Btrfs: Add file data csums back in via hooks in the extent map code

Signed-off-by: Chris Mason <chris.mason@oracle.com>

3 files changed, 193 insertions, 14 deletions

diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index a9c7419615b9..6053f9db0739 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -57,7 +57,7 @@ void extent_map_tree_init(struct extent_map_tree *tree,
 {
         tree->map.rb_node = NULL;
         tree->state.rb_node = NULL;
-        tree->fill_delalloc = NULL;
+        tree->ops = NULL;
         rwlock_init(&tree->lock);
         tree->mapping = mapping;
 }
@@ -93,6 +93,7 @@ struct extent_state *alloc_extent_state(gfp_t mask)
                 return state;
         state->state = 0;
         state->in_tree = 0;
+        state->private = 0;
         atomic_set(&state->refs, 1);
         init_waitqueue_head(&state->wq);
         return state;
@@ -1034,6 +1035,61 @@ int unlock_range(struct extent_map_tree *tree, u64 start, u64 end)
 }
 EXPORT_SYMBOL(unlock_range);
 
+int set_state_private(struct extent_map_tree *tree, u64 start, u64 private)
+{
+        struct rb_node *node;
+        struct extent_state *state;
+        int ret = 0;
+
+        write_lock_irq(&tree->lock);
+        /*
+         * this search will find all the extents that end after
+         * our range starts.
+         */
+        node = tree_search(&tree->state, start);
+        if (!node || IS_ERR(node)) {
+                ret = -ENOENT;
+                goto out;
+        }
+        state = rb_entry(node, struct extent_state, rb_node);
+        if (state->start != start) {
+                ret = -ENOENT;
+                goto out;
+        }
+        state->private = private;
+out:
+        write_unlock_irq(&tree->lock);
+        return ret;
+
+}
+
+int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private)
+{
+        struct rb_node *node;
+        struct extent_state *state;
+        int ret = 0;
+
+        read_lock_irq(&tree->lock);
+        /*
+         * this search will find all the extents that end after
+         * our range starts.
+         */
+        node = tree_search(&tree->state, start);
+        if (!node || IS_ERR(node)) {
+                ret = -ENOENT;
+                goto out;
+        }
+        state = rb_entry(node, struct extent_state, rb_node);
+        if (state->start != start) {
+                ret = -ENOENT;
+                goto out;
+        }
+        *private = state->private;
+out:
+        read_unlock_irq(&tree->lock);
+        return ret;
+}
+
 /*
  * searches a range in the state tree for a given mask.
  * If 'filled' == 1, this returns 1 only if ever extent in the tree
@@ -1185,12 +1241,13 @@ static int end_bio_extent_writepage(struct bio *bio,
 static int end_bio_extent_readpage(struct bio *bio,
                                    unsigned int bytes_done, int err)
 {
-        const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+        int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
         struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
         struct extent_map_tree *tree = bio->bi_private;
         u64 start;
         u64 end;
         int whole_page;
+        int ret;
 
         if (bio->bi_size)
                 return 1;
@@ -1208,6 +1265,11 @@ static int end_bio_extent_readpage(struct bio *bio,
                 if (--bvec >= bio->bi_io_vec)
                         prefetchw(&bvec->bv_page->flags);
 
+                if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) {
+                        ret = tree->ops->readpage_end_io_hook(page, start, end);
+                        if (ret)
+                                uptodate = 0;
+                }
                 if (uptodate) {
                         set_extent_uptodate(tree, start, end, GFP_ATOMIC);
                         if (whole_page)
@@ -1388,9 +1450,16 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page,
                         continue;
                 }
 
-                ret = submit_extent_page(READ, tree, page,
-                                         sector, iosize, page_offset, bdev,
-                                         end_bio_extent_readpage);
+                ret = 0;
+                if (tree->ops && tree->ops->readpage_io_hook) {
+                        ret = tree->ops->readpage_io_hook(page, cur,
+                                                          cur + iosize - 1);
+                }
+                if (!ret) {
+                        ret = submit_extent_page(READ, tree, page,
+                                                 sector, iosize, page_offset,
+                                                 bdev, end_bio_extent_readpage);
+                }
                 if (ret)
                         SetPageError(page);
                 cur = cur + iosize;
@@ -1462,7 +1531,7 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
                                                &delalloc_end,
                                                128 * 1024 * 1024);
         if (nr_delalloc) {
-                tree->fill_delalloc(inode, start, delalloc_end);
+                tree->ops->fill_delalloc(inode, start, delalloc_end);
                 if (delalloc_end >= page_end + 1) {
                         clear_extent_bit(tree, page_end + 1, delalloc_end,
                                          EXTENT_LOCKED | EXTENT_DELALLOC,
@@ -1528,12 +1597,17 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
                         continue;
                 }
                 clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS);
-                set_range_writeback(tree, cur, cur + iosize - 1);
-                ret = submit_extent_page(WRITE, tree, page,
-                                         sector, iosize, page_offset, bdev,
-                                         end_bio_extent_writepage);
+                ret = tree->ops->writepage_io_hook(page, cur, cur + iosize - 1);
                 if (ret)
                         SetPageError(page);
+                else {
+                        set_range_writeback(tree, cur, cur + iosize - 1);
+                        ret = submit_extent_page(WRITE, tree, page, sector,
+                                                 iosize, page_offset, bdev,
+                                                 end_bio_extent_writepage);
+                        if (ret)
+                                SetPageError(page);
+                }
                 cur = cur + iosize;
                 page_offset += iosize;
                 nr++;
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
index e91a2e9619ee..0eae6c420acb 100644
--- a/fs/btrfs/extent_map.h
+++ b/fs/btrfs/extent_map.h
@@ -6,12 +6,19 @@
 #define EXTENT_MAP_INLINE (u64)-2
 #define EXTENT_MAP_DELALLOC (u64)-1
 
+struct extent_map_ops {
+        int (*fill_delalloc)(struct inode *inode, u64 start, u64 end);
+        int (*writepage_io_hook)(struct page *page, u64 start, u64 end);
+        int (*readpage_io_hook)(struct page *page, u64 start, u64 end);
+        int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end);
+};
+
 struct extent_map_tree {
         struct rb_root map;
         struct rb_root state;
         struct address_space *mapping;
         rwlock_t lock;
-        int (*fill_delalloc)(struct inode *inode, u64 start, u64 end);
+        struct extent_map_ops *ops;
 };
 
 /* note, this must start with the same fields as fs/extent_map.c:tree_entry */
@@ -36,6 +43,10 @@ struct extent_state {
         wait_queue_head_t wq;
         atomic_t refs;
         unsigned long state;
+
+        /* for use by the FS */
+        u64 private;
+
         struct list_head list;
 };
 
@@ -89,4 +100,6 @@ int extent_commit_write(struct extent_map_tree *tree,
                         struct inode *inode, struct page *page,
                         unsigned from, unsigned to);
 int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end);
+int set_state_private(struct extent_map_tree *tree, u64 start, u64 private);
+int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private);
 #endif
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 483eea69b0a6..3ee6b2fadf58 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -52,6 +52,7 @@ static struct inode_operations btrfs_file_inode_operations;
 static struct address_space_operations btrfs_aops;
 static struct address_space_operations btrfs_symlink_aops;
 static struct file_operations btrfs_dir_file_operations;
+static struct extent_map_ops btrfs_extent_map_ops;
 
 static struct kmem_cache *btrfs_inode_cachep;
 struct kmem_cache *btrfs_trans_handle_cachep;
@@ -103,6 +104,90 @@ out:
         return ret;
 }
 
+int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end)
+{
+        struct inode *inode = page->mapping->host;
+        struct btrfs_root *root = BTRFS_I(inode)->root;
+        struct btrfs_trans_handle *trans;
+        char *kaddr;
+        int ret;
+        u64 page_start = page->index << PAGE_CACHE_SHIFT;
+        size_t offset = start - page_start;
+
+        mutex_lock(&root->fs_info->fs_mutex);
+        trans = btrfs_start_transaction(root, 1);
+        btrfs_set_trans_block_group(trans, inode);
+        kaddr = kmap(page);
+        btrfs_csum_file_block(trans, root, inode->i_ino,
+                              start, kaddr + offset, end - start + 1);
+        kunmap(page);
+        ret = btrfs_end_transaction(trans, root);
+        BUG_ON(ret);
+        mutex_unlock(&root->fs_info->fs_mutex);
+        return ret;
+}
+
+int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
+{
+        int ret = 0;
+        struct inode *inode = page->mapping->host;
+        struct btrfs_root *root = BTRFS_I(inode)->root;
+        struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+        struct btrfs_csum_item *item;
+        struct btrfs_path *path = NULL;
+        u64 private;
+
+        mutex_lock(&root->fs_info->fs_mutex);
+        path = btrfs_alloc_path();
+        item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
+        if (IS_ERR(item)) {
+                ret = PTR_ERR(item);
+                /* a csum that isn't present is a preallocated region. */
+                if (ret == -ENOENT || ret == -EFBIG)
+                        ret = 0;
+                private = 0;
+                goto out;
+        }
+        memcpy((char *)&private, &item->csum, BTRFS_CRC32_SIZE);
+        set_state_private(em_tree, start, private);
+out:
+        if (path)
+                btrfs_free_path(path);
+        mutex_unlock(&root->fs_info->fs_mutex);
+        return ret;
+}
+
+int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
+{
+        char csum[BTRFS_CRC32_SIZE];
+        size_t offset = start - (page->index << PAGE_CACHE_SHIFT);
+        struct inode *inode = page->mapping->host;
+        struct btrfs_root *root = BTRFS_I(inode)->root;
+        struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+        char *kaddr;
+        u64 private;
+        int ret;
+
+        ret = get_state_private(em_tree, start, &private);
+        kaddr = kmap_atomic(page, KM_IRQ0);
+        if (ret) {
+                goto zeroit;
+        }
+        ret = btrfs_csum_data(root, kaddr + offset, end - start + 1, csum);
+        BUG_ON(ret);
+        if (memcmp(csum, &private, BTRFS_CRC32_SIZE)) {
+                goto zeroit;
+        }
+        kunmap_atomic(kaddr, KM_IRQ0);
+        return 0;
+
+zeroit:
+        printk("btrfs csum failed ino %lu off %llu\n",
+               page->mapping->host->i_ino, (unsigned long long)start);
+        memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page);
+        kunmap_atomic(kaddr, KM_IRQ0);
+        return 0;
+}
 
 void btrfs_read_locked_inode(struct inode *inode)
 {
@@ -155,7 +240,7 @@ void btrfs_read_locked_inode(struct inode *inode)
         switch (inode->i_mode & S_IFMT) {
         case S_IFREG:
                 inode->i_mapping->a_ops = &btrfs_aops;
-                BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range;
+                BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
                 inode->i_fop = &btrfs_file_operations;
                 inode->i_op = &btrfs_file_inode_operations;
                 break;
@@ -1148,7 +1233,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
                 inode->i_op = &btrfs_file_inode_operations;
                 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
                                      inode->i_mapping, GFP_NOFS);
-                BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range;
+                BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
         }
         dir->i_sb->s_dirt = 1;
         btrfs_update_inode_block_group(trans, inode);
@@ -2286,7 +2371,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
                 inode->i_op = &btrfs_file_inode_operations;
                 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
                                      inode->i_mapping, GFP_NOFS);
-                BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range;
+                BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
         }
         dir->i_sb->s_dirt = 1;
         btrfs_update_inode_block_group(trans, inode);
@@ -2362,6 +2447,13 @@ static struct file_operations btrfs_dir_file_operations = {
 #endif
 };
 
+static struct extent_map_ops btrfs_extent_map_ops = {
+        .fill_delalloc = run_delalloc_range,
+        .writepage_io_hook = btrfs_writepage_io_hook,
+        .readpage_io_hook = btrfs_readpage_io_hook,
+        .readpage_end_io_hook = btrfs_readpage_end_io_hook,
+};
+
 static struct address_space_operations btrfs_aops = {
         .readpage       = btrfs_readpage,
         .writepage      = btrfs_writepage,