RPC/RDMA: check selected memory registration mode at runtime.
[safe/jmp/linux-2.6] / fs / ext4 / balloc.c
index 8d59eec..e9fa960 100644 (file)
 #include <linux/capability.h>
 #include <linux/fs.h>
 #include <linux/jbd2.h>
-#include <linux/ext4_fs.h>
-#include <linux/ext4_jbd2.h>
 #include <linux/quotaops.h>
 #include <linux/buffer_head.h>
+#include "ext4.h"
+#include "ext4_jbd2.h"
+#include "group.h"
 
 /*
  * balloc.c contains the blocks allocation and deallocation routines
@@ -28,7 +29,7 @@
  * Calculate the block group number and offset, given a block number
  */
 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
-               unsigned long *blockgrpp, ext4_grpblk_t *offsetp)
+               ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
 {
        struct ext4_super_block *es = EXT4_SB(sb)->s_es;
        ext4_grpblk_t offset;
@@ -42,6 +43,142 @@ void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
 
 }
 
+static int ext4_block_in_group(struct super_block *sb, ext4_fsblk_t block,
+                       ext4_group_t block_group)
+{
+       ext4_group_t actual_group;
+       ext4_get_group_no_and_offset(sb, block, &actual_group, NULL);
+       if (actual_group == block_group)
+               return 1;
+       return 0;
+}
+
+static int ext4_group_used_meta_blocks(struct super_block *sb,
+                               ext4_group_t block_group)
+{
+       ext4_fsblk_t tmp;
+       struct ext4_sb_info *sbi = EXT4_SB(sb);
+       /* block bitmap, inode bitmap, and inode table blocks */
+       int used_blocks = sbi->s_itb_per_group + 2;
+
+       if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
+               struct ext4_group_desc *gdp;
+               struct buffer_head *bh;
+
+               gdp = ext4_get_group_desc(sb, block_group, &bh);
+               if (!ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp),
+                                       block_group))
+                       used_blocks--;
+
+               if (!ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp),
+                                       block_group))
+                       used_blocks--;
+
+               tmp = ext4_inode_table(sb, gdp);
+               for (; tmp < ext4_inode_table(sb, gdp) +
+                               sbi->s_itb_per_group; tmp++) {
+                       if (!ext4_block_in_group(sb, tmp, block_group))
+                               used_blocks -= 1;
+               }
+       }
+       return used_blocks;
+}
+/* Initializes an uninitialized block bitmap if given, and returns the
+ * number of blocks free in the group. */
+unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
+                ext4_group_t block_group, struct ext4_group_desc *gdp)
+{
+       int bit, bit_max;
+       unsigned free_blocks, group_blocks;
+       struct ext4_sb_info *sbi = EXT4_SB(sb);
+
+       if (bh) {
+               J_ASSERT_BH(bh, buffer_locked(bh));
+
+               /* If checksum is bad mark all blocks used to prevent allocation
+                * essentially implementing a per-group read-only flag. */
+               if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) {
+                       ext4_error(sb, __func__,
+                                 "Checksum bad for group %lu\n", block_group);
+                       gdp->bg_free_blocks_count = 0;
+                       gdp->bg_free_inodes_count = 0;
+                       gdp->bg_itable_unused = 0;
+                       memset(bh->b_data, 0xff, sb->s_blocksize);
+                       return 0;
+               }
+               memset(bh->b_data, 0, sb->s_blocksize);
+       }
+
+       /* Check for superblock and gdt backups in this group */
+       bit_max = ext4_bg_has_super(sb, block_group);
+
+       if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
+           block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
+                         sbi->s_desc_per_block) {
+               if (bit_max) {
+                       bit_max += ext4_bg_num_gdb(sb, block_group);
+                       bit_max +=
+                               le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
+               }
+       } else { /* For META_BG_BLOCK_GROUPS */
+               bit_max += ext4_bg_num_gdb(sb, block_group);
+       }
+
+       if (block_group == sbi->s_groups_count - 1) {
+               /*
+                * Even though mke2fs always initialize first and last group
+                * if some other tool enabled the EXT4_BG_BLOCK_UNINIT we need
+                * to make sure we calculate the right free blocks
+                */
+               group_blocks = ext4_blocks_count(sbi->s_es) -
+                       le32_to_cpu(sbi->s_es->s_first_data_block) -
+                       (EXT4_BLOCKS_PER_GROUP(sb) * (sbi->s_groups_count -1));
+       } else {
+               group_blocks = EXT4_BLOCKS_PER_GROUP(sb);
+       }
+
+       free_blocks = group_blocks - bit_max;
+
+       if (bh) {
+               ext4_fsblk_t start, tmp;
+               int flex_bg = 0;
+
+               for (bit = 0; bit < bit_max; bit++)
+                       ext4_set_bit(bit, bh->b_data);
+
+               start = ext4_group_first_block_no(sb, block_group);
+
+               if (EXT4_HAS_INCOMPAT_FEATURE(sb,
+                                             EXT4_FEATURE_INCOMPAT_FLEX_BG))
+                       flex_bg = 1;
+
+               /* Set bits for block and inode bitmaps, and inode table */
+               tmp = ext4_block_bitmap(sb, gdp);
+               if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
+                       ext4_set_bit(tmp - start, bh->b_data);
+
+               tmp = ext4_inode_bitmap(sb, gdp);
+               if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
+                       ext4_set_bit(tmp - start, bh->b_data);
+
+               tmp = ext4_inode_table(sb, gdp);
+               for (; tmp < ext4_inode_table(sb, gdp) +
+                               sbi->s_itb_per_group; tmp++) {
+                       if (!flex_bg ||
+                               ext4_block_in_group(sb, tmp, block_group))
+                               ext4_set_bit(tmp - start, bh->b_data);
+               }
+               /*
+                * Also if the number of blocks within the group is
+                * less than the blocksize * 8 ( which is the size
+                * of bitmap ), set rest of the block bitmap to 1
+                */
+               mark_bitmap_end(group_blocks, sb->s_blocksize * 8, bh->b_data);
+       }
+       return free_blocks - ext4_group_used_meta_blocks(sb, block_group);
+}
+
+
 /*
  * The free blocks are managed by bitmaps.  A file system contains several
  * blocks groups.  Each group contains 1 bitmap block for blocks, 1 bitmap
@@ -64,7 +201,7 @@ void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
  *                     group descriptor
  */
 struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
-                                            unsigned int block_group,
+                                            ext4_group_t block_group,
                                             struct buffer_head ** bh)
 {
        unsigned long group_desc;
@@ -75,7 +212,7 @@ struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
        if (block_group >= sbi->s_groups_count) {
                ext4_error (sb, "ext4_get_group_desc",
                            "block_group >= groups_count - "
-                           "block_group = %d, groups_count = %lu",
+                           "block_group = %lu, groups_count = %lu",
                            block_group, sbi->s_groups_count);
 
                return NULL;
@@ -87,7 +224,7 @@ struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
        if (!sbi->s_group_desc[group_desc]) {
                ext4_error (sb, "ext4_get_group_desc",
                            "Group descriptor not loaded - "
-                           "block_group = %d, group_desc = %lu, desc = %lu",
+                           "block_group = %lu, group_desc = %lu, desc = %lu",
                             block_group, group_desc, offset);
                return NULL;
        }
@@ -100,33 +237,112 @@ struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
        return desc;
 }
 
+static int ext4_valid_block_bitmap(struct super_block *sb,
+                                       struct ext4_group_desc *desc,
+                                       unsigned int block_group,
+                                       struct buffer_head *bh)
+{
+       ext4_grpblk_t offset;
+       ext4_grpblk_t next_zero_bit;
+       ext4_fsblk_t bitmap_blk;
+       ext4_fsblk_t group_first_block;
+
+       if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
+               /* with FLEX_BG, the inode/block bitmaps and itable
+                * blocks may not be in the group at all
+                * so the bitmap validation will be skipped for those groups
+                * or it has to also read the block group where the bitmaps
+                * are located to verify they are set.
+                */
+               return 1;
+       }
+       group_first_block = ext4_group_first_block_no(sb, block_group);
+
+       /* check whether block bitmap block number is set */
+       bitmap_blk = ext4_block_bitmap(sb, desc);
+       offset = bitmap_blk - group_first_block;
+       if (!ext4_test_bit(offset, bh->b_data))
+               /* bad block bitmap */
+               goto err_out;
+
+       /* check whether the inode bitmap block number is set */
+       bitmap_blk = ext4_inode_bitmap(sb, desc);
+       offset = bitmap_blk - group_first_block;
+       if (!ext4_test_bit(offset, bh->b_data))
+               /* bad block bitmap */
+               goto err_out;
+
+       /* check whether the inode table block number is set */
+       bitmap_blk = ext4_inode_table(sb, desc);
+       offset = bitmap_blk - group_first_block;
+       next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
+                               offset + EXT4_SB(sb)->s_itb_per_group,
+                               offset);
+       if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group)
+               /* good bitmap for inode tables */
+               return 1;
+
+err_out:
+       ext4_error(sb, __func__,
+                       "Invalid block bitmap - "
+                       "block_group = %d, block = %llu",
+                       block_group, bitmap_blk);
+       return 0;
+}
 /**
- * read_block_bitmap()
+ * ext4_read_block_bitmap()
  * @sb:                        super block
  * @block_group:       given block group
  *
- * Read the bitmap for a given block_group, reading into the specified
- * slot in the superblock's bitmap cache.
+ * Read the bitmap for a given block_group,and validate the
+ * bits for block/inode/inode tables are set in the bitmaps
  *
  * Return buffer_head on success or NULL in case of failure.
  */
-static struct buffer_head *
-read_block_bitmap(struct super_block *sb, unsigned int block_group)
+struct buffer_head *
+ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
 {
        struct ext4_group_desc * desc;
        struct buffer_head * bh = NULL;
+       ext4_fsblk_t bitmap_blk;
 
-       desc = ext4_get_group_desc (sb, block_group, NULL);
+       desc = ext4_get_group_desc(sb, block_group, NULL);
        if (!desc)
-               goto error_out;
-       bh = sb_bread(sb, ext4_block_bitmap(sb, desc));
-       if (!bh)
-               ext4_error (sb, "read_block_bitmap",
+               return NULL;
+       bitmap_blk = ext4_block_bitmap(sb, desc);
+       bh = sb_getblk(sb, bitmap_blk);
+       if (unlikely(!bh)) {
+               ext4_error(sb, __func__,
+                           "Cannot read block bitmap - "
+                           "block_group = %lu, block_bitmap = %llu",
+                           block_group, bitmap_blk);
+               return NULL;
+       }
+       if (bh_uptodate_or_lock(bh))
+               return bh;
+
+       spin_lock(sb_bgl_lock(EXT4_SB(sb), block_group));
+       if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
+               ext4_init_block_bitmap(sb, bh, block_group, desc);
+               set_buffer_uptodate(bh);
+               unlock_buffer(bh);
+               spin_unlock(sb_bgl_lock(EXT4_SB(sb), block_group));
+               return bh;
+       }
+       spin_unlock(sb_bgl_lock(EXT4_SB(sb), block_group));
+       if (bh_submit_read(bh) < 0) {
+               put_bh(bh);
+               ext4_error(sb, __func__,
                            "Cannot read block bitmap - "
-                           "block_group = %d, block_bitmap = %llu",
-                           block_group,
-                           ext4_block_bitmap(sb, desc));
-error_out:
+                           "block_group = %lu, block_bitmap = %llu",
+                           block_group, bitmap_blk);
+               return NULL;
+       }
+       ext4_valid_block_bitmap(sb, desc, block_group, bh);
+       /*
+        * file system mounted not to panic on error,
+        * continue with corrupt bitmap
+        */
        return bh;
 }
 /*
@@ -191,11 +407,10 @@ restart:
                prev = rsv;
        }
        printk("Window map complete.\n");
-       if (bad)
-               BUG();
+       BUG_ON(bad);
 }
 #define rsv_window_dump(root, verbose) \
-       __rsv_window_dump((root), (verbose), __FUNCTION__)
+       __rsv_window_dump((root), (verbose), __func__)
 #else
 #define rsv_window_dump(root, verbose) do {} while (0)
 #endif
@@ -218,7 +433,7 @@ restart:
  */
 static int
 goal_in_my_reservation(struct ext4_reserve_window *rsv, ext4_grpblk_t grp_goal,
-                       unsigned int group, struct super_block * sb)
+                       ext4_group_t group, struct super_block *sb)
 {
        ext4_fsblk_t group_first_block, group_last_block;
 
@@ -361,7 +576,7 @@ static inline int rsv_is_empty(struct ext4_reserve_window *rsv)
  * when setting the reservation window size through ioctl before the file
  * is open for write (needs block allocation).
  *
- * Needs truncate_mutex protection prior to call this function.
+ * Needs down_write(i_data_sem) protection prior to call this function.
  */
 void ext4_init_block_alloc_info(struct inode *inode)
 {
@@ -412,6 +627,8 @@ void ext4_discard_reservation(struct inode *inode)
        struct ext4_reserve_window_node *rsv;
        spinlock_t *rsv_lock = &EXT4_SB(inode->i_sb)->s_rsv_window_lock;
 
+       ext4_mb_discard_inode_preallocations(inode);
+
        if (!block_i)
                return;
 
@@ -438,7 +655,7 @@ void ext4_free_blocks_sb(handle_t *handle, struct super_block *sb,
 {
        struct buffer_head *bitmap_bh = NULL;
        struct buffer_head *gd_bh;
-       unsigned long block_group;
+       ext4_group_t block_group;
        ext4_grpblk_t bit;
        unsigned long i;
        unsigned long overflow;
@@ -474,7 +691,7 @@ do_more:
                count -= overflow;
        }
        brelse(bitmap_bh);
-       bitmap_bh = read_block_bitmap(sb, block_group);
+       bitmap_bh = ext4_read_block_bitmap(sb, block_group);
        if (!bitmap_bh)
                goto error_return;
        desc = ext4_get_group_desc (sb, block_group, &gd_bh);
@@ -485,11 +702,13 @@ do_more:
            in_range(ext4_inode_bitmap(sb, desc), block, count) ||
            in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) ||
            in_range(block + count - 1, ext4_inode_table(sb, desc),
-                    sbi->s_itb_per_group))
+                    sbi->s_itb_per_group)) {
                ext4_error (sb, "ext4_free_blocks",
                            "Freeing blocks in system zones - "
                            "Block = %llu, count = %lu",
                            block, count);
+               goto error_return;
+       }
 
        /*
         * We are about to start releasing blocks in the bitmap,
@@ -571,7 +790,7 @@ do_more:
                if (!ext4_clear_bit_atomic(sb_bgl_lock(sbi, block_group),
                                                bit + i, bitmap_bh->b_data)) {
                        jbd_unlock_bh_state(bitmap_bh);
-                       ext4_error(sb, __FUNCTION__,
+                       ext4_error(sb, __func__,
                                   "bit already cleared for block %llu",
                                   (ext4_fsblk_t)(block + i));
                        jbd_lock_bh_state(bitmap_bh);
@@ -583,12 +802,18 @@ do_more:
        jbd_unlock_bh_state(bitmap_bh);
 
        spin_lock(sb_bgl_lock(sbi, block_group));
-       desc->bg_free_blocks_count =
-               cpu_to_le16(le16_to_cpu(desc->bg_free_blocks_count) +
-                       group_freed);
+       le16_add_cpu(&desc->bg_free_blocks_count, group_freed);
+       desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc);
        spin_unlock(sb_bgl_lock(sbi, block_group));
        percpu_counter_add(&sbi->s_freeblocks_counter, count);
 
+       if (sbi->s_log_groups_per_flex) {
+               ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
+               spin_lock(sb_bgl_lock(sbi, flex_group));
+               sbi->s_flex_groups[flex_group].free_blocks += count;
+               spin_unlock(sb_bgl_lock(sbi, flex_group));
+       }
+
        /* We dirtied the bitmap block */
        BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
        err = ext4_journal_dirty_metadata(handle, bitmap_bh);
@@ -617,19 +842,29 @@ error_return:
  * @inode:             inode
  * @block:             start physical block to free
  * @count:             number of blocks to count
+ * @metadata:          Are these metadata blocks
  */
 void ext4_free_blocks(handle_t *handle, struct inode *inode,
-                       ext4_fsblk_t block, unsigned long count)
+                       ext4_fsblk_t block, unsigned long count,
+                       int metadata)
 {
        struct super_block * sb;
        unsigned long dquot_freed_blocks;
 
+       /* this isn't the right place to decide whether block is metadata
+        * inode.c/extents.c knows better, but for safety ... */
+       if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode) ||
+                       ext4_should_journal_data(inode))
+               metadata = 1;
+
        sb = inode->i_sb;
-       if (!sb) {
-               printk ("ext4_free_blocks: nonexistent device");
-               return;
-       }
-       ext4_free_blocks_sb(handle, sb, block, count, &dquot_freed_blocks);
+
+       if (!test_opt(sb, MBALLOC) || !EXT4_SB(sb)->s_group_info)
+               ext4_free_blocks_sb(handle, sb, block, count,
+                                               &dquot_freed_blocks);
+       else
+               ext4_mb_free_blocks(handle, inode, block, count,
+                                               metadata, &dquot_freed_blocks);
        if (dquot_freed_blocks)
                DQUOT_FREE_BLOCK(inode, dquot_freed_blocks);
        return;
@@ -817,9 +1052,10 @@ claim_block(spinlock_t *lock, ext4_grpblk_t block, struct buffer_head *bh)
  * ext4_journal_release_buffer(), else we'll run out of credits.
  */
 static ext4_grpblk_t
-ext4_try_to_allocate(struct super_block *sb, handle_t *handle, int group,
-                       struct buffer_head *bitmap_bh, ext4_grpblk_t grp_goal,
-                       unsigned long *count, struct ext4_reserve_window *my_rsv)
+ext4_try_to_allocate(struct super_block *sb, handle_t *handle,
+                       ext4_group_t group, struct buffer_head *bitmap_bh,
+                       ext4_grpblk_t grp_goal, unsigned long *count,
+                       struct ext4_reserve_window *my_rsv)
 {
        ext4_fsblk_t group_first_block;
        ext4_grpblk_t start, end;
@@ -1053,7 +1289,7 @@ static int find_next_reservable_window(
  */
 static int alloc_new_reservation(struct ext4_reserve_window_node *my_rsv,
                ext4_grpblk_t grp_goal, struct super_block *sb,
-               unsigned int group, struct buffer_head *bitmap_bh)
+               ext4_group_t group, struct buffer_head *bitmap_bh)
 {
        struct ext4_reserve_window_node *search_head;
        ext4_fsblk_t group_first_block, group_end_block, start_block;
@@ -1251,7 +1487,7 @@ static void try_to_extend_reservation(struct ext4_reserve_window_node *my_rsv,
  */
 static ext4_grpblk_t
 ext4_try_to_allocate_with_rsv(struct super_block *sb, handle_t *handle,
-                       unsigned int group, struct buffer_head *bitmap_bh,
+                       ext4_group_t group, struct buffer_head *bitmap_bh,
                        ext4_grpblk_t grp_goal,
                        struct ext4_reserve_window_node * my_rsv,
                        unsigned long *count, int *errp)
@@ -1366,23 +1602,38 @@ out:
 
 /**
  * ext4_has_free_blocks()
- * @sbi:               in-core super block structure.
+ * @sbi:       in-core super block structure.
+ * @nblocks:   number of neeed blocks
  *
- * Check if filesystem has at least 1 free block available for allocation.
+ * Check if filesystem has free blocks available for allocation.
+ * Return the number of blocks avaible for allocation for this request
+ * On success, return nblocks
  */
-static int ext4_has_free_blocks(struct ext4_sb_info *sbi)
+ext4_fsblk_t ext4_has_free_blocks(struct ext4_sb_info *sbi,
+                                               ext4_fsblk_t nblocks)
 {
-       ext4_fsblk_t free_blocks, root_blocks;
+       ext4_fsblk_t free_blocks;
+       ext4_fsblk_t root_blocks = 0;
 
        free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter);
-       root_blocks = ext4_r_blocks_count(sbi->s_es);
-       if (free_blocks < root_blocks + 1 && !capable(CAP_SYS_RESOURCE) &&
+
+       if (!capable(CAP_SYS_RESOURCE) &&
                sbi->s_resuid != current->fsuid &&
-               (sbi->s_resgid == 0 || !in_group_p (sbi->s_resgid))) {
+               (sbi->s_resgid == 0 || !in_group_p(sbi->s_resgid)))
+               root_blocks = ext4_r_blocks_count(sbi->s_es);
+#ifdef CONFIG_SMP
+       if (free_blocks - root_blocks < FBC_BATCH)
+               free_blocks =
+                       percpu_counter_sum_and_set(&sbi->s_freeblocks_counter);
+#endif
+       if (free_blocks <= root_blocks)
+               /* we don't have free space */
                return 0;
-       }
-       return 1;
-}
+       if (free_blocks - root_blocks < nblocks)
+               return free_blocks - root_blocks;
+       return nblocks;
+ }
+
 
 /**
  * ext4_should_retry_alloc()
@@ -1398,7 +1649,7 @@ static int ext4_has_free_blocks(struct ext4_sb_info *sbi)
  */
 int ext4_should_retry_alloc(struct super_block *sb, int *retries)
 {
-       if (!ext4_has_free_blocks(EXT4_SB(sb)) || (*retries)++ > 3)
+       if (!ext4_has_free_blocks(EXT4_SB(sb), 1) || (*retries)++ > 3)
                return 0;
 
        jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
@@ -1407,30 +1658,34 @@ int ext4_should_retry_alloc(struct super_block *sb, int *retries)
 }
 
 /**
- * ext4_new_blocks() -- core block(s) allocation function
+ * ext4_old_new_blocks() -- core block bitmap based block allocation function
+ *
  * @handle:            handle to this transaction
  * @inode:             file inode
  * @goal:              given target block(filesystem wide)
  * @count:             target number of blocks to allocate
  * @errp:              error code
  *
- * ext4_new_blocks uses a goal block to assist allocation.  It tries to
- * allocate block(s) from the block group contains the goal block first. If that
- * fails, it will try to allocate block(s) from other block groups without
- * any specific goal block.
+ * ext4_old_new_blocks uses a goal block to assist allocation and look up
+ * the block bitmap directly to do block allocation.  It tries to
+ * allocate block(s) from the block group contains the goal block first. If
+ * that fails, it will try to allocate block(s) from other block groups
+ * without any specific goal block.
+ *
+ * This function is called when -o nomballoc mount option is enabled
  *
  */
-ext4_fsblk_t ext4_new_blocks(handle_t *handle, struct inode *inode,
+ext4_fsblk_t ext4_old_new_blocks(handle_t *handle, struct inode *inode,
                        ext4_fsblk_t goal, unsigned long *count, int *errp)
 {
        struct buffer_head *bitmap_bh = NULL;
        struct buffer_head *gdp_bh;
-       unsigned long group_no;
-       int goal_group;
+       ext4_group_t group_no;
+       ext4_group_t goal_group;
        ext4_grpblk_t grp_target_blk;   /* blockgroup relative goal block */
        ext4_grpblk_t grp_alloc_blk;    /* blockgroup-relative allocated block*/
        ext4_fsblk_t ret_block;         /* filesyetem-wide allocated block */
-       int bgi;                        /* blockgroup iteration index */
+       ext4_group_t bgi;                       /* blockgroup iteration index */
        int fatal = 0, err;
        int performed_allocation = 0;
        ext4_grpblk_t free_blocks;      /* number of free blocks in a group */
@@ -1441,19 +1696,29 @@ ext4_fsblk_t ext4_new_blocks(handle_t *handle, struct inode *inode,
        struct ext4_reserve_window_node *my_rsv = NULL;
        struct ext4_block_alloc_info *block_i;
        unsigned short windowsz = 0;
-#ifdef EXT4FS_DEBUG
-       static int goal_hits, goal_attempts;
-#endif
-       unsigned long ngroups;
+       ext4_group_t ngroups;
        unsigned long num = *count;
 
-       *errp = -ENOSPC;
        sb = inode->i_sb;
        if (!sb) {
+               *errp = -ENODEV;
                printk("ext4_new_block: nonexistent device");
                return 0;
        }
 
+       sbi = EXT4_SB(sb);
+       if (!EXT4_I(inode)->i_delalloc_reserved_flag) {
+               /*
+                * With delalloc we already reserved the blocks
+                */
+               *count = ext4_has_free_blocks(sbi, *count);
+       }
+       if (*count == 0) {
+               *errp = -ENOSPC;
+               return 0;       /*return with ENOSPC error */
+       }
+       num = *count;
+
        /*
         * Check quota for allocation of this block.
         */
@@ -1464,7 +1729,7 @@ ext4_fsblk_t ext4_new_blocks(handle_t *handle, struct inode *inode,
 
        sbi = EXT4_SB(sb);
        es = EXT4_SB(sb)->s_es;
-       ext4_debug("goal=%lu.\n", goal);
+       ext4_debug("goal=%llu.\n", goal);
        /*
         * Allocate a block from reservation only when
         * filesystem is mounted with reservation(default,-o reservation), and
@@ -1477,11 +1742,6 @@ ext4_fsblk_t ext4_new_blocks(handle_t *handle, struct inode *inode,
        if (block_i && ((windowsz = block_i->rsv_window_node.rsv_goal_size) > 0))
                my_rsv = &block_i->rsv_window_node;
 
-       if (!ext4_has_free_blocks(sbi)) {
-               *errp = -ENOSPC;
-               goto out;
-       }
-
        /*
         * First, test whether the goal block is free.
         */
@@ -1505,7 +1765,7 @@ retry_alloc:
                my_rsv = NULL;
 
        if (free_blocks > 0) {
-               bitmap_bh = read_block_bitmap(sb, group_no);
+               bitmap_bh = ext4_read_block_bitmap(sb, group_no);
                if (!bitmap_bh)
                        goto io_error;
                grp_alloc_blk = ext4_try_to_allocate_with_rsv(sb, handle,
@@ -1522,7 +1782,7 @@ retry_alloc:
 
        /*
         * Now search the rest of the groups.  We assume that
-        * i and gdp correctly point to the last group visited.
+        * group_no and gdp correctly point to the last group visited.
         */
        for (bgi = 0; bgi < ngroups; bgi++) {
                group_no++;
@@ -1541,7 +1801,7 @@ retry_alloc:
                        continue;
 
                brelse(bitmap_bh);
-               bitmap_bh = read_block_bitmap(sb, group_no);
+               bitmap_bh = ext4_read_block_bitmap(sb, group_no);
                if (!bitmap_bh)
                        goto io_error;
                /*
@@ -1574,7 +1834,7 @@ retry_alloc:
 
 allocated:
 
-       ext4_debug("using block group %d(%d)\n",
+       ext4_debug("using block group %lu(%d)\n",
                        group_no, gdp->bg_free_blocks_count);
 
        BUFFER_TRACE(gdp_bh, "get_write_access");
@@ -1589,11 +1849,18 @@ allocated:
            in_range(ret_block, ext4_inode_table(sb, gdp),
                     EXT4_SB(sb)->s_itb_per_group) ||
            in_range(ret_block + num - 1, ext4_inode_table(sb, gdp),
-                    EXT4_SB(sb)->s_itb_per_group))
+                    EXT4_SB(sb)->s_itb_per_group)) {
                ext4_error(sb, "ext4_new_block",
                            "Allocating block in system zone - "
                            "blocks from %llu, length %lu",
                             ret_block, num);
+               /*
+                * claim_block marked the blocks we allocated
+                * as in use. So we may want to selectively
+                * mark some of the blocks as free
+                */
+               goto retry_alloc;
+       }
 
        performed_allocation = 1;
 
@@ -1618,7 +1885,7 @@ allocated:
                        if (ext4_test_bit(grp_alloc_blk+i,
                                        bh2jh(bitmap_bh)->b_committed_data)) {
                                printk("%s: block was unexpectedly set in "
-                                       "b_committed_data\n", __FUNCTION__);
+                                       "b_committed_data\n", __func__);
                        }
                }
        }
@@ -1640,14 +1907,21 @@ allocated:
         * list of some description.  We don't know in advance whether
         * the caller wants to use it as metadata or data.
         */
-       ext4_debug("allocating block %lu. Goal hits %d of %d.\n",
-                       ret_block, goal_hits, goal_attempts);
-
        spin_lock(sb_bgl_lock(sbi, group_no));
-       gdp->bg_free_blocks_count =
-                       cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count)-num);
+       if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))
+               gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT);
+       le16_add_cpu(&gdp->bg_free_blocks_count, -num);
+       gdp->bg_checksum = ext4_group_desc_csum(sbi, group_no, gdp);
        spin_unlock(sb_bgl_lock(sbi, group_no));
-       percpu_counter_sub(&sbi->s_freeblocks_counter, num);
+       if (!EXT4_I(inode)->i_delalloc_reserved_flag)
+               percpu_counter_sub(&sbi->s_freeblocks_counter, num);
+
+       if (sbi->s_log_groups_per_flex) {
+               ext4_group_t flex_group = ext4_flex_group(sbi, group_no);
+               spin_lock(sb_bgl_lock(sbi, flex_group));
+               sbi->s_flex_groups[flex_group].free_blocks -= num;
+               spin_unlock(sb_bgl_lock(sbi, flex_group));
+       }
 
        BUFFER_TRACE(gdp_bh, "journal_dirty_metadata for group descriptor");
        err = ext4_journal_dirty_metadata(handle, gdp_bh);
@@ -1680,12 +1954,103 @@ out:
        return 0;
 }
 
-ext4_fsblk_t ext4_new_block(handle_t *handle, struct inode *inode,
-                       ext4_fsblk_t goal, int *errp)
+#define EXT4_META_BLOCK 0x1
+
+static ext4_fsblk_t do_blk_alloc(handle_t *handle, struct inode *inode,
+                               ext4_lblk_t iblock, ext4_fsblk_t goal,
+                               unsigned long *count, int *errp, int flags)
+{
+       struct ext4_allocation_request ar;
+       ext4_fsblk_t ret;
+
+       if (!test_opt(inode->i_sb, MBALLOC)) {
+               return ext4_old_new_blocks(handle, inode, goal, count, errp);
+       }
+
+       memset(&ar, 0, sizeof(ar));
+       /* Fill with neighbour allocated blocks */
+
+       ar.inode = inode;
+       ar.goal = goal;
+       ar.len = *count;
+       ar.logical = iblock;
+
+       if (S_ISREG(inode->i_mode) && !(flags & EXT4_META_BLOCK))
+               /* enable in-core preallocation for data block allocation */
+               ar.flags = EXT4_MB_HINT_DATA;
+       else
+               /* disable in-core preallocation for non-regular files */
+               ar.flags = 0;
+
+       ret = ext4_mb_new_blocks(handle, &ar, errp);
+       *count = ar.len;
+       return ret;
+}
+
+/*
+ * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
+ *
+ * @handle:             handle to this transaction
+ * @inode:              file inode
+ * @goal:               given target block(filesystem wide)
+ * @count:             total number of blocks need
+ * @errp:               error code
+ *
+ * Return 1st allocated block numberon success, *count stores total account
+ * error stores in errp pointer
+ */
+ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
+               ext4_fsblk_t goal, unsigned long *count, int *errp)
+{
+       ext4_fsblk_t ret;
+       ret = do_blk_alloc(handle, inode, 0, goal,
+                               count, errp, EXT4_META_BLOCK);
+       /*
+        * Account for the allocated meta blocks
+        */
+       if (!(*errp)) {
+               spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
+               EXT4_I(inode)->i_allocated_meta_blocks += *count;
+               spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
+       }
+       return ret;
+}
+
+/*
+ * ext4_new_meta_block() -- allocate block for meta data (indexing) blocks
+ *
+ * @handle:             handle to this transaction
+ * @inode:              file inode
+ * @goal:               given target block(filesystem wide)
+ * @errp:               error code
+ *
+ * Return allocated block number on success
+ */
+ext4_fsblk_t ext4_new_meta_block(handle_t *handle, struct inode *inode,
+               ext4_fsblk_t goal, int *errp)
 {
        unsigned long count = 1;
+       return ext4_new_meta_blocks(handle, inode, goal, &count, errp);
+}
 
-       return ext4_new_blocks(handle, inode, goal, &count, errp);
+/*
+ * ext4_new_blocks() -- allocate data blocks
+ *
+ * @handle:             handle to this transaction
+ * @inode:              file inode
+ * @goal:               given target block(filesystem wide)
+ * @count:             total number of blocks need
+ * @errp:               error code
+ *
+ * Return 1st allocated block numberon success, *count stores total account
+ * error stores in errp pointer
+ */
+
+ext4_fsblk_t ext4_new_blocks(handle_t *handle, struct inode *inode,
+                               ext4_lblk_t iblock, ext4_fsblk_t goal,
+                               unsigned long *count, int *errp)
+{
+       return do_blk_alloc(handle, inode, iblock, goal, count, errp, 0);
 }
 
 /**
@@ -1698,8 +2063,8 @@ ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb)
 {
        ext4_fsblk_t desc_count;
        struct ext4_group_desc *gdp;
-       int i;
-       unsigned long ngroups = EXT4_SB(sb)->s_groups_count;
+       ext4_group_t i;
+       ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
 #ifdef EXT4FS_DEBUG
        struct ext4_super_block *es;
        ext4_fsblk_t bitmap_count;
@@ -1718,19 +2083,19 @@ ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb)
                        continue;
                desc_count += le16_to_cpu(gdp->bg_free_blocks_count);
                brelse(bitmap_bh);
-               bitmap_bh = read_block_bitmap(sb, i);
+               bitmap_bh = ext4_read_block_bitmap(sb, i);
                if (bitmap_bh == NULL)
                        continue;
 
                x = ext4_count_free(bitmap_bh, sb->s_blocksize);
-               printk("group %d: stored = %d, counted = %lu\n",
+               printk(KERN_DEBUG "group %lu: stored = %d, counted = %lu\n",
                        i, le16_to_cpu(gdp->bg_free_blocks_count), x);
                bitmap_count += x;
        }
        brelse(bitmap_bh);
        printk("ext4_count_free_blocks: stored = %llu"
                ", computed = %llu, %llu\n",
-              EXT4_FREE_BLOCKS_COUNT(es),
+               ext4_free_blocks_count(es),
                desc_count, bitmap_count);
        return bitmap_count;
 #else
@@ -1747,16 +2112,7 @@ ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb)
 #endif
 }
 
-static inline int
-block_in_use(ext4_fsblk_t block, struct super_block *sb, unsigned char *map)
-{
-       ext4_grpblk_t offset;
-
-       ext4_get_group_no_and_offset(sb, block, NULL, &offset);
-       return ext4_test_bit (offset, map);
-}
-
-static inline int test_root(int a, int b)
+static inline int test_root(ext4_group_t a, int b)
 {
        int num = b;
 
@@ -1765,7 +2121,7 @@ static inline int test_root(int a, int b)
        return num == a;
 }
 
-static int ext4_group_sparse(int group)
+static int ext4_group_sparse(ext4_group_t group)
 {
        if (group <= 1)
                return 1;
@@ -1783,7 +2139,7 @@ static int ext4_group_sparse(int group)
  *     Return the number of blocks used by the superblock (primary or backup)
  *     in this group.  Currently this will be only 0 or 1.
  */
-int ext4_bg_has_super(struct super_block *sb, int group)
+int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
 {
        if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
                                EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
@@ -1792,24 +2148,22 @@ int ext4_bg_has_super(struct super_block *sb, int group)
        return 1;
 }
 
-static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb, int group)
+static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
+                                       ext4_group_t group)
 {
        unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
-       unsigned long first = metagroup * EXT4_DESC_PER_BLOCK(sb);
-       unsigned long last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
+       ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
+       ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
 
        if (group == first || group == first + 1 || group == last)
                return 1;
        return 0;
 }
 
-static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb, int group)
+static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
+                                       ext4_group_t group)
 {
-       if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
-                               EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
-                       !ext4_group_sparse(group))
-               return 0;
-       return EXT4_SB(sb)->s_gdb_count;
+       return ext4_bg_has_super(sb, group) ? EXT4_SB(sb)->s_gdb_count : 0;
 }
 
 /**
@@ -1821,7 +2175,7 @@ static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb, int group)
  *     (primary or backup) in this group.  In the future there may be a
  *     different number of descriptor blocks in each group.
  */
-unsigned long ext4_bg_num_gdb(struct super_block *sb, int group)
+unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
 {
        unsigned long first_meta_bg =
                        le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);