Btrfs: many file_write fixes, inline data
authorChris Mason <chris.mason@oracle.com>
Thu, 19 Apr 2007 17:37:44 +0000 (13:37 -0400)
committerDavid Woodhouse <dwmw2@hera.kernel.org>
Thu, 19 Apr 2007 17:37:44 +0000 (13:37 -0400)
Signed-off-by: Chris Mason <chris.mason@oracle.com>
fs/btrfs/ctree.h
fs/btrfs/disk-io.c
fs/btrfs/extent-tree.c
fs/btrfs/file-item.c
fs/btrfs/super.c

index 8a329d3901a2b80a11bbef3d92fcdaf7c9ccfc0a..f1800008be34d1b29b4116d39288268eff2d8250 100644 (file)
@@ -73,6 +73,9 @@ struct btrfs_header {
                               (sizeof(struct btrfs_disk_key) + sizeof(u64)))
 #define __BTRFS_LEAF_DATA_SIZE(bs) ((bs) - sizeof(struct btrfs_header))
 #define BTRFS_LEAF_DATA_SIZE(r) (__BTRFS_LEAF_DATA_SIZE(r->blocksize))
+#define BTRFS_MAX_INLINE_DATA_SIZE(r) (BTRFS_LEAF_DATA_SIZE(r) - \
+                                       sizeof(struct btrfs_item) - \
+                                       sizeof(struct btrfs_file_extent_item))
 
 struct buffer_head;
 /*
@@ -204,8 +207,12 @@ struct btrfs_root_item {
        __le32 refs;
 } __attribute__ ((__packed__));
 
+#define BTRFS_FILE_EXTENT_REG 0
+#define BTRFS_FILE_EXTENT_INLINE 1
+
 struct btrfs_file_extent_item {
        __le64 generation;
+       u8 type;
        /*
         * disk space consumed by the extent, checksum blocks are included
         * in these numbers
@@ -862,6 +869,34 @@ static inline u8 *btrfs_leaf_data(struct btrfs_leaf *l)
        return (u8 *)l->items;
 }
 
+static inline int btrfs_file_extent_type(struct btrfs_file_extent_item *e)
+{
+       return e->type;
+}
+static inline void btrfs_set_file_extent_type(struct btrfs_file_extent_item *e,
+                                             u8 val)
+{
+       e->type = val;
+}
+
+static inline char *btrfs_file_extent_inline_start(struct
+                                                  btrfs_file_extent_item *e)
+{
+       return (char *)(&e->disk_blocknr);
+}
+
+static inline u32 btrfs_file_extent_calc_inline_size(u32 datasize)
+{
+       return (unsigned long)(&((struct
+                 btrfs_file_extent_item *)NULL)->disk_blocknr) + datasize;
+}
+
+static inline u32 btrfs_file_extent_inline_len(struct btrfs_item *e)
+{
+       struct btrfs_file_extent_item *fe = NULL;
+       return btrfs_item_size(e) - (unsigned long)(&fe->disk_blocknr);
+}
+
 static inline u64 btrfs_file_extent_disk_blocknr(struct btrfs_file_extent_item
                                                 *e)
 {
index a2a3f529cada902ab00724406854c8b9b5fdb730..11e17a2f736bcbce6cba2fb870ed836ecc7ac480 100644 (file)
@@ -108,6 +108,12 @@ int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
 
        int ret;
 
+       if (logical == 0) {
+               bh->b_bdev = NULL;
+               bh->b_blocknr = 0;
+               set_buffer_mapped(bh);
+               return 0;
+       }
        root = root->fs_info->dev_root;
        ret = radix_tree_gang_lookup(&root->fs_info->dev_radix,
                                     (void **)lookup,
index cb04a70eb7e843de939a11f5c4d5cd286ece5d18..b2faad3e8791a0908826fdc5b813f776c767b6da 100644 (file)
@@ -111,6 +111,9 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
                                continue;
                        fi = btrfs_item_ptr(buf_leaf, i,
                                            struct btrfs_file_extent_item);
+                       if (btrfs_file_extent_type(fi) ==
+                           BTRFS_FILE_EXTENT_INLINE)
+                               continue;
                        ret = btrfs_inc_extent_ref(trans, root,
                                    btrfs_file_extent_disk_blocknr(fi),
                                    btrfs_file_extent_disk_num_blocks(fi));
@@ -539,6 +542,8 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans,
                if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
                        continue;
                fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
+               if (btrfs_file_extent_type(fi) == BTRFS_FILE_EXTENT_INLINE)
+                       continue;
                /*
                 * FIXME make sure to insert a trans record that
                 * repeats the snapshot del on crash
index 8cc3c1d15412309de185f34c85d63783012fc475..10e4cf08e9ed1843f65b82995a19a4373480fa80 100644 (file)
@@ -34,6 +34,7 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
        btrfs_set_file_extent_offset(item, 0);
        btrfs_set_file_extent_num_blocks(item, num_blocks);
        btrfs_set_file_extent_generation(item, trans->transid);
+       btrfs_set_file_extent_type(item, BTRFS_FILE_EXTENT_REG);
        btrfs_mark_buffer_dirty(path->nodes[0]);
 
        btrfs_release_path(root, path);
index 583cd87a62e248736244decf82cdca414a035500..1b286bb26f126cb60f5e196ecc63b5a193831e3b 100644 (file)
@@ -317,19 +317,22 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
                        break;
                if (btrfs_disk_key_offset(found_key) < inode->i_size)
                        break;
+               found_extent = 0;
                if (btrfs_disk_key_type(found_key) == BTRFS_EXTENT_DATA_KEY) {
                        fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
                                            path->slots[0],
                                            struct btrfs_file_extent_item);
-                       extent_start = btrfs_file_extent_disk_blocknr(fi);
-                       extent_num_blocks =
-                               btrfs_file_extent_disk_num_blocks(fi);
-                       /* FIXME blocksize != 4096 */
-                       inode->i_blocks -=
-                               btrfs_file_extent_num_blocks(fi) << 3;
-                       found_extent = 1;
-               } else {
-                       found_extent = 0;
+                       if (btrfs_file_extent_type(fi) !=
+                           BTRFS_FILE_EXTENT_INLINE) {
+                               extent_start =
+                                       btrfs_file_extent_disk_blocknr(fi);
+                               extent_num_blocks =
+                                       btrfs_file_extent_disk_num_blocks(fi);
+                               /* FIXME blocksize != 4096 */
+                               inode->i_blocks -=
+                                       btrfs_file_extent_num_blocks(fi) << 3;
+                               found_extent = 1;
+                       }
                }
                ret = btrfs_del_item(trans, root, path);
                BUG_ON(ret);
@@ -1010,9 +1013,9 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
        u64 extent_start = 0;
        u64 extent_end = 0;
        u64 objectid = inode->i_ino;
+       u32 found_type;
        struct btrfs_path *path;
        struct btrfs_root *root = BTRFS_I(inode)->root;
-       struct btrfs_trans_handle *trans = NULL;
        struct btrfs_file_extent_item *item;
        struct btrfs_leaf *leaf;
        struct btrfs_disk_key *found_key;
@@ -1021,13 +1024,12 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
        BUG_ON(!path);
        btrfs_init_path(path);
        if (create) {
-               trans = btrfs_start_transaction(root, 1);
                WARN_ON(1);
        }
 
-       ret = btrfs_lookup_file_extent(trans, root, path,
+       ret = btrfs_lookup_file_extent(NULL, root, path,
                                       inode->i_ino,
-                                      iblock << inode->i_blkbits, create);
+                                      iblock << inode->i_blkbits, 0);
        if (ret < 0) {
                err = ret;
                goto out;
@@ -1036,7 +1038,7 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
        if (ret != 0) {
                if (path->slots[0] == 0) {
                        btrfs_release_path(root, path);
-                       goto allocate;
+                       goto out;
                }
                path->slots[0]--;
        }
@@ -1047,73 +1049,51 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
        blocknr = btrfs_file_extent_disk_blocknr(item);
        blocknr += btrfs_file_extent_offset(item);
 
-       /* exact match found, use it, FIXME, deal with extents
-        * other than the page size
-        */
-       if (0 && ret == 0) {
-               err = 0;
-               if (create &&
-                   btrfs_file_extent_generation(item) != trans->transid) {
-                       struct btrfs_key ins;
-                       ret = btrfs_alloc_extent(trans, root, 1,
-                                                blocknr, (u64)-1, &ins);
-                       BUG_ON(ret);
-                       btrfs_set_file_extent_disk_blocknr(item, ins.objectid);
-                       mark_buffer_dirty(path->nodes[0]);
-                       ret = btrfs_free_extent(trans, root,
-                                               blocknr, 1, 0);
-                       BUG_ON(ret);
-                       blocknr = ins.objectid;
-
-               }
-               btrfs_map_bh_to_logical(root, result, blocknr);
-               goto out;
-       }
-
        /* are we inside the extent that was found? */
        found_key = &leaf->items[path->slots[0]].key;
+       found_type = btrfs_disk_key_type(found_key);
        if (btrfs_disk_key_objectid(found_key) != objectid ||
-           btrfs_disk_key_type(found_key) != BTRFS_EXTENT_DATA_KEY) {
+           found_type != BTRFS_EXTENT_DATA_KEY) {
                extent_end = 0;
                extent_start = 0;
                btrfs_release_path(root, path);
-               goto allocate;
-       }
-
-       extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
-       extent_start = extent_start >> inode->i_blkbits;
-       extent_end = extent_start + btrfs_file_extent_num_blocks(item);
-       if (iblock >= extent_start && iblock < extent_end) {
-               err = 0;
-               btrfs_map_bh_to_logical(root, result, blocknr + iblock -
-                                       extent_start);
-               goto out;
-       }
-allocate:
-       /* ok, create a new extent */
-       if (!create) {
-               err = 0;
                goto out;
        }
-#if 0
-       ret = btrfs_alloc_file_extent(trans, root, objectid,
-                                     iblock << inode->i_blkbits,
-                                     1, extent_end, &blocknr);
-       if (ret) {
-               err = ret;
-               goto out;
+       found_type = btrfs_file_extent_type(item);
+       extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
+       if (found_type == BTRFS_FILE_EXTENT_REG) {
+               extent_start = extent_start >> inode->i_blkbits;
+               extent_end = extent_start + btrfs_file_extent_num_blocks(item);
+               if (iblock >= extent_start && iblock < extent_end) {
+                       err = 0;
+                       btrfs_map_bh_to_logical(root, result, blocknr +
+                                               iblock - extent_start);
+                       goto out;
+               }
+       } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+               char *ptr;
+               char *map;
+               u32 size;
+               size = btrfs_file_extent_inline_len(leaf->items +
+                                                   path->slots[0]);
+               extent_end = (extent_start + size) >> inode->i_blkbits;
+               extent_start >>= inode->i_blkbits;
+               if (iblock < extent_start || iblock > extent_end) {
+                       goto out;
+               }
+               ptr = btrfs_file_extent_inline_start(item);
+               map = kmap(result->b_page);
+               memcpy(map, ptr, size);
+               memset(map + size, 0, PAGE_CACHE_SIZE - size);
+               flush_dcache_page(result->b_page);
+               kunmap(result->b_page);
+               set_buffer_uptodate(result);
+               SetPageChecked(result->b_page);
+               btrfs_map_bh_to_logical(root, result, 0);
        }
-       inode->i_blocks += inode->i_sb->s_blocksize >> 9;
-       set_buffer_new(result);
-       map_bh(result, inode->i_sb, blocknr);
-
-       btrfs_map_bh_to_logical(root, result, blocknr);
-#endif
 out:
        btrfs_release_path(root, path);
        btrfs_free_path(path);
-       if (trans)
-               btrfs_end_transaction(trans, root);
        return err;
 }
 
@@ -1124,7 +1104,6 @@ static int btrfs_get_block(struct inode *inode, sector_t iblock,
        struct btrfs_root *root = BTRFS_I(inode)->root;
        mutex_lock(&root->fs_info->fs_mutex);
        err = btrfs_get_block_lock(inode, iblock, result, create);
-       // err = btrfs_get_block_inline(inode, iblock, result, create);
        mutex_unlock(&root->fs_info->fs_mutex);
        return err;
 }
@@ -1134,11 +1113,6 @@ static int btrfs_prepare_write(struct file *file, struct page *page,
 {
        return nobh_prepare_write(page, from, to, btrfs_get_block);
 }
-static int btrfs_commit_write(struct file *file, struct page *page,
-                              unsigned from, unsigned to)
-{
-       return nobh_commit_write(file, page, from, to);
-}
 
 static void btrfs_write_super(struct super_block *sb)
 {
@@ -1150,9 +1124,227 @@ static int btrfs_readpage(struct file *file, struct page *page)
        return mpage_readpage(page, btrfs_get_block);
 }
 
+/*
+ * While block_write_full_page is writing back the dirty buffers under
+ * the page lock, whoever dirtied the buffers may decide to clean them
+ * again at any time.  We handle that by only looking at the buffer
+ * state inside lock_buffer().
+ *
+ * If block_write_full_page() is called for regular writeback
+ * (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a page which has a
+ * locked buffer.   This only can happen if someone has written the buffer
+ * directly, with submit_bh().  At the address_space level PageWriteback
+ * prevents this contention from occurring.
+ */
+static int __btrfs_write_full_page(struct inode *inode, struct page *page,
+                                  struct writeback_control *wbc)
+{
+       int err;
+       sector_t block;
+       sector_t last_block;
+       struct buffer_head *bh, *head;
+       const unsigned blocksize = 1 << inode->i_blkbits;
+       int nr_underway = 0;
+
+       BUG_ON(!PageLocked(page));
+
+       last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
+
+       if (!page_has_buffers(page)) {
+               create_empty_buffers(page, blocksize,
+                                       (1 << BH_Dirty)|(1 << BH_Uptodate));
+       }
+
+       /*
+        * Be very careful.  We have no exclusion from __set_page_dirty_buffers
+        * here, and the (potentially unmapped) buffers may become dirty at
+        * any time.  If a buffer becomes dirty here after we've inspected it
+        * then we just miss that fact, and the page stays dirty.
+        *
+        * Buffers outside i_size may be dirtied by __set_page_dirty_buffers;
+        * handle that here by just cleaning them.
+        */
+
+       block = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
+       head = page_buffers(page);
+       bh = head;
+
+       /*
+        * Get all the dirty buffers mapped to disk addresses and
+        * handle any aliases from the underlying blockdev's mapping.
+        */
+       do {
+               if (block > last_block) {
+                       /*
+                        * mapped buffers outside i_size will occur, because
+                        * this page can be outside i_size when there is a
+                        * truncate in progress.
+                        */
+                       /*
+                        * The buffer was zeroed by block_write_full_page()
+                        */
+                       clear_buffer_dirty(bh);
+                       set_buffer_uptodate(bh);
+               } else if (!buffer_mapped(bh) && buffer_dirty(bh)) {
+                       WARN_ON(bh->b_size != blocksize);
+                       err = btrfs_get_block(inode, block, bh, 0);
+                       if (err)
+                               goto recover;
+                       if (buffer_new(bh)) {
+                               /* blockdev mappings never come here */
+                               clear_buffer_new(bh);
+                               unmap_underlying_metadata(bh->b_bdev,
+                                                       bh->b_blocknr);
+                       }
+               }
+               bh = bh->b_this_page;
+               block++;
+       } while (bh != head);
+
+       do {
+               if (!buffer_mapped(bh))
+                       continue;
+               /*
+                * If it's a fully non-blocking write attempt and we cannot
+                * lock the buffer then redirty the page.  Note that this can
+                * potentially cause a busy-wait loop from pdflush and kswapd
+                * activity, but those code paths have their own higher-level
+                * throttling.
+                */
+               if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
+                       lock_buffer(bh);
+               } else if (test_set_buffer_locked(bh)) {
+                       redirty_page_for_writepage(wbc, page);
+                       continue;
+               }
+               if (test_clear_buffer_dirty(bh) && bh->b_blocknr != 0) {
+                       mark_buffer_async_write(bh);
+               } else {
+                       unlock_buffer(bh);
+               }
+       } while ((bh = bh->b_this_page) != head);
+
+       /*
+        * The page and its buffers are protected by PageWriteback(), so we can
+        * drop the bh refcounts early.
+        */
+       BUG_ON(PageWriteback(page));
+       set_page_writeback(page);
+
+       do {
+               struct buffer_head *next = bh->b_this_page;
+               if (buffer_async_write(bh)) {
+                       submit_bh(WRITE, bh);
+                       nr_underway++;
+               }
+               bh = next;
+       } while (bh != head);
+       unlock_page(page);
+
+       err = 0;
+done:
+       if (nr_underway == 0) {
+               /*
+                * The page was marked dirty, but the buffers were
+                * clean.  Someone wrote them back by hand with
+                * ll_rw_block/submit_bh.  A rare case.
+                */
+               int uptodate = 1;
+               do {
+                       if (!buffer_uptodate(bh)) {
+                               uptodate = 0;
+                               break;
+                       }
+                       bh = bh->b_this_page;
+               } while (bh != head);
+               if (uptodate)
+                       SetPageUptodate(page);
+               end_page_writeback(page);
+               /*
+                * The page and buffer_heads can be released at any time from
+                * here on.
+                */
+               wbc->pages_skipped++;   /* We didn't write this page */
+       }
+       return err;
+
+recover:
+       /*
+        * ENOSPC, or some other error.  We may already have added some
+        * blocks to the file, so we need to write these out to avoid
+        * exposing stale data.
+        * The page is currently locked and not marked for writeback
+        */
+       bh = head;
+       /* Recovery: lock and submit the mapped buffers */
+       do {
+               if (buffer_mapped(bh) && buffer_dirty(bh)) {
+                       lock_buffer(bh);
+                       mark_buffer_async_write(bh);
+               } else {
+                       /*
+                        * The buffer may have been set dirty during
+                        * attachment to a dirty page.
+                        */
+                       clear_buffer_dirty(bh);
+               }
+       } while ((bh = bh->b_this_page) != head);
+       SetPageError(page);
+       BUG_ON(PageWriteback(page));
+       set_page_writeback(page);
+       do {
+               struct buffer_head *next = bh->b_this_page;
+               if (buffer_async_write(bh)) {
+                       clear_buffer_dirty(bh);
+                       submit_bh(WRITE, bh);
+                       nr_underway++;
+               }
+               bh = next;
+       } while (bh != head);
+       unlock_page(page);
+       goto done;
+}
+
+/*
+ * The generic ->writepage function for buffer-backed address_spaces
+ */
 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
 {
-       return nobh_writepage(page, btrfs_get_block, wbc);
+       struct inode * const inode = page->mapping->host;
+       loff_t i_size = i_size_read(inode);
+       const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
+       unsigned offset;
+       void *kaddr;
+
+       /* Is the page fully inside i_size? */
+       if (page->index < end_index)
+               return __btrfs_write_full_page(inode, page, wbc);
+
+       /* Is the page fully outside i_size? (truncate in progress) */
+       offset = i_size & (PAGE_CACHE_SIZE-1);
+       if (page->index >= end_index+1 || !offset) {
+               /*
+                * The page may have dirty, unmapped buffers.  For example,
+                * they may have been added in ext3_writepage().  Make them
+                * freeable here, so the page does not leak.
+                */
+               block_invalidatepage(page, 0);
+               unlock_page(page);
+               return 0; /* don't care */
+       }
+
+       /*
+        * The page straddles i_size.  It must be zeroed out on each and every
+        * writepage invokation because it may be mmapped.  "A file is mapped
+        * in multiples of the page size.  For a file that is not a multiple of
+        * the  page size, the remaining memory is zeroed when mapped, and
+        * writes to that region are not written out to the file."
+        */
+       kaddr = kmap_atomic(page, KM_USER0);
+       memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
+       flush_dcache_page(page);
+       kunmap_atomic(kaddr, KM_USER0);
+       return __btrfs_write_full_page(inode, page, wbc);
 }
 
 static void btrfs_truncate(struct inode *inode)
@@ -1179,6 +1371,29 @@ static void btrfs_truncate(struct inode *inode)
        mark_inode_dirty(inode);
 }
 
+/*
+ * Make sure any changes to nobh_commit_write() are reflected in
+ * nobh_truncate_page(), since it doesn't call commit_write().
+ */
+static int btrfs_commit_write(struct file *file, struct page *page,
+                             unsigned from, unsigned to)
+{
+       struct inode *inode = page->mapping->host;
+       struct buffer_head *bh;
+       loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
+
+       SetPageUptodate(page);
+       bh = page_buffers(page);
+       if (buffer_mapped(bh) && bh->b_blocknr != 0) {
+               set_page_dirty(page);
+       }
+       if (pos > inode->i_size) {
+               i_size_write(inode, pos);
+               mark_inode_dirty(inode);
+       }
+       return 0;
+}
+
 static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes,
                                struct page **prepared_pages,
                                const char __user * buf)
@@ -1234,6 +1449,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
        int ret;
        int this_write;
        struct inode *inode = file->f_path.dentry->d_inode;
+       struct buffer_head *bh;
+       struct btrfs_file_extent_item *ei;
 
        for (i = 0; i < num_pages; i++) {
                offset = pos & (PAGE_CACHE_SIZE -1);
@@ -1242,16 +1459,47 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 
                mutex_lock(&root->fs_info->fs_mutex);
                trans = btrfs_start_transaction(root, 1);
-               btrfs_csum_file_block(trans, root, inode->i_ino,
+
+               bh = page_buffers(pages[i]);
+               if (buffer_mapped(bh) && bh->b_blocknr == 0) {
+                       struct btrfs_key key;
+                       struct btrfs_path *path;
+                       char *ptr;
+                       u32 datasize;
+
+                       path = btrfs_alloc_path();
+                       BUG_ON(!path);
+                       key.objectid = inode->i_ino;
+                       key.offset = pages[i]->index << PAGE_CACHE_SHIFT;
+                       key.flags = 0;
+                       btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
+                       BUG_ON(write_bytes >= PAGE_CACHE_SIZE);
+                       datasize = offset +
+                               btrfs_file_extent_calc_inline_size(write_bytes);
+                       ret = btrfs_insert_empty_item(trans, root, path, &key,
+                                                     datasize);
+                       BUG_ON(ret);
+                       ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
+                              path->slots[0], struct btrfs_file_extent_item);
+                       btrfs_set_file_extent_generation(ei, trans->transid);
+                       btrfs_set_file_extent_type(ei,
+                                                  BTRFS_FILE_EXTENT_INLINE);
+                       ptr = btrfs_file_extent_inline_start(ei);
+                       memcpy(ptr, bh->b_data, offset + write_bytes);
+                       mark_buffer_dirty(path->nodes[0]);
+                       btrfs_free_path(path);
+               } else {
+                       btrfs_csum_file_block(trans, root, inode->i_ino,
                                      pages[i]->index << PAGE_CACHE_SHIFT,
                                      kmap(pages[i]), PAGE_CACHE_SIZE);
-               kunmap(pages[i]);
+                       kunmap(pages[i]);
+               }
                SetPageChecked(pages[i]);
                ret = btrfs_end_transaction(trans, root);
                BUG_ON(ret);
                mutex_unlock(&root->fs_info->fs_mutex);
 
-               ret = nobh_commit_write(file, pages[i], offset,
+               ret = btrfs_commit_write(file, pages[i], offset,
                                         offset + this_write);
                pos += this_write;
                if (ret) {
@@ -1275,12 +1523,16 @@ static int drop_extents(struct btrfs_trans_handle *trans,
        struct btrfs_leaf *leaf;
        int slot;
        struct btrfs_file_extent_item *extent;
-       u64 extent_end;
+       u64 extent_end = 0;
        int keep;
        struct btrfs_file_extent_item old;
        struct btrfs_path *path;
        u64 search_start = start;
        int bookend;
+       int found_type;
+       int found_extent;
+       int found_inline;
+
        path = btrfs_alloc_path();
        if (!path)
                return -ENOMEM;
@@ -1292,37 +1544,62 @@ static int drop_extents(struct btrfs_trans_handle *trans,
                        goto out;
                if (ret > 0) {
                        if (path->slots[0] == 0) {
-                               ret = -ENOENT;
+                               ret = 0;
                                goto out;
                        }
                        path->slots[0]--;
                }
                keep = 0;
                bookend = 0;
+               found_extent = 0;
+               found_inline = 0;
+               extent = NULL;
                leaf = btrfs_buffer_leaf(path->nodes[0]);
                slot = path->slots[0];
                btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
-               extent = btrfs_item_ptr(leaf, slot,
-                                       struct btrfs_file_extent_item);
-               extent_end = key.offset +
-                       (btrfs_file_extent_num_blocks(extent) <<
-                        inode->i_blkbits);
                if (key.offset >= end || key.objectid != inode->i_ino) {
                        ret = 0;
                        goto out;
                }
-               if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
+               if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) {
+                       ret = 0;
+                       goto out;
+               }
+               extent = btrfs_item_ptr(leaf, slot,
+                                       struct btrfs_file_extent_item);
+               found_type = btrfs_file_extent_type(extent);
+               if (found_type == BTRFS_FILE_EXTENT_REG) {
+                       extent_end = key.offset +
+                               (btrfs_file_extent_num_blocks(extent) <<
+                                inode->i_blkbits);
+                       found_extent = 1;
+               } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+                       found_inline = 1;
+                       extent_end = key.offset +
+                            btrfs_file_extent_inline_len(leaf->items + slot);
+               }
+
+               if (!found_extent && !found_inline) {
+                       ret = 0;
                        goto out;
-               if (search_start >= extent_end)
+               }
+
+               if (search_start >= extent_end) {
+                       ret = 0;
                        goto out;
+               }
+
                search_start = extent_end;
 
                if (end < extent_end && end >= key.offset) {
-                       memcpy(&old, extent, sizeof(old));
-                       ret = btrfs_inc_extent_ref(trans, root,
-                                  btrfs_file_extent_disk_blocknr(&old),
-                                  btrfs_file_extent_disk_num_blocks(&old));
-                       BUG_ON(ret);
+                       if (found_extent) {
+                               memcpy(&old, extent, sizeof(old));
+                               ret = btrfs_inc_extent_ref(trans, root,
+                                     btrfs_file_extent_disk_blocknr(&old),
+                                     btrfs_file_extent_disk_num_blocks(&old));
+                               BUG_ON(ret);
+                       }
+                       WARN_ON(found_inline);
                        bookend = 1;
                }
 
@@ -1332,25 +1609,45 @@ static int drop_extents(struct btrfs_trans_handle *trans,
                        /* truncate existing extent */
                        keep = 1;
                        WARN_ON(start & (root->blocksize - 1));
-                       new_num = (start - key.offset) >> inode->i_blkbits;
-                       old_num = btrfs_file_extent_num_blocks(extent);
-                       inode->i_blocks -= (old_num - new_num) << 3;
-                       btrfs_set_file_extent_num_blocks(extent, new_num);
-                       mark_buffer_dirty(path->nodes[0]);
+                       if (found_extent) {
+                               new_num = (start - key.offset) >>
+                                       inode->i_blkbits;
+                               old_num = btrfs_file_extent_num_blocks(extent);
+                               inode->i_blocks -= (old_num - new_num) << 3;
+                               btrfs_set_file_extent_num_blocks(extent,
+                                                                new_num);
+                               mark_buffer_dirty(path->nodes[0]);
+                       } else {
+                               WARN_ON(1);
+                               /*
+                               ret = btrfs_truncate_item(trans, root, path,
+                                                         start - key.offset);
+                               BUG_ON(ret);
+                               */
+                       }
                }
                if (!keep) {
-                       u64 disk_blocknr;
-                       u64 disk_num_blocks;
-                       disk_blocknr = btrfs_file_extent_disk_blocknr(extent);
-                       disk_num_blocks =
-                               btrfs_file_extent_disk_num_blocks(extent);
+                       u64 disk_blocknr = 0;
+                       u64 disk_num_blocks = 0;
+                       u64 extent_num_blocks = 0;
+                       if (found_extent) {
+                               disk_blocknr =
+                                     btrfs_file_extent_disk_blocknr(extent);
+                               disk_num_blocks =
+                                     btrfs_file_extent_disk_num_blocks(extent);
+                               extent_num_blocks =
+                                     btrfs_file_extent_num_blocks(extent);
+                       }
                        ret = btrfs_del_item(trans, root, path);
                        BUG_ON(ret);
-                       inode->i_blocks -=
-                               btrfs_file_extent_num_blocks(extent) << 3;
                        btrfs_release_path(root, path);
-                       ret = btrfs_free_extent(trans, root, disk_blocknr,
-                                               disk_num_blocks, 0);
+                       if (found_extent) {
+                               inode->i_blocks -=
+                               btrfs_file_extent_num_blocks(extent) << 3;
+                               ret = btrfs_free_extent(trans, root,
+                                                       disk_blocknr,
+                                                       disk_num_blocks, 0);
+                       }
 
                        BUG_ON(ret);
                        if (!bookend && search_start >= end) {
@@ -1360,7 +1657,7 @@ static int drop_extents(struct btrfs_trans_handle *trans,
                        if (!bookend)
                                continue;
                }
-               if (bookend) {
+               if (bookend && found_extent) {
                        /* create bookend */
                        struct btrfs_key ins;
                        ins.objectid = inode->i_ino;
@@ -1390,6 +1687,8 @@ static int drop_extents(struct btrfs_trans_handle *trans,
                                    btrfs_file_extent_num_blocks(&old) -
                                    ((end - key.offset) >> inode->i_blkbits));
 
+                       btrfs_set_file_extent_type(extent,
+                                                  BTRFS_FILE_EXTENT_REG);
                        btrfs_set_file_extent_generation(extent,
                                    btrfs_file_extent_generation(&old));
                        btrfs_mark_buffer_dirty(path->nodes[0]);
@@ -1445,7 +1744,8 @@ static int prepare_pages(struct btrfs_root *root,
                        if (err)
                                goto failed_truncate;
                        bh = bh->b_this_page;
-                       alloc_extent_start++;
+                       if (alloc_extent_start)
+                               alloc_extent_start++;
                } while (bh != head);
                pos += this_write;
                WARN_ON(this_write > write_bytes);
@@ -1543,12 +1843,20 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
                                   start_pos,
                                   (pos + count + root->blocksize -1) &
                                   ~((u64)root->blocksize - 1));
+               BUG_ON(ret);
        }
-       ret = btrfs_alloc_extent(trans, root, num_blocks, 1,
+       if (inode->i_size >= PAGE_CACHE_SIZE || pos + count < inode->i_size ||
+           pos + count - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) {
+               ret = btrfs_alloc_extent(trans, root, num_blocks, 1,
                                 (u64)-1, &ins);
-       BUG_ON(ret);
-       ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
+               BUG_ON(ret);
+               ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
                                       start_pos, ins.objectid, ins.offset);
+               BUG_ON(ret);
+       } else {
+               ins.offset = 0;
+               ins.objectid = 0;
+       }
        BUG_ON(ret);
        alloc_extent_start = ins.objectid;
        ret = btrfs_end_transaction(trans, root);
@@ -1567,7 +1875,8 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
                BUG_ON(ret);
 
                /* FIXME blocks != pagesize */
-               alloc_extent_start += num_pages;
+               if (alloc_extent_start)
+                       alloc_extent_start += num_pages;
                ret = btrfs_copy_from_user(pos, num_pages,
                                           write_bytes, pages, buf);
                BUG_ON(ret);
@@ -1779,10 +2088,11 @@ static int btrfs_read_actor(read_descriptor_t *desc, struct page *page,
        if (!PageChecked(page)) {
                /* FIXME, do it per block */
                struct btrfs_root *root = BTRFS_I(inode)->root;
+
                int ret = btrfs_csum_verify_file_block(root,
-                                         page->mapping->host->i_ino,
-                                         page->index << PAGE_CACHE_SHIFT,
-                                         kmap(page), PAGE_CACHE_SIZE);
+                                 page->mapping->host->i_ino,
+                                 page->index << PAGE_CACHE_SHIFT,
+                                 kmap(page), PAGE_CACHE_SIZE);
                if (ret) {
                        printk("failed to verify ino %lu page %lu\n",
                               page->mapping->host->i_ino,
@@ -2249,6 +2559,16 @@ static int btrfs_get_sb(struct file_system_type *fs_type,
                           btrfs_fill_super, mnt);
 }
 
+
+static int btrfs_getattr(struct vfsmount *mnt,
+                        struct dentry *dentry, struct kstat *stat)
+{
+       struct inode *inode = dentry->d_inode;
+       generic_fillattr(inode, stat);
+       stat->blksize = 256 * 1024;
+       return 0;
+}
+
 static struct file_system_type btrfs_fs_type = {
        .owner          = THIS_MODULE,
        .name           = "btrfs",
@@ -2298,6 +2618,7 @@ static struct address_space_operations btrfs_aops = {
 
 static struct inode_operations btrfs_file_inode_operations = {
        .truncate       = btrfs_truncate,
+       .getattr        = btrfs_getattr,
 };
 
 static struct file_operations btrfs_file_operations = {