Merge branch 'drm-intel-next' of git://git.kernel.org/pub/scm/linux/kernel/git/anholt...
[pandora-kernel.git] / fs / ext4 / extents.c
index 236b834..377309c 100644 (file)
@@ -107,11 +107,8 @@ static int ext4_ext_truncate_extend_restart(handle_t *handle,
        if (err <= 0)
                return err;
        err = ext4_truncate_restart_trans(handle, inode, needed);
-       /*
-        * We have dropped i_data_sem so someone might have cached again
-        * an extent we are going to truncate.
-        */
-       ext4_ext_invalidate_cache(inode);
+       if (err == 0)
+               err = -EAGAIN;
 
        return err;
 }
@@ -185,10 +182,10 @@ static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode,
        if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
                /*
                 * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
-                * block groups per flexgroup, reserve the first block 
-                * group for directories and special files.  Regular 
+                * block groups per flexgroup, reserve the first block
+                * group for directories and special files.  Regular
                 * files will start at the second block group.  This
-                * tends to speed up directory access and improves 
+                * tends to speed up directory access and improves
                 * fsck times.
                 */
                block_group &= ~(flex_size-1);
@@ -439,10 +436,10 @@ static int __ext4_ext_check(const char *function, struct inode *inode,
        return 0;
 
 corrupted:
-       __ext4_error(inode->i_sb, function,
-                       "bad header/extent in inode #%lu: %s - magic %x, "
+       ext4_error_inode(function, inode,
+                       "bad header/extent: %s - magic %x, "
                        "entries %u, max %u(%u), depth %u(%u)",
-                       inode->i_ino, error_msg, le16_to_cpu(eh->eh_magic),
+                       error_msg, le16_to_cpu(eh->eh_magic),
                        le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max),
                        max, le16_to_cpu(eh->eh_depth), depth);
 
@@ -1622,9 +1619,7 @@ int ext4_ext_try_to_merge(struct inode *inode,
                merge_done = 1;
                WARN_ON(eh->eh_entries == 0);
                if (!eh->eh_entries)
-                       ext4_error(inode->i_sb,
-                                  "inode#%lu, eh->eh_entries = 0!",
-                                  inode->i_ino);
+                       EXT4_ERROR_INODE(inode, "eh->eh_entries = 0!");
        }
 
        return merge_done;
@@ -2039,7 +2034,7 @@ ext4_ext_in_cache(struct inode *inode, ext4_lblk_t block,
        struct ext4_ext_cache *cex;
        int ret = EXT4_EXT_CACHE_NO;
 
-       /* 
+       /*
         * We borrow i_block_reservation_lock to protect i_cached_extent
         */
        spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
@@ -2361,7 +2356,7 @@ static int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start)
        int depth = ext_depth(inode);
        struct ext4_ext_path *path;
        handle_t *handle;
-       int i = 0, err = 0;
+       int i, err;
 
        ext_debug("truncate since %u\n", start);
 
@@ -2370,23 +2365,26 @@ static int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start)
        if (IS_ERR(handle))
                return PTR_ERR(handle);
 
+again:
        ext4_ext_invalidate_cache(inode);
 
        /*
         * We start scanning from right side, freeing all the blocks
         * after i_size and walking into the tree depth-wise.
         */
+       depth = ext_depth(inode);
        path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1), GFP_NOFS);
        if (path == NULL) {
                ext4_journal_stop(handle);
                return -ENOMEM;
        }
+       path[0].p_depth = depth;
        path[0].p_hdr = ext_inode_hdr(inode);
        if (ext4_ext_check(inode, path[0].p_hdr, depth)) {
                err = -EIO;
                goto out;
        }
-       path[0].p_depth = depth;
+       i = err = 0;
 
        while (i >= 0 && err == 0) {
                if (i == depth) {
@@ -2480,6 +2478,8 @@ static int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start)
 out:
        ext4_ext_drop_refs(path);
        kfree(path);
+       if (err == -EAGAIN)
+               goto again;
        ext4_journal_stop(handle);
 
        return err;
@@ -2544,7 +2544,7 @@ static void bi_complete(struct bio *bio, int error)
 /* FIXME!! we need to try to merge to left or right after zero-out  */
 static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
 {
-       int ret = -EIO;
+       int ret;
        struct bio *bio;
        int blkbits, blocksize;
        sector_t ee_pblock;
@@ -2568,6 +2568,9 @@ static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
                        len = ee_len;
 
                bio = bio_alloc(GFP_NOIO, len);
+               if (!bio)
+                       return -ENOMEM;
+
                bio->bi_sector = ee_pblock;
                bio->bi_bdev   = inode->i_sb->s_bdev;
 
@@ -2595,22 +2598,20 @@ static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
                submit_bio(WRITE, bio);
                wait_for_completion(&event);
 
-               if (test_bit(BIO_UPTODATE, &bio->bi_flags))
-                       ret = 0;
-               else {
-                       ret = -EIO;
-                       break;
+               if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) {
+                       bio_put(bio);
+                       return -EIO;
                }
                bio_put(bio);
                ee_len    -= done;
                ee_pblock += done  << (blkbits - 9);
        }
-       return ret;
+       return 0;
 }
 
 #define EXT4_EXT_ZERO_LEN 7
 /*
- * This function is called by ext4_ext_get_blocks() if someone tries to write
+ * This function is called by ext4_ext_map_blocks() if someone tries to write
  * to an uninitialized extent. It may result in splitting the uninitialized
  * extent into multiple extents (upto three - one initialized and two
  * uninitialized).
@@ -2620,39 +2621,55 @@ static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
  *   c> Splits in three extents: Somone is writing in middle of the extent
  */
 static int ext4_ext_convert_to_initialized(handle_t *handle,
-                                               struct inode *inode,
-                                               struct ext4_ext_path *path,
-                                               ext4_lblk_t iblock,
-                                               unsigned int max_blocks)
+                                          struct inode *inode,
+                                          struct ext4_map_blocks *map,
+                                          struct ext4_ext_path *path)
 {
        struct ext4_extent *ex, newex, orig_ex;
        struct ext4_extent *ex1 = NULL;
        struct ext4_extent *ex2 = NULL;
        struct ext4_extent *ex3 = NULL;
        struct ext4_extent_header *eh;
-       ext4_lblk_t ee_block;
+       ext4_lblk_t ee_block, eof_block;
        unsigned int allocated, ee_len, depth;
        ext4_fsblk_t newblock;
        int err = 0;
        int ret = 0;
+       int may_zeroout;
+
+       ext_debug("ext4_ext_convert_to_initialized: inode %lu, logical"
+               "block %llu, max_blocks %u\n", inode->i_ino,
+               (unsigned long long)map->m_lblk, map->m_len);
+
+       eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >>
+               inode->i_sb->s_blocksize_bits;
+       if (eof_block < map->m_lblk + map->m_len)
+               eof_block = map->m_lblk + map->m_len;
 
        depth = ext_depth(inode);
        eh = path[depth].p_hdr;
        ex = path[depth].p_ext;
        ee_block = le32_to_cpu(ex->ee_block);
        ee_len = ext4_ext_get_actual_len(ex);
-       allocated = ee_len - (iblock - ee_block);
-       newblock = iblock - ee_block + ext_pblock(ex);
+       allocated = ee_len - (map->m_lblk - ee_block);
+       newblock = map->m_lblk - ee_block + ext_pblock(ex);
+
        ex2 = ex;
        orig_ex.ee_block = ex->ee_block;
        orig_ex.ee_len   = cpu_to_le16(ee_len);
        ext4_ext_store_pblock(&orig_ex, ext_pblock(ex));
 
+       /*
+        * It is safe to convert extent to initialized via explicit
+        * zeroout only if extent is fully insde i_size or new_size.
+        */
+       may_zeroout = ee_block + ee_len <= eof_block;
+
        err = ext4_ext_get_access(handle, inode, path + depth);
        if (err)
                goto out;
        /* If extent has less than 2*EXT4_EXT_ZERO_LEN zerout directly */
-       if (ee_len <= 2*EXT4_EXT_ZERO_LEN) {
+       if (ee_len <= 2*EXT4_EXT_ZERO_LEN && may_zeroout) {
                err =  ext4_ext_zeroout(inode, &orig_ex);
                if (err)
                        goto fix_extent_len;
@@ -2665,10 +2682,10 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
                return allocated;
        }
 
-       /* ex1: ee_block to iblock - 1 : uninitialized */
-       if (iblock > ee_block) {
+       /* ex1: ee_block to map->m_lblk - 1 : uninitialized */
+       if (map->m_lblk > ee_block) {
                ex1 = ex;
-               ex1->ee_len = cpu_to_le16(iblock - ee_block);
+               ex1->ee_len = cpu_to_le16(map->m_lblk - ee_block);
                ext4_ext_mark_uninitialized(ex1);
                ex2 = &newex;
        }
@@ -2677,15 +2694,15 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
         * we insert ex3, if ex1 is NULL. This is to avoid temporary
         * overlap of blocks.
         */
-       if (!ex1 && allocated > max_blocks)
-               ex2->ee_len = cpu_to_le16(max_blocks);
+       if (!ex1 && allocated > map->m_len)
+               ex2->ee_len = cpu_to_le16(map->m_len);
        /* ex3: to ee_block + ee_len : uninitialised */
-       if (allocated > max_blocks) {
+       if (allocated > map->m_len) {
                unsigned int newdepth;
                /* If extent has less than EXT4_EXT_ZERO_LEN zerout directly */
-               if (allocated <= EXT4_EXT_ZERO_LEN) {
+               if (allocated <= EXT4_EXT_ZERO_LEN && may_zeroout) {
                        /*
-                        * iblock == ee_block is handled by the zerouout
+                        * map->m_lblk == ee_block is handled by the zerouout
                         * at the beginning.
                         * Mark first half uninitialized.
                         * Mark second half initialized and zero out the
@@ -2698,7 +2715,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
                        ext4_ext_dirty(handle, inode, path + depth);
 
                        ex3 = &newex;
-                       ex3->ee_block = cpu_to_le32(iblock);
+                       ex3->ee_block = cpu_to_le32(map->m_lblk);
                        ext4_ext_store_pblock(ex3, newblock);
                        ex3->ee_len = cpu_to_le16(allocated);
                        err = ext4_ext_insert_extent(handle, inode, path,
@@ -2711,7 +2728,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
                                ex->ee_len   = orig_ex.ee_len;
                                ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
                                ext4_ext_dirty(handle, inode, path + depth);
-                               /* blocks available from iblock */
+                               /* blocks available from map->m_lblk */
                                return allocated;
 
                        } else if (err)
@@ -2733,8 +2750,8 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
                                 */
                                depth = ext_depth(inode);
                                ext4_ext_drop_refs(path);
-                               path = ext4_ext_find_extent(inode,
-                                                               iblock, path);
+                               path = ext4_ext_find_extent(inode, map->m_lblk,
+                                                           path);
                                if (IS_ERR(path)) {
                                        err = PTR_ERR(path);
                                        return err;
@@ -2754,12 +2771,12 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
                        return allocated;
                }
                ex3 = &newex;
-               ex3->ee_block = cpu_to_le32(iblock + max_blocks);
-               ext4_ext_store_pblock(ex3, newblock + max_blocks);
-               ex3->ee_len = cpu_to_le16(allocated - max_blocks);
+               ex3->ee_block = cpu_to_le32(map->m_lblk + map->m_len);
+               ext4_ext_store_pblock(ex3, newblock + map->m_len);
+               ex3->ee_len = cpu_to_le16(allocated - map->m_len);
                ext4_ext_mark_uninitialized(ex3);
                err = ext4_ext_insert_extent(handle, inode, path, ex3, 0);
-               if (err == -ENOSPC) {
+               if (err == -ENOSPC && may_zeroout) {
                        err =  ext4_ext_zeroout(inode, &orig_ex);
                        if (err)
                                goto fix_extent_len;
@@ -2769,7 +2786,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
                        ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
                        ext4_ext_dirty(handle, inode, path + depth);
                        /* zeroed the full extent */
-                       /* blocks available from iblock */
+                       /* blocks available from map->m_lblk */
                        return allocated;
 
                } else if (err)
@@ -2783,11 +2800,13 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
                 * update the extent length after successful insert of the
                 * split extent
                 */
-               orig_ex.ee_len = cpu_to_le16(ee_len -
-                                               ext4_ext_get_actual_len(ex3));
+               ee_len -= ext4_ext_get_actual_len(ex3);
+               orig_ex.ee_len = cpu_to_le16(ee_len);
+               may_zeroout = ee_block + ee_len <= eof_block;
+
                depth = newdepth;
                ext4_ext_drop_refs(path);
-               path = ext4_ext_find_extent(inode, iblock, path);
+               path = ext4_ext_find_extent(inode, map->m_lblk, path);
                if (IS_ERR(path)) {
                        err = PTR_ERR(path);
                        goto out;
@@ -2801,14 +2820,14 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
                if (err)
                        goto out;
 
-               allocated = max_blocks;
+               allocated = map->m_len;
 
                /* If extent has less than EXT4_EXT_ZERO_LEN and we are trying
                 * to insert a extent in the middle zerout directly
                 * otherwise give the extent a chance to merge to left
                 */
                if (le16_to_cpu(orig_ex.ee_len) <= EXT4_EXT_ZERO_LEN &&
-                                                       iblock != ee_block) {
+                       map->m_lblk != ee_block && may_zeroout) {
                        err =  ext4_ext_zeroout(inode, &orig_ex);
                        if (err)
                                goto fix_extent_len;
@@ -2818,7 +2837,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
                        ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
                        ext4_ext_dirty(handle, inode, path + depth);
                        /* zero out the first half */
-                       /* blocks available from iblock */
+                       /* blocks available from map->m_lblk */
                        return allocated;
                }
        }
@@ -2829,12 +2848,12 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
         */
        if (ex1 && ex1 != ex) {
                ex1 = ex;
-               ex1->ee_len = cpu_to_le16(iblock - ee_block);
+               ex1->ee_len = cpu_to_le16(map->m_lblk - ee_block);
                ext4_ext_mark_uninitialized(ex1);
                ex2 = &newex;
        }
-       /* ex2: iblock to iblock + maxblocks-1 : initialised */
-       ex2->ee_block = cpu_to_le32(iblock);
+       /* ex2: map->m_lblk to map->m_lblk + maxblocks-1 : initialised */
+       ex2->ee_block = cpu_to_le32(map->m_lblk);
        ext4_ext_store_pblock(ex2, newblock);
        ex2->ee_len = cpu_to_le16(allocated);
        if (ex2 != ex)
@@ -2877,7 +2896,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
        goto out;
 insert:
        err = ext4_ext_insert_extent(handle, inode, path, &newex, 0);
-       if (err == -ENOSPC) {
+       if (err == -ENOSPC && may_zeroout) {
                err =  ext4_ext_zeroout(inode, &orig_ex);
                if (err)
                        goto fix_extent_len;
@@ -2904,7 +2923,7 @@ fix_extent_len:
 }
 
 /*
- * This function is called by ext4_ext_get_blocks() from
+ * This function is called by ext4_ext_map_blocks() from
  * ext4_get_blocks_dio_write() when DIO to write
  * to an uninitialized extent.
  *
@@ -2927,9 +2946,8 @@ fix_extent_len:
  */
 static int ext4_split_unwritten_extents(handle_t *handle,
                                        struct inode *inode,
+                                       struct ext4_map_blocks *map,
                                        struct ext4_ext_path *path,
-                                       ext4_lblk_t iblock,
-                                       unsigned int max_blocks,
                                        int flags)
 {
        struct ext4_extent *ex, newex, orig_ex;
@@ -2937,41 +2955,55 @@ static int ext4_split_unwritten_extents(handle_t *handle,
        struct ext4_extent *ex2 = NULL;
        struct ext4_extent *ex3 = NULL;
        struct ext4_extent_header *eh;
-       ext4_lblk_t ee_block;
+       ext4_lblk_t ee_block, eof_block;
        unsigned int allocated, ee_len, depth;
        ext4_fsblk_t newblock;
        int err = 0;
+       int may_zeroout;
+
+       ext_debug("ext4_split_unwritten_extents: inode %lu, logical"
+               "block %llu, max_blocks %u\n", inode->i_ino,
+               (unsigned long long)map->m_lblk, map->m_len);
+
+       eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >>
+               inode->i_sb->s_blocksize_bits;
+       if (eof_block < map->m_lblk + map->m_len)
+               eof_block = map->m_lblk + map->m_len;
 
-       ext_debug("ext4_split_unwritten_extents: inode %lu,"
-                 "iblock %llu, max_blocks %u\n", inode->i_ino,
-                 (unsigned long long)iblock, max_blocks);
        depth = ext_depth(inode);
        eh = path[depth].p_hdr;
        ex = path[depth].p_ext;
        ee_block = le32_to_cpu(ex->ee_block);
        ee_len = ext4_ext_get_actual_len(ex);
-       allocated = ee_len - (iblock - ee_block);
-       newblock = iblock - ee_block + ext_pblock(ex);
+       allocated = ee_len - (map->m_lblk - ee_block);
+       newblock = map->m_lblk - ee_block + ext_pblock(ex);
+
        ex2 = ex;
        orig_ex.ee_block = ex->ee_block;
        orig_ex.ee_len   = cpu_to_le16(ee_len);
        ext4_ext_store_pblock(&orig_ex, ext_pblock(ex));
 
+       /*
+        * It is safe to convert extent to initialized via explicit
+        * zeroout only if extent is fully insde i_size or new_size.
+        */
+       may_zeroout = ee_block + ee_len <= eof_block;
+
        /*
         * If the uninitialized extent begins at the same logical
         * block where the write begins, and the write completely
         * covers the extent, then we don't need to split it.
         */
-       if ((iblock == ee_block) && (allocated <= max_blocks))
+       if ((map->m_lblk == ee_block) && (allocated <= map->m_len))
                return allocated;
 
        err = ext4_ext_get_access(handle, inode, path + depth);
        if (err)
                goto out;
-       /* ex1: ee_block to iblock - 1 : uninitialized */
-       if (iblock > ee_block) {
+       /* ex1: ee_block to map->m_lblk - 1 : uninitialized */
+       if (map->m_lblk > ee_block) {
                ex1 = ex;
-               ex1->ee_len = cpu_to_le16(iblock - ee_block);
+               ex1->ee_len = cpu_to_le16(map->m_lblk - ee_block);
                ext4_ext_mark_uninitialized(ex1);
                ex2 = &newex;
        }
@@ -2980,18 +3012,18 @@ static int ext4_split_unwritten_extents(handle_t *handle,
         * we insert ex3, if ex1 is NULL. This is to avoid temporary
         * overlap of blocks.
         */
-       if (!ex1 && allocated > max_blocks)
-               ex2->ee_len = cpu_to_le16(max_blocks);
+       if (!ex1 && allocated > map->m_len)
+               ex2->ee_len = cpu_to_le16(map->m_len);
        /* ex3: to ee_block + ee_len : uninitialised */
-       if (allocated > max_blocks) {
+       if (allocated > map->m_len) {
                unsigned int newdepth;
                ex3 = &newex;
-               ex3->ee_block = cpu_to_le32(iblock + max_blocks);
-               ext4_ext_store_pblock(ex3, newblock + max_blocks);
-               ex3->ee_len = cpu_to_le16(allocated - max_blocks);
+               ex3->ee_block = cpu_to_le32(map->m_lblk + map->m_len);
+               ext4_ext_store_pblock(ex3, newblock + map->m_len);
+               ex3->ee_len = cpu_to_le16(allocated - map->m_len);
                ext4_ext_mark_uninitialized(ex3);
                err = ext4_ext_insert_extent(handle, inode, path, ex3, flags);
-               if (err == -ENOSPC) {
+               if (err == -ENOSPC && may_zeroout) {
                        err =  ext4_ext_zeroout(inode, &orig_ex);
                        if (err)
                                goto fix_extent_len;
@@ -3001,7 +3033,7 @@ static int ext4_split_unwritten_extents(handle_t *handle,
                        ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
                        ext4_ext_dirty(handle, inode, path + depth);
                        /* zeroed the full extent */
-                       /* blocks available from iblock */
+                       /* blocks available from map->m_lblk */
                        return allocated;
 
                } else if (err)
@@ -3015,11 +3047,13 @@ static int ext4_split_unwritten_extents(handle_t *handle,
                 * update the extent length after successful insert of the
                 * split extent
                 */
-               orig_ex.ee_len = cpu_to_le16(ee_len -
-                                               ext4_ext_get_actual_len(ex3));
+               ee_len -= ext4_ext_get_actual_len(ex3);
+               orig_ex.ee_len = cpu_to_le16(ee_len);
+               may_zeroout = ee_block + ee_len <= eof_block;
+
                depth = newdepth;
                ext4_ext_drop_refs(path);
-               path = ext4_ext_find_extent(inode, iblock, path);
+               path = ext4_ext_find_extent(inode, map->m_lblk, path);
                if (IS_ERR(path)) {
                        err = PTR_ERR(path);
                        goto out;
@@ -3033,7 +3067,7 @@ static int ext4_split_unwritten_extents(handle_t *handle,
                if (err)
                        goto out;
 
-               allocated = max_blocks;
+               allocated = map->m_len;
        }
        /*
         * If there was a change of depth as part of the
@@ -3042,15 +3076,15 @@ static int ext4_split_unwritten_extents(handle_t *handle,
         */
        if (ex1 && ex1 != ex) {
                ex1 = ex;
-               ex1->ee_len = cpu_to_le16(iblock - ee_block);
+               ex1->ee_len = cpu_to_le16(map->m_lblk - ee_block);
                ext4_ext_mark_uninitialized(ex1);
                ex2 = &newex;
        }
        /*
-        * ex2: iblock to iblock + maxblocks-1 : to be direct IO written,
-        * uninitialised still.
+        * ex2: map->m_lblk to map->m_lblk + map->m_len-1 : to be written
+        * using direct I/O, uninitialised still.
         */
-       ex2->ee_block = cpu_to_le32(iblock);
+       ex2->ee_block = cpu_to_le32(map->m_lblk);
        ext4_ext_store_pblock(ex2, newblock);
        ex2->ee_len = cpu_to_le16(allocated);
        ext4_ext_mark_uninitialized(ex2);
@@ -3062,7 +3096,7 @@ static int ext4_split_unwritten_extents(handle_t *handle,
        goto out;
 insert:
        err = ext4_ext_insert_extent(handle, inode, path, &newex, flags);
-       if (err == -ENOSPC) {
+       if (err == -ENOSPC && may_zeroout) {
                err =  ext4_ext_zeroout(inode, &orig_ex);
                if (err)
                        goto fix_extent_len;
@@ -3152,10 +3186,9 @@ static void unmap_underlying_metadata_blocks(struct block_device *bdev,
 
 static int
 ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
-                       ext4_lblk_t iblock, unsigned int max_blocks,
+                       struct ext4_map_blocks *map,
                        struct ext4_ext_path *path, int flags,
-                       unsigned int allocated, struct buffer_head *bh_result,
-                       ext4_fsblk_t newblock)
+                       unsigned int allocated, ext4_fsblk_t newblock)
 {
        int ret = 0;
        int err = 0;
@@ -3163,15 +3196,14 @@ ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
 
        ext_debug("ext4_ext_handle_uninitialized_extents: inode %lu, logical"
                  "block %llu, max_blocks %u, flags %d, allocated %u",
-                 inode->i_ino, (unsigned long long)iblock, max_blocks,
+                 inode->i_ino, (unsigned long long)map->m_lblk, map->m_len,
                  flags, allocated);
        ext4_ext_show_leaf(inode, path);
 
        /* get_block() before submit the IO, split the extent */
        if ((flags & EXT4_GET_BLOCKS_PRE_IO)) {
-               ret = ext4_split_unwritten_extents(handle,
-                                               inode, path, iblock,
-                                               max_blocks, flags);
+               ret = ext4_split_unwritten_extents(handle, inode, map,
+                                                  path, flags);
                /*
                 * Flag the inode(non aio case) or end_io struct (aio case)
                 * that this IO needs to convertion to written when IO is
@@ -3182,7 +3214,7 @@ ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
                else
                        ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
                if (ext4_should_dioread_nolock(inode))
-                       set_buffer_uninit(bh_result);
+                       map->m_flags |= EXT4_MAP_UNINIT;
                goto out;
        }
        /* IO end_io complete, convert the filled extent to written */
@@ -3210,14 +3242,12 @@ ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
                 * the buffer head will be unmapped so that
                 * a read from the block returns 0s.
                 */
-               set_buffer_unwritten(bh_result);
+               map->m_flags |= EXT4_MAP_UNWRITTEN;
                goto out1;
        }
 
        /* buffered write, writepage time, convert*/
-       ret = ext4_ext_convert_to_initialized(handle, inode,
-                                               path, iblock,
-                                               max_blocks);
+       ret = ext4_ext_convert_to_initialized(handle, inode, map, path);
        if (ret >= 0)
                ext4_update_inode_fsync_trans(handle, inode, 1);
 out:
@@ -3226,7 +3256,7 @@ out:
                goto out2;
        } else
                allocated = ret;
-       set_buffer_new(bh_result);
+       map->m_flags |= EXT4_MAP_NEW;
        /*
         * if we allocated more blocks than requested
         * we need to make sure we unmap the extra block
@@ -3234,11 +3264,11 @@ out:
         * unmapped later when we find the buffer_head marked
         * new.
         */
-       if (allocated > max_blocks) {
+       if (allocated > map->m_len) {
                unmap_underlying_metadata_blocks(inode->i_sb->s_bdev,
-                                       newblock + max_blocks,
-                                       allocated - max_blocks);
-               allocated = max_blocks;
+                                       newblock + map->m_len,
+                                       allocated - map->m_len);
+               allocated = map->m_len;
        }
 
        /*
@@ -3252,13 +3282,13 @@ out:
                ext4_da_update_reserve_space(inode, allocated, 0);
 
 map_out:
-       set_buffer_mapped(bh_result);
+       map->m_flags |= EXT4_MAP_MAPPED;
 out1:
-       if (allocated > max_blocks)
-               allocated = max_blocks;
+       if (allocated > map->m_len)
+               allocated = map->m_len;
        ext4_ext_show_leaf(inode, path);
-       bh_result->b_bdev = inode->i_sb->s_bdev;
-       bh_result->b_blocknr = newblock;
+       map->m_pblk = newblock;
+       map->m_len = allocated;
 out2:
        if (path) {
                ext4_ext_drop_refs(path);
@@ -3284,26 +3314,23 @@ out2:
  *
  * return < 0, error case.
  */
-int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
-                       ext4_lblk_t iblock,
-                       unsigned int max_blocks, struct buffer_head *bh_result,
-                       int flags)
+int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
+                       struct ext4_map_blocks *map, int flags)
 {
        struct ext4_ext_path *path = NULL;
        struct ext4_extent_header *eh;
        struct ext4_extent newex, *ex, *last_ex;
        ext4_fsblk_t newblock;
-       int err = 0, depth, ret, cache_type;
+       int i, err = 0, depth, ret, cache_type;
        unsigned int allocated = 0;
        struct ext4_allocation_request ar;
        ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio;
 
-       __clear_bit(BH_New, &bh_result->b_state);
        ext_debug("blocks %u/%u requested for inode %lu\n",
-                       iblock, max_blocks, inode->i_ino);
+                 map->m_lblk, map->m_len, inode->i_ino);
 
        /* check in cache */
-       cache_type = ext4_ext_in_cache(inode, iblock, &newex);
+       cache_type = ext4_ext_in_cache(inode, map->m_lblk, &newex);
        if (cache_type) {
                if (cache_type == EXT4_EXT_CACHE_GAP) {
                        if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
@@ -3316,12 +3343,12 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
                        /* we should allocate requested block */
                } else if (cache_type == EXT4_EXT_CACHE_EXTENT) {
                        /* block is already allocated */
-                       newblock = iblock
+                       newblock = map->m_lblk
                                   - le32_to_cpu(newex.ee_block)
                                   + ext_pblock(&newex);
                        /* number of remaining blocks in the extent */
                        allocated = ext4_ext_get_actual_len(&newex) -
-                                       (iblock - le32_to_cpu(newex.ee_block));
+                               (map->m_lblk - le32_to_cpu(newex.ee_block));
                        goto out;
                } else {
                        BUG();
@@ -3329,7 +3356,7 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
        }
 
        /* find extent for this block */
-       path = ext4_ext_find_extent(inode, iblock, NULL);
+       path = ext4_ext_find_extent(inode, map->m_lblk, NULL);
        if (IS_ERR(path)) {
                err = PTR_ERR(path);
                path = NULL;
@@ -3345,8 +3372,9 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
         */
        if (unlikely(path[depth].p_ext == NULL && depth != 0)) {
                EXT4_ERROR_INODE(inode, "bad extent address "
-                                "iblock: %d, depth: %d pblock %lld",
-                                iblock, depth, path[depth].p_block);
+                                "lblock: %lu, depth: %d pblock %lld",
+                                (unsigned long) map->m_lblk, depth,
+                                path[depth].p_block);
                err = -EIO;
                goto out2;
        }
@@ -3364,12 +3392,12 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
                 */
                ee_len = ext4_ext_get_actual_len(ex);
                /* if found extent covers block, simply return it */
-               if (in_range(iblock, ee_block, ee_len)) {
-                       newblock = iblock - ee_block + ee_start;
+               if (in_range(map->m_lblk, ee_block, ee_len)) {
+                       newblock = map->m_lblk - ee_block + ee_start;
                        /* number of remaining blocks in the extent */
-                       allocated = ee_len - (iblock - ee_block);
-                       ext_debug("%u fit into %u:%d -> %llu\n", iblock,
-                                       ee_block, ee_len, newblock);
+                       allocated = ee_len - (map->m_lblk - ee_block);
+                       ext_debug("%u fit into %u:%d -> %llu\n", map->m_lblk,
+                                 ee_block, ee_len, newblock);
 
                        /* Do not put uninitialized extent in the cache */
                        if (!ext4_ext_is_uninitialized(ex)) {
@@ -3379,8 +3407,8 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
                                goto out;
                        }
                        ret = ext4_ext_handle_uninitialized_extents(handle,
-                                       inode, iblock, max_blocks, path,
-                                       flags, allocated, bh_result, newblock);
+                                       inode, map, path, flags, allocated,
+                                       newblock);
                        return ret;
                }
        }
@@ -3394,7 +3422,7 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
                 * put just found gap into cache to speed up
                 * subsequent requests
                 */
-               ext4_ext_put_gap_in_cache(inode, path, iblock);
+               ext4_ext_put_gap_in_cache(inode, path, map->m_lblk);
                goto out2;
        }
        /*
@@ -3402,11 +3430,11 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
         */
 
        /* find neighbour allocated blocks */
-       ar.lleft = iblock;
+       ar.lleft = map->m_lblk;
        err = ext4_ext_search_left(inode, path, &ar.lleft, &ar.pleft);
        if (err)
                goto out2;
-       ar.lright = iblock;
+       ar.lright = map->m_lblk;
        err = ext4_ext_search_right(inode, path, &ar.lright, &ar.pright);
        if (err)
                goto out2;
@@ -3417,26 +3445,26 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
         * EXT_INIT_MAX_LEN and for an uninitialized extent this limit is
         * EXT_UNINIT_MAX_LEN.
         */
-       if (max_blocks > EXT_INIT_MAX_LEN &&
+       if (map->m_len > EXT_INIT_MAX_LEN &&
            !(flags & EXT4_GET_BLOCKS_UNINIT_EXT))
-               max_blocks = EXT_INIT_MAX_LEN;
-       else if (max_blocks > EXT_UNINIT_MAX_LEN &&
+               map->m_len = EXT_INIT_MAX_LEN;
+       else if (map->m_len > EXT_UNINIT_MAX_LEN &&
                 (flags & EXT4_GET_BLOCKS_UNINIT_EXT))
-               max_blocks = EXT_UNINIT_MAX_LEN;
+               map->m_len = EXT_UNINIT_MAX_LEN;
 
-       /* Check if we can really insert (iblock)::(iblock+max_blocks) extent */
-       newex.ee_block = cpu_to_le32(iblock);
-       newex.ee_len = cpu_to_le16(max_blocks);
+       /* Check if we can really insert (m_lblk)::(m_lblk + m_len) extent */
+       newex.ee_block = cpu_to_le32(map->m_lblk);
+       newex.ee_len = cpu_to_le16(map->m_len);
        err = ext4_ext_check_overlap(inode, &newex, path);
        if (err)
                allocated = ext4_ext_get_actual_len(&newex);
        else
-               allocated = max_blocks;
+               allocated = map->m_len;
 
        /* allocate new block */
        ar.inode = inode;
-       ar.goal = ext4_ext_find_goal(inode, path, iblock);
-       ar.logical = iblock;
+       ar.goal = ext4_ext_find_goal(inode, path, map->m_lblk);
+       ar.logical = map->m_lblk;
        ar.len = allocated;
        if (S_ISREG(inode->i_mode))
                ar.flags = EXT4_MB_HINT_DATA;
@@ -3470,21 +3498,33 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
                                                     EXT4_STATE_DIO_UNWRITTEN);
                }
                if (ext4_should_dioread_nolock(inode))
-                       set_buffer_uninit(bh_result);
+                       map->m_flags |= EXT4_MAP_UNINIT;
        }
 
-       if (unlikely(EXT4_I(inode)->i_flags & EXT4_EOFBLOCKS_FL)) {
+       if (unlikely(ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS))) {
                if (unlikely(!eh->eh_entries)) {
                        EXT4_ERROR_INODE(inode,
-                                        "eh->eh_entries == 0 ee_block %d",
-                                        ex->ee_block);
+                                        "eh->eh_entries == 0 and "
+                                        "EOFBLOCKS_FL set");
                        err = -EIO;
                        goto out2;
                }
                last_ex = EXT_LAST_EXTENT(eh);
-               if (iblock + ar.len > le32_to_cpu(last_ex->ee_block)
-                   + ext4_ext_get_actual_len(last_ex))
-                       EXT4_I(inode)->i_flags &= ~EXT4_EOFBLOCKS_FL;
+               /*
+                * If the current leaf block was reached by looking at
+                * the last index block all the way down the tree, and
+                * we are extending the inode beyond the last extent
+                * in the current leaf block, then clear the
+                * EOFBLOCKS_FL flag.
+                */
+               for (i = depth-1; i >= 0; i--) {
+                       if (path[i].p_idx != EXT_LAST_INDEX(path[i].p_hdr))
+                               break;
+               }
+               if ((i < 0) &&
+                   (map->m_lblk + ar.len > le32_to_cpu(last_ex->ee_block) +
+                    ext4_ext_get_actual_len(last_ex)))
+                       ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
        }
        err = ext4_ext_insert_extent(handle, inode, path, &newex, flags);
        if (err) {
@@ -3500,9 +3540,9 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
        /* previous routine could use block we allocated */
        newblock = ext_pblock(&newex);
        allocated = ext4_ext_get_actual_len(&newex);
-       if (allocated > max_blocks)
-               allocated = max_blocks;
-       set_buffer_new(bh_result);
+       if (allocated > map->m_len)
+               allocated = map->m_len;
+       map->m_flags |= EXT4_MAP_NEW;
 
        /*
         * Update reserved blocks/metadata blocks after successful
@@ -3516,18 +3556,18 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
         * when it is _not_ an uninitialized extent.
         */
        if ((flags & EXT4_GET_BLOCKS_UNINIT_EXT) == 0) {
-               ext4_ext_put_in_cache(inode, iblock, allocated, newblock,
+               ext4_ext_put_in_cache(inode, map->m_lblk, allocated, newblock,
                                                EXT4_EXT_CACHE_EXTENT);
                ext4_update_inode_fsync_trans(handle, inode, 1);
        } else
                ext4_update_inode_fsync_trans(handle, inode, 0);
 out:
-       if (allocated > max_blocks)
-               allocated = max_blocks;
+       if (allocated > map->m_len)
+               allocated = map->m_len;
        ext4_ext_show_leaf(inode, path);
-       set_buffer_mapped(bh_result);
-       bh_result->b_bdev = inode->i_sb->s_bdev;
-       bh_result->b_blocknr = newblock;
+       map->m_flags |= EXT4_MAP_MAPPED;
+       map->m_pblk = newblock;
+       map->m_len = allocated;
 out2:
        if (path) {
                ext4_ext_drop_refs(path);
@@ -3625,7 +3665,7 @@ static void ext4_falloc_update_inode(struct inode *inode,
                 * can proceed even if the new size is the same as i_size.
                 */
                if (new_size > i_size_read(inode))
-                       EXT4_I(inode)->i_flags |= EXT4_EOFBLOCKS_FL;
+                       ext4_set_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
        }
 
 }
@@ -3640,55 +3680,57 @@ static void ext4_falloc_update_inode(struct inode *inode,
 long ext4_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len)
 {
        handle_t *handle;
-       ext4_lblk_t block;
        loff_t new_size;
        unsigned int max_blocks;
        int ret = 0;
        int ret2 = 0;
        int retries = 0;
-       struct buffer_head map_bh;
+       struct ext4_map_blocks map;
        unsigned int credits, blkbits = inode->i_blkbits;
 
        /*
         * currently supporting (pre)allocate mode for extent-based
         * files _only_
         */
-       if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL))
+       if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
                return -EOPNOTSUPP;
 
        /* preallocation to directories is currently not supported */
        if (S_ISDIR(inode->i_mode))
                return -ENODEV;
 
-       block = offset >> blkbits;
+       map.m_lblk = offset >> blkbits;
        /*
         * We can't just convert len to max_blocks because
         * If blocksize = 4096 offset = 3072 and len = 2048
         */
        max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
-                                                       - block;
+               - map.m_lblk;
        /*
         * credits to insert 1 extent into extent tree
         */
        credits = ext4_chunk_trans_blocks(inode, max_blocks);
        mutex_lock(&inode->i_mutex);
+       ret = inode_newsize_ok(inode, (len + offset));
+       if (ret) {
+               mutex_unlock(&inode->i_mutex);
+               return ret;
+       }
 retry:
        while (ret >= 0 && ret < max_blocks) {
-               block = block + ret;
-               max_blocks = max_blocks - ret;
+               map.m_lblk = map.m_lblk + ret;
+               map.m_len = max_blocks = max_blocks - ret;
                handle = ext4_journal_start(inode, credits);
                if (IS_ERR(handle)) {
                        ret = PTR_ERR(handle);
                        break;
                }
-               map_bh.b_state = 0;
-               ret = ext4_get_blocks(handle, inode, block,
-                                     max_blocks, &map_bh,
+               ret = ext4_map_blocks(handle, inode, &map,
                                      EXT4_GET_BLOCKS_CREATE_UNINIT_EXT);
                if (ret <= 0) {
 #ifdef EXT4FS_DEBUG
                        WARN_ON(ret <= 0);
-                       printk(KERN_ERR "%s: ext4_ext_get_blocks "
+                       printk(KERN_ERR "%s: ext4_ext_map_blocks "
                                    "returned error inode#%lu, block=%u, "
                                    "max_blocks=%u", __func__,
                                    inode->i_ino, block, max_blocks);
@@ -3697,14 +3739,14 @@ retry:
                        ret2 = ext4_journal_stop(handle);
                        break;
                }
-               if ((block + ret) >= (EXT4_BLOCK_ALIGN(offset + len,
+               if ((map.m_lblk + ret) >= (EXT4_BLOCK_ALIGN(offset + len,
                                                blkbits) >> blkbits))
                        new_size = offset + len;
                else
-                       new_size = (block + ret) << blkbits;
+                       new_size = (map.m_lblk + ret) << blkbits;
 
                ext4_falloc_update_inode(inode, mode, new_size,
-                                               buffer_new(&map_bh));
+                                        (map.m_flags & EXT4_MAP_NEW));
                ext4_mark_inode_dirty(handle, inode);
                ret2 = ext4_journal_stop(handle);
                if (ret2)
@@ -3733,42 +3775,39 @@ int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
                                    ssize_t len)
 {
        handle_t *handle;
-       ext4_lblk_t block;
        unsigned int max_blocks;
        int ret = 0;
        int ret2 = 0;
-       struct buffer_head map_bh;
+       struct ext4_map_blocks map;
        unsigned int credits, blkbits = inode->i_blkbits;
 
-       block = offset >> blkbits;
+       map.m_lblk = offset >> blkbits;
        /*
         * We can't just convert len to max_blocks because
         * If blocksize = 4096 offset = 3072 and len = 2048
         */
-       max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
-                                                       - block;
+       max_blocks = ((EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) -
+                     map.m_lblk);
        /*
         * credits to insert 1 extent into extent tree
         */
        credits = ext4_chunk_trans_blocks(inode, max_blocks);
        while (ret >= 0 && ret < max_blocks) {
-               block = block + ret;
-               max_blocks = max_blocks - ret;
+               map.m_lblk += ret;
+               map.m_len = (max_blocks -= ret);
                handle = ext4_journal_start(inode, credits);
                if (IS_ERR(handle)) {
                        ret = PTR_ERR(handle);
                        break;
                }
-               map_bh.b_state = 0;
-               ret = ext4_get_blocks(handle, inode, block,
-                                     max_blocks, &map_bh,
+               ret = ext4_map_blocks(handle, inode, &map,
                                      EXT4_GET_BLOCKS_IO_CONVERT_EXT);
                if (ret <= 0) {
                        WARN_ON(ret <= 0);
-                       printk(KERN_ERR "%s: ext4_ext_get_blocks "
+                       printk(KERN_ERR "%s: ext4_ext_map_blocks "
                                    "returned error inode#%lu, block=%u, "
                                    "max_blocks=%u", __func__,
-                                   inode->i_ino, block, max_blocks);
+                                   inode->i_ino, map.m_lblk, map.m_len);
                }
                ext4_mark_inode_dirty(handle, inode);
                ret2 = ext4_journal_stop(handle);
@@ -3898,7 +3937,7 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
        int error = 0;
 
        /* fallback to generic here if not in extents fmt */
-       if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL))
+       if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
                return generic_block_fiemap(inode, fieinfo, start, len,
                        ext4_get_block);