Btrfs: Less aggressive readahead on deletes

[linux-2.6-block.git] / fs / btrfs / disk-io.c

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 5accced2858a94f12af9092054f87d22c25f352c..ebb2db624fddab4c1ffce160322b78e9e0a01c81 100644 (file)
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -43,6 +43,8 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
 }
 #endif
 
+static struct extent_map_ops btree_extent_map_ops;
+
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
                                            u64 bytenr, u32 blocksize)
 {
@@ -159,7 +161,7 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
 {
        struct extent_map_tree *tree;
-       u64 start = page->index << PAGE_CACHE_SHIFT;
+       u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
        u64 found_start;
        int found_level;
        unsigned long len;
@@ -188,15 +190,47 @@ out:
        return 0;
 }
 
-static int btree_writepage(struct page *page, struct writeback_control *wbc)
+static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
 {
-       struct extent_map_tree *tree;
        struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
-       tree = &BTRFS_I(page->mapping->host)->extent_tree;
 
        csum_dirty_buffer(root, page);
+       return 0;
+}
+
+static int btree_writepage(struct page *page, struct writeback_control *wbc)
+{
+       struct extent_map_tree *tree;
+       tree = &BTRFS_I(page->mapping->host)->extent_tree;
        return extent_write_full_page(tree, page, btree_get_extent, wbc);
 }
+
+static int btree_writepages(struct address_space *mapping,
+                           struct writeback_control *wbc)
+{
+       struct extent_map_tree *tree;
+       tree = &BTRFS_I(mapping->host)->extent_tree;
+       if (wbc->sync_mode == WB_SYNC_NONE) {
+               u64 num_dirty;
+               u64 start = 0;
+               unsigned long thresh = 96 * 1024 * 1024;
+
+               if (wbc->for_kupdate)
+                       return 0;
+
+               if (current_is_pdflush()) {
+                       thresh = 96 * 1024 * 1024;
+               } else {
+                       thresh = 8 * 1024 * 1024;
+               }
+               num_dirty = count_range_bits(tree, &start, thresh, EXTENT_DIRTY);
+               if (num_dirty < thresh) {
+                       return 0;
+               }
+       }
+       return extent_writepages(tree, mapping, btree_get_extent, wbc);
+}
+
 int btree_readpage(struct file *file, struct page *page)
 {
        struct extent_map_tree *tree;
@@ -251,6 +285,7 @@ static int btree_writepage(struct page *page, struct writeback_control *wbc)
 static struct address_space_operations btree_aops = {
        .readpage       = btree_readpage,
        .writepage      = btree_writepage,
+       .writepages     = btree_writepages,
        .releasepage    = btree_releasepage,
        .invalidatepage = btree_invalidatepage,
        .sync_page      = block_sync_page,
@@ -320,7 +355,7 @@ int wait_on_tree_block_writeback(struct btrfs_root *root,
 }
 
 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
-                       struct btrfs_root *root,
+                       u32 stripesize, struct btrfs_root *root,
                        struct btrfs_fs_info *fs_info,
                        u64 objectid)
 {
@@ -330,6 +365,7 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
        root->sectorsize = sectorsize;
        root->nodesize = nodesize;
        root->leafsize = leafsize;
+       root->stripesize = stripesize;
        root->ref_cows = 0;
        root->fs_info = fs_info;
        root->objectid = objectid;
@@ -358,7 +394,8 @@ static int find_and_setup_root(struct btrfs_root *tree_root,
        u32 blocksize;
 
        __setup_root(tree_root->nodesize, tree_root->leafsize,
-                    tree_root->sectorsize, root, fs_info, objectid);
+                    tree_root->sectorsize, tree_root->stripesize,
+                    root, fs_info, objectid);
        ret = btrfs_find_last_root(tree_root, objectid,
                                   &root->root_item, &root->root_key);
        BUG_ON(ret);
@@ -395,8 +432,8 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
        }
 
        __setup_root(tree_root->nodesize, tree_root->leafsize,
-                    tree_root->sectorsize, root, fs_info,
-                    location->objectid);
+                    tree_root->sectorsize, tree_root->stripesize,
+                    root, fs_info, location->objectid);
 
        path = btrfs_alloc_path();
        BUG_ON(!path);
@@ -410,6 +447,7 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
        read_extent_buffer(l, &root->root_item,
               btrfs_item_ptr_offset(l, path->slots[0]),
               sizeof(root->root_item));
+       memcpy(&root->root_key, location, sizeof(*location));
        ret = 0;
 out:
        btrfs_release_path(root, path);
@@ -501,6 +539,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
        u32 nodesize;
        u32 leafsize;
        u32 blocksize;
+       u32 stripesize;
        struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
                                                 GFP_NOFS);
        struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
@@ -529,6 +568,8 @@ struct btrfs_root *open_ctree(struct super_block *sb)
        fs_info->tree_root = tree_root;
        fs_info->extent_root = extent_root;
        fs_info->sb = sb;
+       fs_info->mount_opt = 0;
+       fs_info->max_extent = (u64)-1;
        fs_info->btree_inode = new_inode(sb);
        fs_info->btree_inode->i_ino = 1;
        fs_info->btree_inode->i_nlink = 1;
@@ -537,6 +578,8 @@ struct btrfs_root *open_ctree(struct super_block *sb)
        extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
                             fs_info->btree_inode->i_mapping,
                             GFP_NOFS);
+       BTRFS_I(fs_info->btree_inode)->extent_tree.ops = &btree_extent_map_ops;
+
        extent_map_tree_init(&fs_info->free_space_cache,
                             fs_info->btree_inode->i_mapping, GFP_NOFS);
        extent_map_tree_init(&fs_info->block_group_cache,
@@ -549,8 +592,12 @@ struct btrfs_root *open_ctree(struct super_block *sb)
                             fs_info->btree_inode->i_mapping, GFP_NOFS);
        fs_info->do_barriers = 1;
        fs_info->closing = 0;
-
+       fs_info->total_pinned = 0;
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+       INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
+#else
        INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
+#endif
        BTRFS_I(fs_info->btree_inode)->root = tree_root;
        memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
               sizeof(struct btrfs_key));
@@ -568,7 +615,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
                goto fail_iput;
        }
 #endif
-       __setup_root(512, 512, 512, tree_root,
+       __setup_root(512, 512, 512, 512, tree_root,
                     fs_info, BTRFS_ROOT_TREE_OBJECTID);
 
        fs_info->sb_buffer = read_tree_block(tree_root,
@@ -591,9 +638,11 @@ struct btrfs_root *open_ctree(struct super_block *sb)
        nodesize = btrfs_super_nodesize(disk_super);
        leafsize = btrfs_super_leafsize(disk_super);
        sectorsize = btrfs_super_sectorsize(disk_super);
+       stripesize = btrfs_super_stripesize(disk_super);
        tree_root->nodesize = nodesize;
        tree_root->leafsize = leafsize;
        tree_root->sectorsize = sectorsize;
+       tree_root->stripesize = stripesize;
        sb_set_blocksize(sb, sectorsize);
 
        i_size_write(fs_info->btree_inode,
@@ -720,8 +769,18 @@ int close_ctree(struct btrfs_root *root)
 
        btrfs_free_block_groups(root->fs_info);
        del_fs_roots(fs_info);
+
+       filemap_write_and_wait(fs_info->btree_inode->i_mapping);
+
+       extent_map_tree_empty_lru(&fs_info->free_space_cache);
+       extent_map_tree_empty_lru(&fs_info->block_group_cache);
+       extent_map_tree_empty_lru(&fs_info->pinned_extents);
+       extent_map_tree_empty_lru(&fs_info->pending_del);
+       extent_map_tree_empty_lru(&fs_info->extent_ins);
        extent_map_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->extent_tree);
+
        truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
+
        iput(fs_info->btree_inode);
 #if 0
        while(!list_empty(&fs_info->hashers)) {
@@ -831,3 +890,7 @@ int btrfs_read_buffer(struct extent_buffer *buf)
        return read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
                                        buf, 0, 1);
 }
+
+static struct extent_map_ops btree_extent_map_ops = {
+       .writepage_io_hook = btree_writepage_io_hook,
+};