ei = EXT4_I(inode);
  63         struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
  64         __u32 csum;
  65
  66         csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)eh,
  67                            EXT4_EXTENT_TAIL_OFFSET(eh));
  68         return cpu_to_le32(csum);
  69 }
  70
  71 static int ext4_extent_block_csum_verify(struct inode *inode,
  72                                          struct ext4_extent_header *eh)
  73 {
  74         struct ext4_extent_tail *et;
  75
  76         if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
  77                 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
  78                 return 1;
  79
  80         et = find_ext4_extent_tail(eh);
  81         if (et->et_checksum != ext4_extent_block_csum(inode, eh))
  82                 return 0;
  83         return 1;
  84 }
  85
  86 static void ext4_extent_block_csum_set(struct inode *inode,
  87                                        struct ext4_extent_header *eh)
  88 {
  89         struct ext4_extent_tail *et;
  90
  91         if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
  92                 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
  93                 return;
  94
  95         et = find_ext4_extent_tail(eh);
  96         et->et_checksum = ext4_extent_block_csum(inode, eh);
  97 }
  98
  99 static int ext4_split_extent(handle_t *handle,
 100                                 struct inode *inode,
 101                                 struct ext4_ext_path *path,
 102                                 struct ext4_map_blocks *map,
 103                                 int split_flag,
 104                                 int flags);
 105
 106 static int ext4_split_extent_at(handle_t *handle,
 107                              struct inode *inode,
 108                              struct ext4_ext_path *path,
 109                              ext4_lblk_t split,
 110                              int split_flag,
 111                              int flags);
 112
 113 static int ext4_find_delayed_extent(struct inode *inode,
 114                                     struct extent_status *newes);
 115
 116 static int ext4_ext_truncate_extend_restart(handle_t *handle,
 117                                             struct inode *inode,
 118                                             int needed)
 119 {
 120         int err;
 121
 122         if (!ext4_handle_valid(handle))
 123                 return 0;
 124         if (handle->h_buffer_credits > needed)
 125                 return 0;
 126         err = ext4_journal_extend(handle, needed);
 127         if (err <= 0)
 128                 return err;
 129         err = ext4_truncate_restart_trans(handle, inode, needed);
 130         if (err == 0)
 131                 err = -EAGAIN;
 132
 133         return err;
 134 }
 135
 136 /*
 137  * could return:
 138  *  - EROFS
 139  *  - ENOMEM
 140  */
 141 static int ext4_ext_get_access(handle_t *handle, struct inode *inode,
 142                                 struct ext4_ext_path *path)
 143 {
 144         if (path->p_bh) {
 145                 /* path points to block */
 146                 BUFFER_TRACE(path->p_bh, "get_write_access");
 147                 return ext4_journal_get_write_access(handle, path->p_bh);
 148         }
 149         /* path points to leaf/index in inode body */
 150         /* we use in-core data, no need to protect them */
 151         return 0;
 152 }
 153
 154 /*
 155  * could return:
 156  *  - EROFS
 157  *  - ENOMEM
 158  *  - EIO
 159  */
 160 int __ext4_ext_dirty(const char *where, unsigned int line, handle_t *handle,
 161                      struct inode *inode, struct ext4_ext_path *path)
 162 {
 163         int err;
 164
 165         WARN_ON(!rwsem_is_locked(&EXT4_I(inode)->i_data_sem));
 166         if (path->p_bh) {
 167                 ext4_extent_block_csum_set(inode, ext_block_hdr(path->p_bh));
 168                 /* path points to block */
 169                 err = __ext4_handle_dirty_metadata(where, line, handle,
 170                                                    inode, path->p_bh);
 171         } else {
 172                 /* path points to leaf/index in inode body */
 173                 err = ext4_mark_inode_dirty(handle, inode);
 174         }
 175         return err;
 176 }
 177
 178 static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode,
 179                               struct ext4_ext_path *path,
 180                               ext4_lblk_t block)
 181 {
 182         if (path) {
 183                 int depth = path->p_depth;
 184                 struct ext4_extent *ex;
 185
 186                 /*
 187                  * Try to predict block placement assuming that we are
 188                  * filling in a file which will eventually be
 189                  * non-sparse --- i.e., in the case of libbfd writing
 190                  * an ELF object sections out-of-order but in a way
 191                  * the eventually results in a contiguous object or
 192                  * executable file, or some database extending a table
 193                  * space file.  However, this is actually somewhat
 194                  * non-ideal if we are writing a sparse file such as
 195                  * qemu or KVM writing a raw image file that is going
 196                  * to stay fairly sparse, since it will end up
 197                  * fragmenting the file system's free space.  Maybe we
 198                  * should have some hueristics or some way to allow
 199                  * userspace to pass a hint to file system,
 200                  * especially if the latter case turns out to be
 201                  * common.
 202                  */
 203                 ex = path[depth].p_ext;
 204                 if (ex) {
 205                         ext4_fsblk_t ext_pblk = ext4_ext_pblock(ex);
 206                         ext4_lblk_t ext_block = le32_to_cpu(ex->ee_block);
 207
 208                         if (block > ext_block)
 209                                 return ext_pblk + (block - ext_block);
 210                         else
 211                                 return ext_pblk - (ext_block - block);
 212                 }
 213
 214                 /* it looks like index is empty;
 215                  * try to find starting block from index itself */
 216                 if (path[depth].p_bh)
 217                         return path[depth].p_bh->b_blocknr;
 218         }
 219
 220         /* OK. use inode's group */
 221         return ext4_inode_to_goal_block(inode);
 222 }
 223
 224 /*
 225  * Allocation for a meta data block
 226  */
 227 static ext4_fsblk_t
 228 ext4_ext_new_meta_block(handle_t *handle, struct inode *inode,
 229                         struct ext4_ext_path *path,
 230                         struct ext4_extent *ex, int *err, unsigned int flags)
 231 {
 232         ext4_fsblk_t goal, newblock;
 233
 234         goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block));
 235         newblock = ext4_new_meta_blocks(handle, inode, goal, flags,
 236                                         NULL, err);
 237         return newblock;
 238 }
 239
 240 static inline int ext4_ext_space_block(struct inode *inode, int check)
 241 {
 242         int size;
 243
 244         size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
 245                         / sizeof(struct ext4_extent);
 246 #ifdef AGGRESSIVE_TEST
 247         if (!check && size > 6)
 248                 size = 6;
 249 #endif
 250         return size;
 251 }
 252
 253 static inline int ext4_ext_space_block_idx(struct inode *inode, int check)
 254 {
 255         int size;
 256
 257         size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
 258                         / sizeof(struct ext4_extent_idx);
 259 #ifdef AGGRESSIVE_TEST
 260         if (!check && size > 5)
 261                 size = 5;
 262 #endif
 263         return size;
 264 }
 265
 266 static inline int ext4_ext_space_root(struct inode *inode, int check)
 267 {
 268         int size;
 269
 270         size = sizeof(EXT4_I(inode)->i_data);
 271         size -= sizeof(struct ext4_extent_header);
 272         size /= sizeof(struct ext4_extent);
 273 #ifdef AGGRESSIVE_TEST
 274         if (!check && size > 3)
 275                 size = 3;
 276 #endif
 277         return size;
 278 }
 279
 280 static inline int ext4_ext_space_root_idx(struct inode *inode, int check)
 281 {
 282         int size;
 283
 284         size = sizeof(EXT4_I(inode)->i_data);
 285         size -= sizeof(struct ext4_extent_header);
 286         size /= sizeof(struct ext4_extent_idx);
 287 #ifdef AGGRESSIVE_TEST
 288         if (!check && size > 4)
 289                 size = 4;
 290 #endif
 291         return size;
 292 }
 293
 294 /*
 295  * Calculate the number of metadata blocks needed
 296  * to allocate @blocks
 297  * Worse case is one block per extent
 298  */
 299 int ext4_ext_calc_metadata_amount(struct inode *inode, ext4_lblk_t lblock)
 300 {
 301         struct ext4_inode_info *ei = EXT4_I(inode);
 302         int idxs;
 303
 304         idxs = ((inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
 305                 / sizeof(struct ext4_extent_idx));
 306
 307         /*
 308          * If the new delayed allocation block is contiguous with the
 309          * previous da block, it can share index blocks with the
 310          * previous block, so we only need to allocate a new index
 311          * block every idxs leaf blocks.  At ldxs**2 blocks, we need
 312          * an additional index block, and at ldxs**3 blocks, yet
 313          * another index blocks.
 314          */
 315         if (ei->i_da_metadata_calc_len &&
 316             ei->i_da_metadata_calc_last_lblock+1 == lblock) {
 317                 int num = 0;
 318
 319                 if ((ei->i_da_metadata_calc_len % idxs) == 0)
 320                         num++;
 321                 if ((ei->i_da_metadata_calc_len % (idxs*idxs)) == 0)
 322                         num++;
 323                 if ((ei->i_da_metadata_calc_len % (idxs*idxs*idxs)) == 0) {
 324                         num++;
 325                         ei->i_da_metadata_calc_len = 0;
 326                 } else
 327                         ei->i_da_metadata_calc_len++;
 328                 ei->i_da_metadata_calc_last_lblock++;
 329                 return num;
 330         }
 331
 332         /*
 333          * In the worst case we need a new set of index blocks at
 334          * every level of the inode's extent tree.
 335          */
 336         ei->i_da_metadata_calc_len = 1;
 337         ei->i_da_metadata_calc_last_lblock = lblock;
 338         return ext_depth(inode) + 1;
 339 }
 340
 341 static int
 342 ext4_ext_max_entries(struct inode *inode, int depth)
 343 {
 344         int max;
 345
 346         if (depth == ext_depth(inode)) {
 347                 if (depth == 0)
 348                         max = ext4_ext_space_root(inode, 1);
 349                 else
 350                         max = ext4_ext_space_root_idx(inode, 1);
 351         } else {
 352                 if (depth == 0)
 353                         max = ext4_ext_space_block(inode, 1);
 354                 else
 355                         max = ext4_ext_space_block_idx(inode, 1);
 356         }
 357
 358         return max;
 359 }
 360
 361 static int ext4_valid_extent(struct inode *inode, struct ext4_extent *ext)
 362 {
 363         ext4_fsblk_t block = ext4_ext_pblock(ext);
 364         int len = ext4_ext_get_actual_len(ext);
 365         ext4_lblk_t lblock = le32_to_cpu(ext->ee_block);
 366         ext4_lblk_t last = lblock + len - 1;
 367
 368         if (lblock > last)
 369                 return 0;
 370         return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, len);
 371 }
 372
 373 static int ext4_valid_extent_idx(struct inode *inode,
 374                                 struct ext4_extent_idx *ext_idx)
 375 {
 376         ext4_fsblk_t block = ext4_idx_pblock(ext_idx);
 377
 378         return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, 1);
 379 }
 380
 381 static int ext4_valid_extent_entries(struct inode *inode,
 382                                 struct ext4_extent_header *eh,
 383                                 int depth)
 384 {
 385         unsigned short entries;
 386         if (eh->eh_entries == 0)
 387                 return 1;
 388
 389         entries = le16_to_cpu(eh->eh_entries);
 390
 391         if (depth == 0) {
 392                 /* leaf entries */
 393                 struct ext4_extent *ext = EXT_FIRST_EXTENT(eh);
 394                 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
 395                 ext4_fsblk_t pblock = 0;
 396                 ext4_lblk_t lblock = 0;
 397                 ext4_lblk_t prev = 0;
 398                 int len = 0;
 399                 while (entries) {
 400                         if (!ext4_valid_extent(inode, ext))
 401                                 return 0;
 402
 403                         /* Check for overlapping extents */
 404                         lblock = le32_to_cpu(ext->ee_block);
 405                         len = ext4_ext_get_actual_len(ext);
 406                         if ((lblock <= prev) && prev) {
 407                                 pblock = ext4_ext_pblock(ext);
 408                                 es->s_last_error_block = cpu_to_le64(pblock);
 409                                 return 0;
 410                         }
 411                         ext++;
 412                         entries--;
 413                         prev = lblock + len - 1;
 414                 }
 415         } else {
 416                 struct ext4_extent_idx *ext_idx = EXT_FIRST_INDEX(eh);
 417                 while (entries) {
 418                         if (!ext4_valid_extent_idx(inode, ext_idx))
 419                                 return 0;
 420                         ext_idx++;
 421                         entries--;
 422                 }
 423         }
 424         return 1;
 425 }
 426
 427 static int __ext4_ext_check(const char *function, unsigned int line,
 428                             struct inode *inode, struct ext4_extent_header *eh,
 429                             int depth, ext4_fsblk_t pblk)
 430 {
 431         const char *error_msg;
 432         int max = 0;
 433
 434         if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) {
 435                 error_msg = "invalid magic";
 436                 goto corrupted;
 437         }
 438         if (unlikely(le16_to_cpu(eh->eh_depth) != depth)) {
 439                 error_msg = "unexpected eh_depth";
 440                 goto corrupted;
 441         }
 442         if (unlikely(eh->eh_max == 0)) {
 443                 error_msg = "invalid eh_max";
 444                 goto corrupted;
 445         }
 446         max = ext4_ext_max_entries(inode, depth);
 447         if (unlikely(le16_to_cpu(eh->eh_max) > max)) {
 448                 error_msg = "too large eh_max";
 449                 goto corrupted;
 450         }
 451         if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) {
 452                 error_msg = "invalid eh_entries";
 453                 goto corrupted;
 454         }
 455         if (!ext4_valid_extent_entries(inode, eh, depth)) {
 456                 error_msg = "invalid extent entries";
 457                 goto corrupted;
 458         }
 459         /* Verify checksum on non-root extent tree nodes */
 460         if (ext_depth(inode) != depth &&
 461             !ext4_extent_block_csum_verify(inode, eh)) {
 462                 error_msg = "extent tree corrupted";
 463                 goto corrupted;
 464         }
 465         return 0;
 466
 467 corrupted:
 468         ext4_error_inode(inode, function, line, 0,
 469                          "pblk %llu bad header/extent: %s - magic %x, "
 470                          "entries %u, max %u(%u), depth %u(%u)",
 471                          (unsigned long long) pblk, error_msg,
 472                          le16_to_cpu(eh->eh_magic),
 473                          le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max),
 474                          max, le16_to_cpu(eh->eh_depth), depth);
 475         return -EIO;
 476 }
 477
 478 #define ext4_ext_check(inode, eh, depth, pblk)                  \
 479         __ext4_ext_check(__func__, __LINE__, (inode), (eh), (depth), (pblk))
 480
 481 int ext4_ext_check_inode(struct inode *inode)
 482 {
 483         return ext4_ext_check(inode, ext_inode_hdr(inode), ext_depth(inode), 0);
 484 }
 485
 486 static struct buffer_head *
 487 __read_extent_tree_block(const char *function, unsigned int line,
 488                          struct inode *inode, ext4_fsblk_t pblk, int depth,
 489                          int flags)
 490 {
 491         struct buffer_head              *bh;
 492         int                             err;
 493
 494         bh = sb_getblk(inode->i_sb, pblk);
 495         if (unlikely(!bh))
 496                 return ERR_PTR(-ENOMEM);
 497
 498         if (!bh_uptodate_or_lock(bh)) {
 499                 trace_ext4_ext_load_extent(inode, pblk, _RET_IP_);
 500                 err = bh_submit_read(bh);
 501                 if (err < 0)
 502                         goto errout;
 503         }
 504         if (buffer_verified(bh) && !(flags & EXT4_EX_FORCE_CACHE))
 505                 return bh;
 506         err = __ext4_ext_check(function, line, inode,
 507                                ext_block_hdr(bh), depth, pblk);
 508         if (err)
 509                 goto errout;
 510         set_buffer_verified(bh);
 511         /*
 512          * If this is a leaf block, cache all of its entries
 513          */
 514         if (!(flags & EXT4_EX_NOCACHE) && depth == 0) {
 515                 struct ext4_extent_header *eh = ext_block_hdr(bh);
 516                 struct ext4_extent *ex = EXT_FIRST_EXTENT(eh);
 517                 ext4_lblk_t prev = 0;
 518                 int i;
 519
 520                 for (i = le16_to_cpu(eh->eh_entries); i > 0; i--, ex++) {
 521                         unsigned int status = EXTENT_STATUS_WRITTEN;
 522                         ext4_lblk_t lblk = le32_to_cpu(ex->ee_block);
 523                         int len = ext4_ext_get_actual_len(ex);
 524
 525                         if (prev && (prev != lblk))
 526                                 ext4_es_cache_extent(inode, prev,
 527                                                      lblk - prev, ~0,
 528                                                      EXTENT_STATUS_HOLE);
 529
 530                         if (ext4_ext_is_unwritten(ex))
 531                                 status = EXTENT_STATUS_UNWRITTEN;
 532                         ext4_es_cache_extent(inode, lblk, len,
 533                                              ext4_ext_pblock(ex), status);
 534                         prev = lblk + len;
 535                 }
 536         }
 537         return bh;
 538 errout:
 539         put_bh(bh);
 540         return ERR_PTR(err);
 541
 542 }
 543
 544 #define read_extent_tree_block(inode, pblk, depth, flags)               \
 545         __read_extent_tree_block(__func__, __LINE__, (inode), (pblk),   \
 546                                  (depth), (flags))
 547
 548 /*
 549  * This function is called to cache a file's extent information in the
 550  * extent status tree
 551  */
 552 int ext4_ext_precache(struct inode *inode)
 553 {
 554         struct ext4_inode_info *ei = EXT4_I(inode);
 555         struct ext4_ext_path *path = NULL;
 556         struct buffer_head *bh;
 557         int i = 0, depth, ret = 0;
 558
 559         if (!ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
 560                 return 0;       /* not an extent-mapped inode */
 561
 562         down_read(&ei->i_data_sem);
 563         depth = ext_depth(inode);
 564
 565         path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1),
 566                        GFP_NOFS);
 567         if (path == NULL) {
 568                 up_read(&ei->i_data_sem);
 569                 return -ENOMEM;
 570         }
 571
 572         /* Don't cache anything if there are no external extent blocks */
 573         if (depth == 0)
 574                 goto out;
 575         path[0].p_hdr = ext_inode_hdr(inode);
 576         ret = ext4_ext_check(inode, path[0].p_hdr, depth, 0);
 577         if (ret)
 578                 goto out;
 579         path[0].p_idx = EXT_FIRST_INDEX(path[0].p_hdr);
 580         while (i >= 0) {
 581                 /*
 582                  * If this is a leaf block or we've reached the end of
 583                  * the index block, go up
 584                  */
 585                 if ((i == depth) ||
 586                     path[i].p_idx > EXT_LAST_INDEX(path[i].p_hdr)) {
 587                         brelse(path[i].p_bh);
 588                         path[i].p_bh = NULL;
 589                         i--;
 590                         continue;
 591                 }
 592                 bh = read_extent_tree_block(inode,
 593                                             ext4_idx_pblock(path[i].p_idx++),
 594                                             depth - i - 1,
 595                                             EXT4_EX_FORCE_CACHE);
 596                 if (IS_ERR(bh)) {
 597                         ret = PTR_ERR(bh);
 598                         break;
 599                 }
 600                 i++;
 601                 path[i].p_bh = bh;
 602                 path[i].p_hdr = ext_block_hdr(bh);
 603                 path[i].p_idx = EXT_FIRST_INDEX(path[i].p_hdr);
 604         }
 605         ext4_set_inode_state(inode, EXT4_STATE_EXT_PRECACHED);
 606 out:
 607         up_read(&ei->i_data_sem);
 608         ext4_ext_drop_refs(path);
 609         kfree(path);
 610         return ret;
 611 }
 612
 613 #ifdef EXT_DEBUG
 614 static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path)
 615 {
 616         int k, l = path->p_depth;
 617
 618         ext_debug("path:");
 619         for (k = 0; k <= l; k++, path++) {
 620                 if (path->p_idx) {
 621                   ext_debug("  %d->%llu", le32_to_cpu(path->p_idx->ei_block),
 622                             ext4_idx_pblock(path->p_idx));
 623                 } else if (path->p_ext) {
 624                         ext_debug("  %d:[%d]%d:%llu ",
 625                                   le32_to_cpu(path->p_ext->ee_block),
 626                                   ext4_ext_is_unwritten(path->p_ext),
 627                                   ext4_ext_get_actual_len(path->p_ext),
 628                                   ext4_ext_pblock(path->p_ext));
 629                 } else
 630                         ext_debug("  []");
 631         }
 632         ext_debug("\n");
 633 }
 634
 635 static void ext4_ext_show_leaf(struct inode *inode, struct ext4_ext_path *path)
 636 {
 637         int depth = ext_depth(inode);
 638         struct ext4_extent_header *eh;
 639         struct ext4_extent *ex;
 640         int i;
 641
 642         if (!path)
 643                 return;
 644
 645         eh = path[depth].p_hdr;
 646         ex = EXT_FIRST_EXTENT(eh);
 647
 648         ext_debug("Displaying leaf extents for inode %lu\n", inode->i_ino);
 649
 650         for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) {
 651                 ext_debug("%d:[%d]%d:%llu ", le32_to_cpu(ex->ee_block),
 652                           ext4_ext_is_unwritten(ex),
 653                           ext4_ext_get_actual_len(ex), ext4_ext_pblock(ex));
 654         }
 655         ext_debug("\n");
 656 }
 657
 658 static void ext4_ext_show_move(struct inode *inode, struct ext4_ext_path *path,
 659                         ext4_fsblk_t newblock, int level)
 660 {
 661         int depth = ext_depth(inode);
 662         struct ext4_extent *ex;
 663
 664         if (depth != level) {
 665                 struct ext4_extent_idx *idx;
 666                 idx = path[level].p_idx;
 667                 while (idx <= EXT_MAX_INDEX(path[level].p_hdr)) {
 668                         ext_debug("%d: move %d:%llu in new index %llu\n", level,
 669                                         le32_to_cpu(idx->ei_block),
 670                                         ext4_idx_pblock(idx),
 671                                         newblock);
 672                         idx++;
 673                 }
 674
 675                 return;
 676         }
 677
 678         ex = path[depth].p_ext;
 679         while (ex <= EXT_MAX_EXTENT(path[depth].p_hdr)) {
 680                 ext_debug("move %d:%llu:[%d]%d in new leaf %llu\n",
 681                                 le32_to_cpu(ex->ee_block),
 682                                 ext4_ext_pblock(ex),
 683                                 ext4_ext_is_unwritten(ex),
 684                                 ext4_ext_get_actual_len(ex),
 685                                 newblock);
 686                 ex++;
 687         }
 688 }
 689
 690 #else
 691 #define ext4_ext_show_path(inode, path)
 692 #define ext4_ext_show_leaf(inode, path)
 693 #define ext4_ext_show_move(inode, path, newblock, level)
 694 #endif
 695
 696 void ext4_ext_drop_refs(struct ext4_ext_path *path)
 697 {
 698         int depth = path->p_depth;
 699         int i;
 700
 701         for (i = 0; i <= depth; i++, path++)
 702                 if (path->p_bh) {
 703                         brelse(path->p_bh);
 704                         path->p_bh = NULL;
 705                 }
 706 }
 707
 708 /*
 709  * ext4_ext_binsearch_idx:
 710  * binary search for the closest index of the given block
 711  * the header must be checked before calling this
 712  */
 713 static void
 714 ext4_ext_binsearch_idx(struct inode *inode,
 715                         struct ext4_ext_path *path, ext4_lblk_t block)
 716 {
 717         struct ext4_extent_header *eh = path->p_hdr;
 718         struct ext4_extent_idx *r, *l, *m;
 719
 720
 721         ext_debug("binsearch for %u(idx):  ", block);
 722
 723         l = EXT_FIRST_INDEX(eh) + 1;
 724         r = EXT_LAST_INDEX(eh);
 725         while (l <= r) {
 726                 m = l + (r - l) / 2;
 727                 if (block < le32_to_cpu(m->ei_block))
 728                         r = m - 1;
 729                 else
 730                         l = m + 1;
 731                 ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ei_block),
 732                                 m, le32_to_cpu(m->ei_block),
 733                                 r, le32_to_cpu(r->ei_block));
 734         }
 735
 736         path->p_idx = l - 1;
 737         ext_debug("  -> %u->%lld ", le32_to_cpu(path->p_idx->ei_block),
 738                   ext4_idx_pblock(path->p_idx));
 739
 740 #ifdef CHECK_BINSEARCH
 741         {
 742                 struct ext4_extent_idx *chix, *ix;
 743                 int k;
 744
 745                 chix = ix = EXT_FIRST_INDEX(eh);
 746                 for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ix++) {
 747                   if (k != 0 &&
 748                       le32_to_cpu(ix->ei_block) <= le32_to_cpu(ix[-1].ei_block)) {
 749                                 printk(KERN_DEBUG "k=%d, ix=0x%p, "
 750                                        "first=0x%p\n", k,
 751                                        ix, EXT_FIRST_INDEX(eh));
 752                                 printk(KERN_DEBUG "%u <= %u\n",
 753                                        le32_to_cpu(ix->ei_block),
 754                                        le32_to_cpu(ix[-1].ei_block));
 755                         }
 756                         BUG_ON(k && le32_to_cpu(ix->ei_block)
 757                                            <= le32_to_cpu(ix[-1].ei_block));
 758                         if (block < le32_to_cpu(ix->ei_block))
 759                                 break;
 760                         chix = ix;
 761                 }
 762                 BUG_ON(chix != path->p_idx);
 763         }
 764 #endif
 765
 766 }
 767
 768 /*
 769  * ext4_ext_binsearch:
 770  * binary search for closest extent of the given block
 771  * the header must be checked before calling this
 772  */
 773 static void
 774 ext4_ext_binsearch(struct inode *inode,
 775                 struct ext4_ext_path *path, ext4_lblk_t block)
 776 {
 777         struct ext4_extent_header *eh = path->p_hdr;
 778         struct ext4_extent *r, *l, *m;
 779
 780         if (eh->eh_entries == 0) {
 781                 /*
 782                  * this leaf is empty:
 783                  * we get such a leaf in split/add case
 784                  */
 785                 return;
 786         }
 787
 788         ext_debug("binsearch for %u:  ", block);
 789
 790         l = EXT_FIRST_EXTENT(eh) + 1;
 791         r = EXT_LAST_EXTENT(eh);
 792
 793         while (l <= r) {
 794                 m = l + (r - l) / 2;
 795                 if (block < le32_to_cpu(m->ee_block))
 796                         r = m - 1;
 797                 else
 798                         l = m + 1;
 799                 ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ee_block),
 800                                 m, le32_to_cpu(m->ee_block),
 801                                 r, le32_to_cpu(r->ee_block));
 802         }
 803
 804         path->p_ext = l - 1;
 805         ext_debug("  -> %d:%llu:[%d]%d ",
 806                         le32_to_cpu(path->p_ext->ee_block),
 807                         ext4_ext_pblock(path->p_ext),
 808                         ext4_ext_is_unwritten(path->p_ext),
 809                         ext4_ext_get_actual_len(path->p_ext));
 810
 811 #ifdef CHECK_BINSEARCH
 812         {
 813                 struct ext4_extent *chex, *ex;
 814                 int k;
 815
 816                 chex = ex = EXT_FIRST_EXTENT(eh);
 817                 for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ex++) {
 818                         BUG_ON(k && le32_to_cpu(ex->ee_block)
 819                                           <= le32_to_cpu(ex[-1].ee_block));
 820                         if (block < le32_to_cpu(ex->ee_block))
 821                                 break;
 822                         chex = ex;
 823                 }
 824                 BUG_ON(chex != path->p_ext);
 825         }
 826 #endif
 827
 828 }
 829
 830 int ext4_ext_tree_init(handle_t *handle, struct inode *inode)
 831 {
 832         struct ext4_extent_header *eh;
 833
 834         eh = ext_inode_hdr(inode);
 835         eh->eh_depth = 0;
 836         eh->eh_entries = 0;
 837         eh->eh_magic = EXT4_EXT_MAGIC;
 838         eh->eh_max = cpu_to_le16(ext4_ext_space_root(inode, 0));
 839         ext4_mark_inode_dirty(handle, inode);
 840         return 0;
 841 }
 842
 843 struct ext4_ext_path *
 844 ext4_ext_find_extent(struct inode *inode, ext4_lblk_t block,
 845                      struct ext4_ext_path *path, int flags)
 846 {
 847         struct ext4_extent_header *eh;
 848         struct buffer_head *bh;
 849         short int depth, i, ppos = 0, alloc = 0;
 850         int ret;
 851
 852         eh = ext_inode_hdr(inode);
 853         depth = ext_depth(inode);
 854
 855         /* account possible depth increase */
 856         if (!path) {
 857                 path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 2),
 858                                 GFP_NOFS);
 859                 if (!path)
 860                         return ERR_PTR(-ENOMEM);
 861                 alloc = 1;
 862         }
 863         path[0].p_hdr = eh;
 864         path[0].p_bh = NULL;
 865
 866         i = depth;
 867         /* walk through the tree */
 868         while (i) {
 869                 ext_debug("depth %d: num %d, max %d\n",
 870                           ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
 871
 872                 ext4_ext_binsearch_idx(inode, path + ppos, block);
 873                 path[ppos].p_block = ext4_idx_pblock(path[ppos].p_idx);
 874                 path[ppos].p_depth = i;
 875                 path[ppos].p_ext = NULL;
 876
 877                 bh = read_extent_tree_block(inode, path[ppos].p_block, --i,
 878                                             flags);
 879                 if (IS_ERR(bh)) {
 880                         ret = PTR_ERR(bh);
 881                         goto err;
 882                 }
 883
 884                 eh = ext_block_hdr(bh);
 885                 ppos++;
 886                 if (unlikely(ppos > depth)) {
 887                         put_bh(bh);
 888                         EXT4_ERROR_INODE(inode,
 889                                          "ppos %d > depth %d", ppos, depth);
 890                         ret = -EIO;
 891                         goto err;
 892                 }
 893                 path[ppos].p_bh = bh;
 894                 path[ppos].p_hdr = eh;
 895         }
 896
 897         path[ppos].p_depth = i;
 898         path[ppos].p_ext = NULL;
 899         path[ppos].p_idx = NULL;
 900
 901         /* find extent */
 902         ext4_ext_binsearch(inode, path + ppos, block);
 903         /* if not an empty leaf */
 904         if (path[ppos].p_ext)
 905                 path[ppos].p_block = ext4_ext_pblock(path[ppos].p_ext);
 906
 907         ext4_ext_show_path(inode, path);
 908
 909         return path;
 910
 911 err:
 912         ext4_ext_drop_refs(path);
 913         if (alloc)
 914                 kfree(path);
 915         return ERR_PTR(ret);
 916 }
 917
 918 /*
 919  * ext4_ext_insert_index:
 920  * insert new index [@logical;@ptr] into the block at @curp;
 921  * check where to insert: before @curp or after @curp
 922  */
 923 static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
 924                                  struct ext4_ext_path *curp,
 925                                  int logical, ext4_fsblk_t ptr)
 926 {
 927         struct ext4_extent_idx *ix;
 928         int len, err;
 929
 930         err = ext4_ext_get_access(handle, inode, curp);
 931         if (err)
 932                 return err;
 933
 934         if (unlikely(logical == le32_to_cpu(curp->p_idx->ei_block))) {
 935                 EXT4_ERROR_INODE(inode,
 936                                  "logical %d == ei_block %d!",
 937                                  logical, le32_to_cpu(curp->p_idx->ei_block));
 938                 return -EIO;
 939         }
 940
 941         if (unlikely(le16_to_cpu(curp->p_hdr->eh_entries)
 942                              >= le16_to_cpu(curp->p_hdr->eh_max))) {
 943                 EXT4_ERROR_INODE(inode,
 944                                  "eh_entries %d >= eh_max %d!",
 945                                  le16_to_cpu(curp->p_hdr->eh_entries),
 946                                  le16_to_cpu(curp->p_hdr->eh_max));
 947                 return -EIO;
 948         }
 949
 950         if (logical > le32_to_cpu(curp->p_idx->ei_block)) {
 951                 /* insert after */
 952                 ext_debug("insert new index %d after: %llu\n", logical, ptr);
 953                 ix = curp->p_idx + 1;
 954         } else {
 955                 /* insert before */
 956                 ext_debug("insert new index %d before: %llu\n", logical, ptr);
 957                 ix = curp->p_idx;
 958         }
 959
 960         len = EXT_LAST_INDEX(curp->p_hdr) - ix + 1;
 961         BUG_ON(len < 0);
 962         if (len > 0) {
 963                 ext_debug("insert new index %d: "
 964                                 "move %d indices from 0x%p to 0x%p\n",
 965                                 logical, len, ix, ix + 1);
 966                 memmove(ix + 1, ix, len * sizeof(struct ext4_extent_idx));
 967         }
 968
 969         if (unlikely(ix > EXT_MAX_INDEX(curp->p_hdr))) {
 970                 EXT4_ERROR_INODE(inode, "ix > EXT_MAX_INDEX!");
 971                 return -EIO;
 972         }
 973
 974         ix->ei_block = cpu_to_le32(logical);
 975         ext4_idx_store_pblock(ix, ptr);
 976         le16_add_cpu(&curp->p_hdr->eh_entries, 1);
 977
 978         if (unlikely(ix > EXT_LAST_INDEX(curp->p_hdr))) {
 979                 EXT4_ERROR_INODE(inode, "ix > EXT_LAST_INDEX!");
 980                 return -EIO;
 981         }
 982
 983         err = ext4_ext_dirty(handle, inode, curp);
 984         ext4_std_error(inode->i_sb, err);
 985
 986         return err;
 987 }
 988
 989 /*
 990  * ext4_ext_split:
 991  * inserts new subtree into the path, using free index entry
 992  * at depth @at:
 993  * - allocates all needed blocks (new leaf and all intermediate index blocks)
 994  * - makes decision where to split
 995  * - moves remaining extents and index entries (right to the split point)
 996  *   into the newly allocated blocks
 997  * - initializes subtree
 998  */
 999 static int ext4_ext_split(handle_t *handle, struct inode *inode,
1000                           unsigned int flags,
1001                           struct ext4_ext_path *path,
1002                           struct ext4_extent *newext, int at)
1003 {
1004         struct buffer_head *bh = NULL;
1005         int depth = ext_depth(inode);
1006         struct ext4_extent_header *neh;
1007         struct ext4_extent_idx *fidx;
1008         int i = at, k, m, a;
1009         ext4_fsblk_t newblock, oldblock;
1010         __le32 border;
1011         ext4_fsblk_t *ablocks = NULL; /* array of allocated blocks */
1012         int err = 0;
1013
1014         /* make decision: where to split? */
1015         /* FIXME: now decision is simplest: at current extent */
1016
1017         /* if current leaf will be split, then we should use
1018          * border from split point */
1019         if (unlikely(path[depth].p_ext > EXT_MAX_EXTENT(path[depth].p_hdr))) {
1020                 EXT4_ERROR_INODE(inode, "p_ext > EXT_MAX_EXTENT!");
1021                 return -EIO;
1022         }
1023         if (path[depth].p_ext != EXT_MAX_EXTENT(path[depth].p_hdr)) {
1024                 border = path[depth].p_ext[1].ee_block;
1025                 ext_debug("leaf will be split."
1026                                 " next leaf starts at %d\n",
1027                                   le32_to_cpu(border));
1028         } else {
1029                 border = newext->ee_block;
1030                 ext_debug("leaf will be added."
1031                                 " next leaf starts at %d\n",
1032                                 le32_to_cpu(border));
1033         }
1034
1035         /*
1036          * If error occurs, then we break processing
1037          * and mark filesystem read-only. index won't
1038          * be inserted and tree will be in consistent
1039          * state. Next mount will repair buffers too.
1040          */
1041
1042         /*
1043          * Get array to track all allocated blocks.
1044          * We need this to handle errors and free blocks
1045          * upon them.
1046          */
1047         ablocks = kzalloc(sizeof(ext4_fsblk_t) * depth, GFP_NOFS);
1048         if (!ablocks)
1049                 return -ENOMEM;
1050
1051         /* allocate all needed blocks */
1052         ext_debug("allocate %d blocks for indexes/leaf\n", depth - at);
1053         for (a = 0; a < depth - at; a++) {
1054                 newblock = ext4_ext_new_meta_block(handle, inode, path,
1055                                                    newext, &err, flags);
1056                 if (newblock == 0)
1057                         goto cleanup;
1058                 ablocks[a] = newblock;
1059         }
1060
1061         /* initialize new leaf */
1062         newblock = ablocks[--a];
1063         if (unlikely(newblock == 0)) {
1064                 EXT4_ERROR_INODE(inode, "newblock == 0!");
1065                 err = -EIO;
1066                 goto cleanup;
1067         }
1068         bh = sb_getblk(inode->i_sb, newblock);
1069         if (unlikely(!bh)) {
1070                 err = -ENOMEM;
1071                 goto cleanup;
1072         }
1073         lock_buffer(bh);
1074
1075         err = ext4_journal_get_create_access(handle, bh);
1076         if (err)
1077                 goto cleanup;
1078
1079         neh = ext_block_hdr(bh);
1080         neh->eh_entries = 0;
1081         neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0));
1082         neh->eh_magic = EXT4_EXT_MAGIC;
1083         neh->eh_depth = 0;
1084
1085         /* move remainder of path[depth] to the new leaf */
1086         if (unlikely(path[depth].p_hdr->eh_entries !=
1087                      path[depth].p_hdr->eh_max)) {
1088                 EXT4_ERROR_INODE(inode, "eh_entries %d != eh_max %d!",
1089                                  path[depth].p_hdr->eh_entries,
1090                                  path[depth].p_hdr->eh_max);
1091                 err = -EIO;
1092                 goto cleanup;
1093         }
1094         /* start copy from next extent */
1095         m = EXT_MAX_EXTENT(path[depth].p_hdr) - path[depth].p_ext++;
1096         ext4_ext_show_move(inode, path, newblock, depth);
1097         if (m) {
1098                 struct ext4_extent *ex;
1099                 ex = EXT_FIRST_EXTENT(neh);
1100                 memmove(ex, path[depth].p_ext, sizeof(struct ext4_extent) * m);
1101                 le16_add_cpu(&neh->eh_entries, m);
1102         }
1103
1104         ext4_extent_block_csum_set(inode, neh);
1105         set_buffer_uptodate(bh);
1106         unlock_buffer(bh);
1107
1108         err = ext4_handle_dirty_metadata(handle, inode, bh);
1109         if (err)
1110                 goto cleanup;
1111         brelse(bh);
1112         bh = NULL;
1113
1114         /* correct old leaf */
1115         if (m) {
1116                 err = ext4_ext_get_access(handle, inode, path + depth);
1117                 if (err)
1118                         goto cleanup;
1119                 le16_add_cpu(&path[depth].p_hdr->eh_entries, -m);
1120                 err = ext4_ext_dirty(handle, inode, path + depth);
1121                 if (err)
1122                         goto cleanup;
1123
1124         }
1125
1126         /* create intermediate indexes */
1127         k = depth - at - 1;
1128         if (unlikely(k < 0)) {
1129                 EXT4_ERROR_INODE(inode, "k %d < 0!", k);
1130                 err = -EIO;
1131                 goto cleanup;
1132         }
1133         if (k)
1134                 ext_debug("create %d intermediate indices\n", k);
1135         /* insert new index into current index block */
1136         /* current depth stored in i var */
1137         i = depth - 1;
1138         while (k--) {
1139                 oldblock = newblock;
1140                 newblock = ablocks[--a];
1141                 bh = sb_getblk(inode->i_sb, newblock);
1142                 if (unlikely(!bh)) {
1143                         err = -ENOMEM;
1144                         goto cleanup;
1145                 }
1146                 lock_buffer(bh);
1147
1148                 err = ext4_journal_get_create_access(handle, bh);
1149                 if (err)
1150                         goto cleanup;
1151
1152                 neh = ext_block_hdr(bh);
1153                 neh->eh_entries = cpu_to_le16(1);
1154                 neh->eh_magic = EXT4_EXT_MAGIC;
1155                 neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0));
1156                 neh->eh_depth = cpu_to_le16(depth - i);
1157                 fidx = EXT_FIRST_INDEX(neh);
1158                 fidx->ei_block = border;
1159                 ext4_idx_store_pblock(fidx, oldblock);
1160
1161                 ext_debug("int.index at %d (block %llu): %u -> %llu\n",
1162                                 i, newblock, le32_to_cpu(border), oldblock);
1163
1164                 /* move remainder of path[i] to the new index block */
1165                 if (unlikely(EXT_MAX_INDEX(path[i].p_hdr) !=
1166                                         EXT_LAST_INDEX(path[i].p_hdr))) {
1167                         EXT4_ERROR_INODE(inode,
1168                                          "EXT_MAX_INDEX != EXT_LAST_INDEX ee_block %d!",
1169                                          le32_to_cpu(path[i].p_ext->ee_block));
1170                         err = -EIO;
1171                         goto cleanup;
1172                 }
1173                 /* start copy indexes */
1174                 m = EXT_MAX_INDEX(path[i].p_hdr) - path[i].p_idx++;
1175                 ext_debug("cur 0x%p, last 0x%p\n", path[i].p_idx,
1176                                 EXT_MAX_INDEX(path[i].p_hdr));
1177                 ext4_ext_show_move(inode, path, newblock, i);
1178                 if (m) {
1179                         memmove(++fidx, path[i].p_idx,
1180                                 sizeof(struct ext4_extent_idx) * m);
1181                         le16_add_cpu(&neh->eh_entries, m);
1182                 }
1183                 ext4_extent_block_csum_set(inode, neh);
1184                 set_buffer_uptodate(bh);
1185                 unlock_buffer(bh);
1186
1187                 err = ext4_handle_dirty_metadata(handle, inode, bh);
1188                 if (err)
1189                         goto cleanup;
1190                 brelse(bh);
1191                 bh = NULL;
1192
1193                 /* correct old index */
1194                 if (m) {
1195                         err = ext4_ext_get_access(handle, inode, path + i);
1196                         if (err)
1197                                 goto cleanup;
1198                         le16_add_cpu(&path[i].p_hdr->eh_entries, -m);
1199                         err = ext4_ext_dirty(handle, inode, path + i);
1200                         if (err)
1201                                 goto cleanup;
1202                 }
1203
1204                 i--;
1205         }
1206
1207         /* insert new index */
1208         err = ext4_ext_insert_index(handle, inode, path + at,
1209                                     le32_to_cpu(border), newblock);
1210
1211 cleanup:
1212         if (bh) {
1213                 if (buffer_locked(bh))
1214                         unlock_buffer(bh);
1215                 brelse(bh);
1216         }
1217
1218         if (err) {
1219                 /* free all allocated blocks in error case */
1220                 for (i = 0; i < depth; i++) {
1221                         if (!ablocks[i])
1222                                 continue;
1223                         ext4_free_blocks(handle, inode, NULL, ablocks[i], 1,
1224                                          EXT4_FREE_BLOCKS_METADATA);
1225                 }
1226         }
1227         kfree(ablocks);
1228
1229         return err;
1230 }
1231
1232 /*
1233  * ext4_ext_grow_indepth:
1234  * implements tree growing procedure:
1235  * - allocates new block
1236  * - moves top-level data (index block or leaf) into the new block
1237  * - initializes new top-level, creating index that points to the
1238  *   just created block
1239  */
1240 static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
1241                                  unsigned int flags,
1242                                  struct ext4_extent *newext)
1243 {
1244         struct ext4_extent_header *neh;
1245         struct buffer_head *bh;
1246         ext4_fsblk_t newblock;
1247         int err = 0;
1248
1249         newblock = ext4_ext_new_meta_block(handle, inode, NULL,
1250                 newext, &err, flags);
1251         if (newblock == 0)
1252                 return err;
1253
1254         bh = sb_getblk(inode->i_sb, newblock);
1255         if (unlikely(!bh))
1256                 return -ENOMEM;
1257         lock_buffer(bh);
1258
1259         err = ext4_journal_get_create_access(handle, bh);
1260         if (err) {
1261                 unlock_buffer(bh);
1262                 goto out;
1263         }
1264
1265         /* move top-level index/leaf into new block */
1266         memmove(bh->b_data, EXT4_I(inode)->i_data,
1267                 sizeof(EXT4_I(inode)->i_data));
1268
1269         /* set size of new block */
1270         neh = ext_block_hdr(bh);
1271         /* old root could have indexes or leaves
1272          * so calculate e_max right way */
1273         if (ext_depth(inode))
1274                 neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0));
1275         else
1276                 neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0));
1277         neh->eh_magic = EXT4_EXT_MAGIC;
1278         ext4_extent_block_csum_set(inode, neh);
1279         set_buffer_uptodate(bh);
1280         unlock_buffer(bh);
1281
1282         err = ext4_handle_dirty_metadata(handle, inode, bh);
1283         if (err)
1284                 goto out;
1285
1286         /* Update top-level index: num,max,pointer */
1287         neh = ext_inode_hdr(inode);
1288         neh->eh_entries = cpu_to_le16(1);
1289         ext4_idx_store_pblock(EXT_FIRST_INDEX(neh), newblock);
1290         if (neh->eh_depth == 0) {
1291                 /* Root extent block becomes index block */
1292                 neh->eh_max = cpu_to_le16(ext4_ext_space_root_idx(inode, 0));
1293                 EXT_FIRST_INDEX(neh)->ei_block =
1294                         EXT_FIRST_EXTENT(neh)->ee_block;
1295         }
1296         ext_debug("new root: num %d(%d), lblock %d, ptr %llu\n",
1297                   le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max),
1298                   le32_to_cpu(EXT_FIRST_INDEX(neh)->ei_block),
1299                   ext4_idx_pblock(EXT_FIRST_INDEX(neh)));
1300
1301         le16_add_cpu(&neh->eh_depth, 1);
1302         ext4_mark_inode_dirty(handle, inode);
1303 out:
1304         brelse(bh);
1305
1306         return err;
1307 }
1308
1309 /*
1310  * ext4_ext_create_new_leaf:
1311  * finds empty index and adds new leaf.
1312  * if no free index is found, then it requests in-depth growing.
1313  */
1314 static int ext4_ext_create_new_leaf(handle_t *handle, struct inode *inode,
1315                                     unsigned int mb_flags,
1316                                     unsigned int gb_flags,
1317                                     struct ext4_ext_path *path,
1318                                     struct ext4_extent *newext)
1319 {
1320         struct ext4_ext_path *curp;
1321         int depth, i, err = 0;
1322
1323 repeat:
1324         i = depth = ext_depth(inode);
1325
1326         /* walk up to the tree and look for free index entry */
1327         curp = path + depth;
1328         while (i > 0 && !EXT_HAS_FREE_INDEX(curp)) {
1329                 i--;
1330                 curp--;
1331         }
1332
1333         /* we use already allocated block for index block,
1334          * so subsequent data blocks should be contiguous */
1335         if (EXT_HAS_FREE_INDEX(curp)) {
1336                 /* if we found index with free entry, then use that
1337                  * entry: create all needed subtree and add new leaf */
1338                 err = ext4_ext_split(handle, inode, mb_flags, path, newext, i);
1339                 if (err)
1340                         goto out;
1341
1342                 /* refill path */
1343                 ext4_ext_drop_refs(path);
1344                 path = ext4_ext_find_extent(inode,
1345                                     (ext4_lblk_t)le32_to_cpu(newext->ee_block),
1346                                     path, gb_flags);
1347                 if (IS_ERR(path))
1348                         err = PTR_ERR(path);
1349         } else {
1350                 /* tree is full, time to grow in depth */
1351                 err = ext4_ext_grow_indepth(handle, inode, mb_flags, newext);
1352                 if (err)
1353                         goto out;
1354
1355                 /* refill path */
1356                 ext4_ext_drop_refs(path);
1357                 path = ext4_ext_find_extent(inode,
1358                                    (ext4_lblk_t)le32_to_cpu(newext->ee_block),
1359                                     path, gb_flags);
1360                 if (IS_ERR(path)) {
1361                         err = PTR_ERR(path);
1362                         goto out;
1363                 }
1364
1365                 /*
1366                  * only first (depth 0 -> 1) produces free space;
1367                  * in all other cases we have to split the grown tree
1368                  */
1369                 depth = ext_depth(inode);
1370                 if (path[depth].p_hdr->eh_entries == path[depth].p_hdr->eh_max) {
1371                         /* now we need to split */
1372                         goto repeat;
1373                 }
1374         }
1375
1376 out:
1377         return err;
1378 }
1379
1380 /*
1381  * search the closest allocated block to the left for *logical
1382  * and returns it at @logical + it's physical address at @phys
1383  * if *logical is the smallest allocated block, the function
1384  * returns 0 at @phys
1385  * return value contains 0 (success) or error code
1386  */
1387 static int ext4_ext_search_left(struct inode *inode,
1388                                 struct ext4_ext_path *path,
1389                                 ext4_lblk_t *logical, ext4_fsblk_t *phys)
1390 {
1391         struct ext4_extent_idx *ix;
1392         struct ext4_extent *ex;
1393         int depth, ee_len;
1394
1395         if (unlikely(path == NULL)) {
1396                 EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical);
1397                 return -EIO;
1398         }
1399         depth = path->p_depth;
1400         *phys = 0;
1401
1402         if (depth == 0 && path->p_ext == NULL)
1403                 return 0;
1404
1405         /* usually extent in the path covers blocks smaller
1406          * then *logical, but it can be that extent is the
1407          * first one in the file */
1408
1409         ex = path[depth].p_ext;
1410         ee_len = ext4_ext_get_actual_len(ex);
1411         if (*logical < le32_to_cpu(ex->ee_block)) {
1412                 if (unlikely(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex)) {
1413                         EXT4_ERROR_INODE(inode,
1414                                          "EXT_FIRST_EXTENT != ex *logical %d ee_block %d!",
1415                                          *logical, le32_to_cpu(ex->ee_block));
1416                         return -EIO;
1417                 }
1418                 while (--depth >= 0) {
1419                         ix = path[depth].p_idx;
1420                         if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) {
1421                                 EXT4_ERROR_INODE(inode,
1422                                   "ix (%d) != EXT_FIRST_INDEX (%d) (depth %d)!",
1423                                   ix != NULL ? le32_to_cpu(ix->ei_block) : 0,
1424                                   EXT_FIRST_INDEX(path[depth].p_hdr) != NULL ?
1425                 le32_to_cpu(EXT_FIRST_INDEX(path[depth].p_hdr)->ei_block) : 0,
1426                                   depth);
1427                                 return -EIO;
1428                         }
1429                 }
1430                 return 0;
1431         }
1432
1433         if (unlikely(*logical < (le32_to_cpu(ex->ee_block) + ee_len))) {
1434                 EXT4_ERROR_INODE(inode,
1435                                  "logical %d < ee_block %d + ee_len %d!",
1436                                  *logical, le32_to_cpu(ex->ee_block), ee_len);
1437                 return -EIO;
1438         }
1439
1440         *logical = le32_to_cpu(ex->ee_block) + ee_len - 1;
1441         *phys = ext4_ext_pblock(ex) + ee_len - 1;
1442         return 0;
1443 }
1444
1445 /*
1446  * search the closest allocated block to the right for *logical
1447  * and returns it at @logical + it's physical address at @phys
1448  * if *logical is the largest allocated block, the function
1449  * returns 0 at @phys
1450  * return value contains 0 (success) or error code
1451  */
1452 static int ext4_ext_search_right(struct inode *inode,
1453                                  struct ext4_ext_path *path,
1454                                  ext4_lblk_t *logical, ext4_fsblk_t *phys,
1455                                  struct ext4_extent **ret_ex)
1456 {
1457         struct buffer_head *bh = NULL;
1458         struct ext4_extent_header *eh;
1459         struct ext4_extent_idx *ix;
1460         struct ext4_extent *ex;
1461         ext4_fsblk_t block;
1462         int depth;      /* Note, NOT eh_depth; depth from top of tree */
1463         int ee_len;
1464
1465         if (unlikely(path == NULL)) {
1466                 EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical);
1467                 return -EIO;
1468         }
1469         depth = path->p_depth;
1470         *phys = 0;
1471
1472         if (depth == 0 && path->p_ext == NULL)
1473                 return 0;
1474
1475         /* usually extent in the path covers blocks smaller
1476          * then *logical, but it can be that extent is the
1477          * first one in the file */
1478
1479         ex = path[depth].p_ext;
1480         ee_len = ext4_ext_get_actual_len(ex);
1481         if (*logical < le32_to_cpu(ex->ee_block)) {
1482                 if (unlikely(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex)) {
1483                         EXT4_ERROR_INODE(inode,
1484                                          "first_extent(path[%d].p_hdr) != ex",
1485                                          depth);
1486                         return -EIO;
1487                 }
1488                 while (--depth >= 0) {
1489                         ix = path[depth].p_idx;
1490                         if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) {
1491                                 EXT4_ERROR_INODE(inode,
1492                                                  "ix != EXT_FIRST_INDEX *logical %d!",
1493                                                  *logical);
1494                                 return -EIO;
1495                         }
1496                 }
1497                 goto found_extent;
1498         }
1499
1500         if (unlikely(*logical < (le32_to_cpu(ex->ee_block) + ee_len))) {
1501                 EXT4_ERROR_INODE(inode,
1502                                  "logical %d < ee_block %d + ee_len %d!",
1503                                  *logical, le32_to_cpu(ex->ee_block), ee_len);
1504                 return -EIO;
1505         }
1506
1507         if (ex != EXT_LAST_EXTENT(path[depth].p_hdr)) {
1508                 /* next allocated block in this leaf */
1509                 ex++;
1510                 goto found_extent;
1511         }
1512
1513         /* go up and search for index to the right */
1514         while (--depth >= 0) {
1515                 ix = path[depth].p_idx;
1516                 if (ix != EXT_LAST_INDEX(path[depth].p_hdr))
1517                         goto got_index;
1518         }
1519
1520         /* we've gone up to the root and found no index to the right */
1521         return 0;
1522
1523 got_index:
1524         /* we've found index to the right, let's
1525          * follow it and find the closest allocated
1526          * block to the right */
1527         ix++;
1528         block = ext4_idx_pblock(ix);
1529         while (++depth < path->p_depth) {
1530                 /* subtract from p_depth to get proper eh_depth */
1531                 bh = read_extent_tree_block(inode, block,
1532                                             path->p_depth - depth, 0);
1533                 if (IS_ERR(bh))
1534                         return PTR_ERR(bh);
1535                 eh = ext_block_hdr(bh);
1536                 ix = EXT_FIRST_INDEX(eh);
1537                 block = ext4_idx_pblock(ix);
1538                 put_bh(bh);
1539         }
1540
1541         bh = read_extent_tree_block(inode, block, path->p_depth - depth, 0);
1542         if (IS_ERR(bh))
1543                 return PTR_ERR(bh);
1544         eh = ext_block_hdr(bh);
1545         ex = EXT_FIRST_EXTENT(eh);
1546 found_extent:
1547         *logical = le32_to_cpu(ex->ee_block);
1548         *phys = ext4_ext_pblock(ex);
1549         *ret_ex = ex;
1550         if (bh)
1551                 put_bh(bh);
1552         return 0;
1553 }
1554
1555 /*
1556  * ext4_ext_next_allocated_block:
1557  * returns allocated block in subsequent extent or EXT_MAX_BLOCKS.
1558  * NOTE: it considers block number from index entry as
1559  * allocated block. Thus, index entries have to be consistent
1560  * with leaves.
1561  */
1562 static ext4_lblk_t
1563 ext4_ext_next_allocated_block(struct ext4_ext_path *path)
1564 {
1565         int depth;
1566
1567         BUG_ON(path == NULL);
1568         depth = path->p_depth;
1569
1570         if (depth == 0 && path->p_ext == NULL)
1571                 return EXT_MAX_BLOCKS;
1572
1573         while (depth >= 0) {
1574                 if (depth == path->p_depth) {
1575                         /* leaf */
1576                         if (path[depth].p_ext &&
1577                                 path[depth].p_ext !=
1578                                         EXT_LAST_EXTENT(path[depth].p_hdr))
1579                           return le32_to_cpu(path[depth].p_ext[1].ee_block);
1580                 } else {
1581                         /* index */
1582                         if (path[depth].p_idx !=
1583                                         EXT_LAST_INDEX(path[depth].p_hdr))
1584                           return le32_to_cpu(path[depth].p_idx[1].ei_block);
1585                 }
1586                 depth--;
1587         }
1588
1589         return EXT_MAX_BLOCKS;
1590 }
1591
1592 /*
1593  * ext4_ext_next_leaf_block:
1594  * returns first allocated block from next leaf or EXT_MAX_BLOCKS
1595  */
1596 static ext4_lblk_t ext4_ext_next_leaf_block(struct ext4_ext_path *path)
1597 {
1598         int depth;
1599
1600         BUG_ON(path == NULL);
1601         depth = path->p_depth;
1602
1603         /* zero-tree has no leaf blocks at all */
1604         if (depth == 0)
1605                 return EXT_MAX_BLOCKS;
1606
1607         /* go to index block */
1608         depth--;
1609
1610         while (depth >= 0) {
1611                 if (path[depth].p_idx !=
1612                                 EXT_LAST_INDEX(path[depth].p_hdr))
1613                         return (ext4_lblk_t)
1614                                 le32_to_cpu(path[depth].p_idx[1].ei_block);
1615                 depth--;
1616         }
1617
1618         return EXT_MAX_BLOCKS;
1619 }
1620
1621 /*
1622  * ext4_ext_correct_indexes:
1623  * if leaf gets modified and modified extent is first in the leaf,
1624  * then we have to correct all indexes above.
1625  * TODO: do we need to correct tree in all cases?
1626  */
1627 static int ext4_ext_correct_indexes(handle_t *handle, struct inode *inode,
1628                                 struct ext4_ext_path *path)
1629 {
1630         struct ext4_extent_header *eh;
1631         int depth = ext_depth(inode);
1632         struct ext4_extent *ex;
1633         __le32 border;
1634         int k, err = 0;
1635
1636         eh = path[depth].p_hdr;
1637         ex = path[depth].p_ext;
1638
1639         if (unlikely(ex == NULL || eh == NULL)) {
1640                 EXT4_ERROR_INODE(inode,
1641                                  "ex %p == NULL or eh %p == NULL", ex, eh);
1642                 return -EIO;
1643         }
1644
1645         if (depth == 0) {
1646                 /* there is no tree at all */
1647                 return 0;
1648         }
1649
1650         if (ex != EXT_FIRST_EXTENT(eh)) {
1651                 /* we correct tree if first leaf got modified only */
1652                 return 0;
1653         }
1654
1655         /*
1656          * TODO: we need correction if border is smaller than current one
1657          */
1658         k = depth - 1;
1659         border = path[depth].p_ext->ee_block;
1660         err = ext4_ext_get_access(handle, inode, path + k);
1661         if (err)
1662                 return err;
1663         path[k].p_idx->ei_block = border;
1664         err = ext4_ext_dirty(handle, inode, path + k);
1665         if (err)
1666                 return err;
1667
1668         while (k--) {
1669                 /* change all left-side indexes */
1670                 if (path[k+1].p_idx != EXT_FIRST_INDEX(path[k+1].p_hdr))
1671                         break;
1672                 err = ext4_ext_get_access(handle, inode, path + k);
1673                 if (err)
1674                         break;
1675                 path[k].p_idx->ei_block = border;
1676                 err = ext4_ext_dirty(handle, inode, path + k);
1677                 if (err)
1678                         break;
1679         }
1680
1681         return err;
1682 }
1683
1684 int
1685 ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
1686                                 struct ext4_extent *ex2)
1687 {
1688         unsigned short ext1_ee_len, ext2_ee_len;
1689
1690         /*
1691          * Make sure that both extents are initialized. We don't merge
1692          * unwritten extents so that we can be sure that end_io code has
1693          * the extent that was written properly split out and conversion to
1694          * initialized is trivial.
1695          */
1696         if (ext4_ext_is_unwritten(ex1) != ext4_ext_is_unwritten(ex2))
1697                 return 0;
1698
1699         ext1_ee_len = ext4_ext_get_actual_len(ex1);
1700         ext2_ee_len = ext4_ext_get_actual_len(ex2);
1701
1702         if (le32_to_cpu(ex1->ee_block) + ext1_ee_len !=
1703                         le32_to_cpu(ex2->ee_block))
1704                 return 0;
1705
1706         /*
1707          * To allow future support for preallocated extents to be added
1708          * as an RO_COMPAT feature, refuse to merge to extents if
1709          * this can result in the top bit of ee_len being set.
1710          */
1711         if (ext1_ee_len + ext2_ee_len > EXT_INIT_MAX_LEN)
1712                 return 0;
1713         if (ext4_ext_is_unwritten(ex1) &&
1714             (ext4_test_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN) ||
1715              atomic_read(&EXT4_I(inode)->i_unwritten) ||
1716              (ext1_ee_len + ext2_ee_len > EXT_UNWRITTEN_MAX_LEN)))
1717                 return 0;
1718 #ifdef AGGRESSIVE_TEST
1719         if (ext1_ee_len >= 4)
1720                 return 0;
1721 #endif
1722
1723         if (ext4_ext_pblock(ex1) + ext1_ee_len == ext4_ext_pblock(ex2))
1724                 return 1;
1725         return 0;
1726 }
1727
1728 /*
1729  * This function tries to merge the "ex" extent to the next extent in the tree.
1730  * It always tries to merge towards right. If you want to merge towards
1731  * left, pass "ex - 1" as argument instead of "ex".
1732  * Returns 0 if the extents (ex and ex+1) were _not_ merged and returns
1733  * 1 if they got merged.
1734  */
1735 static int ext4_ext_try_to_merge_right(struct inode *inode,
1736                                  struct ext4_ext_path *path,
1737                                  struct ext4_extent *ex)
1738 {
1739         struct ext4_extent_header *eh;
1740         unsigned int depth, len;
1741         int merge_done = 0, unwritten;
1742
1743         depth = ext_depth(inode);
1744         BUG_ON(path[depth].p_hdr == NULL);
1745         eh = path[depth].p_hdr;
1746
1747         while (ex < EXT_LAST_EXTENT(eh)) {
1748                 if (!ext4_can_extents_be_merged(inode, ex, ex + 1))
1749                         break;
1750                 /* merge with next extent! */
1751                 unwritten = ext4_ext_is_unwritten(ex);
1752                 ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex)
1753                                 + ext4_ext_get_actual_len(ex + 1));
1754                 if (unwritten)
1755                         ext4_ext_mark_unwritten(ex);
1756
1757                 if (ex + 1 < EXT_LAST_EXTENT(eh)) {
1758                         len = (EXT_LAST_EXTENT(eh) - ex - 1)
1759                                 * sizeof(struct ext4_extent);
1760                         memmove(ex + 1, ex + 2, len);
1761                 }
1762                 le16_add_cpu(&eh->eh_entries, -1);
1763                 merge_done = 1;
1764                 WARN_ON(eh->eh_entries == 0);
1765                 if (!eh->eh_entries)
1766                         EXT4_ERROR_INODE(inode, "eh->eh_entries = 0!");
1767         }
1768
1769         return merge_done;
1770 }
1771
1772 /*
1773  * This function does a very simple check to see if we can collapse
1774  * an extent tree with a single extent tree leaf block into the inode.
1775  */
1776 static void ext4_ext_try_to_merge_up(handle_t *handle,
1777                                      struct inode *inode,
1778                                      struct ext4_ext_path *path)
1779 {
1780         size_t s;
1781         unsigned max_root = ext4_ext_space_root(inode, 0);
1782         ext4_fsblk_t blk;
1783
1784         if ((path[0].p_depth != 1) ||
1785             (le16_to_cpu(path[0].p_hdr->eh_entries) != 1) ||
1786             (le16_to_cpu(path[1].p_hdr->eh_entries) > max_root))
1787                 return;
1788
1789         /*
1790          * We need to modify the block allocation bitmap and the block
1791          * group descriptor to release the extent tree block.  If we
1792          * can't get the journal credits, give up.
1793          */
1794         if (ext4_journal_extend(handle, 2))
1795                 return;
1796
1797         /*
1798          * Copy the extent data up to the inode
1799          */
1800         blk = ext4_idx_pblock(path[0].p_idx);
1801         s = le16_to_cpu(path[1].p_hdr->eh_entries) *
1802                 sizeof(struct ext4_extent_idx);
1803         s += sizeof(struct ext4_extent_header);
1804
1805         memcpy(path[0].p_hdr, path[1].p_hdr, s);
1806         path[0].p_depth = 0;
1807         path[0].p_ext = EXT_FIRST_EXTENT(path[0].p_hdr) +
1808                 (path[1].p_ext - EXT_FIRST_EXTENT(path[1].p_hdr));
1809         path[0].p_hdr->eh_max = cpu_to_le16(max_root);
1810
1811         brelse(path[1].p_bh);
1812         ext4_free_blocks(handle, inode, NULL, blk, 1,
1813                          EXT4_FREE_BLOCKS_METADATA | EXT4_FREE_BLOCKS_FORGET);
1814 }
1815
1816 /*
1817  * This function tries to merge the @ex extent to neighbours in the tree.
1818  * return 1 if merge left else 0.
1819  */
1820 static void ext4_ext_try_to_merge(handle_t *handle,
1821                                   struct inode *inode,
1822                                   struct ext4_ext_path *path,
1823                                   struct ext4_extent *ex) {
1824         struct ext4_extent_header *eh;
1825         unsigned int depth;
1826         int merge_done = 0;
1827
1828         depth = ext_depth(inode);
1829         BUG_ON(path[depth].p_hdr == NULL);
1830         eh = path[depth].p_hdr;
1831
1832         if (ex > EXT_FIRST_EXTENT(eh))
1833                 merge_done = ext4_ext_try_to_merge_right(inode, path, ex - 1);
1834
1835         if (!merge_done)
1836                 (void) ext4_ext_try_to_merge_right(inode, path, ex);
1837
1838         ext4_ext_try_to_merge_up(handle, inode, path);
1839 }
1840
1841 /*
1842  * check if a portion of the "newext" extent overlaps with an
1843  * existing extent.
1844  *
1845  * If there is an overlap discovered, it updates the length of the newext
1846  * such that there will be no overlap, and then returns 1.
1847  * If there is no overlap found, it returns 0.
1848  */
1849 static unsigned int ext4_ext_check_overlap(struct ext4_sb_info *sbi,
1850                                            struct inode *inode,
1851                                            struct ext4_extent *newext,
1852                                            struct ext4_ext_path *path)
1853 {
1854         ext4_lblk_t b1, b2;
1855         unsigned int depth, len1;
1856         unsigned int ret = 0;
1857
1858         b1 = le32_to_cpu(newext->ee_block);
1859         len1 = ext4_ext_get_actual_len(newext);
1860         depth = ext_depth(inode);
1861         if (!path[depth].p_ext)
1862                 goto out;
1863         b2 = EXT4_LBLK_CMASK(sbi, le32_to_cpu(path[depth].p_ext->ee_block));
1864
1865         /*
1866          * get the next allocated block if the extent in the path
1867          * is before the requested block(s)
1868          */
1869         if (b2 < b1) {
1870                 b2 = ext4_ext_next_allocated_block(path);
1871                 if (b2 == EXT_MAX_BLOCKS)
1872                         goto out;
1873                 b2 = EXT4_LBLK_CMASK(sbi, b2);
1874         }
1875
1876         /* check for wrap through zero on extent logical start block*/
1877         if (b1 + len1 < b1) {
1878                 len1 = EXT_MAX_BLOCKS - b1;
1879                 newext->ee_len = cpu_to_le16(len1);
1880                 ret = 1;
1881         }
1882
1883         /* check for overlap */
1884         if (b1 + len1 > b2) {
1885                 newext->ee_len = cpu_to_le16(b2 - b1);
1886                 ret = 1;
1887         }
1888 out:
1889         return ret;
1890 }
1891
1892 /*
1893  * ext4_ext_insert_extent:
1894  * tries to merge requsted extent into the existing extent or
1895  * inserts requested extent as new one into the tree,
1896  * creating new leaf in the no-space case.
1897  */
1898 int ext4_ext_insert_extent(handle_t *handle, struct inode *inode,
1899                                 struct ext4_ext_path *path,
1900                                 struct ext4_extent *newext, int gb_flags)
1901 {
1902         struct ext4_extent_header *eh;
1903         struct ext4_extent *ex, *fex;
1904         struct ext4_extent *nearex; /* nearest extent */
1905         struct ext4_ext_path *npath = NULL;
1906         int depth, len, err;
1907         ext4_lblk_t next;
1908         int mb_flags = 0, unwritten;
1909
1910         if (unlikely(ext4_ext_get_actual_len(newext) == 0)) {
1911                 EXT4_ERROR_INODE(inode, "ext4_ext_get_actual_len(newext) == 0");
1912                 return -EIO;
1913         }
1914         depth = ext_depth(inode);
1915         ex = path[depth].p_ext;
1916         eh = path[depth].p_hdr;
1917         if (unlikely(path[depth].p_hdr == NULL)) {
1918                 EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
1919                 return -EIO;
1920         }
1921
1922         /* try to insert block into found extent and return */
1923         if (ex && !(gb_flags & EXT4_GET_BLOCKS_PRE_IO)) {
1924
1925                 /*
1926                  * Try to see whether we should rather test the extent on
1927                  * right from ex, or from the left of ex. This is because
1928                  * ext4_ext_find_extent() can return either extent on the
1929                  * left, or on the right from the searched position. This
1930                  * will make merging more effective.
1931                  */
1932                 if (ex < EXT_LAST_EXTENT(eh) &&
1933                     (le32_to_cpu(ex->ee_block) +
1934                     ext4_ext_get_actual_len(ex) <
1935                     le32_to_cpu(newext->ee_block))) {
1936                         ex += 1;
1937                         goto prepend;
1938                 } else if ((ex > EXT_FIRST_EXTENT(eh)) &&
1939                            (le32_to_cpu(newext->ee_block) +
1940                            ext4_ext_get_actual_len(newext) <
1941                            le32_to_cpu(ex->ee_block)))
1942                         ex -= 1;
1943
1944                 /* Try to append newex to the ex */
1945                 if (ext4_can_extents_be_merged(inode, ex, newext)) {
1946                         ext_debug("append [%d]%d block to %u:[%d]%d"
1947                                   "(from %llu)\n",
1948                                   ext4_ext_is_unwritten(newext),
1949                                   ext4_ext_get_actual_len(newext),
1950                                   le32_to_cpu(ex->ee_block),
1951                                   ext4_ext_is_unwritten(ex),
1952                                   ext4_ext_get_actual_len(ex),
1953                                   ext4_ext_pblock(ex));
1954                         err = ext4_ext_get_access(handle, inode,
1955                                                   path + depth);
1956                         if (err)
1957                                 return err;
1958                         unwritten = ext4_ext_is_unwritten(ex);
1959                         ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex)
1960                                         + ext4_ext_get_actual_len(newext));
1961                         if (unwritten)
1962                                 ext4_ext_mark_unwritten(ex);
1963                         eh = path[depth].p_hdr;
1964                         nearex = ex;
1965                         goto merge;
1966                 }
1967
1968 prepend:
1969                 /* Try to prepend newex to the ex */
1970                 if (ext4_can_extents_be_merged(inode, newext, ex)) {
1971                         ext_debug("prepend %u[%d]%d block to %u:[%d]%d"
1972                                   "(from %llu)\n",
1973                                   le32_to_cpu(newext->ee_block),
1974                                   ext4_ext_is_unwritten(newext),
1975                                   ext4_ext_get_actual_len(newext),
1976                                   le32_to_cpu(ex->ee_block),
1977                                   ext4_ext_is_unwritten(ex),
1978                                   ext4_ext_get_actual_len(ex),
1979                                   ext4_ext_pblock(ex));
1980                         err = ext4_ext_get_access(handle, inode,
1981                                                   path + depth);
1982                         if (err)
1983                                 return err;
1984
1985                         unwritten = ext4_ext_is_unwritten(ex);
1986                         ex->ee_block = newext->ee_block;
1987                         ext4_ext_store_pblock(ex, ext4_ext_pblock(newext));
1988                         ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex)
1989                                         + ext4_ext_get_actual_len(newext));
1990                         if (unwritten)
1991                                 ext4_ext_mark_unwritten(ex);
1992                         eh = path[depth].p_hdr;
1993                         nearex = ex;
1994                         goto merge;
1995                 }
1996         }
1997
1998         depth = ext_depth(inode);
1999         eh = path[depth].p_hdr;
2000         if (le16_to_cpu(eh->eh_entries) < le16_to_cpu(eh->eh_max))
2001                 goto has_space;
2002
2003         /* probably next leaf has space for us? */
2004         fex = EXT_LAST_EXTENT(eh);
2005         next = EXT_MAX_BLOCKS;
2006         if (le32_to_cpu(newext->ee_block) > le32_to_cpu(fex->ee_block))
2007                 next = ext4_ext_next_leaf_block(path);
2008         if (next != EXT_MAX_BLOCKS) {
2009                 ext_debug("next leaf block - %u\n", next);
2010                 BUG_ON(npath != NULL);
2011                 npath = ext4_ext_find_extent(inode, next, NULL, 0);
2012                 if (IS_ERR(npath))
2013                         return PTR_ERR(npath);
2014                 BUG_ON(npath->p_depth != path->p_depth);
2015                 eh = npath[depth].p_hdr;
2016                 if (le16_to_cpu(eh->eh_entries) < le16_to_cpu(eh->eh_max)) {
2017                         ext_debug("next leaf isn't full(%d)\n",
2018                                   le16_to_cpu(eh->eh_entries));
2019                         path = npath;
2020                         goto has_space;
2021                 }
2022                 ext_debug("next leaf has no free space(%d,%d)\n",
2023                           le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
2024         }
2025
2026         /*
2027          * There is no free space in the found leaf.
2028          * We're gonna add a new leaf in the tree.
2029          */
2030         if (gb_flags & EXT4_GET_BLOCKS_METADATA_NOFAIL)
2031                 mb_flags = EXT4_MB_USE_RESERVED;
2032         err = ext4_ext_create_new_leaf(handle, inode, mb_flags, gb_flags,
2033                                        path, newext);
2034         if (err)
2035                 goto cleanup;
2036         depth = ext_depth(inode);
2037         eh = path[depth].p_hdr;
2038
2039 has_space:
2040         nearex = path[depth].p_ext;
2041
2042         err = ext4_ext_get_access(handle, inode, path + depth);
2043         if (err)
2044                 goto cleanup;
2045
2046         if (!nearex) {
2047                 /* there is no extent in this leaf, create first one */
2048                 ext_debug("first extent in the leaf: %u:%llu:[%d]%d\n",
2049                                 le32_to_cpu(newext->ee_block),
2050                                 ext4_ext_pblock(newext),
2051                                 ext4_ext_is_unwritten(newext),
2052                                 ext4_ext_get_actual_len(newext));
2053                 nearex = EXT_FIRST_EXTENT(eh);
2054         } else {
2055                 if (le32_to_cpu(newext->ee_block)
2056                            > le32_to_cpu(nearex->ee_block)) {
2057                         /* Insert after */
2058                         ext_debug("insert %u:%llu:[%d]%d before: "
2059                                         "nearest %p\n",
2060                                         le32_to_cpu(newext->ee_block),
2061                                         ext4_ext_pblock(newext),
2062                                         ext4_ext_is_unwritten(newext),
2063                                         ext4_ext_get_actual_len(newext),
2064                                         nearex);
2065                         nearex++;
2066                 } else {
2067                         /* Insert before */
2068                         BUG_ON(newext->ee_block == nearex->ee_block);
2069                         ext_debug("insert %u:%llu:[%d]%d after: "
2070                                         "nearest %p\n",
2071                                         le32_to_cpu(newext->ee_block),
2072                                         ext4_ext_pblock(newext),
2073                                         ext4_ext_is_unwritten(newext),
2074                                         ext4_ext_get_actual_len(newext),
2075                                         nearex);
2076                 }
2077                 len = EXT_LAST_EXTENT(eh) - nearex + 1;
2078                 if (len > 0) {
2079                         ext_debug("insert %u:%llu:[%d]%d: "
2080                                         "move %d extents from 0x%p to 0x%p\n",
2081                                         le32_to_cpu(newext->ee_block),
2082                                         ext4_ext_pblock(newext),
2083                                         ext4_ext_is_unwritten(newext),
2084                                         ext4_ext_get_actual_len(newext),
2085                                         len, nearex, nearex + 1);
2086                         memmove(nearex + 1, nearex,
2087                                 len * sizeof(struct ext4_extent));
2088                 }
2089         }
2090
2091         le16_add_cpu(&eh->eh_entries, 1);
2092         path[depth].p_ext = nearex;
2093         nearex->ee_block = newext->ee_block;
2094         ext4_ext_store_pblock(nearex, ext4_ext_pblock(newext));
2095         nearex->ee_len = newext->ee_len;
2096
2097 merge:
2098         /* try to merge extents */
2099         if (!(gb_flags & EXT4_GET_BLOCKS_PRE_IO))
2100                 ext4_ext_try_to_merge(handle, inode, path, nearex);
2101
2102
2103         /* time to correct all indexes above */
2104         err = ext4_ext_correct_indexes(handle, inode, path);
2105         if (err)
2106                 goto cleanup;
2107
2108         err = ext4_ext_dirty(handle, inode, path + path->p_depth);
2109
2110 cleanup:
2111         if (npath) {
2112                 ext4_ext_drop_refs(npath);
2113                 kfree(npath);
2114         }
2115         return err;
2116 }
2117
2118 static int ext4_fill_fiemap_extents(struct inode *inode,
2119                                     ext4_lblk_t block, ext4_lblk_t num,
2120                                     struct fiemap_extent_info *fieinfo)
2121 {
2122         struct ext4_ext_path *path = NULL;
2123         struct ext4_extent *ex;
2124         struct extent_status es;
2125         ext4_lblk_t next, next_del, start = 0, end = 0;
2126         ext4_lblk_t last = block + num;
2127         int exists, depth = 0, err = 0;
2128         unsigned int flags = 0;
2129         unsigned char blksize_bits = inode->i_sb->s_blocksize_bits;
2130
2131         while (block < last && block != EXT_MAX_BLOCKS) {
2132                 num = last - block;
2133                 /* find extent for this block */
2134                 down_read(&EXT4_I(inode)->i_data_sem);
2135
2136                 if (path && ext_depth(inode) != depth) {
2137                         /* depth was changed. we have to realloc path */
2138                         kfree(path);
2139                         path = NULL;
2140                 }
2141
2142                 path = ext4_ext_find_extent(inode, block, path, 0);
2143                 if (IS_ERR(path)) {
2144                         up_read(&EXT4_I(inode)->i_data_sem);
2145                         err = PTR_ERR(path);
2146                         path = NULL;
2147                         break;
2148                 }
2149
2150                 depth = ext_depth(inode);
2151                 if (unlikely(path[depth].p_hdr == NULL)) {
2152                         up_read(&EXT4_I(inode)->i_data_sem);
2153                         EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
2154                         err = -EIO;
2155                         break;
2156                 }
2157                 ex = path[depth].p_ext;
2158                 next = ext4_ext_next_allocated_block(path);
2159                 ext4_ext_drop_refs(path);
2160
2161                 flags = 0;
2162                 exists = 0;
2163                 if (!ex) {
2164                         /* there is no extent yet, so try to allocate
2165                          * all requested space */
2166                         start = block;
2167                         end = block + num;
2168                 } else if (le32_to_cpu(ex->ee_block) > block) {
2169                         /* need to allocate space before found extent */
2170                         start = block;
2171                         end = le32_to_cpu(ex->ee_block);
2172                         if (block + num < end)
2173                                 end = block + num;
2174                 } else if (block >= le32_to_cpu(ex->ee_block)
2175                                         + ext4_ext_get_actual_len(ex)) {
2176                         /* need to allocate space after found extent */
2177                         start = block;
2178                         end = block + num;
2179                         if (end >= next)
2180                                 end = next;
2181                 } else if (block >= le32_to_cpu(ex->ee_block)) {
2182                         /*
2183                          * some part of requested space is covered
2184                          * by found extent
2185                          */
2186                         start = block;
2187                         end = le32_to_cpu(ex->ee_block)
2188                                 + ext4_ext_get_actual_len(ex);
2189                         if (block + num < end)
2190                                 end = block + num;
2191                         exists = 1;
2192                 } else {
2193                         BUG();
2194                 }
2195                 BUG_ON(end <= start);
2196
2197                 if (!exists) {
2198                         es.es_lblk = start;
2199                         es.es_len = end - start;
2200                         es.es_pblk = 0;
2201                 } else {
2202                         es.es_lblk = le32_to_cpu(ex->ee_block);
2203                         es.es_len = ext4_ext_get_actual_len(ex);
2204                         es.es_pblk = ext4_ext_pblock(ex);
2205                         if (ext4_ext_is_unwritten(ex))
2206                                 flags |= FIEMAP_EXTENT_UNWRITTEN;
2207                 }
2208
2209                 /*
2210                  * Find delayed extent and update es accordingly. We call
2211                  * it even in !exists case to find out whether es is the
2212                  * last existing extent or not.
2213                  */
2214                 next_del = ext4_find_delayed_extent(inode, &es);
2215                 if (!exists && next_del) {
2216                         exists = 1;
2217                         flags |= (FIEMAP_EXTENT_DELALLOC |
2218                                   FIEMAP_EXTENT_UNKNOWN);
2219                 }
2220                 up_read(&EXT4_I(inode)->i_data_sem);
2221
2222                 if (unlikely(es.es_len == 0)) {
2223                         EXT4_ERROR_INODE(inode, "es.es_len == 0");
2224                         err = -EIO;
2225                         break;
2226                 }
2227
2228                 /*
2229                  * This is possible iff next == next_del == EXT_MAX_BLOCKS.
2230                  * we need to check next == EXT_MAX_BLOCKS because it is
2231                  * possible that an extent is with unwritten and delayed
2232                  * status due to when an extent is delayed allocated and
2233                  * is allocated by fallocate status tree will track both of
2234                  * them in a extent.
2235                  *
2236                  * So we could return a unwritten and delayed extent, and
2237                  * its block is equal to 'next'.
2238                  */
2239                 if (next == next_del && next == EXT_MAX_BLOCKS) {
2240                         flags |= FIEMAP_EXTENT_LAST;
2241                         if (unlikely(next_del != EXT_MAX_BLOCKS ||
2242                                      next != EXT_MAX_BLOCKS)) {
2243                                 EXT4_ERROR_INODE(inode,
2244                                                  "next extent == %u, next "
2245                                                  "delalloc extent = %u",
2246                                                  next, next_del);
2247                                 err = -EIO;
2248                                 break;
2249                         }
2250                 }
2251
2252                 if (exists) {
2253                         err = fiemap_fill_next_extent(fieinfo,
2254                                 (__u64)es.es_lblk << blksize_bits,
2255                                 (__u64)es.es_pblk << blksize_bits,
2256                                 (__u64)es.es_len << blksize_bits,
2257                                 flags);
2258                         if (err < 0)
2259                                 break;
2260                         if (err == 1) {
2261                                 err = 0;
2262                                 break;
2263                         }
2264                 }
2265
2266                 block = es.es_lblk + es.es_len;
2267         }
2268
2269         if (path) {
2270                 ext4_ext_drop_refs(path);
2271                 kfree(path);
2272         }
2273
2274         return err;
2275 }
2276
2277 /*
2278  * ext4_ext_put_gap_in_cache:
2279  * calculate boundaries of the gap that the requested block fits into
2280  * and cache this gap
2281  */
2282 static void
2283 ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path,
2284                                 ext4_lblk_t block)
2285 {
2286         int depth = ext_depth(inode);
2287         unsigned long len = 0;
2288         ext4_lblk_t lblock = 0;
2289         struct ext4_extent *ex;
2290
2291         ex = path[depth].p_ext;
2292         if (ex == NULL) {
2293                 /*
2294                  * there is no extent yet, so gap is [0;-] and we
2295                  * don't cache it
2296                  */
2297                 ext_debug("cache gap(whole file):");
2298         } else if (block < le32_to_cpu(ex->ee_block)) {
2299                 lblock = block;
2300                 len = le32_to_cpu(ex->ee_block) - block;
2301                 ext_debug("cache gap(before): %u [%u:%u]",
2302                                 block,
2303                                 le32_to_cpu(ex->ee_block),
2304                                  ext4_ext_get_actual_len(ex));
2305                 if (!ext4_find_delalloc_range(inode, lblock, lblock + len - 1))
2306                         ext4_es_insert_extent(inode, lblock, len, ~0,
2307                                               EXTENT_STATUS_HOLE);
2308         } else if (block >= le32_to_cpu(ex->ee_block)
2309                         + ext4_ext_get_actual_len(ex)) {
2310                 ext4_lblk_t next;
2311                 lblock = le32_to_cpu(ex->ee_block)
2312                         + ext4_ext_get_actual_len(ex);
2313
2314                 next = ext4_ext_next_allocated_block(path);
2315                 ext_debug("cache gap(after): [%u:%u] %u",
2316                                 le32_to_cpu(ex->ee_block),
2317                                 ext4_ext_get_actual_len(ex),
2318                                 block);
2319                 BUG_ON(next == lblock);
2320                 len = next - lblock;
2321                 if (!ext4_find_delalloc_range(inode, lblock, lblock + len - 1))
2322                         ext4_es_insert_extent(inode, lblock, len, ~0,
2323                                               EXTENT_STATUS_HOLE);
2324         } else {
2325                 BUG();
2326         }
2327
2328         ext_debug(" -> %u:%lu\n", lblock, len);
2329 }
2330
2331 /*
2332  * ext4_ext_rm_idx:
2333  * removes index from the index block.
2334  */
2335 static int ext4_ext_rm_idx(handle_t *handle, struct inode *inode,
2336                         struct ext4_ext_path *path, int depth)
2337 {
2338         int err;
2339         ext4_fsblk_t leaf;
2340
2341         /* free index block */
2342         depth--;
2343         path = path + depth;
2344         leaf = ext4_idx_pblock(path->p_idx);
2345         if (unlikely(path->p_hdr->eh_entries == 0)) {
2346                 EXT4_ERROR_INODE(inode, "path->p_hdr->eh_entries == 0");
2347                 return -EIO;
2348         }
2349         err = ext4_ext_get_access(handle, inode, path);
2350         if (err)
2351                 return err;
2352
2353         if (path->p_idx != EXT_LAST_INDEX(path->p_hdr)) {
2354                 int len = EXT_LAST_INDEX(path->p_hdr) - path->p_idx;
2355                 len *= sizeof(struct ext4_extent_idx);
2356                 memmove(path->p_idx, path->p_idx + 1, len);
2357         }
2358
2359         le16_add_cpu(&path->p_hdr->eh_entries, -1);
2360         err = ext4_ext_dirty(handle, inode, path);
2361         if (err)
2362                 return err;
2363         ext_debug("index is empty, remove it, free block %llu\n", leaf);
2364         trace_ext4_ext_rm_idx(inode, leaf);
2365
2366         ext4_free_blocks(handle, inode, NULL, leaf, 1,
2367                          EXT4_FREE_BLOCKS_METADATA | EXT4_FREE_BLOCKS_FORGET);
2368
2369         while (--depth >= 0) {
2370                 if (path->p_idx != EXT_FIRST_INDEX(path->p_hdr))
2371                         break;
2372                 path--;
2373                 err = ext4_ext_get_access(handle, inode, path);
2374                 if (err)
2375                         break;
2376                 path->p_idx->ei_block = (path+1)->p_idx->ei_block;
2377                 err = ext4_ext_dirty(handle, inode, path);
2378                 if (err)
2379                         break;
2380         }
2381         return err;
2382 }
2383
2384 /*
2385  * ext4_ext_calc_credits_for_single_extent:
2386  * This routine returns max. credits that needed to insert an extent
2387  * to the extent tree.
2388  * When pass the actual path, the caller should calculate credits
2389  * under i_data_sem.
2390  */
2391 int ext4_ext_calc_credits_for_single_extent(struct inode *inode, int nrblocks,
2392                                                 struct ext4_ext_path *path)
2393 {
2394         if (path) {
2395                 int depth = ext_depth(inode);
2396                 int ret = 0;
2397
2398                 /* probably there is space in leaf? */
2399                 if (le16_to_cpu(path[depth].p_hdr->eh_entries)
2400                                 < le16_to_cpu(path[depth].p_hdr->eh_max)) {
2401
2402                         /*
2403                          *  There are some space in the leaf tree, no
2404                          *  need to account for leaf block credit
2405                          *
2406                          *  bitmaps and block group descriptor blocks
2407                          *  and other metadata blocks still need to be
2408                          *  accounted.
2409                          */
2410                         /* 1 bitmap, 1 block group descriptor */
2411                         ret = 2 + EXT4_META_TRANS_BLOCKS(inode->i_sb);
2412                         return ret;
2413                 }
2414         }
2415
2416         return ext4_chunk_trans_blocks(inode, nrblocks);
2417 }
2418
2419 /*
2420  * How many index/leaf blocks need to change/allocate to add @extents extents?
2421  *
2422  * If we add a single extent, then in the worse case, each tree level
2423  * index/leaf need to be changed in case of the tree split.
2424  *
2425  * If more extents are inserted, they could cause the whole tree split more
2426  * than once, but this is really rare.
2427  */
2428 int ext4_ext_index_trans_blocks(struct inode *inode, int extents)
2429 {
2430         int index;
2431         int depth;
2432
2433         /* If we are converting the inline data, only one is needed here. */
2434         if (ext4_has_inline_data(inode))
2435                 return 1;
2436
2437         depth = ext_depth(inode);
2438
2439         if (extents <= 1)
2440                 index = depth * 2;
2441         else
2442                 index = depth * 3;
2443
2444         return index;
2445 }
2446
2447 static inline int get_default_free_blocks_flags(struct inode *inode)
2448 {
2449         if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
2450                 return EXT4_FREE_BLOCKS_METADATA | EXT4_FREE_BLOCKS_FORGET;
2451         else if (ext4_should_journal_data(inode))
2452                 return EXT4_FREE_BLOCKS_FORGET;
2453         return 0;
2454 }
2455
2456 static int ext4_remove_blocks(handle_t *handle, struct inode *inode,
2457                               struct ext4_extent *ex,
2458                               long long *partial_cluster,
2459                               ext4_lblk_t from, ext4_lblk_t to)
2460 {
2461         struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
2462         unsigned short ee_len =  ext4_ext_get_actual_len(ex);
2463         ext4_fsblk_t pblk;
2464         int flags = get_default_free_blocks_flags(inode);
2465
2466         /*
2467          * For bigalloc file systems, we never free a partial cluster
2468          * at the beginning of the extent.  Instead, we make a note
2469          * that we tried freeing the cluster, and check to see if we
2470          * need to free it on a subsequent call to ext4_remove_blocks,
2471          * or at the end of the ext4_truncate() operation.
2472          */
2473         flags |= EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER;
2474
2475         trace_ext4_remove_blocks(inode, ex, from, to, *partial_cluster);
2476         /*
2477          * If we have a partial cluster, and it's different from the
2478          * cluster of the last block, we need to explicitly free the
2479          * partial cluster here.
2480          */
2481         pblk = ext4_ext_pblock(ex) + ee_len - 1;
2482         if ((*partial_cluster > 0) &&
2483             (EXT4_B2C(sbi, pblk) != *partial_cluster)) {
2484                 ext4_free_blocks(handle, inode, NULL,
2485                                  EXT4_C2B(sbi, *partial_cluster),
2486                                  sbi->s_cluster_ratio, flags);
2487                 *partial_cluster = 0;
2488         }
2489
2490 #ifdef EXTENTS_STATS
2491         {
2492                 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
2493                 spin_lock(&sbi->s_ext_stats_lock);
2494                 sbi->s_ext_blocks += ee_len;
2495                 sbi->s_ext_extents++;
2496                 if (ee_len < sbi->s_ext_min)
2497                         sbi->s_ext_min = ee_len;
2498                 if (ee_len > sbi->s_ext_max)
2499                         sbi->s_ext_max = ee_len;
2500                 if (ext_depth(inode) > sbi->s_depth_max)
2501                         sbi->s_depth_max = ext_depth(inode);
2502                 spin_unlock(&sbi->s_ext_stats_lock);
2503         }
2504 #endif
2505         if (from >= le32_to_cpu(ex->ee_block)
2506             && to == le32_to_cpu(ex->ee_block) + ee_len - 1) {
2507                 /* tail removal */
2508                 ext4_lblk_t num;
2509                 unsigned int unaligned;
2510
2511                 num = le32_to_cpu(ex->ee_block) + ee_len - from;
2512                 pblk = ext4_ext_pblock(ex) + ee_len - num;
2513                 /*
2514                  * Usually we want to free partial cluster at the end of the
2515                  * extent, except for the situation when the cluster is still
2516                  * used by any other extent (partial_cluster is negative).
2517                  */
2518                 if (*partial_cluster < 0 &&
2519                     -(*partial_cluster) == EXT4_B2C(sbi, pblk + num - 1))
2520                         flags |= EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER;
2521
2522                 ext_debug("free last %u blocks starting %llu partial %lld\n",
2523                           num, pblk, *partial_cluster);
2524                 ext4_free_blocks(handle, inode, NULL, pblk, num, flags);
2525                 /*
2526                  * If the block range to be freed didn't start at the
2527                  * beginning of a cluster, and we removed the entire
2528                  * extent and the cluster is not used by any other extent,
2529                  * save the partial cluster here, since we might need to
2530                  * delete if we determine that the truncate operation has
2531                  * removed all of the blocks in the cluster.
2532                  *
2533                  * On the other hand, if we did not manage to free the whole
2534                  * extent, we have to mark the cluster as used (store negative
2535                  * cluster number in partial_cluster).
2536                  */
2537                 unaligned = EXT4_PBLK_COFF(sbi, pblk);
2538                 if (unaligned && (ee_len == num) &&
2539                     (*partial_cluster != -((long long)EXT4_B2C(sbi, pblk))))
2540                         *partial_cluster = EXT4_B2C(sbi, pblk);
2541                 else if (unaligned)
2542                         *partial_cluster = -((long long)EXT4_B2C(sbi, pblk));
2543                 else if (*partial_cluster > 0)
2544                         *partial_cluster = 0;
2545         } else
2546                 ext4_error(sbi->s_sb, "strange request: removal(2) "
2547                            "%u-%u from %u:%u\n",
2548                            from, to, le32_to_cpu(ex->ee_block), ee_len);
2549         return 0;
2550 }
2551
2552
2553 /*
2554  * ext4_ext_rm_leaf() Removes the extents associated with the
2555  * blocks appearing between "start" and "end", and splits the extents
2556  * if "start" and "end" appear in the same extent
2557  *
2558  * @handle: The journal handle
2559  * @inode:  The files inode
2560  * @path:   The path to the leaf
2561  * @partial_cluster: The cluster which we'll have to free if all extents
2562  *                   has been released from it. It gets negative in case
2563  *                   that the cluster is still used.
2564  * @start:  The first block to remove
2565  * @end:   The last block to remove
2566  */
2567 static int
2568 ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
2569                  struct ext4_ext_path *path,
2570                  long long *partial_cluster,
2571                  ext4_lblk_t start, ext4_lblk_t end)
2572 {
2573         struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
2574         int err = 0, correct_index = 0;
2575         int depth = ext_depth(inode), credits;
2576         struct ext4_extent_header *eh;
2577         ext4_lblk_t a, b;
2578         unsigned num;
2579         ext4_lblk_t ex_ee_block;
2580         unsigned short ex_ee_len;
2581         unsigned unwritten = 0;
2582         struct ext4_extent *ex;
2583         ext4_fsblk_t pblk;
2584
2585         /* the header must be checked already in ext4_ext_remove_space() */
2586         ext_debug("truncate since %u in leaf to %u\n", start, end);
2587         if (!path[depth].p_hdr)
2588                 path[depth].p_hdr = ext_block_hdr(path[depth].p_bh);
2589         eh = path[depth].p_hdr;
2590         if (unlikely(path[depth].p_hdr == NULL)) {
2591                 EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
2592                 return -EIO;
2593         }
2594         /* find where to start removing */
2595         ex = path[depth].p_ext;
2596         if (!ex)
2597                 ex = EXT_LAST_EXTENT(eh);
2598
2599         ex_ee_block = le32_to_cpu(ex->ee_block);
2600         ex_ee_len = ext4_ext_get_actual_len(ex);
2601
2602         /*
2603          * If we're starting with an extent other than the last one in the
2604          * node, we need to see if it shares a cluster with the extent to
2605          * the right (towards the end of the file). If its leftmost cluster
2606          * is this extent's rightmost cluster and it is not cluster aligned,
2607          * we'll mark it as a partial that is not to be deallocated.
2608          */
2609
2610         if (ex != EXT_LAST_EXTENT(eh)) {
2611                 ext4_fsblk_t current_pblk, right_pblk;
2612                 long long current_cluster, right_cluster;
2613
2614                 current_pblk = ext4_ext_pblock(ex) + ex_ee_len - 1;
2615                 current_cluster = (long long)EXT4_B2C(sbi, current_pblk);
2616                 right_pblk = ext4_ext_pblock(ex + 1);
2617                 right_cluster = (long long)EXT4_B2C(sbi, right_pblk);
2618                 if (current_cluster == right_cluster &&
2619                         EXT4_PBLK_COFF(sbi, right_pblk))
2620                         *partial_cluster = -right_cluster;
2621         }
2622
2623         trace_ext4_ext_rm_leaf(inode, start, ex, *partial_cluster);
2624
2625         while (ex >= EXT_FIRST_EXTENT(eh) &&
2626                         ex_ee_block + ex_ee_len > start) {
2627
2628                 if (ext4_ext_is_unwritten(ex))
2629                         unwritten = 1;
2630                 else
2631                         unwritten = 0;
2632
2633                 ext_debug("remove ext %u:[%d]%d\n", ex_ee_block,
2634                           unwritten, ex_ee_len);
2635                 path[depth].p_ext = ex;
2636
2637                 a = ex_ee_block > start ? ex_ee_block : start;
2638                 b = ex_ee_block+ex_ee_len - 1 < end ?
2639                         ex_ee_block+ex_ee_len - 1 : end;
2640
2641                 ext_debug("  border %u:%u\n", a, b);
2642
2643                 /* If this extent is beyond the end of the hole, skip it */
2644                 if (end < ex_ee_block) {
2645                         /*
2646                          * We're going to skip this extent and move to another,
2647                          * so if this extent is not cluster aligned we have
2648                          * to mark the current cluster as used to avoid
2649                          * accidentally freeing it later on
2650                          */
2651                         pblk = ext4_ext_pblock(ex);
2652                         if (EXT4_PBLK_COFF(sbi, pblk))
2653                                 *partial_cluster =
2654                                         -((long long)EXT4_B2C(sbi, pblk));
2655                         ex--;
2656                         ex_ee_block = le32_to_cpu(ex->ee_block);
2657                         ex_ee_len = ext4_ext_get_actual_len(ex);
2658                         continue;
2659                 } else if (b != ex_ee_block + ex_ee_len - 1) {
2660                         EXT4_ERROR_INODE(inode,
2661                                          "can not handle truncate %u:%u "
2662                                          "on extent %u:%u",
2663                                          start, end, ex_ee_block,
2664                                          ex_ee_block + ex_ee_len - 1);
2665                         err = -EIO;
2666                         goto out;
2667                 } else if (a != ex_ee_block) {
2668                         /* remove tail of the extent */
2669                         num = a - ex_ee_block;
2670                 } else {
2671                         /* remove whole extent: excellent! */
2672                         num = 0;
2673                 }
2674                 /*
2675                  * 3 for leaf, sb, and inode plus 2 (bmap and group
2676                  * descriptor) for each block group; assume two block
2677                  * groups plus ex_ee_len/blocks_per_block_group for
2678                  * the worst case
2679                  */
2680                 credits = 7 + 2*(ex_ee_len/EXT4_BLOCKS_PER_GROUP(inode->i_sb));
2681                 if (ex == EXT_FIRST_EXTENT(eh)) {
2682                         correct_index = 1;
2683                         credits += (ext_depth(inode)) + 1;
2684                 }
2685                 credits += EXT4_MAXQUOTAS_TRANS_BLOCKS(inode->i_sb);
2686
2687                 err = ext4_ext_truncate_extend_restart(handle, inode, credits);
2688                 if (err)
2689                         goto out;
2690
2691                 err = ext4_ext_get_access(handle, inode, path + depth);
2692                 if (err)
2693                         goto out;
2694
2695                 err = ext4_remove_blocks(handle, inode, ex, partial_cluster,
2696                                          a, b);
2697                 if (err)
2698                         goto out;
2699
2700                 if (num == 0)
2701                         /* this extent is removed; mark slot entirely unused */
2702                         ext4_ext_store_pblock(ex, 0);
2703
2704                 ex->ee_len = cpu_to_le16(num);
2705                 /*
2706                  * Do not mark unwritten if all the blocks in the
2707                  * extent have been removed.
2708                  */
2709                 if (unwritten && num)
2710                         ext4_ext_mark_unwritten(ex);
2711                 /*
2712                  * If the extent was completely released,
2713                  * we need to remove it from the leaf
2714                  */
2715                 if (num == 0) {
2716                         if (end != EXT_MAX_BLOCKS - 1) {
2717                                 /*
2718                                  * For hole punching, we need to scoot all the
2719                                  * extents up when an extent is removed so that
2720                                  * we dont have blank extents in the middle
2721                                  */
2722                                 memmove(ex, ex+1, (EXT_LAST_EXTENT(eh) - ex) *
2723                                         sizeof(struct ext4_extent));
2724
2725                                 /* Now get rid of the one at the end */
2726                                 memset(EXT_LAST_EXTENT(eh), 0,
2727                                         sizeof(struct ext4_extent));
2728                         }
2729                         le16_add_cpu(&eh->eh_entries, -1);
2730                 } else if (*partial_cluster > 0)
2731                         *partial_cluster = 0;
2732
2733                 err = ext4_ext_dirty(handle, inode, path + depth);
2734                 if (err)
2735                         goto out;
2736
2737                 ext_debug("new extent: %u:%u:%llu\n", ex_ee_block, num,
2738                                 ext4_ext_pblock(ex));
2739                 ex--;
2740                 ex_ee_block = le32_to_cpu(ex->ee_block);
2741                 ex_ee_len = ext4_ext_get_actual_len(ex);
2742         }
2743
2744         if (correct_index && eh->eh_entries)
2745                 err = ext4_ext_correct_indexes(handle, inode, path);
2746
2747         /*
2748          * If there's a partial cluster and at least one extent remains in
2749          * the leaf, free the partial cluster if it isn't shared with the
2750          * current extent.  If there's a partial cluster and no extents
2751          * remain in the leaf, it can't be freed here.  It can only be
2752          * freed when it's possible to determine if it's not shared with
2753          * any other extent - when the next leaf is processed or when space
2754          * removal is complete.
2755          */
2756         if (*partial_cluster > 0 && eh->eh_entries &&
2757             (EXT4_B2C(sbi, ext4_ext_pblock(ex) + ex_ee_len - 1) !=
2758              *partial_cluster)) {
2759                 int flags = get_default_free_blocks_flags(inode);
2760
2761                 ext4_free_blocks(handle, inode, NULL,
2762                                  EXT4_C2B(sbi, *partial_cluster),
2763                                  sbi->s_cluster_ratio, flags);
2764                 *partial_cluster = 0;
2765         }
2766
2767         /* if this leaf is free, then we should
2768          * remove it from index block above */
2769         if (err == 0 && eh->eh_entries == 0 && path[depth].p_bh != NULL)
2770                 err = ext4_ext_rm_idx(handle, inode, path, depth);
2771
2772 out:
2773         return err;
2774 }
2775
2776 /*
2777  * ext4_ext_more_to_rm:
2778  * returns 1 if current index has to be freed (even partial)
2779  */
2780 static int
2781 ext4_ext_more_to_rm(struct ext4_ext_path *path)
2782 {
2783         BUG_ON(path->p_idx == NULL);
2784
2785         if (path->p_idx < EXT_FIRST_INDEX(path->p_hdr))
2786                 return 0;
2787
2788         /*
2789          * if truncate on deeper level happened, it wasn't partial,
2790          * so we have to consider current index for truncation
2791          */
2792         if (le16_to_cpu(path->p_hdr->eh_entries) == path->p_block)
2793                 return 0;
2794         return 1;
2795 }
2796
2797 int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
2798                           ext4_lblk_t end)
2799 {
2800         struct super_block *sb = inode->i_sb;
2801         int depth = ext_depth(inode);
2802         struct ext4_ext_path *path = NULL;
2803         long long partial_cluster = 0;
2804         handle_t *handle;
2805         int i = 0, err = 0;
2806
2807         ext_debug("truncate since %u to %u\n", start, end);
2808
2809         /* probably first extent we're gonna free will be last in block */
2810         handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE, depth + 1);
2811         if (IS_ERR(handle))
2812                 return PTR_ERR(handle);
2813
2814 again:
2815         trace_ext4_ext_remove_space(inode, start, end, depth);
2816
2817         /*
2818          * Check if we are removing extents inside the extent tree. If that
2819          * is the case, we are going to punch a hole inside the extent tree
2820          * so we have to check whether we need to split the extent covering
2821          * the last block to remove so we can easily remove the part of it
2822          * in ext4_ext_rm_leaf().
2823          */
2824         if (end < EXT_MAX_BLOCKS - 1) {
2825                 struct ext4_extent *ex;
2826                 ext4_lblk_t ee_block;
2827
2828                 /* find extent for this block */
2829                 path = ext4_ext_find_extent(inode, end, NULL, EXT4_EX_NOCACHE);
2830                 if (IS_ERR(path)) {
2831                         ext4_journal_stop(handle);
2832                         return PTR_ERR(path);
2833                 }
2834                 depth = ext_depth(inode);
2835                 /* Leaf not may not exist only if inode has no blocks at all */
2836                 ex = path[depth].p_ext;
2837                 if (!ex) {
2838                         if (depth) {
2839                                 EXT4_ERROR_INODE(inode,
2840                                                  "path[%d].p_hdr == NULL",
2841                                                  depth);
2842                                 err = -EIO;
2843                         }
2844                         goto out;
2845                 }
2846
2847                 ee_block = le32_to_cpu(ex->ee_block);
2848
2849                 /*
2850                  * See if the last block is inside the extent, if so split
2851                  * the extent at 'end' block so we can easily remove the
2852                  * tail of the first part of the split extent in
2853                  * ext4_ext_rm_leaf().
2854                  */
2855                 if (end >= ee_block &&
2856                     end < ee_block + ext4_ext_get_actual_len(ex) - 1) {
2857                         int split_flag = 0;
2858
2859                         if (ext4_ext_is_unwritten(ex))
2860                                 split_flag = EXT4_EXT_MARK_UNWRIT1 |
2861                                              EXT4_EXT_MARK_UNWRIT2;
2862
2863                         /*
2864                          * Split the extent in two so that 'end' is the last
2865                          * block in the first new extent. Also we should not
2866                          * fail removing space due to ENOSPC so try to use
2867                          * reserved block if that happens.
2868                          */
2869                         err = ext4_split_extent_at(handle, inode, path,
2870                                         end + 1, split_flag,
2871                                         EXT4_EX_NOCACHE |
2872                                         EXT4_GET_BLOCKS_PRE_IO |
2873                                         EXT4_GET_BLOCKS_METADATA_NOFAIL);
2874
2875                         if (err < 0)
2876                                 goto out;
2877                 }
2878         }
2879         /*
2880          * We start scanning from right side, freeing all the blocks
2881          * after i_size and walking into the tree depth-wise.
2882          */
2883         depth = ext_depth(inode);
2884         if (path) {
2885                 int k = i = depth;
2886                 while (--k > 0)
2887                         path[k].p_block =
2888                                 le16_to_cpu(path[k].p_hdr->eh_entries)+1;
2889         } else {
2890                 path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1),
2891                                GFP_NOFS);
2892                 if (path == NULL) {
2893                         ext4_journal_stop(handle);
2894                         return -ENOMEM;
2895                 }
2896                 path[0].p_depth = depth;
2897                 path[0].p_hdr = ext_inode_hdr(inode);
2898                 i = 0;
2899
2900                 if (ext4_ext_check(inode, path[0].p_hdr, depth, 0)) {
2901                         err = -EIO;
2902                         goto out;
2903                 }
2904         }
2905         err = 0;
2906
2907         while (i >= 0 && err == 0) {
2908                 if (i == depth) {
2909                         /* this is leaf block */
2910                         err = ext4_ext_rm_leaf(handle, inode, path,
2911                                                &partial_cluster, start,
2912                                                end);
2913                         /* root level has p_bh == NULL, brelse() eats this */
2914                         brelse(path[i].p_bh);
2915                         path[i].p_bh = NULL;
2916                         i--;
2917                         continue;
2918                 }
2919
2920                 /* this is index block */
2921                 if (!path[i].p_hdr) {
2922                         ext_debug("initialize header\n");
2923                         path[i].p_hdr = ext_block_hdr(path[i].p_bh);
2924                 }
2925
2926                 if (!path[i].p_idx) {
2927                         /* this level hasn't been touched yet */
2928                         path[i].p_idx = EXT_LAST_INDEX(path[i].p_hdr);
2929                         path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries)+1;
2930                         ext_debug("init index ptr: hdr 0x%p, num %d\n",
2931                                   path[i].p_hdr,
2932                                   le16_to_cpu(path[i].p_hdr->eh_entries));
2933                 } else {
2934                         /* we were already here, see at next index */
2935                         path[i].p_idx--;
2936                 }
2937
2938                 ext_debug("level %d - index, first 0x%p, cur 0x%p\n",
2939                                 i, EXT_FIRST_INDEX(path[i].p_hdr),
2940                                 path[i].p_idx);
2941                 if (ext4_ext_more_to_rm(path + i)) {
2942                         struct buffer_head *bh;
2943                         /* go to the next level */
2944                         ext_debug("move to level %d (block %llu)\n",
2945                                   i + 1, ext4_idx_pblock(path[i].p_idx));
2946                         memset(path + i + 1, 0, sizeof(*path));
2947                         bh = read_extent_tree_block(inode,
2948                                 ext4_idx_pblock(path[i].p_idx), depth - i - 1,
2949                                 EXT4_EX_NOCACHE);
2950                         if (IS_ERR(bh)) {
2951                                 /* should we reset i_size? */
2952                                 err = PTR_ERR(bh);
2953                                 break;
2954                         }
2955                         /* Yield here to deal with large extent trees.
2956                          * Should be a no-op if we did IO above. */
2957                         cond_resched();
2958                         if (WARN_ON(i + 1 > depth)) {
2959                                 err = -EIO;
2960                                 break;
2961                         }
2962                         path[i + 1].p_bh = bh;
2963
2964                         /* save actual number of indexes since this
2965                          * number is changed at the next iteration */
2966                         path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries);
2967                         i++;
2968                 } else {
2969                         /* we finished processing this index, go up */
2970                         if (path[i].p_hdr->eh_entries == 0 && i > 0) {
2971                                 /* index is empty, remove it;
2972                                  * handle must be already prepared by the
2973                                  * truncatei_leaf() */
2974                                 err = ext4_ext_rm_idx(handle, inode, path, i);
2975                         }
2976                         /* root level has p_bh == NULL, brelse() eats this */
2977                         brelse(path[i].p_bh);
2978                         path[i].p_bh = NULL;
2979                         i--;
2980                         ext_debug("return to level %d\n", i);
2981                 }
2982         }
2983
2984         trace_ext4_ext_remove_space_done(inode, start, end, depth,
2985                         partial_cluster, path->p_hdr->eh_entries);
2986
2987         /* If we still have something in the partial cluster and we have removed
2988          * even the first extent, then we should free the blocks in the partial
2989          * cluster as well. */
2990         if (partial_cluster > 0 && path->p_hdr->eh_entries == 0) {
2991                 int flags = get_default_free_blocks_flags(inode);
2992
2993                 ext4_free_blocks(handle, inode, NULL,
2994                                  EXT4_C2B(EXT4_SB(sb), partial_cluster),
2995                                  EXT4_SB(sb)->s_cluster_ratio, flags);
2996                 partial_cluster = 0;
2997         }
2998
2999         /* TODO: flexible tree reduction should be here */
3000         if (path->p_hdr->eh_entries == 0) {
3001                 /*
3002                  * truncate to zero freed all the tree,
3003                  * so we need to correct eh_depth
3004                  */
3005                 err = ext4_ext_get_access(handle, inode, path);
3006                 if (err == 0) {
3007                         ext_inode_hdr(inode)->eh_depth = 0;
3008                         ext_inode_hdr(inode)->eh_max =
3009                                 cpu_to_le16(ext4_ext_space_root(inode, 0));
3010                         err = ext4_ext_dirty(handle, inode, path);
3011                 }
3012         }
3013 out:
3014         ext4_ext_drop_refs(path);
3015         kfree(path);
3016         if (err == -EAGAIN) {
3017                 path = NULL;
3018                 goto again;
3019         }
3020         ext4_journal_stop(handle);
3021
3022         return err;
3023 }
3024
3025 /*
3026  * called at mount time
3027  */
3028 void ext4_ext_init(struct super_block *sb)
3029 {
3030         /*
3031          * possible initialization would be here
3032          */
3033
3034         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
3035 #if defined(AGGRESSIVE_TEST) || defined(CHECK_BINSEARCH) || defined(EXTENTS_STATS)
3036                 printk(KERN_INFO "EXT4-fs: file extents enabled"
3037 #ifdef AGGRESSIVE_TEST
3038                        ", aggressive tests"
3039 #endif
3040 #ifdef CHECK_BINSEARCH
3041                        ", check binsearch"
3042 #endif
3043 #ifdef EXTENTS_STATS
3044                        ", stats"
3045 #endif
3046                        "\n");
3047 #endif
3048 #ifdef EXTENTS_STATS
3049                 spin_lock_init(&EXT4_SB(sb)->s_ext_stats_lock);
3050                 EXT4_SB(sb)->s_ext_min = 1 << 30;
3051                 EXT4_SB(sb)->s_ext_max = 0;
3052 #endif
3053         }
3054 }
3055
3056 /*
3057  * called at umount time
3058  */
3059 void ext4_ext_release(struct super_block *sb)
3060 {
3061         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3062                 return;
3063
3064 #ifdef EXTENTS_STATS
3065         if (EXT4_SB(sb)->s_ext_blocks && EXT4_SB(sb)->s_ext_extents) {
3066                 struct ext4_sb_info *sbi = EXT4_SB(sb);
3067                 printk(KERN_ERR "EXT4-fs: %lu blocks in %lu extents (%lu ave)\n",
3068                         sbi->s_ext_blocks, sbi->s_ext_extents,
3069                         sbi->s_ext_blocks / sbi->s_ext_extents);
3070                 printk(KERN_ERR "EXT4-fs: extents: %lu min, %lu max, max depth %lu\n",
3071                         sbi->s_ext_min, sbi->s_ext_max, sbi->s_depth_max);
3072         }
3073 #endif
3074 }
3075
3076 static int ext4_zeroout_es(struct inode *inode, struct ext4_extent *ex)
3077 {
3078         ext4_lblk_t  ee_block;
3079         ext4_fsblk_t ee_pblock;
3080         unsigned int ee_len;
3081
3082         ee_block  = le32_to_cpu(ex->ee_block);
3083         ee_len    = ext4_ext_get_actual_len(ex);
3084         ee_pblock = ext4_ext_pblock(ex);
3085
3086         if (ee_len == 0)
3087                 return 0;
3088
3089         return ext4_es_insert_extent(inode, ee_block, ee_len, ee_pblock,
3090                                      EXTENT_STATUS_WRITTEN);
3091 }
3092
3093 /* FIXME!! we need to try to merge to left or right after zero-out  */
3094 static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
3095 {
3096         ext4_fsblk_t ee_pblock;
3097         unsigned int ee_len;
3098         int ret;
3099
3100         ee_len    = ext4_ext_get_actual_len(ex);
3101         ee_pblock = ext4_ext_pblock(ex);
3102
3103         ret = sb_issue_zeroout(inode->i_sb, ee_pblock, ee_len, GFP_NOFS);
3104         if (ret > 0)
3105                 ret = 0;
3106
3107         return ret;
3108 }
3109
3110 /*
3111  * ext4_split_extent_at() splits an extent at given block.
3112  *
3113  * @handle: the journal handle
3114  * @inode: the file inode
3115  * @path: the path to the extent
3116  * @split: the logical block where the extent is splitted.
3117  * @split_flags: indicates if the extent could be zeroout if split fails, and
3118  *               the states(init or unwritten) of new extents.
3119  * @flags: flags used to insert new extent to extent tree.
3120  *
3121  *
3122  * Splits extent [a, b] into two extents [a, @split) and [@split, b], states
3123  * of which are deterimined by split_flag.
3124  *
3125  * There are two cases:
3126  *  a> the extent are splitted into two extent.
3127  *  b> split is not needed, and just mark the extent.
3128  *
3129  * return 0 on success.
3130  */
3131 static int ext4_split_extent_at(handle_t *handle,
3132                              struct inode *inode,
3133                              struct ext4_ext_path *path,
3134                              ext4_lblk_t split,
3135                              int split_flag,
3136                              int flags)
3137 {
3138         ext4_fsblk_t newblock;
3139         ext4_lblk_t ee_block;
3140         struct ext4_extent *ex, newex, orig_ex, zero_ex;
3141         struct ext4_extent *ex2 = NULL;
3142         unsigned int ee_len, depth;
3143         int err = 0;
3144
3145         BUG_ON((split_flag & (EXT4_EXT_DATA_VALID1 | EXT4_EXT_DATA_VALID2)) ==
3146                (EXT4_EXT_DATA_VALID1 | EXT4_EXT_DATA_VALID2));
3147
3148         ext_debug("ext4_split_extents_at: inode %lu, logical"
3149                 "block %llu\n", inode->i_ino, (unsigned long long)split);
3150
3151         ext4_ext_show_leaf(inode, path);
3152
3153         depth = ext_depth(inode);
3154         ex = path[depth].p_ext;
3155         ee_block = le32_to_cpu(ex->ee_block);
3156         ee_len = ext4_ext_get_actual_len(ex);
3157         newblock = split - ee_block + ext4_ext_pblock(ex);
3158
3159         BUG_ON(split < ee_block || split >= (ee_block + ee_len));
3160         BUG_ON(!ext4_ext_is_unwritten(ex) &&
3161                split_flag & (EXT4_EXT_MAY_ZEROOUT |
3162                              EXT4_EXT_MARK_UNWRIT1 |
3163                              EXT4_EXT_MARK_UNWRIT2));
3164
3165         err = ext4_ext_get_access(handle, inode, path + depth);
3166         if (err)
3167                 goto out;
3168
3169         if (split == ee_block) {
3170                 /*
3171                  * case b: block @split is the block that the extent begins with
3172                  * then we just change the state of the extent, and splitting
3173                  * is not needed.
3174                  */
3175                 if (split_flag & EXT4_EXT_MARK_UNWRIT2)
3176                         ext4_ext_mark_unwritten(ex);
3177                 else
3178                         ext4_ext_mark_initialized(ex);
3179
3180                 if (!(flags & EXT4_GET_BLOCKS_PRE_IO))
3181                         ext4_ext_try_to_merge(handle, inode, path, ex);
3182
3183                 err = ext4_ext_dirty(handle, inode, path + path->p_depth);
3184                 goto out;
3185         }
3186
3187         /* case a */
3188         memcpy(&orig_ex, ex, sizeof(orig_ex));
3189         ex->ee_len = cpu_to_le16(split - ee_block);
3190         if (split_flag & EXT4_EXT_MARK_UNWRIT1)
3191                 ext4_ext_mark_unwritten(ex);
3192
3193         /*
3194          * path may lead to new leaf, not to original leaf any more
3195          * after ext4_ext_insert_extent() returns,
3196          */
3197         err = ext4_ext_dirty(handle, inode, path + depth);
3198         if (err)
3199                 goto fix_extent_len;
3200
3201         ex2 = &newex;
3202         ex2->ee_block = cpu_to_le32(split);
3203         ex2->ee_len   = cpu_to_le16(ee_len - (split - ee_block));
3204         ext4_ext_store_pblock(ex2, newblock);
3205         if (split_flag & EXT4_EXT_MARK_UNWRIT2)
3206                 ext4_ext_mark_unwritten(ex2);
3207
3208         err = ext4_ext_insert_extent(handle, inode, path, &newex, flags);
3209         if (err == -ENOSPC && (EXT4_EXT_MAY_ZEROOUT & split_flag)) {
3210                 if (split_flag & (EXT4_EXT_DATA_VALID1|EXT4_EXT_DATA_VALID2)) {
3211                         if (split_flag & EXT4_EXT_DATA_VALID1) {
3212                                 err = ext4_ext_zeroout(inode, ex2);
3213                                 zero_ex.ee_block = ex2->ee_block;
3214                                 zero_ex.ee_len = cpu_to_le16(
3215                                                 ext4_ext_get_actual_len(ex2));
3216                                 ext4_ext_store_pblock(&zero_ex,
3217                                                       ext4_ext_pblock(ex2));
3218                         } else {
3219                                 err = ext4_ext_zeroout(inode, ex);
3220                                 zero_ex.ee_block = ex->ee_block;
3221                                 zero_ex.ee_len = cpu_to_le16(
3222                                                 ext4_ext_get_actual_len(ex));
3223                                 ext4_ext_store_pblock(&zero_ex,
3224                                                       ext4_ext_pblock(ex));
3225                         }
3226                 } else {
3227                         err = ext4_ext_zeroout(inode, &orig_ex);
3228                         zero_ex.ee_block = orig_ex.ee_block;
3229                         zero_ex.ee_len = cpu_to_le16(
3230                                                 ext4_ext_get_actual_len(&orig_ex));
3231                         ext4_ext_store_pblock(&zero_ex,
3232                                               ext4_ext_pblock(&orig_ex));
3233                 }
3234
3235                 if (err)
3236                         goto fix_extent_len;
3237                 /* update the extent length and mark as initialized */
3238                 ex->ee_len = cpu_to_le16(ee_len);
3239                 ext4_ext_try_to_merge(handle, inode, path, ex);
3240                 err = ext4_ext_dirty(handle, inode, path + path->p_depth);
3241                 if (err)
3242                         goto fix_extent_len;
3243
3244                 /* update extent status tree */
3245                 err = ext4_zeroout_es(inode, &zero_ex);
3246
3247                 goto out;
3248         } else if (err)
3249                 goto fix_extent_len;
3250
3251 out:
3252         ext4_ext_show_leaf(inode, path);
3253         return err;
3254
3255 fix_extent_len:
3256         ex->ee_len = orig_ex.ee_len;
3257         ext4_ext_dirty(handle, inode, path + path->p_depth);
3258         return err;
3259 }
3260
3261 /*
3262  * ext4_split_extents() splits an extent and mark extent which is covered
3263  * by @map as split_flags indicates
3264  *
3265  * It may result in splitting the extent into multiple extents (up to three)
3266  * There are three possibilities:
3267  *   a> There is no split required
3268  *   b> Splits in two extents: Split is happening at either end of the extent
3269  *   c> Splits in three extents: Somone is splitting in middle of the extent
3270  *
3271  */
3272 static int ext4_split_extent(handle_t *handle,
3273                               struct inode *inode,
3274                               struct ext4_ext_path *path,
3275                               struct ext4_map_blocks *map,
3276                               int split_flag,
3277                               int flags)
3278 {
3279         ext4_lblk_t ee_block;
3280         struct ext4_extent *ex;
3281         unsigned int ee_len, depth;
3282         int err = 0;
3283         int unwritten;
3284         int split_flag1, flags1;
3285         int allocated = map->m_len;
3286
3287         depth = ext_depth(inode);
3288         ex = path[depth].p_ext;
3289         ee_block = le32_to_cpu(ex->ee_block);
3290         ee_len = ext4_ext_get_actual_len(ex);
3291         unwritten = ext4_ext_is_unwritten(ex);
3292
3293         if (map->m_lblk + map->m_len < ee_block + ee_len) {
3294                 split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT;
3295                 flags1 = flags | EXT4_GET_BLOCKS_PRE_IO;
3296                 if (unwritten)
3297                         split_flag1 |= EXT4_EXT_MARK_UNWRIT1 |
3298                                        EXT4_EXT_MARK_UNWRIT2;
3299                 if (split_flag & EXT4_EXT_DATA_VALID2)
3300                         split_flag1 |= EXT4_EXT_DATA_VALID1;
3301                 err = ext4_split_extent_at(handle, inode, path,
3302                                 map->m_lblk + map->m_len, split_flag1, flags1);
3303                 if (err)
3304                         goto out;
3305         } else {
3306                 allocated = ee_len - (map->m_lblk - ee_block);
3307         }
3308         /*
3309          * Update path is required because previous ext4_split_extent_at() may
3310          * result in split of original leaf or extent zeroout.
3311          */
3312         ext4_ext_drop_refs(path);
3313         path = ext4_ext_find_extent(inode, map->m_lblk, path, 0);
3314         if (IS_ERR(path))
3315                 return PTR_ERR(path);
3316         depth = ext_depth(inode);
3317         ex = path[depth].p_ext;
3318         if (!ex) {
3319                 EXT4_ERROR_INODE(inode, "unexpected hole at %lu",
3320                                  (unsigned long) map->m_lblk);
3321                 return -EIO;
3322         }
3323         unwritten = ext4_ext_is_unwritten(ex);
3324         split_flag1 = 0;
3325
3326         if (map->m_lblk >= ee_block) {
3327                 split_flag1 = split_flag & EXT4_EXT_DATA_VALID2;
3328                 if (unwritten) {
3329                         split_flag1 |= EXT4_EXT_MARK_UNWRIT1;
3330                         split_flag1 |= split_flag & (EXT4_EXT_MAY_ZEROOUT |
3331                                                      EXT4_EXT_MARK_UNWRIT2);
3332                 }
3333                 err = ext4_split_extent_at(handle, inode, path,
3334                                 map->m_lblk, split_flag1, flags);
3335                 if (err)
3336                         goto out;
3337         }
3338
3339         ext4_ext_show_leaf(inode, path);
3340 out:
3341         return err ? err : allocated;
3342 }
3343
3344 /*
3345  * This function is called by ext4_ext_map_blocks() if someone tries to write
3346  * to an unwritten extent. It may result in splitting the unwritten
3347  * extent into multiple extents (up to three - one initialized and two
3348  * unwritten).
3349  * There are three possibilities:
3350  *   a> There is no split required: Entire extent should be initialized
3351  *   b> Splits in two extents: Write is happening at either end of the extent
3352  *   c> Splits in three extents: Somone is writing in middle of the extent
3353  *
3354  * Pre-conditions:
3355  *  - The extent pointed to by 'path' is unwritten.
3356  *  - The extent pointed to by 'path' contains a superset
3357  *    of the logical span [map->m_lblk, map->m_lblk + map->m_len).
3358  *
3359  * Post-conditions on success:
3360  *  - the returned value is the number of blocks beyond map->l_lblk
3361  *    that are allocated and initialized.
3362  *    It is guaranteed to be >= map->m_len.
3363  */
3364 static int ext4_ext_convert_to_initialized(handle_t *handle,
3365                                            struct inode *inode,
3366                                            struct ext4_map_blocks *map,
3367                                            struct ext4_ext_path *path,
3368                                            int flags)
3369 {
3370         struct ext4_sb_info *sbi;
3371         struct ext4_extent_header *eh;
3372         struct ext4_map_blocks split_map;
3373         struct ext4_extent zero_ex;
3374         struct ext4_extent *ex, *abut_ex;
3375         ext4_lblk_t ee_block, eof_block;
3376         unsigned int ee_len, depth, map_len = map->m_len;
3377         int allocated = 0, max_zeroout = 0;
3378         int err = 0;
3379         int split_flag = 0;
3380
3381         ext_debug("ext4_ext_convert_to_initialized: inode %lu, logical"
3382                 "block %llu, max_blocks %u\n", inode->i_ino,
3383                 (unsigned long long)map->m_lblk, map_len);
3384
3385         sbi = EXT4_SB(inode->i_sb);
3386         eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >>
3387                 inode->i_sb->s_blocksize_bits;
3388         if (eof_block < map->m_lblk + map_len)
3389                 eof_block = map->m_lblk + map_len;
3390
3391         depth = ext_depth(inode);
3392         eh = path[depth].p_hdr;
3393         ex = path[depth].p_ext;
3394         ee_block = le32_to_cpu(ex->ee_block);
3395         ee_len = ext4_ext_get_actual_len(ex);
3396         zero_ex.ee_len = 0;
3397
3398         trace_ext4_ext_convert_to_initialized_enter(inode, map, ex);
3399
3400         /* Pre-conditions */
3401         BUG_ON(!ext4_ext_is_unwritten(ex));
3402         BUG_ON(!in_range(map->m_lblk, ee_block, ee_len));
3403
3404         /*
3405          * Attempt to transfer newly initialized blocks from the currently
3406          * unwritten extent to its neighbor. This is much cheaper
3407          * than an insertion followed by a merge as those involve costly
3408          * memmove() calls. Transferring to the left is the common case in
3409          * steady state for workloads doing fallocate(FALLOC_FL_KEEP_SIZE)
3410          * followed by append writes.
3411          *
3412          * Limitations of the current logic:
3413          *  - L1: we do not deal with writes covering the whole extent.
3414          *    This would require removing the extent if the transfer
3415          *    is possible.
3416          *  - L2: we only attempt to merge with an extent stored in the
3417          *    same extent tree node.
3418          */
3419         if ((map->m_lblk == ee_block) &&
3420                 /* See if we can merge left */
3421                 (map_len < ee_len) &&           /*L1*/
3422                 (ex > EXT_FIRST_EXTENT(eh))) {  /*L2*/
3423                 ext4_lblk_t prev_lblk;
3424                 ext4_fsblk_t prev_pblk, ee_pblk;
3425                 unsigned int prev_len;
3426
3427                 abut_ex = ex - 1;
3428                 prev_lblk = le32_to_cpu(abut_ex->ee_block);
3429                 prev_len = ext4_ext_get_actual_len(abut_ex);
3430                 prev_pblk = ext4_ext_pblock(abut_ex);
3431                 ee_pblk = ext4_ext_pblock(ex);
3432
3433                 /*
3434                  * A transfer of blocks from 'ex' to 'abut_ex' is allowed
3435                  * upon those conditions:
3436                  * - C1: abut_ex is initialized,
3437                  * - C2: abut_ex is logically abutting ex,
3438                  * - C3: abut_ex is physically abutting ex,
3439                  * - C4: abut_ex can receive the additional blocks without
3440                  *   overflowing the (initialized) length limit.
3441                  */
3442                 if ((!ext4_ext_is_unwritten(abut_ex)) &&                /*C1*/
3443                         ((prev_lblk + prev_len) == ee_block) &&         /*C2*/
3444                         ((prev_pblk + prev_len) == ee_pblk) &&          /*C3*/
3445                         (prev_len < (EXT_INIT_MAX_LEN - map_len))) {    /*C4*/
3446                         err = ext4_ext_get_access(handle, inode, path + depth);
3447                         if (err)
3448                                 goto out;
3449
3450                         trace_ext4_ext_convert_to_initialized_fastpath(inode,
3451                                 map, ex, abut_ex);
3452
3453                         /* Shift the start of ex by 'map_len' blocks */
3454                         ex->ee_block = cpu_to_le32(ee_block + map_len);
3455                         ext4_ext_store_pblock(ex, ee_pblk + map_len);
3456                         ex->ee_len = cpu_to_le16(ee_len - map_len);
3457                         ext4_ext_mark_unwritten(ex); /* Restore the flag */
3458
3459                         /* Extend abut_ex by 'map_len' blocks */
3460                         abut_ex->ee_len = cpu_to_le16(prev_len + map_len);
3461
3462                         /* Result: number of initialized blocks past m_lblk */
3463                         allocated = map_len;
3464                 }
3465         } else if (((map->m_lblk + map_len) == (ee_block + ee_len)) &&
3466                    (map_len < ee_len) &&        /*L1*/
3467                    ex < EXT_LAST_EXTENT(eh)) {  /*L2*/
3468                 /* See if we can merge right */
3469                 ext4_lblk_t next_lblk;
3470                 ext4_fsblk_t next_pblk, ee_pblk;
3471                 unsigned int next_len;
3472
3473                 abut_ex = ex + 1;
3474                 next_lblk = le32_to_cpu(abut_ex->ee_block);
3475                 next_len = ext4_ext_get_actual_len(abut_ex);
3476                 next_pblk = ext4_ext_pblock(abut_ex);
3477                 ee_pblk = ext4_ext_pblock(ex);
3478
3479                 /*
3480                  * A transfer of blocks from 'ex' to 'abut_ex' is allowed
3481                  * upon those conditions:
3482                  * - C1: abut_ex is initialized,
3483                  * - C2: abut_ex is logically abutting ex,
3484                  * - C3: abut_ex is physically abutting ex,
3485                  * - C4: abut_ex can receive the additional blocks without
3486                  *   overflowing the (initialized) length limit.
3487                  */
3488                 if ((!ext4_ext_is_unwritten(abut_ex)) &&                /*C1*/
3489                     ((map->m_lblk + map_len) == next_lblk) &&           /*C2*/
3490                     ((ee_pblk + ee_len) == next_pblk) &&                /*C3*/
3491                     (next_len < (EXT_INIT_MAX_LEN - map_len))) {        /*C4*/
3492                         err = ext4_ext_get_access(handle, inode, path + depth);
3493                         if (err)
3494                                 goto out;
3495
3496                         trace_ext4_ext_convert_to_initialized_fastpath(inode,
3497                                 map, ex, abut_ex);
3498
3499                         /* Shift the start of abut_ex by 'map_len' blocks */
3500                         abut_ex->ee_block = cpu_to_le32(next_lblk - map_len);
3501                         ext4_ext_store_pblock(abut_ex, next_pblk - map_len);
3502                         ex->ee_len = cpu_to_le16(ee_len - map_len);
3503                         ext4_ext_mark_unwritten(ex); /* Restore the flag */
3504
3505                         /* Extend abut_ex by 'map_len' blocks */
3506                         abut_ex->ee_len = cpu_to_le16(next_len + map_len);
3507
3508                         /* Result: number of initialized blocks past m_lblk */
3509                         allocated = map_len;
3510                 }
3511         }
3512         if (allocated) {
3513                 /* Mark the block containing both extents as dirty */
3514                 ext4_ext_dirty(handle, inode, path + depth);
3515
3516                 /* Update path to point to the right extent */
3517                 path[depth].p_ext = abut_ex;
3518                 goto out;
3519         } else
3520                 allocated = ee_len - (map->m_lblk - ee_block);
3521
3522         WARN_ON(map->m_lblk < ee_block);
3523         /*
3524          * It is safe to convert extent to initialized via explicit
3525          * zeroout only if extent is fully inside i_size or new_size.
3526          */
3527         split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0;
3528
3529         if (EXT4_EXT_MAY_ZEROOUT & split_flag)
3530                 max_zeroout = sbi->s_extent_max_zeroout_kb >>
3531                         (inode->i_sb->s_blocksize_bits - 10);
3532
3533         /* If extent is less than s_max_zeroout_kb, zeroout directly */
3534         if (max_zeroout && (ee_len <= max_zeroout)) {
3535                 err = ext4_ext_zeroout(inode, ex);
3536                 if (err)
3537                         goto out;
3538                 zero_ex.ee_block = ex->ee_block;
3539                 zero_ex.ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex));
3540                 ext4_ext_store_pblock(&zero_ex, ext4_ext_pblock(ex));
3541
3542                 err = ext4_ext_get_access(handle, inode, path + depth);
3543                 if (err)
3544                         goto out;
3545                 ext4_ext_mark_initialized(ex);
3546                 ext4_ext_try_to_merge(handle, inode, path, ex);
3547                 err = ext4_ext_dirty(handle, inode, path + path->p_depth);
3548                 goto out;
3549         }
3550
3551         /*
3552          * four cases:
3553          * 1. split the extent into three extents.
3554          * 2. split the extent into two extents, zeroout the first half.
3555          * 3. split the extent into two extents, zeroout the second half.
3556          * 4. split the extent into two extents with out zeroout.
3557          */
3558         split_map.m_lblk = map->m_lblk;
3559         split_map.m_len = map->m_len;
3560
3561         if (max_zeroout && (allocated > map->m_len)) {
3562                 if (allocated <= max_zeroout) {
3563                         /* case 3 */
3564                         zero_ex.ee_block =
3565                                          cpu_to_le32(map->m_lblk);
3566                         zero_ex.ee_len = cpu_to_le16(allocated);
3567                         ext4_ext_store_pblock(&zero_ex,
3568                                 ext4_ext_pblock(ex) + map->m_lblk - ee_block);
3569                         err = ext4_ext_zeroout(inode, &zero_ex);
3570                         if (err)
3571                                 goto out;
3572                         split_map.m_lblk = map->m_lblk;
3573                         split_map.m_len = allocated;
3574                 } else if (map->m_lblk - ee_block + map->m_len < max_zeroout) {
3575                         /* case 2 */
3576                         if (map->m_lblk != ee_block) {
3577                                 zero_ex.ee_block = ex->ee_block;
3578                                 zero_ex.ee_len = cpu_to_le16(map->m_lblk -
3579                                                         ee_block);
3580                                 ext4_ext_store_pblock(&zero_ex,
3581                                                       ext4_ext_pblock(ex));
3582                                 err = ext4_ext_zeroout(inode, &zero_ex);
3583                                 if (err)
3584                                         goto out;
3585                         }
3586
3587                         split_map.m_lblk = ee_block;
3588                         split_map.m_len = map->m_lblk - ee_block + map->m_len;
3589                         allocated = map->m_len;
3590                 }
3591         }
3592
3593         allocated = ext4_split_extent(handle, inode, path,
3594                                       &split_map, split_flag, flags);
3595         if (allocated < 0)
3596                 err = allocated;
3597
3598 out:
3599         /* If we have gotten a failure, don't zero out status tree */
3600         if (!err)
3601                 err = ext4_zeroout_es(inode, &zero_ex);
3602         return err ? err : allocated;
3603 }
3604
3605 /*
3606  * This function is called by ext4_ext_map_blocks() from
3607  * ext4_get_blocks_dio_write() when DIO to write
3608  * to an unwritten extent.
3609  *
3610  * Writing to an unwritten extent may result in splitting the unwritten
3611  * extent into multiple initialized/unwritten extents (up to three)
3612  * There are three possibilities:
3613  *   a> There is no split required: Entire extent should be unwritten
3614  *   b> Splits in two extents: Write is happening at either end of the extent
3615  *   c> Splits in three extents: Somone is writing in middle of the extent
3616  *
3617  * This works the same way in the case of initialized -> unwritten conversion.
3618  *
3619  * One of more index blocks maybe needed if the extent tree grow after
3620  * the unwritten extent split. To prevent ENOSPC occur at the IO
3621  * complete, we need to split the unwritten extent before DIO submit
3622  * the IO. The unwritten extent called at this time will be split
3623  * into three unwritten extent(at most). After IO complete, the part
3624  * being filled will be convert to initialized by the end_io callback function
3625  * via ext4_convert_unwritten_extents().
3626  *
3627  * Returns the size of unwritten extent to be written on success.
3628  */
3629 static int ext4_split_convert_extents(handle_t *handle,
3630                                         struct inode *inode,
3631                                         struct ext4_map_blocks *map,
3632                                         struct ext4_ext_path *path,
3633                                         int flags)
3634 {
3635         ext4_lblk_t eof_block;
3636         ext4_lblk_t ee_block;
3637         struct ext4_extent *ex;
3638         unsigned int ee_len;
3639         int split_flag = 0, depth;
3640
3641         ext_debug("%s: inode %lu, logical block %llu, max_blocks %u\n",
3642                   __func__, inode->i_ino,
3643                   (unsigned long long)map->m_lblk, map->m_len);
3644
3645         eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >>
3646                 inode->i_sb->s_blocksize_bits;
3647         if (eof_block < map->m_lblk + map->m_len)
3648                 eof_block = map->m_lblk + map->m_len;
3649         /*
3650          * It is safe to convert extent to initialized via explicit
3651          * zeroout only if extent is fully insde i_size or new_size.
3652          */
3653         depth = ext_depth(inode);
3654         ex = path[depth].p_ext;
3655         ee_block = le32_to_cpu(ex->ee_block);
3656         ee_len = ext4_ext_get_actual_len(ex);
3657
3658         /* Convert to unwritten */
3659         if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN) {
3660                 split_flag |= EXT4_EXT_DATA_VALID1;
3661         /* Convert to initialized */
3662         } else if (flags & EXT4_GET_BLOCKS_CONVERT) {
3663                 split_flag |= ee_block + ee_len <= eof_block ?
3664                               EXT4_EXT_MAY_ZEROOUT : 0;
3665                 split_flag |= (EXT4_EXT_MARK_UNWRIT2 | EXT4_EXT_DATA_VALID2);
3666         }
3667         flags |= EXT4_GET_BLOCKS_PRE_IO;
3668         return ext4_split_extent(handle, inode, path, map, split_flag, flags);
3669 }
3670
3671 static int ext4_convert_initialized_extents(handle_t *handle,
3672                                             struct inode *inode,
3673                                             struct ext4_map_blocks *map,
3674                                             struct ext4_ext_path *path)
3675 {
3676         struct ext4_extent *ex;
3677         ext4_lblk_t ee_block;
3678         unsigned int ee_len;
3679         int depth;
3680         int err = 0;
3681
3682         depth = ext_depth(inode);
3683         ex = path[depth].p_ext;
3684         ee_block = le32_to_cpu(ex->ee_block);
3685         ee_len = ext4_ext_get_actual_len(ex);
3686
3687         ext_debug("%s: inode %lu, logical"
3688                 "block %llu, max_blocks %u\n", __func__, inode->i_ino,
3689                   (unsigned long long)ee_block, ee_len);
3690
3691         if (ee_block != map->m_lblk || ee_len > map->m_len) {
3692                 err = ext4_split_convert_extents(handle, inode, map, path,
3693                                 EXT4_GET_BLOCKS_CONVERT_UNWRITTEN);
3694                 if (err < 0)
3695                         goto out;
3696                 ext4_ext_drop_refs(path);
3697                 path = ext4_ext_find_extent(inode, map->m_lblk, path, 0);
3698                 if (IS_ERR(path)) {
3699                         err = PTR_ERR(path);
3700                         goto out;
3701                 }
3702                 depth = ext_depth(inode);
3703                 ex = path[depth].p_ext;
3704                 if (!ex) {
3705                         EXT4_ERROR_INODE(inode, "unexpected hole at %lu",
3706                                          (unsigned long) map->m_lblk);
3707                         err = -EIO;
3708                         goto out;
3709                 }
3710         }
3711
3712         err = ext4_ext_get_access(handle, inode, path + depth);
3713         if (err)
3714                 goto out;
3715         /* first mark the extent as unwritten */
3716         ext4_ext_mark_unwritten(ex);
3717
3718         /* note: ext4_ext_correct_indexes() isn't needed here because
3719          * borders are not changed
3720          */
3721         ext4_ext_try_to_merge(handle, inode, path, ex);
3722
3723         /* Mark modified extent as dirty */
3724         err = ext4_ext_dirty(handle, inode, path + path->p_depth);
3725 out:
3726         ext4_ext_show_leaf(inode, path);
3727         return err;
3728 }
3729
3730
3731 static int ext4_convert_unwritten_extents_endio(handle_t *handle,
3732                                                 struct inode *inode,
3733                                                 struct ext4_map_blocks *map,
3734                                                 struct ext4_ext_path *path)
3735 {
3736         struct ext4_extent *ex;
3737         ext4_lblk_t ee_block;
3738         unsigned int ee_len;
3739         int depth;
3740         int err = 0;
3741
3742         depth = ext_depth(inode);
3743         ex = path[depth].p_ext;
3744         ee_block = le32_to_cpu(ex->ee_block);
3745         ee_len = ext4_ext_get_actual_len(ex);
3746
3747         ext_debug("ext4_convert_unwritten_extents_endio: inode %lu, logical"
3748                 "block %llu, max_blocks %u\n", inode->i_ino,
3749                   (unsigned long long)ee_block, ee_len);
3750
3751         /* If extent is larger than requested it is a clear sign that we still
3752          * have some extent state machine issues left. So extent_split is still
3753          * required.
3754          * TODO: Once all related issues will be fixed this situation should be
3755          * illegal.
3756          */
3757         if (ee_block != map->m_lblk || ee_len > map->m_len) {
3758 #ifdef EXT4_DEBUG
3759                 ext4_warning("Inode (%ld) finished: extent logical block %llu,"
3760                              " len %u; IO logical block %llu, len %u\n",
3761                              inode->i_ino, (unsigned long long)ee_block, ee_len,
3762                              (unsigned long long)map->m_lblk, map->m_len);
3763 #endif
3764                 err = ext4_split_convert_extents(handle, inode, map, path,
3765                                                  EXT4_GET_BLOCKS_CONVERT);
3766                 if (err < 0)
3767                         goto out;
3768                 ext4_ext_drop_refs(path);
3769                 path = ext4_ext_find_extent(inode, map->m_lblk, path, 0);
3770                 if (IS_ERR(path)) {
3771                         err = PTR_ERR(path);
3772                         goto out;
3773                 }
3774                 depth = ext_depth(inode);
3775                 ex = path[depth].p_ext;
3776         }
3777
3778         err = ext4_ext_get_access(handle, inode, path + depth);
3779         if (err)
3780                 goto out;
3781         /* first mark the extent as initialized */
3782         ext4_ext_mark_initialized(ex);
3783
3784         /* note: ext4_ext_correct_indexes() isn't needed here because
3785          * borders are not changed
3786          */
3787         ext4_ext_try_to_merge(handle, inode, path, ex);
3788
3789         /* Mark modified extent as dirty */
3790         err = ext4_ext_dirty(handle, inode, path + path->p_depth);
3791 out:
3792         ext4_ext_show_leaf(inode, path);
3793         return err;
3794 }
3795
3796 static void unmap_underlying_metadata_blocks(struct block_device *bdev,
3797                         sector_t block, int count)
3798 {
3799         int i;
3800         for (i = 0; i < count; i++)
3801                 unmap_underlying_metadata(bdev, block + i);
3802 }
3803
3804 /*
3805  * Handle EOFBLOCKS_FL flag, clearing it if necessary
3806  */
3807 static int check_eofblocks_fl(handle_t *handle, struct inode *inode,
3808                               ext4_lblk_t lblk,
3809                               struct ext4_ext_path *path,
3810                               unsigned int len)
3811 {
3812         int i, depth;
3813         struct ext4_extent_header *eh;
3814         struct ext4_extent *last_ex;
3815
3816         if (!ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS))
3817                 return 0;
3818
3819         depth = ext_depth(inode);
3820         eh = path[depth].p_hdr;
3821
3822         /*
3823          * We're going to remove EOFBLOCKS_FL entirely in future so we
3824          * do not care for this case anymore. Simply remove the flag
3825          * if there are no extents.
3826          */
3827         if (unlikely(!eh->eh_entries))
3828                 goto out;
3829         last_ex = EXT_LAST_EXTENT(eh);
3830         /*
3831          * We should clear the EOFBLOCKS_FL flag if we are writing the
3832          * last block in the last extent in the file.  We test this by
3833          * first checking to see if the caller to
3834          * ext4_ext_get_blocks() was interested in the last block (or
3835          * a block beyond the last block) in the current extent.  If
3836          * this turns out to be false, we can bail out from this
3837          * function immediately.
3838          */
3839         if (lblk + len < le32_to_cpu(last_ex->ee_block) +
3840             ext4_ext_get_actual_len(last_ex))
3841                 return 0;
3842         /*
3843          * If the caller does appear to be planning to write at or
3844          * beyond the end of the current extent, we then test to see
3845          * if the current extent is the last extent in the file, by
3846          * checking to make sure it was reached via the rightmost node
3847          * at each level of the tree.
3848          */
3849         for (i = depth-1; i >= 0; i--)
3850                 if (path[i].p_idx != EXT_LAST_INDEX(path[i].p_hdr))
3851                         return 0;
3852 out:
3853         ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
3854         return ext4_mark_inode_dirty(handle, inode);
3855 }
3856
3857 /**
3858  * ext4_find_delalloc_range: find delayed allocated block in the given range.
3859  *
3860  * Return 1 if there is a delalloc block in the range, otherwise 0.
3861  */
3862 int ext4_find_delalloc_range(struct inode *inode,
3863                              ext4_lblk_t lblk_start,
3864                              ext4_lblk_t lblk_end)
3865 {
3866         struct extent_status es;
3867
3868         ext4_es_find_delayed_extent_range(inode, lblk_start, lblk_end, &es);
3869         if (es.es_len == 0)
3870                 return 0; /* there is no delay extent in this tree */
3871         else if (es.es_lblk <= lblk_start &&
3872                  lblk_start < es.es_lblk + es.es_len)
3873                 return 1;
3874         else if (lblk_start <= es.es_lblk && es.es_lblk <= lblk_end)
3875                 return 1;
3876         else
3877                 return 0;
3878 }
3879
3880 int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk)
3881 {
3882         struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
3883         ext4_lblk_t lblk_start, lblk_end;
3884         lblk_start = EXT4_LBLK_CMASK(sbi, lblk);
3885         lblk_end = lblk_start + sbi->s_cluster_ratio - 1;
3886
3887         return ext4_find_delalloc_range(inode, lblk_start, lblk_end);
3888 }
3889
3890 /**
3891  * Determines how many complete clusters (out of those specified by the 'map')
3892  * are under delalloc and were reserved quota for.
3893  * This function is called when we are writing out the blocks that were
3894  * originally written with their allocation delayed, but then the space was
3895  * allocated using fallocate() before the delayed allocation could be resolved.
3896  * The cases to look for are:
3897  * ('=' indicated delayed allocated blocks
3898  *  '-' indicates non-delayed allocated blocks)
3899  * (a) partial clusters towards beginning and/or end outside of allocated range
3900  *     are not delalloc'ed.
3901  *      Ex:
3902  *      |----c---=|====c====|====c====|===-c----|
3903  *               |++++++ allocated ++++++|
3904  *      ==> 4 complete clusters in above example
3905  *
3906  * (b) partial cluster (outside of allocated range) towards either end is
3907  *     marked for delayed allocation. In this case, we will exclude that
3908  *     cluster.
3909  *      Ex:
3910  *      |----====c========|========c========|
3911  *           |++++++ allocated ++++++|
3912  *      ==> 1 complete clusters in above example
3913  *
3914  *      Ex:
3915  *      |================c================|
3916  *            |++++++ allocated ++++++|
3917  *      ==> 0 complete clusters in above example
3918  *
3919  * The ext4_da_update_reserve_space will be called only if we
3920  * determine here that there were some "entire" clusters that span
3921  * this 'allocated' range.
3922  * In the non-bigalloc case, this function will just end up returning num_blks
3923  * without ever calling ext4_find_delalloc_range.
3924  */
3925 static unsigned int
3926 get_reserved_cluster_alloc(struct inode *inode, ext4_lblk_t lblk_start,
3927                            unsigned int num_blks)
3928 {
3929         struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
3930         ext4_lblk_t alloc_cluster_start, alloc_cluster_end;
3931         ext4_lblk_t lblk_from, lblk_to, c_offset;
3932         unsigned int allocated_clusters = 0;
3933
3934         alloc_cluster_start = EXT4_B2C(sbi, lblk_start);
3935         alloc_cluster_end = EXT4_B2C(sbi, lblk_start + num_blks - 1);
3936
3937         /* max possible clusters for this allocation */
3938         allocated_clusters = alloc_cluster_end - alloc_cluster_start + 1;
3939
3940         trace_ext4_get_reserved_cluster_alloc(inode, lblk_start, num_blks);
3941
3942         /* Check towards left side */
3943         c_offset = EXT4_LBLK_COFF(sbi, lblk_start);
3944         if (c_offset) {
3945                 lblk_from = EXT4_LBLK_CMASK(sbi, lblk_start);
3946                 lblk_to = lblk_from + c_offset - 1;
3947
3948                 if (ext4_find_delalloc_range(inode, lblk_from, lblk_to))
3949                         allocated_clusters--;
3950         }
3951
3952         /* Now check towards right. */
3953         c_offset = EXT4_LBLK_COFF(sbi, lblk_start + num_blks);
3954         if (allocated_clusters && c_offset) {
3955                 lblk_from = lblk_start + num_blks;
3956                 lblk_to = lblk_from + (sbi->s_cluster_ratio - c_offset) - 1;
3957
3958                 if (ext4_find_delalloc_range(inode, lblk_from, lblk_to))
3959                         allocated_clusters--;
3960         }
3961
3962         return allocated_clusters;
3963 }
3964
3965 static int
3966 ext4_ext_convert_initialized_extent(handle_t *handle, struct inode *inode,
3967                         struct ext4_map_blocks *map,
3968                         struct ext4_ext_path *path, int flags,
3969                         unsigned int allocated, ext4_fsblk_t newblock)
3970 {
3971         int ret = 0;
3972         int err = 0;
3973
3974         /*
3975          * Make sure that the extent is no bigger than we support with
3976          * unwritten extent
3977          */
3978         if (map->m_len > EXT_UNWRITTEN_MAX_LEN)
3979                 map->m_len = EXT_UNWRITTEN_MAX_LEN / 2;
3980
3981         ret = ext4_convert_initialized_extents(handle, inode, map,
3982                                                 path);
3983         if (ret >= 0) {
3984                 ext4_update_inode_fsync_trans(handle, inode, 1);
3985                 err = check_eofblocks_fl(handle, inode, map->m_lblk,
3986                                          path, map->m_len);
3987         } else
3988                 err = ret;
3989         map->m_flags |= EXT4_MAP_UNWRITTEN;
3990         if (allocated > map->m_len)
3991                 allocated = map->m_len;
3992         map->m_len = allocated;
3993
3994         return err ? err : allocated;
3995 }
3996
3997 static int
3998 ext4_ext_handle_unwritten_extents(handle_t *handle, struct inode *inode,
3999                         struct ext4_map_blocks *map,
4000                         struct ext4_ext_path *path, int flags,
4001                         unsigned int allocated, ext4_fsblk_t newblock)
4002 {
4003         int ret = 0;
4004         int err = 0;
4005         ext4_io_end_t *io = ext4_inode_aio(inode);
4006
4007         ext_debug("ext4_ext_handle_unwritten_extents: inode %lu, logical "
4008                   "block %llu, max_blocks %u, flags %x, allocated %u\n",
4009                   inode->i_ino, (unsigned long long)map->m_lblk, map->m_len,
4010                   flags, allocated);
4011         ext4_ext_show_leaf(inode, path);
4012
4013         /*
4014          * When writing into unwritten space, we should not fail to
4015          * allocate metadata blocks for the new extent block if needed.
4016          */
4017         flags |= EXT4_GET_BLOCKS_METADATA_NOFAIL;
4018
4019         trace_ext4_ext_handle_unwritten_extents(inode, map, flags,
4020                                                     allocated, newblock);
4021
4022         /* get_block() before submit the IO, split the extent */
4023         if (flags & EXT4_GET_BLOCKS_PRE_IO) {
4024                 ret = ext4_split_convert_extents(handle, inode, map,
4025                                          path, flags | EXT4_GET_BLOCKS_CONVERT);
4026                 if (ret <= 0)
4027                         goto out;
4028                 /*
4029                  * Flag the inode(non aio case) or end_io struct (aio case)
4030                  * that this IO needs to conversion to written when IO is
4031                  * completed
4032                  */
4033                 if (io)
4034                         ext4_set_io_unwritten_flag(inode, io);
4035                 else
4036                         ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
4037                 map->m_flags |= EXT4_MAP_UNWRITTEN;
4038                 goto out;
4039         }
4040         /* IO end_io complete, convert the filled extent to written */
4041         if (flags & EXT4_GET_BLOCKS_CONVERT) {
4042                 ret = ext4_convert_unwritten_extents_endio(handle, inode, map,
4043                                                         path);
4044                 if (ret >= 0) {
4045                         ext4_update_inode_fsync_trans(handle, inode, 1);
4046                         err = check_eofblocks_fl(handle, inode, map->m_lblk,
4047                                                  path, map->m_len);
4048                 } else
4049                         err = ret;
4050                 map->m_flags |= EXT4_MAP_MAPPED;
4051                 map->m_pblk = newblock;
4052                 if (allocated > map->m_len)
4053                         allocated = map->m_len;
4054                 map->m_len = allocated;
4055                 goto out2;
4056         }
4057         /* buffered IO case */
4058         /*
4059          * repeat fallocate creation request
4060          * we already have an unwritten extent
4061          */
4062         if (flags & EXT4_GET_BLOCKS_UNWRIT_EXT) {
4063                 map->m_flags |= EXT4_MAP_UNWRITTEN;
4064                 goto map_out;
4065         }
4066
4067         /* buffered READ or buffered write_begin() lookup */
4068         if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
4069                 /*
4070                  * We have blocks reserved already.  We
4071                  * return allocated blocks so that delalloc
4072                  * won't do block reservation for us.  But
4073                  * the buffer head will be unmapped so that
4074                  * a read from the block returns 0s.
4075                  */
4076                 map->m_flags |= EXT4_MAP_UNWRITTEN;
4077                 goto out1;
4078         }
4079
4080         /* buffered write, writepage time, convert*/
4081         ret = ext4_ext_convert_to_initialized(handle, inode, map, path, flags);
4082         if (ret >= 0)
4083                 ext4_update_inode_fsync_trans(handle, inode, 1);
4084 out:
4085         if (ret <= 0) {
4086                 err = ret;
4087                 goto out2;
4088         } else
4089                 allocated = ret;
4090         map->m_flags |= EXT4_MAP_NEW;
4091         /*
4092          * if we allocated more blocks than requested
4093          * we need to make sure we unmap the extra block
4094          * allocated. The actual needed block will get
4095          * unmapped later when we find the buffer_head marked
4096          * new.
4097          */
4098         if (allocated > map->m_len) {
4099                 unmap_underlying_metadata_blocks(inode->i_sb->s_bdev,
4100                                         newblock + map->m_len,
4101                                         allocated - map->m_len);
4102                 allocated = map->m_len;
4103         }
4104         map->m_len = allocated;
4105
4106         /*
4107          * If we have done fallocate with the offset that is already
4108          * delayed allocated, we would have block reservation
4109          * and quota reservation done in the delayed write path.
4110          * But fallocate would have already updated quota and block
4111          * count for this offset. So cancel these reservation
4112          */
4113         if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) {
4114                 unsigned int reserved_clusters;
4115                 reserved_clusters = get_reserved_cluster_alloc(inode,
4116                                 map->m_lblk, map->m_len);
4117                 if (reserved_clusters)
4118                         ext4_da_update_reserve_space(inode,
4119                                                      reserved_clusters,
4120                                                      0);
4121         }
4122
4123 map_out:
4124         map->m_flags |= EXT4_MAP_MAPPED;
4125         if ((flags & EXT4_GET_BLOCKS_KEEP_SIZE) == 0) {
4126                 err = check_eofblocks_fl(handle, inode, map->m_lblk, path,
4127                                          map->m_len);
4128                 if (err < 0)
4129                         goto out2;
4130         }
4131 out1:
4132         if (allocated > map->m_len)
4133                 allocated = map->m_len;
4134         ext4_ext_show_leaf(inode, path);
4135         map->m_pblk = newblock;
4136         map->m_len = allocated;
4137 out2:
4138         return err ? err : allocated;
4139 }
4140
4141 /*
4142  * get_implied_cluster_alloc - check to see if the requested
4143  * allocation (in the map structure) overlaps with a cluster already
4144  * allocated in an extent.
4145  *      @sb     The filesystem superblock structure
4146  *      @map    The requested lblk->pblk mapping
4147  *      @ex     The extent structure which might contain an implied
4148  *                      cluster allocation
4149  *
4150  * This function is called by ext4_ext_map_blocks() after we failed to
4151  * find blocks that were already in the inode's extent tree.  Hence,
4152  * we know that the beginning of the requested region cannot overlap
4153  * the extent from the inode's extent tree.  There are three cases we
4154  * want to catch.  The first is this case:
4155  *
4156  *               |--- cluster # N--|
4157  *    |--- extent ---|  |---- requested region ---|
4158  *                      |==========|
4159  *
4160  * The second case that we need to test for is this one:
4161  *
4162  *   |--------- cluster # N ----------------|
4163  *         |--- requested region --|   |------- extent ----|
4164  *         |=======================|
4165  *
4166  * The third case is when the requested region lies between two extents
4167  * within the same cluster:
4168  *          |------------- cluster # N-------------|
4169  * |----- ex -----|                  |---- ex_right ----|
4170  *                  |------ requested region ------|
4171  *                  |================|
4172  *
4173  * In each of the above cases, we need to set the map->m_pblk and
4174  * map->m_len so it corresponds to the return the extent labelled as
4175  * "|====|" from cluster #N, since it is already in use for data in
4176  * cluster EXT4_B2C(sbi, map->m_lblk).  We will then return 1 to
4177  * signal to ext4_ext_map_blocks() that map->m_pblk should be treated
4178  * as a new "allocated" block region.  Otherwise, we will return 0 and
4179  * ext4_ext_map_blocks() will then allocate one or more new clusters
4180  * by calling ext4_mb_new_blocks().
4181  */
4182 static int get_implied_cluster_alloc(struct super_block *sb,
4183                                      struct ext4_map_blocks *map,
4184                                      struct ext4_extent *ex,
4185                                      struct ext4_ext_path *path)
4186 {
4187         struct ext4_sb_info *sbi = EXT4_SB(sb);
4188         ext4_lblk_t c_offset = EXT4_LBLK_COFF(sbi, map->m_lblk);
4189         ext4_lblk_t ex_cluster_start, ex_cluster_end;
4190         ext4_lblk_t rr_cluster_start;
4191         ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block);
4192         ext4_fsblk_t ee_start = ext4_ext_pblock(ex);
4193         unsigned short ee_len = ext4_ext_get_actual_len(ex);
4194
4195         /* The extent passed in that we are trying to match */
4196         ex_cluster_start = EXT4_B2C(sbi, ee_block);
4197         ex_cluster_end = EXT4_B2C(sbi, ee_block + ee_len - 1);
4198
4199         /* The requested region passed into ext4_map_blocks() */
4200         rr_cluster_start = EXT4_B2C(sbi, map->m_lblk);
4201
4202         if ((rr_cluster_start == ex_cluster_end) ||
4203             (rr_cluster_start == ex_cluster_start)) {
4204                 if (rr_cluster_start == ex_cluster_end)
4205                         ee_start += ee_len - 1;
4206                 map->m_pblk = EXT4_PBLK_CMASK(sbi, ee_start) + c_offset;
4207                 map->m_len = min(map->m_len,
4208                                  (unsigned) sbi->s_cluster_ratio - c_offset);
4209                 /*
4210                  * Check for and handle this case:
4211                  *
4212                  *   |--------- cluster # N-------------|
4213                  *                     |------- extent ----|
4214                  *         |--- requested region ---|
4215                  *         |===========|
4216                  */
4217
4218                 if (map->m_lblk < ee_block)
4219                         map->m_len = min(map->m_len, ee_block - map->m_lblk);
4220
4221                 /*
4222                  * Check for the case where there is already another allocated
4223                  * block to the right of 'ex' but before the end of the cluster.
4224                  *
4225                  *          |------------- cluster # N-------------|
4226                  * |----- ex -----|                  |---- ex_right ----|
4227                  *                  |------ requested region ------|
4228                  *                  |================|
4229                  */
4230                 if (map->m_lblk > ee_block) {
4231                         ext4_lblk_t next = ext4_ext_next_allocated_block(path);
4232                         map->m_len = min(map->m_len, next - map->m_lblk);
4233                 }
4234
4235                 trace_ext4_get_implied_cluster_alloc_exit(sb, map, 1);
4236                 return 1;
4237         }
4238
4239         trace_ext4_get_implied_cluster_alloc_exit(sb, map, 0);
4240         return 0;
4241 }
4242
4243
4244 /*
4245  * Block allocation/map/preallocation routine for extents based files
4246  *
4247  *
4248  * Need to be called with
4249  * down_read(&EXT4_I(inode)->i_data_sem) if not allocating file system block
4250  * (ie, create is zero). Otherwise down_write(&EXT4_I(inode)->i_data_sem)
4251  *
4252  * return > 0, number of of blocks already mapped/allocated
4253  *          if create == 0 and these are pre-allocated blocks
4254  *              buffer head is unmapped
4255  *          otherwise blocks are mapped
4256  *
4257  * return = 0, if plain look up failed (blocks have not been allocated)
4258  *          buffer head is unmapped
4259  *
4260  * return < 0, error case.
4261  */
4262 int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
4263                         struct ext4_map_blocks *map, int flags)
4264 {
4265         struct ext4_ext_path *path = NULL;
4266         struct ext4_extent newex, *ex, *ex2;
4267         struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
4268         ext4_fsblk_t newblock = 0;
4269         int free_on_err = 0, err = 0, depth, ret;
4270         unsigned int allocated = 0, offset = 0;
4271         unsigned int allocated_clusters = 0;
4272         struct ext4_allocation_request ar;
4273         ext4_io_end_t *io = ext4_inode_aio(inode);
4274         ext4_lblk_t cluster_offset;
4275         int set_unwritten = 0;
4276
4277         ext_debug("blocks %u/%u requested for inode %lu\n",
4278                   map->m_lblk, map->m_len, inode->i_ino);
4279         trace_ext4_ext_map_blocks_enter(inode, map->m_lblk, map->m_len, flags);
4280
4281         /* find extent for this block */
4282         path = ext4_ext_find_extent(inode, map->m_lblk, NULL, 0);
4283         if (IS_ERR(path)) {
4284                 err = PTR_ERR(path);
4285                 path = NULL;
4286                 goto out2;
4287         }
4288
4289         depth = ext_depth(inode);
4290
4291         /*
4292          * consistent leaf must not be empty;
4293          * this situation is possible, though, _during_ tree modification;
4294          * this is why assert can't be put in ext4_ext_find_extent()
4295          */
4296         if (unlikely(path[depth].p_ext == NULL && depth != 0)) {
4297                 EXT4_ERROR_INODE(inode, "bad extent address "
4298                                  "lblock: %lu, depth: %d pblock %lld",
4299                                  (unsigned long) map->m_lblk, depth,
4300                                  path[depth].p_block);
4301                 err = -EIO;
4302                 goto out2;
4303         }
4304
4305         ex = path[depth].p_ext;
4306         if (ex) {
4307                 ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block);
4308                 ext4_fsblk_t ee_start = ext4_ext_pblock(ex);
4309                 unsigned short ee_len;
4310
4311
4312                 /*
4313                  * unwritten extents are treated as holes, except that
4314                  * we split out initialized portions during a write.
4315                  */
4316                 ee_len = ext4_ext_get_actual_len(ex);
4317
4318                 trace_ext4_ext_show_extent(inode, ee_block, ee_start, ee_len);
4319
4320                 /* if found extent covers block, simply return it */
4321                 if (in_range(map->m_lblk, ee_block, ee_len)) {
4322                         newblock = map->m_lblk - ee_block + ee_start;
4323                         /* number of remaining blocks in the extent */
4324                         allocated = ee_len - (map->m_lblk - ee_block);
4325                         ext_debug("%u fit into %u:%d -> %llu\n", map->m_lblk,
4326                                   ee_block, ee_len, newblock);
4327
4328                         /*
4329                          * If the extent is initialized check whether the
4330                          * caller wants to convert it to unwritten.
4331                          */
4332                         if ((!ext4_ext_is_unwritten(ex)) &&
4333                             (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN)) {
4334                                 allocated = ext4_ext_convert_initialized_extent(
4335                                                 handle, inode, map, path, flags,
4336                                                 allocated, newblock);
4337                                 goto out2;
4338                         } else if (!ext4_ext_is_unwritten(ex))
4339                                 goto out;
4340
4341                         ret = ext4_ext_handle_unwritten_extents(
4342                                 handle, inode, map, path, flags,
4343                                 allocated, newblock);
4344                         if (ret < 0)
4345                                 err = ret;
4346                         else
4347                                 allocated = ret;
4348                         goto out2;
4349                 }
4350         }
4351
4352         if ((sbi->s_cluster_ratio > 1) &&
4353             ext4_find_delalloc_cluster(inode, map->m_lblk))
4354                 map->m_flags |= EXT4_MAP_FROM_CLUSTER;
4355
4356         /*
4357          * requested block isn't allocated yet;
4358          * we couldn't try to create block if create flag is zero
4359          */
4360         if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
4361                 /*
4362                  * put just found gap into cache to speed up
4363                  * subsequent requests
4364                  */
4365                 if ((flags & EXT4_GET_BLOCKS_NO_PUT_HOLE) == 0)
4366                         ext4_ext_put_gap_in_cache(inode, path, map->m_lblk);
4367                 goto out2;
4368         }
4369
4370         /*
4371          * Okay, we need to do block allocation.
4372          */
4373         map->m_flags &= ~EXT4_MAP_FROM_CLUSTER;
4374         newex.ee_block = cpu_to_le32(map->m_lblk);
4375         cluster_offset = EXT4_LBLK_COFF(sbi, map->m_lblk);
4376
4377         /*
4378          * If we are doing bigalloc, check to see if the extent returned
4379          * by ext4_ext_find_extent() implies a cluster we can use.
4380          */
4381         if (cluster_offset && ex &&
4382             get_implied_cluster_alloc(inode->i_sb, map, ex, path)) {
4383                 ar.len = allocated = map->m_len;
4384                 newblock = map->m_pblk;
4385                 map->m_flags |= EXT4_MAP_FROM_CLUSTER;
4386                 goto got_allocated_blocks;
4387         }
4388
4389         /* find neighbour allocated blocks */
4390         ar.lleft = map->m_lblk;
4391         err = ext4_ext_search_left(inode, path, &ar.lleft, &ar.pleft);
4392         if (err)
4393                 goto out2;
4394         ar.lright = map->m_lblk;
4395         ex2 = NULL;
4396         err = ext4_ext_search_right(inode, path, &ar.lright, &ar.pright, &ex2);
4397         if (err)
4398                 goto out2;
4399
4400         /* Check if the extent after searching to the right implies a
4401          * cluster we can use. */
4402         if ((sbi->s_cluster_ratio > 1) && ex2 &&
4403             get_implied_cluster_alloc(inode->i_sb, map, ex2, path)) {
4404                 ar.len = allocated = map->m_len;
4405                 newblock = map->m_pblk;
4406                 map->m_flags |= EXT4_MAP_FROM_CLUSTER;
4407                 goto got_allocated_blocks;
4408         }
4409
4410         /*
4411          * See if request is beyond maximum number of blocks we can have in
4412          * a single extent. For an initialized extent this limit is
4413          * EXT_INIT_MAX_LEN and for an unwritten extent this limit is
4414          * EXT_UNWRITTEN_MAX_LEN.
4415          */
4416         if (map->m_len > EXT_INIT_MAX_LEN &&
4417             !(flags & EXT4_GET_BLOCKS_UNWRIT_EXT))
4418                 map->m_len = EXT_INIT_MAX_LEN;
4419         else if (map->m_len > EXT_UNWRITTEN_MAX_LEN &&
4420                  (flags & EXT4_GET_BLOCKS_UNWRIT_EXT))
4421                 map->m_len = EXT_UNWRITTEN_MAX_LEN;
4422
4423         /* Check if we can really insert (m_lblk)::(m_lblk + m_len) extent */
4424         newex.ee_len = cpu_to_le16(map->m_len);
4425         err = ext4_ext_check_overlap(sbi, inode, &newex, path);
4426         if (err)
4427                 allocated = ext4_ext_get_actual_len(&newex);
4428         else
4429                 allocated = map->m_len;
4430
4431         /* allocate new block */
4432         ar.inode = inode;
4433         ar.goal = ext4_ext_find_goal(inode, path, map->m_lblk);
4434         ar.logical = map->m_lblk;
4435         /*
4436          * We calculate the offset from the beginning of the cluster
4437          * for the logical block number, since when we allocate a
4438          * physical cluster, the physical block should start at the
4439          * same offset from the beginning of the cluster.  This is
4440          * needed so that future calls to get_implied_cluster_alloc()
4441          * work correctly.
4442          */
4443         offset = EXT4_LBLK_COFF(sbi, map->m_lblk);
4444         ar.len = EXT4_NUM_B2C(sbi, offset+allocated);
4445         ar.goal -= offset;
4446         ar.logical -= offset;
4447         if (S_ISREG(inode->i_mode))
4448                 ar.flags = EXT4_MB_HINT_DATA;
4449         else
4450                 /* disable in-core preallocation for non-regular files */
4451                 ar.flags = 0;
4452         if (flags & EXT4_GET_BLOCKS_NO_NORMALIZE)
4453                 ar.flags |= EXT4_MB_HINT_NOPREALLOC;
4454         newblock = ext4_mb_new_blocks(handle, &ar, &err);
4455         if (!newblock)
4456                 goto out2;
4457         ext_debug("allocate new block: goal %llu, found %llu/%u\n",
4458                   ar.goal, newblock, allocated);
4459         free_on_err = 1;
4460         allocated_clusters = ar.len;
4461         ar.len = EXT4_C2B(sbi, ar.len) - offset;
4462         if (ar.len > allocated)
4463                 ar.len = allocated;
4464
4465 got_allocated_blocks:
4466         /* try to insert new extent into found leaf and return */
4467         ext4_ext_store_pblock(&newex, newblock + offset);
4468         newex.ee_len = cpu_to_le16(ar.len);
4469         /* Mark unwritten */
4470         if (flags & EXT4_GET_BLOCKS_UNWRIT_EXT){
4471                 ext4_ext_mark_unwritten(&newex);
4472                 map->m_flags |= EXT4_MAP_UNWRITTEN;
4473                 /*
4474                  * io_end structure was created for every IO write to an
4475                  * unwritten extent. To avoid unnecessary conversion,
4476                  * here we flag the IO that really needs the conversion.
4477                  * For non asycn direct IO case, flag the inode state
4478                  * that we need to perform conversion when IO is done.
4479                  */
4480                 if (flags & EXT4_GET_BLOCKS_PRE_IO)
4481                         set_unwritten = 1;
4482         }
4483
4484         err = 0;
4485         if ((flags & EXT4_GET_BLOCKS_KEEP_SIZE) == 0)
4486                 err = check_eofblocks_fl(handle, inode, map->m_lblk,
4487                                          path, ar.len);
4488         if (!err)
4489                 err = ext4_ext_insert_extent(handle, inode, path,
4490                                              &newex, flags);
4491
4492         if (!err && set_unwritten) {
4493                 if (io)
4494                         ext4_set_io_unwritten_flag(inode, io);
4495                 else
4496                         ext4_set_inode_state(inode,
4497                                              EXT4_STATE_DIO_UNWRITTEN);
4498         }
4499
4500         if (err && free_on_err) {
4501                 int fb_flags = flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE ?
4502                         EXT4_FREE_BLOCKS_NO_QUOT_UPDATE : 0;
4503                 /* free data blocks we just allocated */
4504                 /* not a good idea to call discard here directly,
4505                  * but otherwise we'd need to call it every free() */
4506                 ext4_discard_preallocations(inode);
4507                 ext4_free_blocks(handle, inode, NULL, newblock,
4508                                  EXT4_C2B(sbi, allocated_clusters), fb_flags);
4509                 goto out2;
4510         }
4511
4512         /* previous routine could use block we allocated */
4513         newblock = ext4_ext_pblock(&newex);
4514         allocated = ext4_ext_get_actual_len(&newex);
4515         if (allocated > map->m_len)
4516                 allocated = map->m_len;
4517         map->m_flags |= EXT4_MAP_NEW;
4518
4519         /*
4520          * Update reserved blocks/metadata blocks after successful
4521          * block allocation which had been deferred till now.
4522          */
4523         if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) {
4524                 unsigned int reserved_clusters;
4525                 /*
4526                  * Check how many clusters we had reserved this allocated range
4527                  */
4528                 reserved_clusters = get_reserved_cluster_alloc(inode,
4529                                                 map->m_lblk, allocated);
4530                 if (map->m_flags & EXT4_MAP_FROM_CLUSTER) {
4531                         if (reserved_clusters) {
4532                                 /*
4533                                  * We have clusters reserved for this range.
4534                                  * But since we are not doing actual allocation
4535                                  * and are simply using blocks from previously
4536                                  * allocated cluster, we should release the
4537                                  * reservation and not claim quota.
4538                                  */
4539                                 ext4_da_update_reserve_space(inode,
4540                                                 reserved_clusters, 0);
4541                         }
4542                 } else {
4543                         BUG_ON(allocated_clusters < reserved_clusters);
4544                         if (reserved_clusters < allocated_clusters) {
4545                                 struct ext4_inode_info *ei = EXT4_I(inode);
4546                                 int reservation = allocated_clusters -
4547                                                   reserved_clusters;
4548                                 /*
4549                                  * It seems we claimed few clusters outside of
4550                                  * the range of this allocation. We should give
4551                                  * it back to the reservation pool. This can
4552                                  * happen in the following case:
4553                                  *
4554                                  * * Suppose s_cluster_ratio is 4 (i.e., each
4555                                  *   cluster has 4 blocks. Thus, the clusters
4556                                  *   are [0-3],[4-7],[8-11]...
4557                                  * * First comes delayed allocation write for
4558                                  *   logical blocks 10 & 11. Since there were no
4559                                  *   previous delayed allocated blocks in the
4560                                  *   range [8-11], we would reserve 1 cluster
4561                                  *   for this write.
4562                                  * * Next comes write for logical blocks 3 to 8.
4563                                  *   In this case, we will reserve 2 clusters
4564                                  *   (for [0-3] and [4-7]; and not for [8-11] as
4565                                  *   that range has a delayed allocated blocks.
4566                                  *   Thus total reserved clusters now becomes 3.
4567                                  * * Now, during the delayed allocation writeout
4568                                  *   time, we will first write blocks [3-8] and
4569                                  *   allocate 3 clusters for writing these
4570                                  *   blocks. Also, we would claim all these
4571                                  *   three clusters above.
4572                                  * * Now when we come here to writeout the
4573                                  *   blocks [10-11], we would expect to claim
4574                                  *   the reservation of 1 cluster we had made
4575                                  *   (and we would claim it since there are no
4576                                  *   more delayed allocated blocks in the range
4577                                  *   [8-11]. But our reserved cluster count had
4578                                  *   already gone to 0.
4579                                  *
4580                                  *   Thus, at the step 4 above when we determine
4581                                  *   that there are still some unwritten delayed
4582                                  *   allocated blocks outside of our current
4583                                  *   block range, we should increment the
4584                                  *   reserved clusters count so that when the
4585                                  *   remaining blocks finally gets written, we
4586                                  *   could claim them.
4587                                  */
4588                                 dquot_reserve_block(inode,
4589                                                 EXT4_C2B(sbi, reservation));
4590                                 spin_lock(&ei->i_block_reservation_lock);
4591                                 ei->i_reserved_data_blocks += reservation;
4592                                 spin_unlock(&ei->i_block_reservation_lock);
4593                         }
4594                         /*
4595                          * We will claim quota for all newly allocated blocks.
4596                          * We're updating the reserved space *after* the
4597                          * correction above so we do not accidentally free
4598                          * all the metadata reservation because we might
4599                          * actually need it later on.
4600                          */
4601                         ext4_da_update_reserve_space(inode, allocated_clusters,
4602                                                         1);
4603                 }
4604         }
4605
4606         /*
4607          * Cache the extent and update transaction to commit on fdatasync only
4608          * when it is _not_ an unwritten extent.
4609          */
4610         if ((flags & EXT4_GET_BLOCKS_UNWRIT_EXT) == 0)
4611                 ext4_update_inode_fsync_trans(handle, inode, 1);
4612         else
4613                 ext4_update_inode_fsync_trans(handle, inode, 0);
4614 out:
4615         if (allocated > map->m_len)
4616                 allocated = map->m_len;
4617         ext4_ext_show_leaf(inode, path);
4618         map->m_flags |= EXT4_MAP_MAPPED;
4619         map->m_pblk = newblock;
4620         map->m_len = allocated;
4621 out2:
4622         if (path) {
4623                 ext4_ext_drop_refs(path);
4624                 kfree(path);
4625         }
4626
4627         trace_ext4_ext_map_blocks_exit(inode, flags, map,
4628                                        err ? err : allocated);
4629         ext4_es_lru_add(inode);
4630         return err ? err : allocated;
4631 }
4632
4633 void ext4_ext_truncate(handle_t *handle, struct inode *inode)
4634 {
4635         struct super_block *sb = inode->i_sb;
4636         ext4_lblk_t last_block;
4637         int err = 0;
4638
4639         /*
4640          * TODO: optimization is possible here.
4641          * Probably we need not scan at all,
4642          * because page truncation is enough.
4643          */
4644
4645         /* we have to know where to truncate from in crash case */
4646         EXT4_I(inode)->i_disksize = inode->i_size;
4647         ext4_mark_inode_dirty(handle, inode);
4648
4649         last_block = (inode->i_size + sb->s_blocksize - 1)
4650                         >> EXT4_BLOCK_SIZE_BITS(sb);
4651 retry:
4652         err = ext4_es_remove_extent(inode, last_block,
4653                                     EXT_MAX_BLOCKS - last_block);
4654         if (err == -ENOMEM) {
4655                 cond_resched();
4656                 congestion_wait(BLK_RW_ASYNC, HZ/50);
4657                 goto retry;
4658         }
4659         if (err) {
4660                 ext4_std_error(inode->i_sb, err);
4661                 return;
4662         }
4663         err = ext4_ext_remove_space(inode, last_block, EXT_MAX_BLOCKS - 1);
4664         ext4_std_error(inode->i_sb, err);
4665 }
4666
4667 static int ext4_alloc_file_blocks(struct file *file, ext4_lblk_t offset,
4668                                   ext4_lblk_t len, loff_t new_size,
4669                                   int flags, int mode)
4670 {
4671         struct inode *inode = file_inode(file);
4672         handle_t *handle;
4673         int ret = 0;
4674         int ret2 = 0;
4675         int retries = 0;
4676         struct ext4_map_blocks map;
4677         unsigned int credits;
4678         loff_t epos;
4679
4680         map.m_lblk = offset;
4681         map.m_len = len;
4682         /*
4683          * Don't normalize the request if it can fit in one extent so
4684          * that it doesn't get unnecessarily split into multiple
4685          * extents.
4686          */
4687         if (len <= EXT_UNWRITTEN_MAX_LEN)
4688                 flags |= EXT4_GET_BLOCKS_NO_NORMALIZE;
4689
4690         /*
4691          * credits to insert 1 extent into extent tree
4692          */
4693         credits = ext4_chunk_trans_blocks(inode, len);
4694
4695 retry:
4696         while (ret >= 0 && len) {
4697                 handle = ext4_journal_start(inode, EXT4_HT_MAP_BLOCKS,
4698                                             credits);
4699                 if (IS_ERR(handle)) {
4700                         ret = PTR_ERR(handle);
4701                         break;
4702                 }
4703                 ret = ext4_map_blocks(handle, inode, &map, flags);
4704                 if (ret <= 0) {
4705                         ext4_debug("inode #%lu: block %u: len %u: "
4706                                    "ext4_ext_map_blocks returned %d",
4707                                    inode->i_ino, map.m_lblk,
4708                                    map.m_len, ret);
4709                         ext4_mark_inode_dirty(handle, inode);
4710                         ret2 = ext4_journal_stop(handle);
4711                         break;
4712                 }
4713                 map.m_lblk += ret;
4714                 map.m_len = len = len - ret;
4715                 epos = (loff_t)map.m_lblk << inode->i_blkbits;
4716                 inode->i_ctime = ext4_current_time(inode);
4717                 if (new_size) {
4718                         if (epos > new_size)
4719                                 epos = new_size;
4720                         if (ext4_update_inode_size(inode, epos) & 0x1)
4721                                 inode->i_mtime = inode->i_ctime;
4722                 } else {
4723                         if (epos > inode->i_size)
4724                                 ext4_set_inode_flag(inode,
4725                                                     EXT4_INODE_EOFBLOCKS);
4726                 }
4727                 ext4_mark_inode_dirty(handle, inode);
4728                 ret2 = ext4_journal_stop(handle);
4729                 if (ret2)
4730                         break;
4731         }
4732         if (ret == -ENOSPC &&
4733                         ext4_should_retry_alloc(inode->i_sb, &retries)) {
4734                 ret = 0;
4735                 goto retry;
4736         }
4737
4738         return ret > 0 ? ret2 : ret;
4739 }
4740
4741 static long ext4_zero_range(struct file *file, loff_t offset,
4742                             loff_t len, int mode)
4743 {
4744         struct inode *inode = file_inode(file);
4745         handle_t *handle = NULL;
4746         unsigned int max_blocks;
4747         loff_t new_size = 0;
4748         int ret = 0;
4749         int flags;
4750         int credits;
4751         int partial_begin, partial_end;
4752         loff_t start, end;
4753         ext4_lblk_t lblk;
4754         struct address_space *mapping = inode->i_mapping;
4755         unsigned int blkbits = inode->i_blkbits;
4756
4757         trace_ext4_zero_range(inode, offset, len, mode);
4758
4759         if (!S_ISREG(inode->i_mode))
4760                 return -EINVAL;
4761
4762         /* Call ext4_force_commit to flush all data in case of data=journal. */
4763         if (ext4_should_journal_data(inode)) {
4764                 ret = ext4_force_commit(inode->i_sb);
4765                 if (ret)
4766                         return ret;
4767         }
4768
4769         /*
4770          * Write out all dirty pages to avoid race conditions
4771          * Then release them.
4772          */
4773         if (mapping->nrpages && mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) {
4774                 ret = filemap_write_and_wait_range(mapping, offset,
4775                                                    offset + len - 1);
4776                 if (ret)
4777                         return ret;
4778         }
4779
4780         /*
4781          * Round up offset. This is not fallocate, we neet to zero out
4782          * blocks, so convert interior block aligned part of the range to
4783          * unwritten and possibly manually zero out unaligned parts of the
4784          * range.
4785          */
4786         start = round_up(offset, 1 << blkbits);
4787         end = round_down((offset + len), 1 << blkbits);
4788
4789         if (start < offset || end > offset + len)
4790                 return -EINVAL;
4791         partial_begin = offset & ((1 << blkbits) - 1);
4792         partial_end = (offset + len) & ((1 << blkbits) - 1);
4793
4794         lblk = start >> blkbits;
4795         max_blocks = (end >> blkbits);
4796         if (max_blocks < lblk)
4797                 max_blocks = 0;
4798         else
4799                 max_blocks -= lblk;
4800
4801         flags = EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT |
4802                 EXT4_GET_BLOCKS_CONVERT_UNWRITTEN;
4803         if (mode & FALLOC_FL_KEEP_SIZE)
4804                 flags |= EXT4_GET_BLOCKS_KEEP_SIZE;
4805
4806         mutex_lock(&inode->i_mutex);
4807
4808         /*
4809          * Indirect files do not support unwritten extnets
4810          */
4811         if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) {
4812                 ret = -EOPNOTSUPP;
4813                 goto out_mutex;
4814         }
4815
4816         if (!(mode & FALLOC_FL_KEEP_SIZE) &&
4817              offset + len > i_size_read(inode)) {
4818                 new_size = offset + len;
4819                 ret = inode_newsize_ok(inode, new_size);
4820                 if (ret)
4821                         goto out_mutex;
4822                 /*
4823                  * If we have a partial block after EOF we have to allocate
4824                  * the entire block.
4825                  */
4826                 if (partial_end)
4827                         max_blocks += 1;
4828         }
4829
4830         if (max_blocks > 0) {
4831
4832                 /* Now release the pages and zero block aligned part of pages*/
4833                 truncate_pagecache_range(inode, start, end - 1);
4834                 inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
4835
4836                 /* Wait all existing dio workers, newcomers will block on i_mutex */
4837                 ext4_inode_block_unlocked_dio(inode);
4838                 inode_dio_wait(inode);
4839
4840                 /*
4841                  * Remove entire range from the extent status tree.
4842                  */
4843                 ret = ext4_es_remove_extent(inode, lblk, max_blocks);
4844                 if (ret)
4845                         goto out_dio;
4846
4847                 ret = ext4_alloc_file_blocks(file, lblk, max_blocks, new_size,
4848                                              flags, mode);
4849                 if (ret)
4850                         goto out_dio;
4851         }
4852         if (!partial_begin && !partial_end)
4853                 goto out_dio;
4854
4855         /*
4856          * In worst case we have to writeout two nonadjacent unwritten
4857          * blocks and update the inode
4858          */
4859         credits = (2 * ext4_ext_index_trans_blocks(inode, 2)) + 1;
4860         if (ext4_should_journal_data(inode))
4861                 credits += 2;
4862         handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
4863         if (IS_ERR(handle)) {
4864                 ret = PTR_ERR(handle);
4865                 ext4_std_error(inode->i_sb, ret);
4866                 goto out_dio;
4867         }
4868
4869         inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
4870         if (new_size) {
4871                 ext4_update_inode_size(inode, new_size);
4872         } else {
4873                 /*
4874                 * Mark that we allocate beyond EOF so the subsequent truncate
4875                 * can proceed even if the new size is the same as i_size.
4876                 */
4877                 if ((offset + len) > i_size_read(inode))
4878                         ext4_set_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
4879         }
4880         ext4_mark_inode_dirty(handle, inode);
4881
4882         /* Zero out partial block at the edges of the range */
4883         ret = ext4_zero_partial_blocks(handle, inode, offset, len);
4884
4885         if (file->f_flags & O_SYNC)
4886                 ext4_handle_sync(handle);
4887
4888         ext4_journal_stop(handle);
4889 out_dio:
4890         ext4_inode_resume_unlocked_dio(inode);
4891 out_mutex:
4892         mutex_unlock(&inode->i_mutex);
4893         return ret;
4894 }
4895
4896 /*
4897  * preallocate space for a file. This implements ext4's fallocate file
4898  * operation, which gets called from sys_fallocate system call.
4899  * For block-mapped files, posix_fallocate should fall back to the method
4900  * of writing zeroes to the required new blocks (the same behavior which is
4901  * expected for file systems which do not support fallocate() system call).
4902  */
4903 long ext4_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
4904 {
4905         struct inode *inode = file_inode(file);
4906         loff_t new_size = 0;
4907         unsigned int max_blocks;
4908         int ret = 0;
4909         int flags;
4910         ext4_lblk_t lblk;
4911         unsigned int blkbits = inode->i_blkbits;
4912
4913         /* Return error if mode is not supported */
4914         if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
4915                      FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE))
4916                 return -EOPNOTSUPP;
4917
4918         if (mode & FALLOC_FL_PUNCH_HOLE)
4919                 return ext4_punch_hole(inode, offset, len);
4920
4921         ret = ext4_convert_inline_data(inode);
4922         if (ret)
4923                 return ret;
4924
4925         /*
4926          * currently supporting (pre)allocate mode for extent-based
4927          * files _only_
4928          */
4929         if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
4930                 return -EOPNOTSUPP;
4931
4932         if (mode & FALLOC_FL_COLLAPSE_RANGE)
4933                 return ext4_collapse_range(inode, offset, len);
4934
4935         if (mode & FALLOC_FL_ZERO_RANGE)
4936                 return ext4_zero_range(file, offset, len, mode);
4937
4938         trace_ext4_fallocate_enter(inode, offset, len, mode);
4939         lblk = offset >> blkbits;
4940         /*
4941          * We can't just convert len to max_blocks because
4942          * If blocksize = 4096 offset = 3072 and len = 2048
4943          */
4944         max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
4945                 - lblk;
4946
4947         flags = EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT;
4948         if (mode & FALLOC_FL_KEEP_SIZE)
4949                 flags |= EXT4_GET_BLOCKS_KEEP_SIZE;
4950
4951         mutex_lock(&inode->i_mutex);
4952
4953         if (!(mode & FALLOC_FL_KEEP_SIZE) &&
4954              offset + len > i_size_read(inode)) {
4955                 new_size = offset + len;
4956                 ret = inode_newsize_ok(inode, new_size);
4957                 if (ret)
4958                         goto out;
4959         }
4960
4961         ret = ext4_alloc_file_blocks(file, lblk, max_blocks, new_size,
4962                                      flags, mode);
4963         if (ret)
4964                 goto out;
4965
4966         if (file->f_flags & O_SYNC && EXT4_SB(inode->i_sb)->s_journal) {
4967                 ret = jbd2_complete_transaction(EXT4_SB(inode->i_sb)->s_journal,
4968                                                 EXT4_I(inode)->i_sync_tid);
4969         }
4970 out:
4971         mutex_unlock(&inode->i_mutex);
4972         trace_ext4_fallocate_exit(inode, offset, max_blocks, ret);
4973         return ret;
4974 }
4975
4976 /*
4977  * This function convert a range of blocks to written extents
4978  * The caller of this function will pass the start offset and the size.
4979  * all unwritten extents within this range will be converted to
4980  * written extents.
4981  *
4982  * This function is called from the direct IO end io call back
4983  * function, to convert the fallocated extents after IO is completed.
4984  * Returns 0 on success.
4985  */
4986 int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
4987                                    loff_t offset, ssize_t len)
4988 {
4989         unsigned int max_blocks;
4990         int ret = 0;
4991         int ret2 = 0;
4992         struct ext4_map_blocks map;
4993         unsigned int credits, blkbits = inode->i_blkbits;
4994
4995         map.m_lblk = offset >> blkbits;
4996         /*
4997          * We can't just convert len to max_blocks because
4998          * If blocksize = 4096 offset = 3072 and len = 2048
4999          */
5000         max_blocks = ((EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) -
5001                       map.m_lblk);
5002         /*
5003          * This is somewhat ugly but the idea is clear: When transaction is
5004          * reserved, everything goes into it. Otherwise we rather start several
5005          * smaller transactions for conversion of each extent separately.
5006          */
5007         if (handle) {
5008                 handle = ext4_journal_start_reserved(handle,
5009                                                      EXT4_HT_EXT_CONVERT);
5010                 if (IS_ERR(handle))
5011                         return PTR_ERR(handle);
5012                 credits = 0;
5013         } else {
5014                 /*
5015                  * credits to insert 1 extent into extent tree
5016                  */
5017                 credits = ext4_chunk_trans_blocks(inode, max_blocks);
5018         }
5019         while (ret >= 0 && ret < max_blocks) {
5020                 map.m_lblk += ret;
5021                 map.m_len = (max_blocks -= ret);
5022                 if (credits) {
5023                         handle = ext4_journal_start(inode, EXT4_HT_MAP_BLOCKS,
5024                                                     credits);
5025                         if (IS_ERR(handle)) {
5026                                 ret = PTR_ERR(handle);
5027                                 break;
5028                         }
5029                 }
5030                 ret = ext4_map_blocks(handle, inode, &map,
5031                                       EXT4_GET_BLOCKS_IO_CONVERT_EXT);
5032                 if (ret <= 0)
5033                         ext4_warning(inode->i_sb,
5034                                      "inode #%lu: block %u: len %u: "
5035                                      "ext4_ext_map_blocks returned %d",
5036                                      inode->i_ino, map.m_lblk,
5037                                      map.m_len, ret);
5038                 ext4_mark_inode_dirty(handle, inode);
5039                 if (credits)
5040                         ret2 = ext4_journal_stop(handle);
5041                 if (ret <= 0 || ret2)
5042                         break;
5043         }
5044         if (!credits)
5045                 ret2 = ext4_journal_stop(handle);
5046         return ret > 0 ? ret2 : ret;
5047 }
5048
5049 /*
5050  * If newes is not existing extent (newes->ec_pblk equals zero) find
5051  * delayed extent at start of newes and update newes accordingly and
5052  * return start of the next delayed extent.
5053  *
5054  * If newes is existing extent (newes->ec_pblk is not equal zero)
5055  * return start of next delayed extent or EXT_MAX_BLOCKS if no delayed
5056  * extent found. Leave newes unmodified.
5057  */
5058 static int ext4_find_delayed_extent(struct inode *inode,
5059                                     struct extent_status *newes)
5060 {
5061         struct extent_status es;
5062         ext4_lblk_t block, next_del;
5063
5064         if (newes->es_pblk == 0) {
5065                 ext4_es_find_delayed_extent_range(inode, newes->es_lblk,
5066                                 newes->es_lblk + newes->es_len - 1, &es);
5067
5068                 /*
5069                  * No extent in extent-tree contains block @newes->es_pblk,
5070                  * then the block may stay in 1)a hole or 2)delayed-extent.
5071                  */
5072                 if (es.es_len == 0)
5073                         /* A hole found. */
5074                         return 0;
5075
5076                 if (es.es_lblk > newes->es_lblk) {
5077                         /* A hole found. */
5078                         newes->es_len = min(es.es_lblk - newes->es_lblk,
5079                                             newes->es_len);
5080                         return 0;
5081                 }
5082
5083                 newes->es_len = es.es_lblk + es.es_len - newes->es_lblk;
5084         }
5085
5086         block = newes->es_lblk + newes->es_len;
5087         ext4_es_find_delayed_extent_range(inode, block, EXT_MAX_BLOCKS, &es);
5088         if (es.es_len == 0)
5089                 next_del = EXT_MAX_BLOCKS;
5090         else
5091                 next_del = es.es_lblk;
5092
5093         return next_del;
5094 }
5095 /* fiemap flags we can handle specified here */
5096 #define EXT4_FIEMAP_FLAGS       (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
5097
5098 static int ext4_xattr_fiemap(struct inode *inode,
5099                                 struct fiemap_extent_info *fieinfo)
5100 {
5101         __u64 physical = 0;
5102         __u64 length;
5103         __u32 flags = FIEMAP_EXTENT_LAST;
5104         int blockbits = inode->i_sb->s_blocksize_bits;
5105         int error = 0;
5106
5107         /* in-inode? */
5108         if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
5109                 struct ext4_iloc iloc;
5110                 int offset;     /* offset of xattr in inode */
5111
5112                 error = ext4_get_inode_loc(inode, &iloc);
5113                 if (error)
5114                         return error;
5115                 physical = (__u64)iloc.bh->b_blocknr << blockbits;
5116                 offset = EXT4_GOOD_OLD_INODE_SIZE +
5117                                 EXT4_I(inode)->i_extra_isize;
5118                 physical += offset;
5119                 length = EXT4_SB(inode->i_sb)->s_inode_size - offset;
5120                 flags |= FIEMAP_EXTENT_DATA_INLINE;
5121                 brelse(iloc.bh);
5122         } else { /* external block */
5123                 physical = (__u64)EXT4_I(inode)->i_file_acl << blockbits;
5124                 length = inode->i_sb->s_blocksize;
5125         }
5126
5127         if (physical)
5128                 error = fiemap_fill_next_extent(fieinfo, 0, physical,
5129                                                 length, flags);
5130         return (error < 0 ? error : 0);
5131 }
5132
5133 int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
5134                 __u64 start, __u64 len)
5135 {
5136         ext4_lblk_t start_blk;
5137         int error = 0;
5138
5139         if (ext4_has_inline_data(inode)) {
5140                 int has_inline = 1;
5141
5142                 error = ext4_inline_data_fiemap(inode, fieinfo, &has_inline);
5143
5144                 if (has_inline)
5145                         return error;
5146         }
5147
5148         if (fieinfo->fi_flags & FIEMAP_FLAG_CACHE) {
5149                 error = ext4_ext_precache(inode);
5150                 if (error)
5151                         return error;
5152         }
5153
5154         /* fallback to generic here if not in extents fmt */
5155         if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
5156                 return generic_block_fiemap(inode, fieinfo, start, len,
5157                         ext4_get_block);
5158
5159         if (fiemap_check_flags(fieinfo, EXT4_FIEMAP_FLAGS))
5160                 return -EBADR;
5161
5162         if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
5163                 error = ext4_xattr_fiemap(inode, fieinfo);
5164         } else {
5165                 ext4_lblk_t len_blks;
5166                 __u64 last_blk;
5167
5168                 start_blk = start >> inode->i_sb->s_blocksize_bits;
5169                 last_blk = (start + len - 1) >> inode->i_sb->s_blocksize_bits;
5170                 if (last_blk >= EXT_MAX_BLOCKS)
5171                         last_blk = EXT_MAX_BLOCKS-1;
5172                 len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1;
5173
5174                 /*
5175                  * Walk the extent tree gathering extent information
5176                  * and pushing extents back to the user.
5177                  */
5178                 error = ext4_fill_fiemap_extents(inode, start_blk,
5179                                                  len_blks, fieinfo);
5180         }
5181         ext4_es_lru_add(inode);
5182         return error;
5183 }
5184
5185 /*
5186  * ext4_access_path:
5187  * Function to access the path buffer for marking it dirty.
5188  * It also checks if there are sufficient credits left in the journal handle
5189  * to update path.
5190  */
5191 static int
5192 ext4_access_path(handle_t *handle, struct inode *inode,
5193                 struct ext4_ext_path *path)
5194 {
5195         int credits, err;
5196
5197         if (!ext4_handle_valid(handle))
5198                 return 0;
5199
5200         /*
5201          * Check if need to extend journal credits
5202          * 3 for leaf, sb, and inode plus 2 (bmap and group
5203          * descriptor) for each block group; assume two block
5204          * groups
5205          */
5206         if (handle->h_buffer_credits < 7) {
5207                 credits = ext4_writepage_trans_blocks(inode);
5208                 err = ext4_ext_truncate_extend_restart(handle, inode, credits);
5209                 /* EAGAIN is success */
5210                 if (err && err != -EAGAIN)
5211                         return err;
5212         }
5213
5214         err = ext4_ext_get_access(handle, inode, path);
5215         return err;
5216 }
5217
5218 /*
5219  * ext4_ext_shift_path_extents:
5220  * Shift the extents of a path structure lying between path[depth].p_ext
5221  * and EXT_LAST_EXTENT(path[depth].p_hdr) downwards, by subtracting shift
5222  * from starting block for each extent.
5223  */
5224 static int
5225 ext4_ext_shift_path_extents(struct ext4_ext_path *path, ext4_lblk_t shift,
5226                             struct inode *inode, handle_t *handle,
5227                             ext4_lblk_t *start)
5228 {
5229         int depth, err = 0;
5230         struct ext4_extent *ex_start, *ex_last;
5231         bool update = 0;
5232         depth = path->p_depth;
5233
5234         while (depth >= 0) {
5235                 if (depth == path->p_depth) {
5236                         ex_start = path[depth].p_ext;
5237                         if (!ex_start)
5238                                 return -EIO;
5239
5240                         ex_last = EXT_LAST_EXTENT(path[depth].p_hdr);
5241                         if (!ex_last)
5242                                 return -EIO;
5243
5244                         err = ext4_access_path(handle, inode, path + depth);
5245                         if (err)
5246                                 goto out;
5247
5248                         if (ex_start == EXT_FIRST_EXTENT(path[depth].p_hdr))
5249                                 update = 1;
5250
5251                         *start = le32_to_cpu(ex_last->ee_block) +
5252                                 ext4_ext_get_actual_len(ex_last);
5253
5254                         while (ex_start <= ex_last) {
5255                                 le32_add_cpu(&ex_start->ee_block, -shift);
5256                                 /* Try to merge to the left. */
5257                                 if ((ex_start >
5258                                      EXT_FIRST_EXTENT(path[depth].p_hdr)) &&
5259                                     ext4_ext_try_to_merge_right(inode,
5260                                                         path, ex_start - 1))
5261                                         ex_last--;
5262                                 else
5263                                         ex_start++;
5264                         }
5265                         err = ext4_ext_dirty(handle, inode, path + depth);
5266                         if (err)
5267                                 goto out;
5268
5269                         if (--depth < 0 || !update)
5270                                 break;
5271                 }
5272
5273                 /* Update index too */
5274                 err = ext4_access_path(handle, inode, path + depth);
5275                 if (err)
5276                         goto out;
5277
5278                 le32_add_cpu(&path[depth].p_idx->ei_block, -shift);
5279                 err = ext4_ext_dirty(handle, inode, path + depth);
5280                 if (err)
5281                         goto out;
5282
5283                 /* we are done if current index is not a starting index */
5284                 if (path[depth].p_idx != EXT_FIRST_INDEX(path[depth].p_hdr))
5285                         break;
5286
5287                 depth--;
5288         }
5289
5290 out:
5291         return err;
5292 }
5293
5294 /*
5295  * ext4_ext_shift_extents:
5296  * All the extents which lies in the range from start to the last allocated
5297  * block for the file are shifted downwards by shift blocks.
5298  * On success, 0 is returned, error otherwise.
5299  */
5300 static int
5301 ext4_ext_shift_extents(struct inode *inode, handle_t *handle,
5302                        ext4_lblk_t start, ext4_lblk_t shift)
5303 {
5304         struct ext4_ext_path *path;
5305         int ret = 0, depth;
5306         struct ext4_extent *extent;
5307         ext4_lblk_t stop_block, current_block;
5308         ext4_lblk_t ex_start, ex_end;
5309
5310         /* Let path point to the last extent */
5311         path = ext4_ext_find_extent(inode, EXT_MAX_BLOCKS - 1, NULL, 0);
5312         if (IS_ERR(path))
5313                 return PTR_ERR(path);
5314
5315         depth = path->p_depth;
5316         extent = path[depth].p_ext;
5317         if (!extent) {
5318                 ext4_ext_drop_refs(path);
5319                 kfree(path);
5320                 return ret;
5321         }
5322
5323         stop_block = le32_to_cpu(extent->ee_block) +
5324                         ext4_ext_get_actual_len(extent);
5325         ext4_ext_drop_refs(path);
5326         kfree(path);
5327
5328         /* Nothing to shift, if hole is at the end of file */
5329         if (start >= stop_block)
5330                 return ret;
5331
5332         /*
5333          * Don't start shifting extents until we make sure the hole is big
5334          * enough to accomodate the shift.
5335          */
5336         path = ext4_ext_find_extent(inode, start - 1, NULL, 0);
5337         if (IS_ERR(path))
5338                 return PTR_ERR(path);
5339         depth = path->p_depth;
5340         extent =  path[depth].p_ext;
5341         if (extent) {
5342                 ex_start = le32_to_cpu(extent->ee_block);
5343                 ex_end = le32_to_cpu(extent->ee_block) +
5344                         ext4_ext_get_actual_len(extent);
5345         } else {
5346                 ex_start = 0;
5347                 ex_end = 0;
5348         }
5349         ext4_ext_drop_refs(path);
5350         kfree(path);
5351
5352         if ((start == ex_start && shift > ex_start) ||
5353             (shift > start - ex_end))
5354                 return -EINVAL;
5355
5356         /* Its safe to start updating extents */
5357         while (start < stop_block) {
5358                 path = ext4_ext_find_extent(inode, start, NULL, 0);
5359                 if (IS_ERR(path))
5360                         return PTR_ERR(path);
5361                 depth = path->p_depth;
5362                 extent = path[depth].p_ext;
5363                 if (!extent) {
5364                         EXT4_ERROR_INODE(inode, "unexpected hole at %lu",
5365                                          (unsigned long) start);
5366                         return -EIO;
5367                 }
5368
5369                 current_block = le32_to_cpu(extent->ee_block);
5370                 if (start > current_block) {
5371                         /* Hole, move to the next extent */
5372                         ret = mext_next_extent(inode, path, &extent);
5373                         if (ret != 0) {
5374                                 ext4_ext_drop_refs(path);
5375                                 kfree(path);
5376                                 if (ret == 1)
5377                                         ret = 0;
5378                                 break;
5379                         }
5380                 }
5381                 ret = ext4_ext_shift_path_extents(path, shift, inode,
5382                                 handle, &start);
5383                 ext4_ext_drop_refs(path);
5384                 kfree(path);
5385                 if (ret)
5386                         break;
5387         }
5388
5389         return ret;
5390 }
5391
5392 /*
5393  * ext4_collapse_range:
5394  * This implements the fallocate's collapse range functionality for ext4
5395  * Returns: 0 and non-zero on error.
5396  */
5397 int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len)
5398 {
5399         struct super_block *sb = inode->i_sb;
5400         ext4_lblk_t punch_start, punch_stop;
5401         handle_t *handle;
5402         unsigned int credits;
5403         loff_t new_size, ioffset;
5404         int ret;
5405
5406         /* Collapse range works only on fs block size aligned offsets. */
5407         if (offset & (EXT4_CLUSTER_SIZE(sb) - 1) ||
5408             len & (EXT4_CLUSTER_SIZE(sb) - 1))
5409                 return -EINVAL;
5410
5411         if (!S_ISREG(inode->i_mode))
5412                 return -EINVAL;
5413
5414         trace_ext4_collapse_range(inode, offset, len);
5415
5416         punch_start = offset >> EXT4_BLOCK_SIZE_BITS(sb);
5417         punch_stop = (offset + len) >> EXT4_BLOCK_SIZE_BITS(sb);
5418
5419         /* Call ext4_force_commit to flush all data in case of data=journal. */
5420         if (ext4_should_journal_data(inode)) {
5421                 ret = ext4_force_commit(inode->i_sb);
5422                 if (ret)
5423                         return ret;
5424         }
5425
5426         /*
5427          * Need to round down offset to be aligned with page size boundary
5428          * for page size > block size.
5429          */
5430         ioffset = round_down(offset, PAGE_SIZE);
5431
5432         /* Write out all dirty pages */
5433         ret = filemap_write_and_wait_range(inode->i_mapping, ioffset,
5434                                            LLONG_MAX);
5435         if (ret)
5436                 return ret;
5437
5438         /* Take mutex lock */
5439         mutex_lock(&inode->i_mutex);
5440
5441         /*
5442          * There is no need to overlap collapse range with EOF, in which case
5443          * it is effectively a truncate operation
5444          */
5445         if (offset + len >= i_size_read(inode)) {
5446                 ret = -EINVAL;
5447                 goto out_mutex;
5448         }
5449
5450         /* Currently just for extent based files */
5451         if (!ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
5452                 ret = -EOPNOTSUPP;
5453                 goto out_mutex;
5454         }
5455
5456         truncate_pagecache(inode, ioffset);
5457
5458         /* Wait for existing dio to complete */
5459         ext4_inode_block_unlocked_dio(inode);
5460         inode_dio_wait(inode);
5461
5462         credits = ext4_writepage_trans_blocks(inode);
5463         handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE, credits);
5464         if (IS_ERR(handle)) {
5465                 ret = PTR_ERR(handle);
5466                 goto out_dio;
5467         }
5468
5469         down_write(&EXT4_I(inode)->i_data_sem);
5470         ext4_discard_preallocations(inode);
5471
5472         ret = ext4_es_remove_extent(inode, punch_start,
5473                                     EXT_MAX_BLOCKS - punch_start);
5474         if (ret) {
5475                 up_write(&EXT4_I(inode)->i_data_sem);
5476                 goto out_stop;
5477         }
5478
5479         ret = ext4_ext_remove_space(inode, punch_start, punch_stop - 1);
5480         if (ret) {
5481                 up_write(&EXT4_I(inode)->i_data_sem);
5482                 goto out_stop;
5483         }
5484         ext4_discard_preallocations(inode);
5485
5486         ret = ext4_ext_shift_extents(inode, handle, punch_stop,
5487                                      punch_stop - punch_start);
5488         if (ret) {
5489                 up_write(&EXT4_I(inode)->i_data_sem);
5490                 goto out_stop;
5491         }
5492
5493         new_size = i_size_read(inode) - len;
5494         i_size_write(inode, new_size);
5495         EXT4_I(inode)->i_disksize = new_size;
5496
5497         up_write(&EXT4_I(inode)->i_data_sem);
5498         if (IS_SYNC(inode))
5499                 ext4_handle_sync(handle);
5500         inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
5501         ext4_mark_inode_dirty(handle, inode);
5502
5503 out_stop:
5504         ext4_journal_stop(handle);
5505 out_dio:
5506         ext4_inode_resume_unlocked_dio(inode);
5507 out_mutex:
5508         mutex_unlock(&inode->i_mutex);
5509         return ret;
5510 }